- Wednesday, 10 February 2010
- INQ Mobile
- RSS
BathOS: We've heard of it. Your mobile phone works while you are bathing
A look at the Nvidia GT300 architecture
THERE'S A LOT of fake news going around about the upcoming GPUish chip called the GT300. Let's clear some air on this Larrabee-lite architecture.
First of all, almost everything you have heard about the two upcoming DX11 architectures is wrong. There is a single source making up news, and second rate sites are parroting it left and right. The R870 news is laughably inaccurate, and the GT300 info is quite curious too. Either ATI figured out a way to break the laws of physics with memory speed and Nvidia managed to almost double its transistor density - do the math on purported numbers, they aren't even in the ballpark - or someone is blatantly making up numbers.
That said, lets get on with what we know, and delve into the architectures a bit. The GT300 is going to lose, badly, in the GPU game, and we will go over why and how.
First a little background science and math. There are three fabrication processes out there that ATI and Nvidia use, all from TSMC, 65nm, 55nm and 40nm. They are each a 'half step' from the next, and 65nm to 40nm is a full step. If you do the math, the shrink from 65nm to 55nm ((55 * 55) / (65 *65) ~= 0.72) saves you about 1/4 the area, that is, 55nm is 0.72 of the area of 65nm for the same transistor count. 55nm shrunk to 40nm gives you 0.53 of the area, and 65nm shrunk to 40nm gives you 0.38 of the area. We will be using these later.
Second is the time it takes to do things. We will use the best case scenarios, with a hot lot from TSMC taking a mere six weeks, and the time from wafers in to boards out of an AIB being 12 weeks. Top it off with test and debug times of two weeks for first silicon and one week for each subsequent spin. To simplify rough calculations, all months will be assumed to have 4 weeks.
Okay, ATI stated that it will have DX11 GPUs on sale when Windows 7 launches, purportedly October 23, 2009. Since this was done in a financial conference call, SEC rules applying, you can be pretty sure ATI is serious about this. Nvidia on the other hand basically dodged the question, hard, in its conference call the other day.
At least you should know why Nvidia picked the farcical date of October 15 for its partners. Why farcical? Lets go over the numbers once again.
According to sources in Satan Clara, GT300 has not taped out yet, as of last week. It is still set for June, which means best case, June 1st. Add six weeks for first silicon, two more for initial debug, and you are at eight weeks, minimum. That means the go or no-go decision might be made as early as August 1st. If everything goes perfectly, and there is no second spin required, you would have to add 90 days to that, meaning November 1st, before you could see any boards.
So, if all the stars align, and everything goes perfectly, Nvidia could hit Q4 of 2009. But that won't happen.
Why not? There is a concept called risk when doing chips, and the GT300 is a high risk part. GT300 is the first chip of a new architecture, or so Nvidia claims. It is also going to be the first GDDR5 part, and moreover, it will be Nvidia's first 'big' chip on the 40nm process.
Nvidia chipmaking of late has been laughably bad. GT200 was slated for November of 2007 and came out in May or so in 2008, two quarters late. We are still waiting for the derivative parts. The shrink, GT206/GT200b is technically a no-brainer, but instead of arriving in August of 2008, it trickled out in January, 2009. The shrink of that to 40nm, the GT212/GT200c was flat out canceled, Nvidia couldn't do it.
The next largest 40nm part, the GT214 also failed, and it was redone as the GT215. The next smallest parts, the GT216 and GT218, very small chips, are hugely delayed, perhaps to finally show up in late June. Nvidia can't make a chip that is one-quarter of the purported size of the GT300 on the TSMC 40nm process. That is, make it at all, period - making it profitably is, well, a humorous concept for now.
GT300 is also the first DX11 part from the green team, and it didn't even have DX10.1 parts. Between the new process, larger size, bleeding-edge memory technology, dysfunctional design teams, new feature sets and fab partners trashed at every opportunity, you could hardly imagine ways to have more risk in a new chip design than Nvidia has with the GT300.
If everything goes perfectly and Nvidia puts out a GT300 with zero bugs, or easy fix minor bugs, then it could be out in November. Given that there is only one GPU that we have heard of that hit this milestone, a derivative part, not a new architecture, it is almost assuredly not going to happen. No OEM is going to bet their Windows 7 launch vehicles on Nvidia's track record. They remember the 9400, GT200, and well, everything else.
If there is only one respin, you are into 2010. If there is a second respin, then you might have a hard time hitting Q1 of 2010. Of late, we can't think of any Nvidia product that hasn't had at least two respins, be they simple optical shrinks or big chips.
Conversely, the ATI R870 is a low risk part. ATI has a functional 40nm part on the market with the RV740/HD4770, and has had GDDR5 on cards since last June. Heck, it basically developed GDDR5. The RV740 - again, a part already on the market - is rumored to be notably larger than either the GT216 or 218, and more or less the same size as the GT215 that Nvidia can't seem to make.
DX11 is a much funnier story. The DX10 feature list was quite long when it was first proposed. ATI dutifully worked with Microsoft to get it implemented, and did so with the HD2900. Nvidia stomped around like a petulant child and refused to support most of those features, and Microsoft stupidly capitulated and removed large tracts of DX10 functionality.
This had several effects, the most notable being that the now castrated DX10 was a pretty sad API, barely moving anything forward. It also meant that ATI spent a lot of silicon area implementing things that would never be used. DX10.1 put some of those back, but not the big ones.
DX11 is basically what DX10 was meant to be with a few minor additions. That means ATI has had a mostly DX11 compliant part since the HD2900. The R870/HD5870 effectively will be the fourth generation DX11 GPU from the red team. Remember the tessellator? Been there, done that since 80nm parts.
This is not to say that is will be easy for either side, TSMC has basically come out and said that its 40nm process basically is horrid, an assertion backed up by everyone that uses it. That said, both the GT300 and R870 are designed for the process, so they are stuck with it. If yields can't be made economically viable, you will be in a situation of older 55nm parts going head to head for all of 2010. Given Nvidia's total lack of cost competitiveness on that node, it would be more a question of them surviving the year.
That brings us to the main point, what is GT300? If you recall Jen-Hsun's mocking jabs about Laughabee, you might find it ironic that GT300 is basically a Larrabee clone. Sadly though, it doesn't have the process tech, software support, or architecture behind it to make it work, but then again, this isn't the first time that Nvidia's grand prognostications have landed on its head.
The basic structure of GT300 is the same as Larrabee. Nvidia is going to use general purpose 'shaders' to do compute tasks, and the things that any sane company would put into dedicated hardware are going to be done in software. Basically DX11 will be shader code on top of a generic CPU-like structure. Just like Larrabee, but from the look of it, Larrabee got the underlying hardware right.
Before you jump up and down, and before all the Nvidiots start drooling, this is a massive problem for Nvidia. The chip was conceived at a time when Nvidia thought GPU compute was actually going to bring it some money, and it was an exit strategy for the company when GPUs went away.
It didn't happen that way, partially because of buggy hardware, partially because of over-promising and under-delivering, and then came the deathblows from Larrabee and Fusion. Nvidia's grand ambitions were stuffed into the dirt, and rightly so.
Nvidia Investor Relations tells people that between five to ten per cent of the GT200 die area is dedicated to GPU compute tasks. The GT300 goes way farther here, but let's be charitable and call it 10 per cent. This puts Nvidia at a 10 per cent areal disadvantage to ATI on the DX11 front, and that is before you talk about anything else. Out of the gate in second place.
On 55nm, the ATI RV790 basically ties the GT200b in performance, but does it in about 60 per cent of the area, and that means less than 60 per cent of the cost. Please note, we are not taking board costs into account, and if you look at yield too, things get very ugly for Nvidia. Suffice it to say that architecturally, GT200 is a dog, a fat, bloated dog.
Rather than go lean and mean for GT300, possibly with a multi-die strategy like ATI, Nvidia is going for bigger and less areally efficient. They are giving up GPU performance to chase a market that doesn't exist, but was a nice fantasy three years ago. Also, remember that part about ATI's DX10 being the vast majority of the current DX11? ATI is not going to have to bloat its die size to get to DX11, but Nvidia will be forced to, one way or another. Step 1) Collect Underpants. Step 2) ??? Step 3) Profit!
On the shrink from 55nm to 40nm, you about double your transistor count, but due to current leakage, doing so will hit a power wall. Let's assume that both sides can double their transistor counts and stay within their power budgets though, that is the best case for Nvidia.
If AMD doubles its transistor count, it could almost double performance. If it does, Nvidia will have to as well. But, because Nvidia has to add in all the DX11 features, or additional shaders to essentially dedicate to them, its chips' areal efficiency will likely go down. Meanwhile, ATI has those features already in place, and it will shrink its chip sizes to a quarter of what they were in the 2900, or half of what they were in the R770.
Nvidia will gain some area back when it goes to GDDR5. Then the open question will be how wide the memory interface will have to be to support a hugely inefficient GPGPU strategy. That code has to be loaded, stored and flushed, taking bandwidth and memory.
In the end, what you will end up with is ATI that can double performance if it choses to double shader count, while Nvidia can double shader count, but it will lose a lot of real world performance if it does.
In the R870, if you compare the time it takes to render 1 Million triangles from 250K using the tesselator, it will take a bit longer than running those same 1 Million triangles through without the tesselator. Tesselation takes no shader time, so other than latency and bandwidth, there is essentially zero cost. If ATI implemented things right, and remember, this is generation four of the technology, things should be almost transparent.
Contrast that with the GT300 approach. There is no dedicated tesselator, and if you use that DX11 feature, it will take large amounts of shader time, used inefficiently as is the case with general purpose hardware. You will then need the same shaders again to render the triangles. 250K to 1 Million triangles on the GT300 should be notably slower than straight 1 Million triangles.
The same should hold true for all DX11 features, ATI has dedicated hardware where applicable, Nvidia has general purpose shaders roped into doing things far less efficiently. When you turn on DX11 features, the GT300 will take a performance nosedive, the R870 won't.
Worse yet, when the derivatives come out, the proportion of shaders needed to run DX11 will go up for Nvidia, but the dedicated hardware won't change for ATI. It is currently selling parts on the low end of the market that have all the "almost DX11" features, and is doing so profitably. Nvidia will have a situation on its hands in the low end that will make the DX10 performance of the 8600 and 8400 class parts look like drag racers.
In the end, Nvidia architecturally did just about everything wrong with this part. It is chasing a market that doesn't exist, and skewing its parts away from their core purpose, graphics, to fulfill that pipe dream. Meanwhile, ATI will offer you an x86 hybrid Fusion part if that is what you want to do, and Intel will have Larrabee in the same time frame.
GT300 is basically Larrabee done wrong for the wrong reasons. Amusingly though, it misses both of the attempted targets. R870 should pummel it in DX10/DX11 performance, but if you buy a $400-600 GPU for ripping DVDs to your Ipod, Nvidia has a card for you. Maybe. Yield problems notwithstanding.
GT300 will be quarters late, and without a miracle, miss back to school, the Windows 7 launch, and Christmas. It won't come close to R870 in graphics performance, and it will cost much more to make. This is not an architecture that will dig Nvidia out of its hole, but instead will dig it deeper. It made a Laughabee. µ
Share this:
ShareBusiness Intelligence Software
Customer Relationship Management (CRM)
Enterprise Accounting Software
Enterprise Resource Planning (ERP)
Enterprise Systems Management
waiting for the tirade of angry fanbou posts calling for Charlie's head on a spike.
Looks like Nvidia is a bit screwed.
I hope Nvidia survivie and come back fitter because if we only have ATI and Intel then we are going to get seriously shafted for price-performance. Intel have a terrible track record for this, and ATI's isnt much better.
I reject this reality and substitute my own. NVidia will win this next generation!!!!
I'm looking forward to more Charlie Versus series, maybe it can expand to Charlie Vs Google, Demerjian Vs Branson, Charlie Vs Robots or Charlie Vs Bears
not interested in what this "writer" has to say about nv.
This make logical since and would put Nvidia behind Ati.
LETS GO AMD/ATI
approach to this. I remember articles being written about how the 2900 would be the bees knees and it's architecture didn't work well until you went for high bandwidth usage on memory. The HD3850 and HD3870 and the follow up part of the 4800 series did a lot to correct this. Just remember though, NVidia does still have really fast parts, that aren't terrible on the power consumption. Just because their next gen product isn't shaping up to what everyone wants doesn't mean that it's going to suck. I will wait until both products come out and then I will read reviews (hopefully unbiased ones), and then I will purchase whichever card is right for me. Being that I don't have a huge monitor on my desktop, I'm not terribly concerned with the 1920x1080 resolution scores as my monitor doesn't support that res. If the card performs better than my current card, has low power usage (come on lower than what I have right now with the 4850!), and has the price range I can afford (not more the $300), I will buy it and be happy. I have owned both cards over the years and I have had good luck with NVidia and ATI parts, so I have nothing to complain about. Here's to hoping that the next gen cards are better in every way!
OMG, what will happedn now? will Nvidia gio out of business, retroactively, meaning they will never even have existed in the first place!?
I have always admired Charlie's sources and resources (e.g. access to independent semiconductor labs that are on-par to what most in-house labs can offer) especially with regard to nVidia. Its true that there are no secrets in semiconductor, especially in Taiwan where TSMC and UMC practically know what the other is doing at any one time, but it is very seldom that we get to hear those outside the Taiwan circle. A big thanks on that.
The other thing is the clear explanation of semiconductor physics (I suppose that it is still true that an Engineering education does prepare one well for many things). I wish some of my colleagues can be that well verse.
As for GPGPU, well, most people I know are not interested in that - the general comment from the people I know (thankfully no movie or TV uberpirate among them) is that a dual core is more than adequate, a quad core is overkill and why would anyone wants the GPU to do general compute as well?
Hehe, Charlie vs Chuck Norris, the ultimate showdown!
That aside, where are the graphs? The only way to impress your point upon the people who matter is with shiny graphs and slideshows.
Why are you saying that Charlie? Remember Nvidia first chipset actually is for AMD? Intel have been bullied nvidia. Nvidia is only like bastard child of AMD. Nvidia still and will always support for future generations of AMD processor. And, the fact is the adopted child (ATI) and the bastard child (NVIDIA) have been doing nicely in Graphics Card Market (are you still remember that booth have been accused for anti trust case?).
You are a silly writer about them and as SPIntel you are the most intelligent Intel Inside.
Any chance of a video interview with Jen-Hsun and Charlie sometime ?:P
You wanna learn something from someone or some people who ACTUALLY HAVE A CLUE AS TO WHAT THEY ARE TALKING ABOUT, go to www.beyond3d.com You'll learn two things, charlie is a complete and utter farce and joke and his head is so far up AMDs ass that he actually tastes Ruizs' food for him before he gets to eat it. The seconded thing you'll learn is that Charlie is should be fired for his out right his out right lies of "what he knows".
Charlie, if your brave enough, I dare you to come to beyond3d.com and spout this same nonsence there and defend it.
Dont worry folks, he wont cause he knows he can't.
"It won't come close to R870 in graphics performance,"
Same qoute, but only 4 years after the R600 where he said the same thing back then. Anyone wanna care to guess how correct he was then?
Here is a little hint, HE WASN"T EVEN REMOTELY RIGHT AT ANY LEVEL!
So basicly, Charlie is throwing sheit at nvidia and hoping one sticks eventually. And ati fanboys are eating it up. I have owned cards from both and they are roughly the same. You take some and you give some... but roughly the same. Well, i guess time will tell if nvidia is going to fuc this one up or come out a winner...
Glad to see there still are some smart/honest reporter/writers out there! No other site (posting rumors of Nvidia being first to market with DX11) seem to have point these two known facts...
I have heard rumors how MS neutered DX10 so Nvidia's parts are compliant but that aside.
From early previews of DX11, it looks like it will indeed be a super-set of DX10.1 and in turn DX10.1 is a super-set of DX10. And AMD/ATI has 2 generation of DX10.1 (Radeon HD 3xxx, 4xxx) parts out. Since Charlie is couting DX10 parts, it is "3", I guess (HD 2xxx). By any account AMD has the advantage here.
And AMD does have a 40nm part out already (HD 4770).
Nvidia could have started earlier than AMD's DX11 generation or threw down more resources but then they need extra time and resource just catch up...
GPGPU aside, I seriously doubt the back-to-school target I have been hearing. Christmas? Maybe? We'll know soon enough when June rounds around. :)
Charlie does seem like ranting now and then but he has been right so far so my money (for a small snack) is on him. :)
I really like you charlie and I think you have a vast amount of knowledge and a great instinct with regard to these parts of this industry but I think that there is a decline in your writing of these articles.
there is seldom a sentence in this article that does not contain a harshly demeaning attack. I don't even care about nvidia and love my ATI card - It isn't a case of an nvidia fan - It's just that halfway reading your articles in the past months I start getting the feeling that I am in a blog and I am not reading an article of news.
I can see you have a personal spite with nvidia and you may as well - I just think you should try making your articles less personal and redundant - for the sake of us getting good news from a person who has the knowledge and skill required to do so.
Thanks and good luck,
Roy
infantile rubbish as usual
the technological process ( e.g. 65 nm, 55nm, 40 nm) refers to the width of the transistor channel. the length doesn't necessarily have to respect the same ratio over different technological processes.
this being said the math at the being of the article isn't perfect but it's close to the real figures
First of all, I want to say that i'm not a Nvidia fanboy, I have a HD4890 and i'm happy with it. But I am an electronic engineer and know a thing or two about this. To the writer: In the article you say that we know nothing about GT300 but here you are knowing all the details about it and how it is like Larrabee but worse and that it doesnt have hardware tesselators. In the DX11 specification, it says that the GPU MUST HAVE hardware implemented tesselators, so Nvidia must have them to say that it supports DX11. And it can't be like Larrabee because it's shaders are a enhanced version of the GT200 and have nothing in common with the x86 modified CPUs that Larrabee has. Besides, Nvidia's GPU have been able to execute GPGPU code since G80, and the gaming performance was great (or you would say that the 8800GTX was a failure too??) If the news about this chip are really true, then Nvidia's GT300 will have 512 shader processors and 512bit wide GDDR5 memory interface. That will be at least twice as fast compared to GT200b with the same die area. You may be right about the launch date being delayed maybe to Q1 2010 but performancewise i'm sure it'll be great. Is because articles like this that The Inquirer has lost the credibility and the respect of the readers. Really pathetic
Lol u said go to www.beyond3d.com
ok i went their and i saw
Latest Reviews with
# AMD R6xx: Image Quality Analysis w/ AA focus
# Sapphire Ultimate Radeon HD 2600 XT - Early Peek
# AMD Radeon HD 3870 and 3850 First Look
Really looks like beyond3d is really good up to date site.
I must say that I'm impressed by this article. Charlie keeps getting better and better at infusing every single word with bitterness and hatred for the green goblin.
While it's probably true that Nvidia will continue pushing GPGPU and AMD will continue to focus more narrowly on just 3D performance Charlie doesn't have a clue how it all turns out in the end. But at least now we all know how he wishes it turns out - with Nvidia burned to ashes.
With regard to the hogwash about software emulation it's unfortunate that Charlie didn't share his wealth of technical insights that informed this opinion that Nvidia's architecture will be inferior to Larrabee. I'm sure that would be an amusing read!
PhysX's byproduct, the performance penalty on the GT200 is huge, something with threads and doing two things at the same time. The dedicated Badaboom is working fine though. It would be somewhat of an irony if all this extra performance is to be the reason for an untimely demise of Nvidia; giving Intel a view for another monopoly.
If there would be one GPGPU believer it should've been AMD. I know Intel is.
I always appreciate when a writer puts together the pieces. An interesting read with a very solid opinion behind it. Only a fool would write off what was said here. Charlie literally explains what he knows and then explains how he believes things will happen. It's spelled out for you.
To those questioning the validity of his information... This is TheInquirer.net. A site with mixed news and rumors. You take what you read here with a grain of salt and enjoy it.
Thank you for real journalism.
Remember the INQ promoting the 2900HD aginst nv's counterpart then? How about the INQ promoting the 3xxxHD series? It was the first in everything... GDDR4, DX10.1, etc...
When it was released, it was a hugely underpowered power-hungry card.
This article don't write any news and appear to be an ati fanboy comment. With errors too. For example:
-9400 and g200 problem for producers... (g200 had any problem)
-nvidia not ready for windows seven...
Other sources claim nvidia to be ahead with 7, with stable drivers ready from the first beta
-40nm tmsc problem...ati and nvidia will suffer in the same way
If you go GPU to GPU, 4890 to 285 as an example, ATI hasn't been able to catch up yet in performance. The difference isn't overlarge (not like the 6/7 Nvidia heyday or the ATI 9700 heyday) so it really comes down to whether you want the fastest, or close but cheaper. The whining of the ATI fanboi's notwithstanding, and for the good of graphical computing, it is fortunate that Nvidia is still in the game.
I'm not too interested in the ramblings of a bunch of fanbois, salivating like pavlov's dogs after a bell, completely ignorant of business and technology. Those of you on both sides are ridiculous.
As far as the GT300, from my info (and I fed Fuad a bit of info in my earlier years) I think Charlie got info from a monkey with some outdated knowledge. Happens, I know, but it's no more reliable than his recent Tegra info.
I really think that Tegra itself is a good idea, if the companies implement it well. There is a fine balance with it, but I think it could easily outdo the iPod touch. The real problem is the reference is too expensive (IMHO), but, the capabilities are greater. Still, expense is a heavy factor in consumer purchasing. Oh and they've got to find a better touchscreen. I'm not saying Synaptics is incompetent, but there is a continuing issue and the reference uses that. I think they'd be better off to at least try someone else even if only to put pressure on the current.
The GT300 will stopgap beat the RV870 at a guess, but not by much. One thing I've learned about AMD/ATI is they manage to feed victory to the jaws of defeat. Nvidia isn't much better, however, they tend to focus more on brute force than elegance. Then again, most ATI solutions are often also far from elegant, regardless of what a columnist thinks. Why will the GT300 probably still win? (again, not by much) For the same reason that ATI doesn't win right now, nobody really has the software in the pipeline to fully utilise DX11. You can quote all sorts of archaic functions but, just like when Nvidia did it to ATI when the 6 series came out, it doesn't make it magically faster. Oh it'll help in the "checkbox" wars. AMD/ATI's OEM shipments in graphics might tick up a few tenths of a percent from it, don't get me wrong, but overall it won't amount to anything for the consumer until after the refresh (if then). Heck, programmers are just getting a good handle on the current "castrated DX10" as is.
As far as Jen's comment on Intel's upcoming part, I don't agree with it all. He's defusing a potential drag with bluster, but not poorly placed bluster. Intel has shown itself to be incapable of graphics in gaming terms. The only card that really made a splash was the i740, and we who remember know Intel didn't really design it. Heck, we know Intel didn't really bother to keep it up to date! (The base design of the GMA series is the i740)
Interesting mix of truth, fiction, and outright speculation Charlie. I suggest updating your sources (or at least vetting them better) however the AMD fanbois here will fellate you and the Nvidia fanbois will flame you, driving up page hits I guess.
I guess that's the important thing anyway, the truth be damned. I've still not seen Charlie announce he was wrong again and again. Even Fuad would admit, grudgingly mind you, when he was wrong. I still haven't seen a single chip ATI solution that beats a single chip Nvidia solution across the board. Then again, to the Nvidia fanboi chagrine, Nvidia hasn't exactly distinguished itself either...
For you fanbois, that's because what Charlie often promotes is impossible fluff based on well timed leaks. Truth? He believes in page hits. Fanbois always have this problem of identifying with their purchases. Kinda like the Blu-ray / HD-DVD "format war" (really a patent war with Toshiba vs the rest of the CE industry) where each side's fanbois fellated their own side and flamed the other.
@interested_party:
Yeah, whenever one truly dominates the other the prices go stratospheric. Should Nvidia (or ATI) ever truly fail, the consumer will pay for it. Why? Who's left? Intel? (snicker) Matrox? (guffaw) Should only one of these two competitors remain, that competitor will crush us with mega-prices.
Well, if even half of what Charlie says turns out to be true, then NV is in some seriously deep shit. Personally, I think NV can pull off a near perfect execution of their future chip, if they really put some effort into it. But I don't know if that's going to be enough to compete, considering the GT300's concept. For sure, it's going to be an interesting winter for GPU fans.
Oh charlie you certainly make for some interesting reading indeed :) Your actually right though for all intents and purposes. My friend is an engineer that works at TSMC and he said there is no way they will have Nvidia's tape out of the GT300 by june simply because they have heard nothing from yet at all when by now they definitely should of. Oh by the way TSMC are shitting themselves because global foundries is going to smash them in 32nm TSMC is huge problems there. GF are right on the mark with 32nm. Everyone negates to factor in the what the foundries are doing.
Excellent article, but it doesn't take into account the fact that since architectures are going to change with the new generation, that also will affect performance, not just the transistor count. Still, Nvidia are in a big trouble it seems. Intel is also all set to launch it's Larrabee (or Laughabee, and Jen-Hsun would call it, but we'll see about that). I'd like to see a detailed analysis of Larrabee's performance.
If I were Nvidia I'd go out and hire Manny Pacquiao right NOW.
He's champion across 4 weight divisions and consistently performs beyond expectations.
Sounds like what they need...
I am confident Nvidia can manage to get G300 out of the door before xmass if that is their goal. Maybe a overvolted, leaking monster with a 3-slot cooler but it would do its task, make Nvidia the Q4 champ, trampling RV870 in performance.
Charlie is always fun to read for me, difficult point is understanding the actual "leaks" from the speculation.
The "do it in software" way GT300 seems to follow often backfired (let's remember RV670 AA performance) but offers more flexibility. "software" since it is actually done in hardware, simply by non dedicated parts.
Fun part is that GT300 starts from GT200 and goes a bit towards RV7xx and a bit towards Larrabee.
Hmmm how come its best "outdated" cards beat ATI's "4th generation" chips by large margin in idle power consumption (much more important than load) and are also faster.
Good article and analysis.
Looks like nVidia need to drop their arrogance, and get serious and change their ways or they are down the toilet for good. Can't say i'll miss them - i've been a victim of bumpgate myself.
I'm a ATI fanboy...but i'm also a realistic person...and i refuse to think that all those smart people at NV didn't learn one bit from the past...
We'll have to see if you are right Charlie...and i do hope you are not...competition is what the industry needs...:)
And i do tend to think...INQ loves ATI...witch is not a bad thing...:P
This is all pure speculation presented by known NVidia hateboy Charlie Demerjan, yet the last paragraph sounds like all this are known facts that have happened already.
I wish NVidia to survive and prosper as much as I wish AMD/ATI to do so. We need competition in the CPU and GPU markets.
I won't go as far as I would telling Charlie what I have become to think of him, but someone should make this obnoxious NVidia hating clown shut up.
Btw, I am using exclusively ATI graphics hardware since the days of the Radeon 9700 and never had problems with it, so I love it. Yet I don't want to see NVidia go down.
Or maybe an @idiot.
The nVidia architecture is far better than Ati then.
If ATI has hardware tesselator, what happens to that hardware when there is no need to tesselate in the graphics pipeline? It will be standing around.
An GPGPU is gaining momentum. Just look at Tokio Institute of Technology.
Atidiot.
Yes ... it is official.
Duke Nukem Forever has been designed from the ground up to run at 150FPS on a GT300.
It will also max out Crysis.
I checked at:
www.beyond3d.com
P.S. Charlie is god.
that did not need a respin?
Thank you very much Charlie, keep them coming ;-)
Ed6
So basically what the author is saying, nvidia just managed to make themselves an Itanium? Blazingly fast, with huge theoretical computational power, but terrible power characteristic and poor performance because of the heavy use of software emulation.
Orignally Larrabe Was Scorned as cpu, then migrated to integrated gpu. Now What?
Larrabee is moving to germany, for mere $6 million. Hide Cash & Program. If truth be known, Intel Might Fall. If Larrabee comes out, it will be stick on name from another hard worked team, like cpugpu, team. NOT Larrabee. Same Story, Only Confirmed.
criminal Manipulation of Funded Developement for 30 years left larrabee with absolutely NO Clue as to its Whereabouts.
Charlie does it again. Don't listen to the haters, your articles are awesome. I don't get half the hard figures OR enjoyment from other articles. Keep up the good work.
Don't know if you bother to read the comments Charlie but keep it up, your ranting brings me amusement!
Has it been confirmed that DX10 functionality initially in the specification (but subsequently removed by Microsoft after Nvidia objected, per Charlie) was still incorporated into ATI's 3xxx and 4xxx series GPUs? Charlie indicated DX10.1 added back some of the functionality that had been dropped from DX10 following Nvidia objections, but not the important items.
I have a hard time believing ATI would continue to "waste" significant die space and transistors on functionality that wasn't going to be supported by software vendors in its 3xxx and 4xxx series GPUs. So, if ATI didn't include the dropped DX10 functionality that has (apparently) been reinstated in DX11, it will have to add transistors and die space to its R8xx GPUs.
Just a thought.
I think the main major piece of functionality he's talking about is the hardware tessellation unit, which was indeed incorporated into the 3800 and 4800 series. That said, maybe you're on to something, the transistor count from R600 to RV670 did drop by about 44 million transistors. I guess 44 million isn't too much to be worried about, though, given we're nearing 1 billion with RV770 anyway, and are probably looking at more like 1.8 or so with RV870.
Charlie. Make a fan club! You really have a LOT of Charlinian Fanboy here. Don’t be modest.
Compete to win Charles hart.
The ATIdiot Vs the NVidiots lol.
And yes the First NV chipset was for AMD and was SUPER crappy (The chipset NF2 was good).
To everyone that is complaning that this article is no news, that this is not a proper way of telling news...
What the f..k !? Are you all blind ???
On the second line of the article it reads: "Analysis Compromised by wrong vision"...
"Analysis" !!!! Do you know the word? Do you know what it means??? bahhhhhhh
Charlie, thank you for sharing your thoughts... keep'em coming !
ATi just rules.
Just like your old friend Fuad. http://www.fudzilla.com/index.php?option=com_content&task=view&id=13689&Itemid=1 That's my uneducated guess. Good to know GT300 is a uber GPGPU chip. Should be nice for Win7 or Snow Leopard or anything OpenCL. Don't need Crysis R800 anyway, I'd rather prefer e. g. Starcraft II. Well informed sources (me and myself) tell us we are buying Nvidia stock. Cheers
That was a very good reading.Charlie wasn't wrong very often like Fuad when it came about news on nVIDIA side.
Remember when FUAD said rv790 will have 960SP and charlie said it will be only a redesigned rv770? Who believed that shit? Almost nobody... but it was true
When you say "....second rate sites", you mean the pathetically biased ones or the equally represented ones.
Up your dosage, once and for all.
What the hell are you doing as a news writer with this kind of horribly sloppy bias? Get the hell out.
We just witnessed another crusade from this writer against NVIDIA because a long time ago they stopped inviting him to meetings and stopped sending him samples. Now he is all "ATI" and the bad part is that he doesn't even hide it. His sayings are all "What If this and What if that?" without solid evidence or proof. However since i am a nice guy i can understand how nice "fiction" can be at times but enough with this grudge against NVIDIA, it got old a long time ago.
This entire article is a "fabrication process".
Everything in the article makes sense but ignores one factor - x86. Nvidia has made noises before about entering this market and if they are making a Larrabee x86 clone the technology they develop now will greatly help them do this.
Nvidia = suckage
AMD / ATi = ownage
Folks need to understand that Charlie isn't an AMD fanboy. Not in the slightest.
What it is, is that Charlie hates the living shite out of Nvidia. Seriously, this is pure unbridled hatred. AMD just happens to be the "other team". I've got a friend who wears a Detroit baseball cap entirely because he hates the living shit out of the Cubs (he lives in Chicago).
That aside, don't call an article "a look at the Nvidia gt300 architecture" when it's really "The GT300 isn't going to ship this year". I mean this is blatantly bait/switching headlining and it'll piss of anyone who hasn't seen your articles before (do a google search for the inquirer under Charlie's full name, and relegate results to the last two years. EVERYTHING is Nvidia hatred. Even AMD praise is only there to put in perspective how bad Nvidia is)
Oh, and quit the "two cores is more than enough" nonsense. For business apps sure, but that's been the case since the P4D. For gaming and iPod transcoding, nobody's going to complain about having to wait less. GTAIV, for instance, is the first game to run substancially nicer on a quad core, probably because all the console code was designed to be as threadable as Rockstar could manage.
Per Charlie's article:
"...This had several effects, the most notable being that the now castrated DX10 was a pretty sad API, barely moving anything forward. It also meant that ATI spent a lot of silicon area implementing things that would never be used. DX10.1 put some of those back, but not the big ones."
In re-reading the article, it appears ATI might have continued with the original (pre-Microsoft pruning) DX10 specificiations on at least the 3xxx series GPU, given time constraints.
If ATI incorporated those (unimplemented DX10) features in the 4xxx series GPU and still managed to achieve the price/performance benchmarks that have been revealed, as well as the cost structure advantage that have been attributed to it, wow, that is quite an accomplishment. And if some of those unimplemented DX10 features that ATI has included are a part of DX11, then their efforts will be rewarded (albeit a bit later than originally anticipated).
That said, I refuse to discount the possiblity of Nvidia making a strong comeback in the future, especially when that company still has some tremendous engineering resources at its disposal. However, Nvidia managment needs to be more honest with and responsible to its customer base, end users included. It's not a smart strategy for them to alienate HP, Apple, Dell, Sony, Microsoft, and a host of AIB manufacturers, let alone the customers who actually end up purchasing the finished products.
Then tell goodbye to your money cos nvidia is going to plummet. Muahahaa... :)
These are just some of the words I've read echoing around different forums from various individuals who are neutral towards this article. In a nutshell people understand that this information may or may not be true.
However, I've observed something a tad more deeper. I've read every response here. Only one person was long winded and in all no one (and I do mean no one) was able to counter or offer a rebuttal to this article.
That's what makes this article (AKA opinion) such a wonderful piece to read. No on has successfully been able to refute any of the information said in this article. Because it's with the understanding that some of it was literally explanation of what he knows. At other times it's an explanation of how he believes things will happen based on the facts he's gathered so far. It's spelled out for you.
So, in the end the masses who've committed on this article here and around web come to an agreement that it may come true or it may not. But they can't or they refuse to counter the information presented.
That IMHO is what I call flawless.
If Charlie has that much hatred for nvidia, it's clear what he writes should not be taken seriously in any way.
Charlie has a disorder, a type for Obsessive Compulsive in reverse, we can only wish him luck with it. His articles are simply a form of 'reaching out'
Unfortunately, he chose the wrong field.
Charlie,
I typically take what you say about nvidia with a grain of salt. This article however impressed me. If what you wrote comes true, I will be going ATI for my next set of GPU's. I could care less how fast my card encodes video. I have a Core i7 Xeon CPU for that. Thanks for this, it's some food for thought for sure.
If you consider how Charlie allows his hatred and biased views reflect is professionalism, ask yourself this question:
Do you think Charlie is the kind of person who has the journalistic integrety to keep himself from posting in this comments thread?
Anyone care to bet on how many posts originates from our Charlie boy?
heres AI thing:
software is essentially a Larrabee emulator that compiles and “runs” the code on the primary system CPU. The current setup will allow developers to start playing with as much as a 16-wide SIMD and to see how the programming model for Larrabee is similar to, and differs, from traditional x86 programming.
its C&P from:
http://www.pcper.com/article.php?aid=683
Where detailed photo exists.also fuzzy one,too.
Maybe larrabee will work. Maybe its end for old larrabee tradition from ibm to intel in selling taxcollectors "Graphics" in era noone heard name: SlingBox. (swearing?).
Worse than swear word in larrabee garjon. Don't lose your hearing, forysthe, you might need it for larrabees graphics enhancment. HAHAHA.
Sorry got old article above, heres todays article with photos & stuff on AI Grand Opening for Intel:
http://www.pcper.com/#NewsID-7154
Larrabee Who or What Are You?
Charlie, i remember fragging you about 87 times in Quakeworld one night with my TNT2 when you had your Rage 128. Why did you have to send me that ping of death?
@John, you are so right. Charlie is prophesizing the end of Nvidia and their inferiority for years- and he always FAILS.
Right NOW NVIDIA has 90% of all the enthusiast hearts, because they simply make faster hardware, period.
GTX295 in QUAD SLI ist the fastest solution right now. ATI has nothing.
I don't give a sh*t how big a chip is, or how much current it sucks up- it has to calculate fast and do what i want HOW i want it. Nvidia just delivers that, easy as that. And it did so for years.
And it is of NO MATTER how superior the ATI chips "internally" might be- because nobody codes for them anyway ;)
Nvidia kind of said: "10.1 is dead" and so it came- you can't even find 5 AAA games that show any advantage because of 10.1.
Nvidia's Phys-X on the other side is very common now and brings huge benefits.
THAT must hurt :)
Last Comment. Nvidia NV70 or G300 straight from theJENS Man:
GT300 or G300 or NV70 is consisted out of 512 cores [or Shader Processors, whatever you like], features a 512-bit memory controller connecting to GDDR5 memory and well, it is a beast. nVidia packed everything in around 2.4 billion transistors using TSMCs 40nm high-performance process
ATI MOST RECENT 4770 is 800,000 tranies with 4890 hefty 1.3 nillion while gtx200 is 1.3 billion. So 300 series Is doing something beyond 40 nm in scale or is it Magic? Ahso, ASUS put out pics of Crosshair III Main today.
Now think 1 dunnington bloomfield @ 2.4 billion tranies plus GTX390X2 @ 4.8 Billion in 3 way sli & its 17 BILLION TRANSISTORS TOTAL. not too shaby....for Game Machine.
"I hope Nvidia survivie and come back fitter because if we only have ATI and Intel then we are going to get seriously shafted for price-performance. Intel have a terrible track record for this, and ATI's isnt much better."
Are you kidding me? You do know that ATI was annexed by AMD a few years ago right? AMD and Intel are mortal enemies. Prices will drop even more if Intel succeeds with Larrabee.
that all your rants about nVidia are kinda without any credibility. (ref: Nvidia hoodwinks reviewers again with mythical GT275s; http://www.theinquirer.net/inquirer/news/1051599/nvidia-hoodwinks-reviewers-mythical-gt275s)
I still read them for their entertainment value, though. I also find the various psychos and "monsters" in US TV Series are the most fun characters.
Have you tried taking a deep breath, a cold shower, or relaxing pills, lately ?
nvidia always guilty...till proven innocent. But it won't matter even then for Charlie D(Mr. anti-nvidia) will be slinging mud about something else.
Let's look at Charlie's track record, yes?
ATI R600 cards will have a built-in sound card that will be an nVidia-crushing killer app for home theater
(http://www.theinquirer.net/inquirer/news/1014862/r600s-secret-weapon-revealed)
G80 shaping up to be a "patchy part"
(http://www.theinquirer.net/inquirer/news/1044075/nvidia-g80-mystery-starts-thickening)
G80 and R600 will hit the market at "about the same time"
(http://www.theinquirer.net/inquirer/news/1006681/nvidia-g80-misses-tape-out)
G80 SLI is "borked", cards have "a lot broken":
(http://www.theinquirer.net/inquirer/news/1012954/nvidias-g80-has-problems)
Nvidia hoodwinks reviewers again with mythical GT275s (http://www.theinquirer.net/inquirer/news/1051599/nvidia-hoodwinks-reviewers-mythical-gt275s)
GT200 has "near-fatal" problems, will sell for $500 (260) and $650 (280)
(http://www.theinquirer.net/inquirer/news/1049121/nvidia-gt200-sucessor-tapes)
GTX 280 will "lose badly" to R700 but cost much more
(http://www.theinquirer.net/inquirer/news/1029372/gtx260-280-revealed)
GT200 is "botched badly", except now is "barely faster" than R700 but will still cost much more. And you'll never see a dual-GPU card, like, say, the GTX 295
(http://www.theinquirer.net/inquirer/news/1040792/gt200-scores-revealed)
I could go on, but it's pretty obvious. Charlie is a loon. He has no idea what he's talking about, he claims to cite unnamed insider sources that are consistently wrong, and his childish mudslinging is both comedically inaccurate and pathetically unprofessional.
As long as this idiot is a part of 'Inq, it's nothing but a cheap tabloid. When I want actual tech news, there are plenty of sites where the writers have both dignity and credibility — qualities lost on Demerjian.
And somehow, despite Charlie's best efforts to sling FUD into cyberspace, nVidia still has the fastest GPUs on the market, and absolutely dominates ATI in the discreet GPU market (http://www.bit-tech.net/news/hardware/2009/04/30/nvidia-increases-market-share/1).
No wonder Charlie's panties are in such a wad. It must be hard being so wrong so often.
you lay into nvidia's upcoming gt300 tech saying basically its going to be crap because of nvidias current track record. you can never judge a product/company because of the companies past/present status. remember when intel was producing the prescott's that were utter crap and getting trounced by amd. a few years later Intel is at the top of their game and amd's struggling just to survive. Nvidia also dug themselves out of the hole they put themselves in when they produced the fx's with the 6800 series. just because a company has a hard few years doesnt mean they cant change. your entire arguement revolves around it and its weak as hell. nevermind the fact your judging a product that doesnt even exist. if youre going to whine at least have more proof than the company isnt at their greatest at the moment. i would love to see the next gen ati card to flop and nvidia to blow it out of the water just because i know it will piss you off to no ends.
I remember the time when Charlie nearly swore that Nvidia will never succeed in designing a dual chip out of the GT200 processor because of heat and power consumption limitations .
his argument was so sound and so rational that every ignorant fanboy here believed it and received it as the only truth .
of course Nvidia released GTX295 and it had lower temperature and power consumption than 4870X2 , and it gave back the crown of performance to Nvidia .
not to mention how the GTX275 would be made out of GTX260 parts or GTX280 parts , how pathetic is that ? he is making stuff up !
my point is , Charlie's bulls are always wrong , because he is always trying to prove one thing , he is always looking at things from a biased view , and that costs him his credibility .
you know why I think charlie is left ranting without consequences , I think he is a tool in the hands of Nvidia to fool the competitors .
good luck with that , tool !
Nvidia was able to re-make the GTX295 on a single PCB .
http://www.techpowerup.com/94054/First_Single-PCB_GeForce_GTX_295_Accelerator_Spotted.html
if they managed to do that , then the manufacturing of the GT300 chip will be just as easy and doable !
you are WRONG for Charlie , 50 times wrong .
I do not agree with the premise that the path nvidia is taking is the wrong one, game developers around the world is at the moment franticly trying to take advantage of things like cuda and physx, the reason for it is simple, doubling the processing power does not increase the visual quality that much.
Simply put with my GF9600gt i can play crysis reasonably well, it's a middle ground hardware, use a gt200 gpu and it runs as good as needed, anything above that is basically overkill, the gameplay in this case would not benefit from more and better looking pixels but rather from more physics and more dynamics.
Graphics quality should be directly linearly relational to the rest of the game processing, and the cpu just can't keep up anymore.
Now i don't believe for a second that nvidia are a bunch of time optimists given their recent history, my guess is that they have an ace up a sleeve somewhere.
And just because ATI had tessellation built in doesn't mean nvidia has to build an entire tessellator extra, instead they would probably opt for a "software" solution.
My first PC had an ATI integrated solution back in '99. I upgraded it to a discrete PCI card to improve performance. Built my first PC in '02 with the AIW8500DV and a AthlonXP CPU. I wasn't really a PC gamer (I had my consoles for that) but I was interested in the multimedia capabilities of the card I subsequently bought a number ATI cards.
That is until I became a dedicated PC gamer. With an FX-60@2.8Gghz and a x800XT stock and eventually the x1950XT. I began to experience issues with drivers and performance. Granted I was still on the AGP platform with a 22 widescreen LCD trying to run games at native resolution with all the eye candy. I guess I started to associate my performance issues with the company.
With my next build I made the jump to the darkside. After salivating over the legendary 8800 series and the Intel Core Duo I invested in a Quad core chip and a 8800gt. At the time the price to performance was great and I could finally game the way I wanted to.
My bias toward ATI/AMD may have been unfounded as I later learned about the limitations of the AGP platform, GPU and CPU limited performance, screen resolution related performance, refresh rates and image quality impacting performance.
For now for me that stigma has been attached to ATI. Right now I have 8GB of DD2800 the Q6600 OC'd@ 3.6GHZ along with a factory OC'd gtx260 both water cooled running on Windows 7 RC x64..
You can't prove god exists either. However that doesnt cause him/it/her to exist. That's what's cool about most future predictions, you can't really prove them wrong until it gets here. By then everyone forgets if you were wrong, and if you were right, you get to rub it in everyones faces. :
Thanks for that lovely article! nVidians are a bunch of wankers.
"A can of whipped-ass," more like. Mmmm.
CAN'T WAIT FOR RADEON HD 5000! BUYING ONE ASAP! PUSH PUSH PUUUSH!
If you want to read real news, go and visit "Bright Side Of News" http://www.brightsideofnews.com/.
It's author is the former inquirer man, Theo Valich.
Also you can visit "Fudzilla" http://www.fudzilla.com/. This one belongs to a former inq man too, Fuad Abazovic.
And they are very good at their job and most of the time correct too.
Everything you say are lies.
The next generation will have a bigger GPU wasting more energy than ever. I'll spare some money on heating, because i live in Alaska and Sara says it's ok because we burn more oil with Nvidia gfx than with regular heating systems.
I'll buy 3 Nvidia gfx just to win my 3dmark because 3 is a cool number and sounds better than ATI, Oh ! OOps it's AMD now !!!! hehehe ! (duh)
Btw , I own 3 Hummers ...
Go Nvidia !!!!
UPS Charlie, your guesswork appears to be way off.
"BSN tells us that NVIDIA’s next flagship GPU, the GT300, has already been taped out and that they’ve got A1 silicon at their office in Santa Clara."
http://enthusiast.hardocp.com/news.html?news=Mzk2MjAsLCxoZW50aHVzaWFzdCwsLDE=
For me it all amounts to "The best bang for the buck$", and not who's penis is bigger than who's ! So Just hope you get your monies worth !
Good Info Charlie.
Common people, when has the inquierer (a seriously reputable new source, LMFAO) ever had anything good to say about Nvidia. Oh wait thats right, they never have, why pray tell? They hate Nvidia, this is not an impartial news source and never has been. All those who believe this filth I hope you believe everything you read on the interwebs. Have a good day.
Another biased diatribe based on thin assumptions and extrapolations that probably won't come to pass.
He's the tech reporting version of Bill O'Reilly or Rush Limbaugh, and NV is the Democratic Party. They can't do no right and are doomed to fail according to that angle.
The Inquirer talking about fake coverage? That's laughable. Inq is the most biased site I've ever visited. Lies, lies, and AMD fanboys. You don't deserve a web site.
There a pattern here with Charlie, ATI good, Nvidia bad. Sad that he seems to have lost all his objectivity as a journalist. All we ever get is the same thing over & over. Sure Nvdia seems to have dug itself into a hole, but its not in anyones interest that they stay there (apart from Intels) so for the sake of competition, and lower prices for all comsumers now & in the future, lets hope they sort themselves out soon.
Spouting this 'we need nvidia for lower prices' crap.
Newsflash - Nvidia never lowered prices when they had a gpu tech lead did they? The whole market pricing currently is ALL because of ATI pricing their cards at reasonable and decent prices.
Its got feck all to do with nvidia, they'd be screwing you for as much as they could, if they could.
@jamahl: I really didn't want to post anything to this rubbish flame war but you obviously have no clue on how a market works. The reason that ATI cards are priced lower is because they perform slightly worse then the high end nvidia cards. Therefore their seg group is the performance conscious comsumer that doesn't want to spend top dollar. Nvidia has no reason to lower their prices they have a segment that wants to spend ludicrious amounts of money on crazy GPUs.
Consider a Merecedes S65 AMG vs a Lambo Gallardo. Both have similar specs although the Lambo takes about .8 second off it's 0-60. They are both beautiful machines but the higher end lambo costs ~$70,000 U.S. more. It's meeting a different segment demand just like ATI is (I know a bad analogy considering the amount of choices in automobiles, but...)
Now imagine no competition in these markets. ATI no longer is trying to differentiate it's product from Nvidia by charging slightly less and can charge whatever they choose. The cards become much more expensive since they don't have competition and a market group they are targeting. They are not only top end, but only option. Same as if there were one automobile manufacturer, it could charge anything for it's high end models. Without choice and competition the market will no longer automatically adjust itself. That S65 AMG would go from 150K to 350K overnight.
So after my short ramble (explained like I would to a 3 year old) I would like to tell you that, you sir and anyone wishing for the demise of a competitive market are foolish and an idiot. Take a basic economics class.
By the way I own a X1650PRO, X1950 XTX, 8800GT, and will decide on my next upgrade when the next gen cards arrive. No Discrimination here, just wanted to point out how foolish you and all fanboys are. Buy what fits your budget/performance needs, it's how the market works. Who cares who made it?
This article is based on rumors, so even he can't see what will happen. But he is right, the rumors/news whats coming from nvidia is not really good.....
This is an interesting article but it's mere speculation at this point. It's filled with lopsided bias and it comes off as a rooter for ATI.
However, I'm an ATI eventhough I have an Nvidia card. I prefer both competitors to be successful regardless and hold no bias against either.
Writing without bias makes for better credibility... and a journalist should always have that in mind before typing one letter.
It will be interesting (as always) to sit it out as the consumer to see what gets offered to us.
On one hand there are partial truths about ATI having DX10.1 hardware / Nvidia not having the DX10.1 hardware, and the various reasons for this. On the other hand there is speculation about Nvidia capabilities to produce the GT300 within the timescale available.
Hopefully they'll be able to resolve the "fair" problem of getting a good 40nm production set in place in time.
Regardless of ATI having "almost" DX11 parts in place for a couple of gens (they didn't get time to remove the parts Nvidia couldn't cope with at the time due to different production cycles between ATI/Nvidia), it will be interesting to see if that lands in their favour for DX11 release.
Equally said, it'll be interesting to see how Nvidia/Intel square up in GPGPU, and if OpenCL will open the door wide for ATI/AMD to regain some ground against both companies.
Definitely shaping up to be a good solid fight between all 3 companies, regardless of any bias shown in the article. The benefits end up with the consumer when THAT happens.
@FUDO - ATI X1950XTX was first DDR4 graphics card iirc ;)
im getting so bored of this site now, originally it was informative with a edge of opinion that was sarcastically brutal but honeswt, now all i ever see is the most insanely imaginative spins aginst nvidia, i dont have a company preference i just buy what is best when i want a system but heres a questions.... WHY ARE YOU WRITING ARTICLES WITH THE IMMATURITY OF A 8 YR OLD BOY!
This whole article is a marketing stunt by ATI, obviously. To anyone with a brain or insight into marketing tactics, this is a grand example. ATI is just hoping to sell as many DX11 cards before Nvidia comes out with their DX11 card. That is their only hope, albeit a short time-frame one.
Notice the guy didn't give any concrete evidence to what the GT300 is made up of. He only speculated that Nvidia may be a quarter or two behind ATI's launch. That doesn't matter. There will not be any applications for DX11 technology until the beginning to mid of 2010, so what makes the difference.
The only thing that should be noted is the performance. Which one will perform better in gaming and other new-aged aspects of GPUs. I can tell you that I am confident that Nvidia will win hands-down on both fronts.
With Nvidia, you also receive Cuda technology which I enjoy most "right now" in rendering 480 movies into 1080 resolutions HD in "real-time". I enjoy PhysX technology which ATI doesn't even have a grasp on now. ATI needs to realize that there are more applications for GPUs in the 21st century than solely "gaming". What is even more disturbing is that ATI can't even win at that. So, in the end, ATI loses and will continue to lose in gaming and all other applications that graphics cards are becoming utilized for nowadays.
ATI is leaps and bounds behind Nvidia. We are living in an age where GPUs are becoming more and more like CPUs. Nvidia does them both. ATI simply does not. Nvidia always "pummels" ATI in real-world applications and benchmarks.
Sincerely,
An informed user who is a smart shopper and "waits to see real-world benchmarks before making purchases".
Charlie, let's address your points one at a time.
1. nVIDIA can't double their shader count from G200 with their next part.
Going from 65nm to 40nm processor technology means you can fit 2.64 times as many transistors in the same area. Your own math shows this. This may come as a shock to you, but 2.64 2. You are wrong. QED.
2. nVIDIA won't have a DX11 part available before Windows 7 launches.
Possibly true, but so what? Games take 18 months+ to develop, so even if AMD finished validation testing today and shipped samples to developers, we shouldn't expect any DX11 games until late 2010 or early 2011.
3. nVIDIA doesn't have a DX10.1 part.
True. Charlie, why would nVIDIA create a DX10.1 part to comply with ATi's chosen feature set? You do remember SM2.0a and SM2.0b, right? DirectX 10 was supposed to solve this problem by making compatible=compliant and eliminating the caps bits. Microsoft seems to have backpedaled in creating DX10.1. Suppose nVIDIA had gone off in another direction, choosing a different feature set to support after their first DX10 part. Would Microsoft have created DX10.2 just for nVIDIA? Probably not as future DX versions are supposed to be supersets of functionality. Hey maybe Larabee will force DirectX 11.1! If not, well then the lesson here is make sure your part is in the next Xbox so you get preferential treatment...
4. G200 was horribly late. Two quarters late according to you.
False. G80 shipped Nov. 2006. Historically major GPUs are released approximately 18 months from each other with one or two refresh parts in-between. G200 shipped June 2008, exactly 19 months later.
5. GT212 was a high end part due mid 2009, to be superseded by GT300 late 2009. Oh, wait, Charlie changed his mind. GT212 is now a 40nm shrink due about now.
Sorry, Charlie. GT212 was a figment of your own imagination. Oh, and nVIDIA uses G for codenames not GT.
6. GT206 was late! Uh... and I (Charlie) guess its GT200b not GT206.
Actually it's G200b, see above. Second, as a refresh it should be expected about 9 months after the initial part. The GTX260 and GTX280 were released June 2008. That means G200b should arrive around March 2009. GTX285 arrived in January, and GTX275 arrived in April. Sounds early, if anything, to me.
7. GT300 has the same "basic structure" as Larabee.
I invite you to say that on Beyond3D, Charlie. I really do. *snicker* Oh, and, yet again, it's G300, not GT300.
8. GT300 will be late/buggy/slow.
Who knows? Certainly, not Charlie Demerjian. AMD faces the same risks. They managed to tape out first silicon a few days ago. Will it go into production, or will they need to respin. Does it run hot? At what frequency? Will their strategy of small dies which helps in the mid-range hurt them too heavily on the high-end where the higher profit margins are? Will using fewer functional units with more ALUs per unit make them less competitive with nVIDIA's greater independency of shaders? Will ATi get spun off like GLOBALFOUNDRIES to save AMD from its crippling debt before their next part even comes to market? See, it's easy to make FUD up, Charlie; you don't even have to be technically inaccurate with it.
Cheers!
(Possibly) Off Topic: Is there any reason I get a picture of a furry on the first page of hits when I do a google image search for Charlie Demerjian? Inquiring minds want to know.
Your web script for handling comments ate my greater than sign in my previous post. It should read "This may come as a shock to you, but 2.64 is greater than 2." on the last line of my reply to Charlie's first point.
My God! This guy is a real moron. I feel nothing but incredible pity for the idiot who wrote this worthless article.
If you got beef with Charlie why do you even bother to read his articles?some of you can't even write proper English commenting on the article so they probably didn't even understand what he's talking bout.he deems nVidia a ripoff and that's what they actually are!if it weren't for nVidia neutering DX 10, we would be having some awesome looking games by now but we the consumers ended up losing out for the sake of their ripoff schemes.
Charlies telling it like it is so keep your mouths shut and avoid his articles if you got beef.
...but I got so bored of you slagging off nvidia towards the end. My loyalty is with speed, not manufacturer. I will buy the fastest card, regardless of whether it is red or green so I like to read forward looking articles such as this. I also like impartial journalism, not biased fanboyism. Surely an esteemed person such as yourself could articulate your arguments without sounding so juvenile.
"If you do the math, the shrink from 65nm to 55nm ((55 * 55) / (65 *65) ~= 0.72) saves you about 1/4 the area, that is, 55nm is 0.72 of the area of 65nm for the same transistor count."
what stupid logic do you use. yep math is right but charlie isnt 55nm is "HALF-NODE" of 65nm process. And yes you could put 50% thats 1.5 times or 1/3 of the previous area for the transistors if density is the sam
40nm is 45nm half-node and roughly you can put 2.0 times more transistors than on previous 55nm half node. just in case that you dont confuse others and yourself.
Hope, four our benefit, that newer hear-say that 40nm is heavily ramping up now is not just another bubble talk you misunderstaood and put on web :D
I have to agree with SiUnit.I`m the same as him and couldn`t give a toss who i buy from,so long as it`s a faster product, with the right price tag on it of course.
However,the article quickly gains pace in it`s hate of nvidia and clumsily fails to hide it the more it goes on.I soon lost interest as that became more and more obvious.
It sounds like a rant from some angry kid that got his arse handed to him too many times in game.Bit silly really.
you sir are clearly an ati fan boy.nvidiots?seriously?clearly it looks as though you beg on your knees to ATI patiently waiting for them to squeeze one off in your mouth.your so called facts should not be allowed to be posted here simply for the fact that its biased in every way shape and form.quit giving biased opinions and "facts".i highly doubt nvidia would put their balls on the line without some confidence in the 40nm process.i remember when ATI had no answer for the 8800 series.the 3870 was a joke at best.the 8800gt still wipes the floor with to this day.my pair of gt's can hit 20k on 3dmark 06.ive never heard of any 3800 series coming close without a quad core.honestly i think ATI will flop on their faces once again with driver support.their drivers are still buggy today despite many improvements.heres my advice to you,start giving unbiased opinions and learn the facts instead of giving crap info.the inquerer has a trac k record for giving falso info,hmm go figure.you sir suffer from ATI fanboyism.dont worry theres a remedy for you.GREEN means go,RED means stop.
July 27 , 2009. No GT300 has taped out as of yet. Back to School with DX11 from ATI and a happy 2010 to you.
You are calling the GT300 a Larrabee-lite now?
LOL! Do you even know the substantial design, theoretical, conceptual and physical differences between a CPU with an on-die GPU, and a highly parrallel Computing GPU? In the field of microelectronics, theoretical computer science, and mathematics, they are as different as chalk and cheese.
My GOD! He fails in the second line. I'm no NVidia fanboi, but I pity the poor fools who have to rely on idiots such as these for their information.
You sir, are no engineer. And as such should not be allowed to diseminate your worthless opinion as 'authority' on any respectable online media source. And if you are, I just lost all my respect for the university and engineering regulating body that vetted you.
But hey, welcome to the internet, where idiots are scholars, and opinionated morons pass themselves off as philosophers.
Throw a bit of 5th grade math around and pretend to know a thing or two. But failing this hard, calls your credibility into question.
At any rate, NVidia that GT300 has taped out already, 2 weeks behind schedule, but it happened.
for those of you who look to this Dermjian guy for your information. Get professional help.
a few fair points, but it does show somewhat of a bias. You are assuming the worst case scenario with nvidia and the best case with ati. and then making the assumption that none of these engineers and business strategists at these companies know what they're talking about.
the tesselator in 10.1 for one, is not even DX11 compatible. But yes, there are also some points for concern. however, the gt300 has finally taped out, and I feel fairly confident we'll have the product en masse by early 2010.
you also pin many concerns about nvidia that are actually unique to the foundry and the 40nm process rather than the VLSI silicon design. many of the benefits of ATI managing to get the 4770 is likely to carry over to nvidia.
but like i said, while many points are valid, you do show a heavy ati bias. by interpreting facts and data in the best light possible when dealing with ati, and the worst when it comes to nvidia.
i think that's what leads many to act outraged (my flatmate included) when reading your article
anyway, time will tell.
i hope to do an article on my site soon with some updates.
good day.
Nvidia still does have the fastest cards out there hands down. With me it comes to who has the fastest product. This has been nvidia for years now. ATI is always just a step behind nvidia so nvidia must be doing something right.
gt280 4870
As if we needed yet another example of how line width guys are totally clueless about software, along comes this rant from Charles. He can be very right about so many things yet at the end totally wrong by missing the single, most important point.
Charles forgets (or perhaps has never internalized) the decisive role of software. NVIDIA "gets" software in a way that Intel and ATI do not. It's no accident that since the 786 in the 80's Intel has been terrified of graphics parts, because they are terrified of software. They just don't understand it. They are so afraid of it that for decades now whenever they run their internal jihads against each other the clarion call for settling bureaucratic scores with each other is "Oh, he's a software guy. Better get him under control."
NVIDIA's investment of real estate for GPGPU is not chasing a market that doesn't exist, it is generating the software support for the only market there is. It's a shame Charles doesn't understand there are hundreds of GPGPU applications now, all of which run on NVIDIA, virtually none of which run on AMD parts and absolutely none of which run on... what was that Intel part again?
For people who run those applications, using anything but NVIDIA is non-negotiable. No one gives up a factor of 50 speed up. If you are doing state of the art graphics arts work with Adobe products that's all you will use.
For those folks, they'll happily configure systems to plug in six or eight GPUs to run faster. The cost is nothing compared to the gain.
Do those applications add up to hundreds of millions of units? Not yet. But as parallelization becomes ubiquitous it will, and it will become ubiquitous for those who have the software base. NVIDIA is invested in that. AMD and Intel are not.
Whether or not NVIDIA can survive to see that happy day is a fair question. Odds look good, because they appear to have achieved enough critical mass within GPGPU for that to start contributing small but highly recession-resistant earnings while still doing well enough in graphics/gaming to stay alive. But only NVIDIA has both of those irons truly heated up and working.
Ok so you got every card wrong since the 8800GTX launched, you still dont seem to understand whats going on in the industry. Nvidia cards are bigger and faster, ATI has to OC to match the preformance, and its been this way since the G80. ATI has a long way to go to make up for its design, there using a design that isnt accepted widly, those 800 shaders equal 200 actual shaders and 600 dedicated to other crap.
This might be true in some way, and somethings here are correct. Keeping an open mind... Not too open to let your brains fall out.
You are stating that nvidia failed to meet the original dx10 specs, while ati did. So microsoft downgraded the specs on dx10 to get both players onboard.
And DX11 is mostly what DX10 should of been. Does this mean that the features that have been unused on the ATI 2900-4800 cards will available in DX11? Will I have a mostly DX11 card in Windows 7? (Mine is the 4870)
Or.. will these features be on the card, but unavailable? due to some bs.
Will we see games that have options available.. to ati cards, and nvidias cards be grayed out? "last gen" That would be hilarious.
Leave it to Microsoft to mess it all up!
http://en.wikipedia.org/wiki/Direct3D#Direct3D_11
Tessellation was earlier considered for Direct3D 10, but was later abandoned. GPUs such as Radeon R600 feature a tessellation engine that can be used with Direct3D 9/10[20] and OpenGL, but it's not compatible with Direct3D 11 (according to Microsoft).
I like the Nvidia card it allows me to rip my gangbangvideo's fast to my Iphone.
CUDA tech here i come!
Christ I already bought a future brick, a HP with the 8400 GS card that will fail in the future. No more please. What I want is Intel inside, along with a ATI graphics card with it for my next laptop with a 2.0 web cam. No Blu ray as there is no content worth going out to rent, let alone buy. Something fast, and RELIABLE to last a long while.
Well, the FERMI benches have leaked and it's 20%-40% faster than ati's sorry 5000 series top core.
Oh well, better loser fantasies next time for the lying, in the red, billions in debt, red rooster raging red retard crew.
what a bunch of crap.
Get your head straigh you moron
.. right. They have leaked.. But who leaked them? Oh, I know! It was nVidia! Well that's a reliable source!
And you have to read the article instead of jumping to conclusions.