BathOS: We've heard of it. Your mobile phone works while you are bathing
|
|
|
Fermi GPUs to pack less of a punch
THE UNFASHIONABLY LATE next-gen graphics core from Nvidia, Fermi, also known as the GF100, also known as GTX480 and GTX470, will pack a smaller punch than initially expected.
According to Digitimes, Nvidia's own board partners have put forward a theory that TSMC’s 40nm isn’t good enough to produce cores containing the full 512 CUDA cores that should shock and awe consumers.
The Green Goblin now seems poised to release cut-down versions of both the GTX480 and GTX470 that should have 480 and 448cores, respectively.
Anyone in the business will tell you this makes little sense.
Nvidia has to have made a choice - either it can have a full 512 cores with a lower yield and output a smaller number of GPUs, or it can in fact produce to meet volume demands (and campaign promises), only at less than optimum quality and power, as part of a more dastardly plan.
So, the fact of the matter is, there is a good chance that the GTX480 that Nvidia will release on March 26th through April 6th will not perform on the same levels as those about which certain benchmarks have leaked throughout this last month.
What it does feel like is that Nvidia is actually redoing its marketing plan and going ahead with a slightly less powerful first attempt at Fermi, just so it can later launch cards with the promised 512 cores and sell a few more chips on the upper-extreme high end after that, eventually. GTX490, anyone?
After all, what die hard Nvidia enthusiast worthy of the name won't pay twice, if not both twice and twice as much, for the same technology? µ
Share this:
Share
Remember the the original (192) GTX 260 it was a slug and even the 4870 512Mb was was all over it when launched for what $150 less!
The green side of the market is so "pent-up" right now anything at any price that Nvidia will sell they'll eat-up. At least this time (Nvidia) AIB's won't need to send them checks to keep them happy like what happen back July 2008!
@ Psihomodo..
Do tell which side i`m a fanboy of because i would really like to know?
I currently own 2 pc`s, one with a gtx280 the other with an x1950 pro.In the past i`ve also owned a 6800GT and a 9800pro amongst others.
So c`mon tell me which side i`m a fanboy of ya clueless muppet.
Look, why make things so difficult. This algorithm has always worked for me:
IF CardA faster than CardB THEN
buy CardA
ELSE
buy CardB
END IF
I don't give a flying f*** which logo is on the card. Yes that's right. I am an fps whore. I will go to bed with whomever has the biggest fps. ;-)
first GTX470 is 607 Mhz/s Unit. next gtx 480 didn't have frame rates done in Dx11, merely in Dx9, which left Ati 5870 looking less in Tez & Fps, yet in fact, Not so. here:
numbers were indeed too good to be true. If you run the Dirt2 demo, the GTX470 and GTX480 drop back to DX9 mode, as you can see it in the picture above. Since they are doing far less work, frame rates go up. Running the DX11 code path drop frames by around 25 to 40 percent for the same work. In the game itself, DX11 works just fine on the GTX480, so it is likely that the demo lacked the correct profile for the then unreleased and 6 months out GTX4x0s.
For some odd reason, that point wasn't mentioned in the Nvidia slides
So when Gets into Wild, Truth will be told, as Jens is apparently direct ancestor of theKHAN, rip, furry bunny.
STeWie
....next week.
Nothing to see but tech site after tech site with the same dull benchmark graphs over and over.
Then all the fanbois arguing over the fact that the pro Nvidia sites used the 10.2 drivers for the 5870 and the pro ATI sites used the 10.3a drivers.
@JooEffOh
Funny how you didn't do well in trying to hide your obvious fanboy-ism into a block of text, which is pretty false especially with all the hard evidence around the web.
I`m fed up of the rants of people with their own agenda and would just like some news without the speculation and crystal ball stuff.The trouble with half these idiotic comments is that they assume people are as clueless as they are and will spunk their money on the next product coming from their favourate gpu company.Give us a bit of credit would ya and realise that most people don`t give a flying f*ck about ati or nvidia and just want a decent gpu at a decent price.You can keep all your propaganda and bullshit talk because ultimately your just wasting your breath.
Since ati launched their latest and greatest gpu they have been mostly unavailable and certainly overpriced for the performance they give compared to the last gen nvidia.Now anyone that isn`t in need of a major upgrade (like me) is just waiting to see what the response is from nvidia.If it`s not better bang per buck then prices of the ati stuff aren`t gonna drop either so i guess we will forget about current gen and wait till the next one that does make our jaw drop and our wallets not hurt so much.
Means customers end up losing more than winning.
It means ATI can keep its prices up and innovation slows down a little.
-It's a non-profit funneling all proceeds to its leader
-It is based solely on myths, legends, lies and fanboy apparitions
-A steady stream of wackos tithe regularly (see above samples)
-Fermi generates enough heat to rekindle hell itself
"The Green Goblin now seems poised to release cut-down versions of both the GTX480 and GTX470 that should have 480 and 448cores, respectively."
We've known that since, like, the end of January. You're the one who told us, in fact. How can you possibly pretend that this is news?
Is this another Geforce FX 5800 vs Radeon 9700 scenario?
Agreeing with everyone else, please do some of your own research, there's been no 512 core edition in anyone's hands but nVidia's, any of the leaked benchmarks are going to be based on the reduced core model to being with, and yes, they were impressive, their tessellation which AMD had been touting about far outshone the competition and normal performance appeared to be on par even though the model was several months pre-release.
OMG Just as stealing another article, razor let article be & now heres Intel :
ONE DOWNSIDE of Intel’s Extreme Edition CPUs that makes most people shy away from them is the sky high price tag, but it looks like Intel is getting ready to take a leaf out of AMD's book by releasing what will hopefully be more cost conscious overclocking CPUs with unlocked multiplier carrying the K suffix.
The Core i7 875K has been confirmed by Gigabyte in its CPU support lists of multiple motherboard models using the P55, H55 and H57 chipsets. The other CPU will be the 655K, although opinions go apart with regards to the branding here, as XFastests suggests Core i3 while TechConnect Magazine suggests a Core i5 branding. The latter makes more sense to us, especially as the Core i5’s are part of the 650-series already.
The quad core Core i7 875K will be clocked at the same 2.93GHz as the Core i7 870 and it also features the same default multiplier, 8MB L3 cache and 95W TDP. However, with no BCLK multiplier lock this CPU should appeal to overclockers, despite predictions of a starting price that will be higher than that of the Core i7 870 which can be had for about $550 online.
still less chache, with google above, Clock stopped here in main floor at 2:49 EDT, computer time still standard, so another hour behind meaning Google is gone, by any time, yet warp was about several maybe 4 hours ago.
655K is of intrest to blodget minded, about $200, yet only 4 mb cache. hummm.
Back to Fermi:
In tessselation testing Fermi did good STABLE 60-70 FPS. while 5870 did beat fermi bit, here & there, mostly averaged about 30-40 FPS Today. about 3% of time 5870 actually was above fermi & 70 fps. also, done on frames NOT that hard, by Nvidia own confession. so working & gettting some strong Golds, yet not ALL gold.
Mr. Cheesewright.
theRED'R heard Roar of Victory, Now Read Crippling Story Here:
Hong Kong is a special administrative region of China and operates its own economic and political systems. By redirecting Chinese search traffic through servers in Hong Kong, to http://google.com.hk, the firm said it had effectively transferred the jurisdiction of its search business off the mainland.
Still Deeper Cuts Are Coming, Much Deeper, As Republic of China Looks over event. yet, DOWN WENT google.cn. Up With RD Star of EAST. Supreme world Commander: DRASHEK Lives.
Happiness Fills Land wiith Bloom & Results: u b Judge.
C STeWie @ Associated file Grinders.
Maybe you prefer the kind of "journalism" that simply reprints manufacturers' press releases (cough*dailytech*cough). Personally I like journalists that a) investigate and report things that are actually *new*, b) care more about informing the public than getting gifts from the industry and c) call companies' bullshit when they see it.
If that's "bile", I think we need more journalists with a functioning liver (and a couple of other organs as well).
This wouldnt be the first time nvidia has given the press cards they wont be selling..
Not did Charlie say it months ago but ya people still talk like he is just bashing Nvidia and everything he says is wrong.
WEll I got news for all you Nvidia Fanboyz... Jensen huang LIED to all of you months ago when he stated " This puppy here is Fermi " uh ya right, only it was a FAKE.
Now Nvidia are claiming and claiming and claiming and, well you get the picture.
No Cards,No Open Benchmarks ,No Answeres
And I am suppose to BUY a card from this Company ? lol Are you guys kidding me ?
For gods sake are people really Retardted ?
I am willing to bet that the Fermi Cards that get shipped to some Sites for Testing and Benching will not be the same Cards Nvidia will sell.
Thats right. The Cards Nvidia ship to selective sites to Review will NOT BE the SAME cards that are sold at Retail.
Wait and see. I suspect the retail ones will somehow not show the same performance.
Problem with Charlie is that all his articles on NVIDIA are so full of his own anti-NVIDIA bile that it undermines everything else he says about the company. It doesn't matter whether the info he's got is correct or not - whatever he says now is an instant "Charlie + NVIDIA = truck-load of salt".
Charlie from Semiaccurate has been reporting on this for months and some people STILL don't believe it. The 512 core version will not be available until the B1 stepping because Nvidia screwed up the fundamental architecture that would allow a fault-tolerant design. The current yield is not even an option. This means no 512 core variant for 6 months at least. He also predicted that 512 core samples would be sent for benchmarking.
None of this matters too much as Nvidia is now more of a marketing firm than GPU company. They could just rename the GT300 series (GT 200) to GTX 400 and be done with it. They would still appear in OEM systems and unsuspecting customers (Nvidia's target demo) would continue to buy.