This is the sort of English up with which I shall not put - Winston Churchill
|
|
|
CPU gigahertz battle resumes
DROWNED OUT by the recent buzz was the fact that, for the first time in a long time, AMD has announced a desktop CPU at a clock rate faster than the highest frequency Intel chip of the moment.
Yes, the new Phenom II X4 965 Black Edition runs at 3.4GHz out of the box, a tad tick higher than the 3.33GHz Intel Core i7 975 Extreme Edition. Of course, that overlooks the automatic Turbo mode on the i7 that ups its clock rate to 3.46GHz or more by default, plus of course the average 15 to 20 per cent clock-for-clock performance advantage that Intel claims its chips deliver over AMD's.
Nevertheless, it is a welcome change for the benchmarking community. At least there is something upon which to compare the two on similar terms, for the first time in a while.
Furthermore, if you pour on some liquid helium after using liquid nitrogen for a first round of freezing, the Phenom will do a Phenomenal 7+GHz for a short while, while Intel's top Core i7 can't quite reach that level, it falls short by a full gigglehurts in fact. Now that must hurt, until you realise that the number of users on the planet who will run such a configuration for longer than a few hours is, well, roughly zero.
What are the prospects for the competition here 'til the year end? The unlocked Phenom II X4 965BE will do well against the $300-class Core i7 860 Lynnfield part, until the user lets the latter's Turbo go all the way, clocking all four cores at 3+GHz, at which point the performance battle will even out. The $550+ Core i7 870 parts will still have a bit of an edge over the fastest Phenom here, and remember, these are mid range parts, mind you.
But then, AMD's Phenom II BE parts are unlocked, unlike most of Intel's new "Lynnfield" parts.
A stepping and a process tune more, I guess, and AMD should easily be able to release a 3.6GHz or even 3.8GHz flavour before year end, although a L3 cache latency decrease or size increase - remember, AMD does have some nasty dense cache cell technology - will do as much for a performance boost as a clock step up. Intel can do that as well.
But AMD can do something Intel can't before at least February 2010, that is, repackage its six core Istanbul chip into a desktop package and pick the best dies to let the six cores run at somewhere around 3.2GHz or higher. Now, in some benchmarks at least, that baby would seriously threaten even the Core i7 975XE. Yes, at the cost of drawing 150W TDP, but then, hey, people are willing to risk their lungs with oxygen-eating liquid helium, so what's a few more watts? In the end, the Core i7 975XE also guzzles power too.
Since we won't see anything from AMD's Bulldozer stable until the middle of next year at the earliest, not far from the first Sandy Bridge parts, tuning its six core parts for the extreme desktop may work well for AMD to grab some benchmark wins and, at least, effectively counter Intel's hyper-threading, which is helping Intel well in quite a few tests. And now, that feature is in Intel's mainstream Lynnfields, too.
AMD, here's a challenge. Can we see 3.2GHz six core desktop parts by Intel IDF Fall? Or at least the 'lucky' 3.8GHz extreme desktop ones? The Chinese do like anything with an '8', and that's the biggest market now. µ
Share this:
Share
"Furthermore, if you pour on some liquid helium after using liquid nitrogen..."
Don't get me wrong, I like your work, but be realistic, only a few strange people on the entire planet care or indeed can pour LN or LHe on their CPUs. The rest of us worry about stock clocks or air-cooled OC.
There's no mention of a "965" anywhere on AMD's price list, news space or processor comparer. Does it even exist?
Technically, it's not supposed to be out yet. But if you look around on the internet you can find some etailers that are already selling it. I don't think it's supposed to come out until middle or late August? If it's affordable and you are already using AMD's phenom II platform, I guess it would be interesting. I have yet to upgrade my PC this year, and I am weighing with going with an i7 or a phenom II. Phenom will save me money, but i7 will give me performance. Not certain what I will do. I know! Let's put the presidents of both companies in a steel cage and let them fight to the death!! Whoever wins, gets my money. :)
Oxygen/Helium nuclear fusion?
Noble helium would never deign to be "eaten" through a chemical process....
:-)
If they want to be taken seriously they need to increase the clock rate. The IPC is lower so they need to make up for it. AMD better hurry up with 32nm and 22nm though.
To bad for AMD, clock speed doesn't matter. It's real world benchmarking that matters. Sorry AMD fanboys. Intel still rules and will for quite some time. Intel has tricks up their sleeve at all times.
You are correct on the current software front. Remember as applications are developed to be more multi-threaded then then single thread I don't believe that the differences will be all that much.
I will say that Intel did make a great stop gap design with their current iteration of their Pentium CPU but that doesn't mean it will scale well on the multi-thread applications. This is really the only reason why AMD still has strong says on the server/clustering side of things. AMD just scales better then Intel does, the performance increase is almost linear.
Like I said they made a great stop gap design knowing that most software will take a few years to become multi-threaded. Also you know you can't blame design on AMD so much since Hector RuinIT totally F%#$K AMD big time and lost a lot of its very good talent by being such an money hungry asshole.
Intel didn't waist time in taking advantage of that. AMD was and will be a smaller company for a long time to come and dumbass Hector was on his high horse riding the wave, not realizing that the wave would eventually subside and there would be no go. I really hope that dipshit never gets hired again by anyone to run their company.
here is where you can find the extreme overclocking results, validated:
http://service.futuremark.com/search/form.action
Try 3DMark06, 3dmark05 ... well, top is Nehalem. If you go for the bench that really matters, 3DMark Vantage (last version is always the reference) ... Nehalem wins by FAR.
Try PcMark vantage ... well, Nehalem again by Far ... Sysmark ... cinebench? SuperPI? all Nehalem wins
If you are looking for frequency only, and do not care about performance, I recommand that you go for a Celeron Cedarmill based .. http://www.theinquirer.net/inquirer/news/1137478/celeron-overclocked-1ghz
If you look for Pure performance, then Nehalem is far ahead.
And if you go with Stock frequency, then the gap is even bigger (What matter to non overclocking customers)
This is my personal opinion, based on DATA. hwbot.org will confirm the same data set.
Peace!
Can't wait for the 6 core blackies to come out @ least 3.4Ghz - We have waited tooo long for 4Ghz proc. But many cores is cool - SETI will like this new upgrade. AMD will have 6 core 1st but Intel Will OFFER 8 Core expensive eddition proc maybe @ 1st Q 09 - No?
Regards dudes
Intel currently is the fastest x86 available. But definitively Istanbul can be a threat for Nehalem. Nehalem is a Quad Core with Hyper Threading, which means that 8 threads shares the 4 execution engines available, but Istanbul is a Six Core processor which can process 6 threads in which each one will have its own execution engine, and since AMD server processors usually scale better toward multi threading, probably we can see some competition, so we all can have lower prices from both camps, like currently with ATi and nVidia, so much power for so much little money.
2 AMD chips....
In one DS motherboard kicking CoreI7 around like a red headed stepchild.
Intel desktop solutions don't scale, ever since the end of SMP P3s. But when they repent of their evil ways I'll look at Intel again.
Until then you can look the scores up. wPrime 32 & 1024M. For a real Bang-per-buck and core&clock tick; dualsocket AMDs are the real CoreI7 killers.
the scaling of nehalem is really good, i don t know where you got this idea.
i invite you to go and check the recommanded spec_fp rate scores (bench recommanded by AMD from 2002 to 2008)
you will be able to calculate the scaling per core by yourself. get the spec_fp rate score and divide by the number of cores ... then, look at the overall score, nehalem scores are wayyyyy bigger.
www.spec.org
who? why are you feeding the trolls? ;-)
On a more serious note, you are an exceptional software engineer. How well do you think Istanbul would perform on SPEC if it ran the exact same codepath as a Core2 or an i7 whether desktop or XEON? Also, how well would Core2 or i7 perform if they were to run a P3/K7 optimized version?
I ask this question only because while there are dedicated compiler support teams from ODMs like Intel, IBM, & SUN optimizing for SPEC, most applications outside of a narrow few receive that much effort and are compiled for good performance on the lowest common denominator platform.
I look forward to your reply. Thank you.
Best regards,
Daniel Rwizi
PS You can on respond at Aceshardware if that would be better. If you can also pm me there. Thanks again.
Cmon guys, benchmarks aren't everything. Maybe in the IT arena and servers, but (most) "normal" people just need a PC for decent multimedia performance, good response speed, and the pack should be INEXPENSIVE. You guys are forgetting the $$$ factor.
And even if we take that out of the equation, Intel is not faaaaaar better than AMD. Its just better, but not by faaaaar.
Also, AMD leads in the platform war. Why you say... Well, where is intel's integrated graphics? have you ever seen the GMA4500 running? oh my god it cant even handle winamp visualisations right. C'mon! zero games are playable. I wouldn't be surprised if solitaire run slow on that piece of crap.
And Nvidia's integrated solutions while being far superior than intel's, it still can't beet AMD own and cheaper 780G or 790GX chipset. Specially against this last one.
So, is people nuts? Are they willing to spend an extra 100$ (or more!) just to make a CD to MP3 convertion 10% faster? I hope you all spend good time with that extra minute you saved there. Cmon. Gimme a break. I prefer saving money to pay my bills...
I won't spend 100$ more in no way. I'll stick with the mighty AMD+integrated formula. As a system builder all I can say is that you wont find anything more equilibrated. And also worked for last christmas. You should have seen all kids smiling while trying Far Cry 2 for the first time, running perfectly fine on medium graphics, on a simple Athlon 64 X2 7850 with 780G powered motherboard. Thanks AMD, you saved my day. And customers saved money!
You make very good points with which I completely agree.
Best regards,
Daniel Rwizi
True, per core the Corei7 is liek miles and miles ahead of AMD. The thing is that Corei7 as a desktop chip dosn't come in a 2 socketed flavour. Thats where the win comes from the AMD corner.
Example: Just did an upgrade of an 2 socket Socket F to a year old Barcelona chips and get more juice out of those 2 chip then one Corei7 chip, for the paltry price of 100 euros @yer local ebay.
I hope you get my point. When Intel goes back to allowing dualsocket desktop chips I'm back onto Intel', but the current platform that I have right now ( socket f ) I bought that Q406 when QC Intels were just avaliable for the now-dead socket 754. To-day I plug in 2 quadcores tomorrow 2 sixcores. I hope you understand; Intel desktop solution don't scale. They die ( too soon ).
We covered the W5580 before, that's the dual CPU Xeon flavour of Core i7 965, and has 6 memory channels in total (yes you can use the desktop non ECC memory too!). Watch out for an even faster version reviewed next week... and yes, Francois is the man working on the dual-CPU desktops in Intel too :)
Since when are synthetic benchmarks like most of the ones you listed a good indication of real world performance? I have some swamp land I could sell you, it benched well........
@MrGiggleNutz
SPECs and all the consorsiums benchmarks are the best effort to produce a smart and realistic benchmark. It is very hard to produce a realistic workload, and having most of the players of the industry in one consorsium and drive it though vote is the best way to make sure that "corporate agendas" go away.
In the case of Spec_fp rate, it was recommanded for 7 years by AMD, you can t blame Intel for using it now that they have a missive lead on it.
The best effort to build benchmark is very important to the industry, and who ever joined the developement of one of those benchmark knows that think are done through votes, with all votes Equal for each company.
I see a lot of people dancing around, myself, I don t think that some part of SPEC FP rate are nice to my taste, but I bend to the opinion of the group, and respect it.
If you do not agree on something about Spec, or others, feel free to voice it and join the developement group, all of those consorsiums have an open door policy, it is possible to join for very little membership.
Let's stop spitting on some of the best benchmark effort ever make, and join the group to make it better if you know something smarter about it.
try to put some many REAL workload into a benchmark, you ll figure out that it is hard! very hard!
@hoohoo, stop taking crap out of context, he said basically no one does that to run apps.
Congratulations are in order for Nebo-Nova .... after all, he stayed objective for two whole introductory sentences before descending into his customary Intel fanboy screeding.
Alas, with Charlie gone, look we must to dear Sylvie, as it is she only who possesses enough courage to hold up the silver AMD/GloFo cross and banish the Nebo-Nova Intel Bloodsucker into the burning rays of dawn.
Thank you both (and all) for your passion and invaluable contribution for the industry.
"In one DS motherboard kicking CoreI7 around like a red headed stepchild"
Speacking as a redhead and a descendant of readheads, I must protest.
"Alright, alright I apologize. I'm really, really sorry. I apologize unreservedly. I offer a complete and utter retraction. The imputation was totally without basic in fact and was in no way fair comment and was motivated purely by malice and I deeply regret any distress that my comments may have caused you or your family and I hereby undertake not to repeat any such slander at any time in the future."
__________
Quote from John Cleese in "A fish called wanda" link: http://www.youtube.com/watch?v=m7mIy97_rlo
For those of us who have wives, kids, bills etc to pay then I'd say an AMD setup is a great solution. Not only does it offer pretty good performance it's also affordable too.
Sure, I'd love to have a top of the range rig with a Core i7 CPU in it, but ONE I wouldn't use half the features of it and TWO I can't afford it (I mean I could get a Phenom II X4 and motherboard for the cost of a Core I7 motherboard alone!).
I guess one thing though, all this competition from AMD on the slightly lower end may cause Intel to drop the price of the Core 2 Quad CPU a tad, but then I still think I may be better off going for an AM3 system with DDR3 memory and maybe a low end Phenom II (or even Athlon II) and dropping in a quicker CPU when I can smuggle it in past the wife.
Rob
(Before anyone asks, I'm not really a gamer, the XBOX 360 does my gaming needs, I'm more into watching movies etc which I can do on my overclocked Pentium Dual Core machine quite happily anyway).
Now you have a little high end AM3 mobo on the cheap. The ones with 785G IGP and SB710.
Gigabyte MA785GMT UD2H
http://www.newegg.com/Product/Product.aspx?Item=N82E16813128397
Perfect choice if you don't need a dedicated GPU and when a 770T UD3P is just useless asuming that.
Because, he is the man that responsible to make a false benchmark in the past that occur when Intel still uses the netburst and amd holds the champion with athlon 64. He is an Intel Spawn. Intel admits they uses unethically business practices. This man (Francois) is behind of misleading benchnarketing apps with BAPCo. I think this man must be punished for cheating and fraud to mark up the performances of Intel Pentium 4.
@Surya
As usual, you are so not accurate, it is not even funny anymore.
The last benchmark I worked on as an active member was WebMark, in 2000 ... that is 9 years ago.
By the way, you seems very much able to do a lot of personal attacks, why don't you tell us what do you do in life, and for who you are working for, at least, I sign with my real name, and Nova can tell you that I am very very honnest in the way I work, I am direct, and always provide data to support my claims.
who are you?
thanks for being sooooo innacurate, and incredibly not credible.
PS: I use the 1st Amendement of the USA constitution, and the 11th article of the french constitution, I use my freedom of speech as a French citizen you prove your claim false, I do not represent intel in anyway in this posting.
I'm just a plain computer user. I have found your's linked page that you are working for Intel. Based on previous comments in the different articles you are a stubborn man that defend misleading BAPCO benchmark. So, i think you the mastermind man that behind the BAPCO current or past benchmark or I thought you an Intel representative for BAPCO that Intel using their powers to make BAPCO benchmark prefers Intel systems. I think it is right your comments not an Intel opinion but what have you say and where are you working for is the prove that you will be always biased towards AMD and will defend Intel that they always have performance lead even if AMD holds true performance lead like in the past with athlon before mid 2006. If you prefer amd systems you will be fired by Intel totaliter management.
You fail. Hard.
Nice half-baked opinion, retard! I've heard better formed arguments(with more research to back them up) literally come farting out of a monkey's ass.
Yes, I compared your intelligence to a monkey fart, and you lost.
AMD has for the first time in a last 2 years a choice what they want to do and launch for next year. The current roadmap looks quite empty but i doubt AMD will be sleeping at the job for a whole year i rather beleve they haven't reveald there marketing plans yet and they wait to see were the market is going based on that they will launch new products. For example if the market really wants a 6 core part with 3,2ghz and AMD sees a profit in it i bet they reswap the istanbul to the desktop market. if they dont need it you can expact faster phenom II models to appear. 3,8ghz should be doable. if not 4ghz for the win. other then that i do beleve the hydra is an desktop optimized version of istanbul with a die shrink to 32nm. This should grand AMD the advantage they need when it comes to overclocking. AMD dous not want to do any big transistor parts anymore for the desktop market and we know it. AMD will make there die shrinks faster for the upcomming years then they used too. Global foundries has enough benefit from it if they manage to reach the die shrinks around the same time intel get theres. AMD has indeed some tricks upon its sleeve and we will just have to wait what it is going to turn out to become.