Analysis on IT trends and competitive strategies, with emphasis on micro processors, computer systems and networks. Based on latest news, backed up with real data, this site intends to provide a true and realtime picture of the fast changing IT landscape. This journal strives to be accurate on facts and sharp on criticisms. You may email your opinion to sharikou@yahoo.com or post comments here, be cool and intelligent.
Freelance journalist on IT matters. Some of my writings have been published on online IT journals. Any original content on this journal is Copyrighted, but it's free for non-commercial use. Any Trademarks used on this site belong to their respective owners. Some of the pictures are links. If there is any issue with the content of this site, please email sharikou@yahoo.com .
2.5GHZ K10 is 33% to 66% faster than 3GHZ QX6800 quad.
posted by Sharikou, Ph. D. @ 4/24/2007 12:05:00 AM
60 Comments:
AMD returns..
AMD is going further into debt just to finance it's operations.
http://www.theinquirer.net/default.aspx?article=39130
Yeah, AMD is back on track! Go go AMD! Intel in finished one and for all! Thank GOD...
K10 = 1.5 x QX
K10 = [(QX + QX) / 2] * 1.5
First, here is a working link.
"2.5GHZ K10 is 33% to 66% faster than 3GHZ QX6800 quad."
I'm not sure how but did Fuad got his math right or am I looking at wrong benchmarks?
clicky:
Dhrystone ~500k
Whetstone ~350k
rathor
"K10 = [(QX + QX) / 2] * 1.5"
In math there is a thing called simplification. Are you trying to to the opposite?
2*QX = 2QX
2QX/2 = QX
QX*1.5 = 1.5 QX
Oh, this math.....
K10 = 1.5 x QX <=> K10 = [(QX + QX) / 2] * 1.5
Am I right ?
[(QX + QX) / 2] * 1.5 =
(2 * QX) / 2 * 1.5 = QX * 1.5
What was the point of your calculations? It is exactly the same as (1 + 1) / 2 = 1
oh man, of course is the same thing, i'm just offering you two examples (which are the same thing)... i'm just messing around with calculations ;) for fun, you know.
"I'm not sure how but did Fuad got his math right or am I looking at wrong benchmarks?
clicky:
Dhrystone ~500k
Whetstone ~350k"
Look at the benchmark carefully bro!!
Dhrystone ~50k
Whetstone ~35k
its clear now that intelers has a lower IQ.
Hehe, too many zeros there :)
It also seems that I was looking at wrong benchmark then.
Funny how AMD has to use an Intel platform to benchmark their new R600,
Onto the benchmarks, the tests were conducted on an Intel D975XBX2 BadAxe2, Intel Core 2 Extreme QX6700 and 2x1GB DDR2-800 MHz. The operating system on the test system was Windows XP, with a fresh install before benchmarking each card. Testing of the AMD ATI Radeon HD 2900 XT was performed using the 8.361 Catalyst RC4 drivers, while the GeForce 8800 GTS used ForceWare 158.19 drivers.
http://www.dailytech.com/article.aspx?newsid=7043
Even AMD knows their K8 is garbage.
"Even AMD knows their K8 is garbage"
you should says, without ATI/AMD, core2 is garbage. Intel engineers has no expertise in producing 3D card to challenge AMD.
when barcelona is officially launched later, I suggest Intel sells all the garbage Core2 to the scrap iron companies. At least the garbage still can returned money back to intel. Wakakakakkaka... very funny..
The early news on the K10 looks impressive; however I’m not quite ready to do an act of worship yet until the main stream acknowledges this extraordinary accomplishment with proven benchmarks. Thrashing the competition by 66% doesn’t seem possible, especially when considering 2.5 ghz clock. I’m an AMD fan but lets get real.
pezal
"Intel engineers has no expertise in producing 3D card to challenge AMD"
Fact:
12 months ago AMD had developed exaactly ZERO 3D cards. Intel had developed them for years.
AMD had no internal knowledge about GPUs, they just bought the brains.
r
"Thrashing the competition by 66% doesn’t seem possible"
It is when you choose your benchmarks carefully
But where is the whole intelers stuff :))) ??? Why I can't see them now? A... for the well job done, perhaps now they are fired.
12 months ago AMD had developed exaactly ZERO 3D cards. Intel had developed them for years.
Im talking about the current situation not the past. BTW, Intel is far bigger compare to AMD+ATI, why still cant produce best 3d card to challenge AMD+ATI?? Isn't intel R&D focusing on how to make 3d card as well??
Sharikou
Intel's X6800 had about a 66% advantage in Integer, and about a 33% advantage in Floating point over the FX62 when it launched.
Explain how these benchmarks effect the real world abilities of these processors.
And again where are the real benchmarks ? I'm not interested in synthetic tests especially from fudzilla.
At the very least I want to see something like New ATI card test. It's not a complete test but it does include real applications
pezel
"BTW, Intel is far bigger compare to AMD+ATI, why still cant produce best 3d card to challenge AMD+ATI??"
If you define "best" as having biggest revenue then Intel has been dominating the market for years.
"Isn't intel R&D focusing on how to make 3d card as well??"
It just started with developing high-end GPUs and so far it seems as it will go way beyond any current rasterizing GPU. Past years Intel has shown a great interest in ray tracing and from what I know about labarre it will be a ray tracing monster.
middle_of_the_road
"At the very least I want to see something like New ATI card test"
Too bad HL2 results are badly wrong.
No comment from the AMD fanboi's of AMD's new loan for a cool 1.8 BILLION dollars.
Thats BILLION with a B.
So AMD sinks deeper into debt.
Meanwhile, AMD fanbois need to remember that:
1.) AMD bought ATI. Its the ONLY reason AMD has anything video related. They bought ATI because they had no choice as Intel is still currently the leader sales for video components. Notice I said SALES not performance fanbois.
2.) Intel is already working on faster performance integrated video. What happens with Intel's performance is close to that of ATI's? Whats the need for ATI then?
3.) AMD fanboi's also fail to take notice on exactly who Intel has working for them in the Video market... and that Bearlake and Larabee are quickly approaching.
4.) AMD isn't back on track. Is the K10 out yet? is it selling yet? When will it be available? Once again AMD fanboi's still keep comparing the K10 to Intel's CURRENT chips...
The HL2 results are out but the website admits that as well (dodgy driver, incorrectly setup PC, who knows). Heck of a lot better than the X is Y% better than Z junk we have been getting so far.
middle_of_the_road
"The HL2 results are out but the website admits that as well (dodgy driver, incorrectly setup PC, who knows)."
If one benchmark is so off (1920x1200 4xAA is twice as fast as their 1280x1024 test) then how can we know that those other numbers are correct?
Btw, that Quadro they compare with is based on old G71 architecture and is actually not much faster than 7900GTX. For one its memory bandwidth is roughly 1/3 of 8800GTX.
I wouldn't be drawing any final conclusions based on that single report.
Btw, that Quadro they compare with is based on old G71 architecture and is actually not much faster than 7900GTX. For one its memory bandwidth is roughly 1/3 of 8800GTX.
Not only that, Quadro cards are nowhere near as good for games as Geforce cards. Quadro cards are optimised for products like Maya, not games. Any good review of a Quadro card will show that.
P.S. Good to know AMD is using the world's fastest desktop processors to benchmark R600, the Core 2 Extreme. They know the scores will be embarrassingly low if they use an old K8 spaceheater garbage.
giant
"Not only that, Quadro cards are nowhere near as good for games as Geforce cards. Quadro cards are optimised for products like Maya, not games. Any good review of a Quadro card will show that."
Had you read it you would have seen that Quadro was only used for workstation programs. Games were tested against 640M 8800GTS
Still no comment from Ph(ake)d or the AMD fanbois...
Didn't think so... they more or less scatter when presented facts.
Kinda like cockroaches and the light...
Fact:
12 months ago AMD had developed exaactly ZERO 3D cards. Intel had developed them for years.
Not really. Intel never developed a graphics card, the only one was i740 that in reality is one REAL 3D STARFIGHTER, Intel acquired the company Real 3D. Intel has never done anything.
If you are talking of IGP than you do know about RD690, right?
Good to know AMD is using the world's fastest desktop processors to benchmark R600
Not AMD, but dailytech. AMD did not provide the computer.
It was not done in an Intel controllable environment.
aguia
"If you are talking of IGP than you do know about RD690, right?"
Yes, I am talking about IGP's, they are GPUs too, you know. Are Intel IGP's not made by themselves?
RD690 is designed by ATI, just that AMD bought them before it was finished. Also the original RD690 was canceled.
If the 2.5GHz K10 is 66% faster than the 3.0GHz quad, this means it is a full 200% faster in a clock for clock comparison. Sharikou's initial 200% calculation was correct all along. AMD now has Intel beat in every respect.
PENIX said... "If the 2.5GHz K10 is 66% faster than the 3.0GHz quad, this means it is a full 200% faster in a clock for clock comparison. Sharikou's initial 200% calculation was correct all along. AMD now has Intel beat in every respect."
So you & Sharikou went to the same School of Misapplied Maths.
Is 2.5 Ghz K10 is 66% faster than a 3Ghz quad, then at 3Ghz (1.2x more) it would be 79.2% faster.
Both are wrong. A 3 GHz K10 would be up to 99.2% faster than a 3 GHz "One Pony Trick".
And the winner is symbiansn!
Too bad though that you have to wait about a year before you can get 3GHz Barcelona. By that time you should also be able to get ~3.8GHz Penryn.
Penryn will have lowest-latency SSE instructions and better bit manipulation so as K10. Too bad Intel's monolythic FSB will hamper its multimedia prowess... AMD's engineering skills are just too much for them to follow...
Of course that could get fixed with more and more cache. Intel should use something like ReadyBoost, sort of a modular cache... their FSB stinks ass!
symbiansn
"Too bad Intel's monolythic FSB will hamper its multimedia prowess"
It might be surprising to you but vast majority of usage scenarios are not bandwidth limited and 1.33GHz is enough for most of those tasks. I wouldn't be surprised to see 1.6GHz on desktop PC's by next Christmas.
Of cource lower latency of IMC would boost Intel performance quite a bit and it could replace some of its cache with additional cores. That would be a boost AMD can't have any more, it already had it with K8.
symbiansn said...
"Both are wrong. A 3 GHz K10 would be up to 99.2% faster than a 3 GHz "One Pony Trick". "
If a K10 running at 2.5Ghz is 66% faster then if it is running at 3Ghz (1.2x faster clockspeed) then it is 66% x 1.2 = 79.2% faster than a 3Ghz quad (assuming that the benchmarks are realistic in any way)
I believe Intel may have tradeoffs to do between their Core (Nehalem) smart cache manager and IMC enhancements.
For the first time ever, AMD is implementing a full-force North Bridge onto K10, with built-in advanced prefetchers and modular MC, just like those of late era AXP chipsets such as Nforce2.
K10's L1 cache is the sole responsible for prefetching data. One cache set per core is allowed to do that, very possibly. That's a very balanced approach, because the four cores won't have to contend for data.
Intel superseded their Core 2 with cache management policies which can't do well with a lower-latency IMC. They need to revamp everything related to memory speed-ups in Nehalem.
In a nutshell, Intel can't priviledge caching and reach massive speed-ups with an IMC at the same time.
middle_of_the_road
"If a K10 running at 2.5Ghz is 66% faster then if it is running at 3Ghz (1.2x faster clockspeed) then it is 66% x 1.2 = 79.2% faster than a 3Ghz quad "
Time for some math 101!
2.5 - 166%
3.0 - 100%
2.5 * 100 * X = 3.0 * 166
250 * X = 498
X = 498 / 250
X = 1.992
IIRC I learnt that in seventh grade or earlier.
synbiansn
"In a nutshell, Intel can't priviledge caching and reach massive speed-ups with an IMC at the same time."
That is irrevelant since Core2 won't be the one that would use IMC. Or do you honestly believe Intel will use Core2 in H2 2008?
In response to middle_of_the_road, symbiansn and ho ho...
For my logic I assumed a true 2/3 in place of the 66% estimate. This calculates to an even 2, or 200%. If your numbers differentiate, please revisit your elementary school math class.
penix
"This calculates to an even 2, or 200%"
Please either show where I made a mistake or provide your own calculations. One of us must be wrong.
Ahhh, I see my mistake. It should have been 166*1.2 not 66*1.2
Always happens when you try to take a short cut
According to Mercury, AMD was left with 19 per cent of the x86 market in the first quarter of 2007, while Via took 1 per cent.
AMD's glory days ran throughout 2006 when it ate up more than 20 per cent of the x86 market and as much as 25 per cent of the market in the fourth quarter. A resurgent Intel, however, has punished the smaller rival via better product and brutal pricing.
Executives at AMD have claimed the company needs at least 30 per cent of the x86 market to post a profit on a regular basis.
Proving this point, AMD reported a $600m loss during its most recent quarter, as processor sales plummeted close to 40 per cent.
AMD lost ground in the notebook (four points of share), desktop (eight points) and server segments (seven points) quarter-on-quarter, according to Mercury
Sharikou's initial 200% calculation was correct all along. AMD now has Intel beat in every respect.
My numbers were derived analysis of the two architectures. I also stated that K10 has 40% integer lead, which is quite close also.
People have to distinguish an application benchmark from a synthetic benchmark, while AMD has 40% integer advantage and smashing FP advantage, in real applications, such as games, the lead may not be so dramatic. But in simple minded media tasks, expect to see Intel crushed like a scumbug.
AMD was left with 19 per cent of the x86 market in the first quarter of 2007
Expect AMD to come back in a big way in 2Q07, as far as market share is concerned. AMD has a lot of good processors to sell at very competitive prices.
Expect AMD to come back in a big way in 2Q07
So that's it? AMD will BK Intel in one quarter?
Everything is going perfectly for the architecture of K10 and r600.
I again just can't believe these ridicules Intels fanboys, they really just don't understand!
2.5Ghz K10 is faster then 3Ghz Core 2 by 45%.
In 3dMark06 Intel Core 2 has a scaling of 5% per 270Mhz
in Windows media encoding and lame
the difference is from 2 to 3% between QX6700 and QX6800
in Cinebench and POV-Ray is 10% the difference
Sisoft sandra integer and floating point again 10% the difference for 270Mhz
ecc, ecc...
If Intel Penryn wants to compete with 2.5Ghz K10 in integer and floating point, Penryn needs approximately 4.62Ghz
Imagine the difference of K10 at 1.8Ghz overclocked at 3Ghz what will do to the fastest Penryn
What to say more.
Go AMD-ATI!
Lol
Wow, leecooper.
You're such a moron that you copy-paste posts between threads.
Here's a clue, if you can't get that pea sized brain of yours to think of something original, you should keep your hands in your pants.
copy-paste posts?
What are you talking beetle?
sharikou
"Expect AMD to come back in a big way in 2Q07"
What would be the thing to push a lot of people to buying K8's? There is no Barcelona in Q2 and R600 is negligble. Let me guess, AMD pulled back so far in Q1 that now it has insane amount of impuse to punch Intel into BK?
leecooper, if you'd check you'd see that QX6800 is 2.93GHz, not 3GHz. 2.93/2.66=10.2% clock speed difference.
leecooper
What are you talking beetle?
He is talking about you posting exactly the same things to as two comments on this blog.
Ho Ho said...
Please either show where I made a mistake or provide your own calculations. One of us must be wrong.
My response is aimed more at the first responses. Your calculations are correct when using 66%.
He said 07?!??!
OMG, I thought he said 08!
Forget it, something like then isn't even worth mentioning.
:)hm,
Yes, I made a mistake to post a comment below "AMD = performance", Sharikou can delete one of my posts, but I wasn't wrong about Intel ;).
What's wrong with this guy 'bubba'?
Amd is knocking on his head.
He maybe realized some things in his life.
Maybe he is enlighten.
All of you Intels fanboys must be enlighten ;).
Lol
Yawn...............
Another "Ph"ony post from the "Ph"ony "D"octorate
Lets put this in prospective.
66% and 33% look reasonable. I'm sure for a very specific application they numbers are probably close enough to real performance. Its funny that it didn't come with some independent observation. Any annoucement by AMD under their current duress is suspect. Actuall I'm dissapointed they didn't release any more detail application benchmarks. What a dissapointment for AMD and good news for INTEL. There is nothing to fear from INTEL. Penrym is out, it ain't going to change much except to fix bugs and speed improvements. Nehalem is locked and loaded, it ain't going to change much no matter what AMD shows. To now show is to confirm your weakness. And what might that weakness be? If you were hurting for cash do you go out and show how crappy your product is in real world applications? NO and what is AMD doing right now?
"Ph"ony did you hear about that new 1.8 billion debt placement, that tells you where AMD is and why you don't see more Barcebologna.
Bottom line: Barcebolgna is simply NOT enough. It simply doesen't change the overall business outlook for AMD. As such the "Ph"onys continued talk about INTEL collapse, BK, and Patty stock selling, INTEL Marketshare collapse are so laughable and confirms he is a "Ph"ony "D"octorate.
If everyone in the world wanted a Barcebologna over a INTEL Clovertown; Google, and every server OEM and gamer, AMD couldn't supply them. The demand could drive prices to above 1K but again as the price rises the value drops and a balance will be achieved and people will turn to buying INTEL for the value proposition.
The fundamental issue for AMD and one they simply the can never fix is that they have but ONE palty 65nm fab that can't even support 30% marketshare of the quadcore server space. That Barcelbologna is a honking 283mm^2. Selling those units at some high price will be too late to resecue AMD finances as they won't have enoughvolume! Selling at the the lower price once Nehalme arrives will result in a loss as the die is too big to make money on AMD 65nm wiht their current yields.
The consumer and mobile space which represents 90% of the unit volume will be owned by INTEL with their amoritzed 65nm factories and their new 45nm factories which will flood the market with C2D and Penrym CPUs in the 2nd half of 2007. Nehalem is arond the corner in 2008 so Barcleona will only have a 9 month window and even at thousand dollar prices won't provide enough revenue from AMD's one fab to sustain them. Let me remind you fanbois Penrym is 107mm^2 compared to 283mm^2 for Barcebologna. INTEL has 4 factories to AMD's ONE. Add to the fact you have 4 factories running C2D for INTEL at 143mm^2 / chip. The capacity to pump competitive die at a profit says INTEL will move from 80% MS to 85% MS by Q3.
"Ph"ony any comments about them Q1 Marketshare numbers. What happened to your earlier predictions heah there?
AMD will be going to the bank another time in Q3 as they are torn between spending their new money on 65nm capacity while still running in negative cash flow with little to show for it as Barcebalogna won't provide any signficant revenue.
Sorry your "Ph"ony "D"octorate and "Ph"ony thinking are so easily exposed. You really could help yourself and be more on topic and tell us a little about your qualifications to continue to use the PhD title.
INTEL ain't going BK, INTEL cpus don't cause fires. INTEL cpus don't get frag, INTEL isn't losing Market Share.
Intel is gaining Market Share, Growing Revenue, Making billions of Profits.
Looks like independent R600 benches will be put up within the month or so and it would be nice for K10 benches to be released about the same time so we can finally judge these two entrants.
AMD has done a good job of keeping K10 under wraps but if it is to be brought out in Q2 then that is June latest.
"2.5Ghz K10 is faster then 3Ghz Core 2 by 45%."
This is ridiculous... there's no way such benchmark metric could be accurate or reflective of real-world results.
K10 might be 20-40% faster than Core 2 at the same clock. However, we really don't know how K10 would fare when compared to a Core 2 that has 20% clock advantage, like described above.
S W E E T
This comment has been removed by the author.
Sharikou,
just remember what you've said. we'll all be watching in Q208.
I would say by the end of Q2 this year he'll be changing his tune...
well he already claimed Intel will BK in 4Q07, while suffering losses as early as 3Q06.
http://sharikou.blogspot.com/2006/
06/intel-may-suffer-losses-in-2q06-
3q06.html
so you know how credible his prediction is.
From the link BK article:
2Q07: Intel server market share drops to 40% as AMD ramps Rev H quadcore. Bulldozer hits hard.
So this quarter some time, AMD's quad core will be released and ramp up incredibly quickly, taking half of Intel's 80% of the server market! Wow!
Post a Comment
<< Home