SpecInt/SpecFP - Intel vs AMD

Page 7 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
To be honest I hadn't noticed the bios before. I had forgotten about the information that was filled out when I joined this forum. At any rate, the information is only as trustworthy as the person who fills it in. There could be 10 AMD marketing guys frequenting this forum and noone would be the wiser.

-Raystonn

-- The center of your digital world --
 
But I once again digress, can we know get back to the heart of the discusion before Mr Intel OEM got us of on a tangent. I do beleive we were discusing the merits/ or lack there of one specific benchmark that you opened this thread with. Seeing I have neither the time nor the resources to continue this debate on two seperate fronts, shall we continue with the original and leave the information Cyberimage has brought to fold for another thread?

A little bit of knowledge is a dangerous thing!
 
thanks for the support ..
and the other 2 should follow in 3rd qtr hopefully..
best
CAMERON

CYBERIMAGE
<A HREF="http://www.4CyberImage.com " target="_new">http://www.4CyberImage.com </A>
Ultra High Performance Computers-
 
Exactly..Intel has starting making processors at .13 micron at one plant....Whoopty doo. Can this one plant meet the demands for this processor alone? Well maybe at its current demand rate 😉

A little bit of knowledge is a dangerous thing!
 
Until we find out more about the exact specs on that AMD system, there's not much more to debate on the subject. At 1.3GHz the P4 and Athlon are both pretty close in Integer and the P4 always takes the lead in floating point. This is the slower speed at which the P4 is available. At higher speeds the P4 just takes off beyond the Athlon's current offerings.

Price is really the only barrier to the P4 at this point. If you want something that performs better than the best Athlon, there really is no alternative other than the P4, it's just expensive at the moment. This will be changing very soon.

-Raystonn

-- The center of your digital world --
 
I'll make it simple for you
http://news.cnet.com/news/0-1006-200-5530934.html#

here is a link, goto page and click on the video link 1\2 way down the page click the CPU die picture
entitles KEVIN KREWELL..
HE WAS A CPU ENGINEER AT AMD LAST YEAR SO YOU THINK HE WOULD KNOW HUH

play the video and LISTEN AS HE WOULD KNOW FOR SURE

CAMERON

CYBERIMAGE
<A HREF="http://www.4CyberImage.com " target="_new">http://www.4CyberImage.com </A>
Ultra High Performance Computers-
 
This precise version of the P4 is not yet for sale. It's the one with the new socket that will be released in Q3.

-Raystonn

-- The center of your digital world --
 
I see you carefully avoided the Intel/Cnet investment arragement. Never did I question the fact Intel had in fact in one plant started to manufactorer at .13. What I would like you to prove is the claim you made as stateing the source I made as to be bogus.

A little bit of knowledge is a dangerous thing!
 
well the number of retractions and lawsuits and embarrassing
misinfo they have reported in teh last year, not to mention when you ask an AMD or INTEL or MS engineer about the truth
and what they think of the register, they laugh out loud
usually


I can tell you comparing the news credibility of the register and saying it is on par with CNET, EETIMES< or MICROPROCESSOR REPORT is the stupidest thing anyone who knows anything about the IT industry will ever hear
quit while you are behind PLEASE
its embarrassing

CAMERON

CYBERIMAGE
<A HREF="http://www.4CyberImage.com " target="_new">http://www.4CyberImage.com </A>
Ultra High Performance Computers-
 
the person I referenced as saying what I said and proving my argument is KEVIN KREWELL amoung others..

see his video on cnet about INTELs manufacturing advantages..

here are his qualifications in case you doubt them..

Kevin Krewell, a 20-year veteran of the electronics industry, comes to MDR from Advanced Micro Devices, where he served in technical marketing and field application engineer roles over a 10-1/2 year period, working directly with customers such as IBM, Compaq, Acer America, 3Com, NEC, and Symbol Technology. At MDR, he will focus on the PC industry he has worked in for the last decade. He has recently updated and improved the model MDR uses for the Intel Microprocessor Forecast.

At AMD, Mr. Krewell provided technical support for the launch of the AMD-K6-2 and 3D-Now technology and the AMD Athlon processor, the first 7-th generation processor. He has presented at various forums including WINHEC, Comdex, PC Expo and AMD'S Developer Forum and represented AMD at various Industry and Government forums including EPA's Energy Star Program and DoD MTOPs. He was a contributor to a book on AMD - The AMD-K7 3D Processor.

now you are trying to tell me that your register article
written by some 3rd rate gossip columnist which cites no credible source or proof or expert is on par with what I cited.. stricly amateur gossip, not in the pro league
give me a [-peep-] break
CAMERON


CYBERIMAGE
<A HREF="http://www.4CyberImage.com " target="_new">http://www.4CyberImage.com </A>
Ultra High Performance Computers-
 
"Until we find out more about the exact specs on that AMD system, there's not much more to debate on the subject. At 1.3GHz the P4 and Athlon are both pretty close in Integer and the P4 always takes the lead in floating point. This is the slower speed at which the P4 is available. At higher speeds the P4 just takes off beyond the Athlon's current offerings."

Actually the Athlon has shown to be quite superior in FPU strength. The P4 gets its strength from software that is sse2 optimized. Personally, I do not have a problem with this and I hope that more SSE2 software is developed. I have yet to see the blazing speed in developement of sse2 optimizations, of which everyone pro-intel suggested at the onset of the p4's release happen.

Now in the test you provided as proof of your argument, the test's fail to accuratly describe the test bed platform of the athlon CPU. Furthermore, you made a claim that:

"This is a highly respected, industry accepted benchmark. Are you trying to make the libelous statement that Intel fudged the numbers? I don't think so. Look at the numbers submitted by other companies using the same CPUs, they are close to the same numbers. Are you now saying that they _all_ fudged their numbers? I wouldn't bet on that."

Well, I would not say that myself at all ( as a matter of fact I never did :0) ) It does appear however that if one were to try to fudge there numbers it would not be all that difficult otherwise the results posted by what appears to be a gigabyte motherboard and DDR ram surely would not be listed as an asus motherboard with sdr ram on a slower processor. Can we say quality control? Now before you go and argue the fact that It was AMD's fault ( which may or may not be true) it would certainly appear that SPEC should at least check the validity of the results before they were posted on there website, they are afterall charging for there software.
I would also at this time like to revisit my contention that if the athlon was placed on a different, better performing platform ( in this case the asus board) and given the benefit of the faster cas2 DDR Ram the results might in fact be totally different. Or, shall we instead show results using the p4 coupled with the slower pc600 RDRAM?
Unfortuantly the only way we will know if one of us is to buy the software suite in acess of $500.00 to do these actual test.
So, in summary, you have made a claim to the superority of the p4 vs the athlon. To support your claim you have choosen an AMD platform with whose specs we do not know and can not validate. Additionally, we can not do the same tests on or own, to test any possible improvements that may have come to fold. Now, in my opinion, to which I am sure you will take issue, this makes your opening statement rather week. Perhaps we should move to another benchmark that we ourselves can easily duplicate?


A little bit of knowledge is a dangerous thing!
 
Ok now this is taken from your own link in which you are so adimate about:

While more speed boosts sales, it is only one measure of performance, and on many standard applications, Athlon outpaces the Pentium 4. The Pentium 4, however, does better on video compression and many multimedia applications.

"On certain things like streaming media, the Pentium 4 does extremely well. On the standard benchmarks, the Athlon does extremely well," said MicroDesign Resources analyst Kevin Krewell.

Clearly this is showing us that the p4 is not the best in all applications. Furthermore Kevin goes on to explain that with the release of the p4 at 1.7 gig Intel is finally able to meet the performance of a much more slower clocked athlon in standard apps. Finally, at 1.7 gig the p4 can possibly lay claim to being the fastest. At least until a faster clocked T-bird is released. The palomino has already been debuted at CEBIT at 1.5. current athlons are reaching speeds ( overclocked) in excess of 1.6 gig. Two weeks from now perhaps we should revisit this to see just exactly how the price cuts has effected the overall scenario. You still have the RDRAM cost to content with.
In my opinion Intel will not have a real threat to AMD until the next version of the p4 (northwood) is released. Then I believe, AMD has possibly a major problem to contend with. With the northwood, Intel has the ability to re-incorporate the casterated fpu it ripped out of the P4 along with adding back the cache. If intel chooses to go this route, whch I sincerly hope they do, I myself might be boosting Intel inside once again.


A little bit of knowledge is a dangerous thing!
 
Furthermore, I would like to add that before you came to fold in this thread we were in actuality discussing a clock for clock comparison. The links you provide us with is comparing the 1.7 p4 to the 1.33 athlon. Intel does indeed have the current advantage in scalability. A crown AMD held until the release of the p4 as shown by the p3 1.13. If then scalability is the only measurement of better archtiture, in which I do not agree, you can make the claim of the p4 being superior. My contention is if the p4 does indeed have better architecture, it should beet the athlon on a clock for clock basis and not need a 400 mhz advantage.

A little bit of knowledge is a dangerous thing!
 
Yes the figures are impressive and look very good indeed but what are they telling us? HOW is the P4 better?

The figures for FP are bandwidth limited, a modern CPU can process the numbers faster than the system bus can ship them to it. The on-die caches for CPUs are not much help, the large arrays of data used by the applications in SPEC don't fit into them. When it comes down to it the massive bandwidth of the P4 platform provides a real benefit.

In addition the hardware prefetcher helps by starting to fetch some data much earlier than it naturally would be and thus reduces the latency. Finally some SSE-2 instructions are used which can provide significant performance benefits. The only problem with SSE-2 is that only the Intel compiler currently supports it and requires that the code is written in a certain way to make use of it. It has also been reported that the Intel compiler is quite flaky and fails to produce reasonable code in many other circumstances (some have suggested that many of the optimisations in the compiler have been put there purely to improve the SPEC score). There are no figures to suggest how much each of these strengths contribute to the P4s great score.

It has to be said then that if you want to (and can) compile your heavy FP apps with the Intel compiler, rather than the MS one used by 90%+ of developers, then you will probably see an increase in performance, so long as you don't mind creating an executable which runs only on a P4. I would say the market for this sort of app is rather small and will remain so for some time to come though some companies may produce P4 specific binaries for some consumer video and audio apps.

Also bear in mind that the Palomino will include a hardware prefetcher which will provide significant benefits for FP code and increase efficiency.

As far as integer apps go I am amazed at how you can suggest averaging the base and peak scores as though it is valid in any way. The scores are very good for both platforms and show what is theoretically possible. Looking at current benchmarks show that in the majority of apps the Athlon performs better but the P4 could perform better if they were all recompiled for it. Again the Palomino will have a better branch prediction unit so integer performance should improve noticeably.

You have some valid points but your attitude and lack of proper understanding of what is happening mean that the impact of your post is somewhat limited

Bottom line, P4 is good on many (not all) programs if you want to heavily optimise for it but Athlon is not far behind when using non-optimised code and would dominate at the same clock speed.

L
 
All profesional users owe a debt of gratitude to a company like yours. After I while you tend to buy things you can trust, like Intel cpu's and seagate hard drives. Most of the people here don't do much besides surf the net. They don't understand quality or reliability. Good on you and stick to your guns, because the profesional comunity needs you.

"Cock-a-doodle-do" is what I say to my girl when I wake her UP in the morning!!
 
"this makes your opening statement rather week"

Not really, if AMD wants to use subpar equipment in their benchmark results, it only shows how stupid they really are.

"Perhaps we should move to another benchmark that we ourselves can easily duplicate?"

If you'd like we can run other benchmarks in addition to this. They must, however, be optimizaed for all processors, including the P4.

-Raystonn

-- The center of your digital world --
 
Not really, if AMD wants to use subpar equipment in their benchmark results, it only shows how stupid they really are.

Not at all, it only shows that at the time of testing this may have been the best equipment available to them. In todays world a month can bring quite a change in equipment.


"If you'd like we can run other benchmarks in addition to this. They must, however, be optimizaed for all processors, including the P4."

Why is this? If the p4 is superior as you claim, it should not need optimizations specifically designed for it. If you want to make the claim that when using sse2 optimizations, the p4 is the better processor, I do not think I would have so adimattly defended the Athlon. Superiority should also mean cross-compatability ( a better term at the moment escapes me), something that seems to be lacking in the current version of the p4.


A little bit of knowledge is a dangerous thing!
 
"Why is this? If the p4 is superior as you claim, it should not need optimizations specifically designed for it."

Incorrect. Today's CPUs are optimizing by pushing for operating as much in parallel as possible. This means optimizing for the P4's pipeline as well as the others. If you're going to produce code targetted at reducing pipeline stalls and branch mispredictions on one CPU, you'd best make sure the code does this for all CPUs. Otherwise you give one set of CPUs an advantage. Note that this has nothing to do with SSE/SSE2.

-Raystonn

-- The center of your digital world --
 
Well good then, you are saying that the p4 can compare with the athlon clock for clock without sse2 optimizations? This bet I would love to take up with you. Otherwise, the p4 will only excell when using apps written with sse2.

A little bit of knowledge is a dangerous thing!
 
"so long as you don't mind creating an executable which runs only on a P4."

Not true at all. Standard practice for incorporating CPU specific optimizations is to use a set of function pointers or a class hierarchy that allows you to break the important code into multiple source files. You then build each source file targetting a different CPU, and link them together into the final binary executable. This allows you to pick which optimized function set to use at runtime based on what CPU you are using.

"I would say the market for this sort of app is rather small and will remain so for some time to come"

Not at all. All DirectX/OpenGL drivers will have optimizations for all the CPUs. In fact, they've had optimizations for all previous CPUs for quite a while now. This will directly and immediately impact all applications released for these APIs without having to recompile any of them. MMX and 3DNow worked the same way.

-Raystonn


-- The center of your digital world --
 
I wouldn't know how the P4 would fare in such a test. Name a single benchmark that doesn't take full advantage of AMD's 3DNow offerings. If you're going to remove extra instructions from the P4, do the same for the Athlon.

-Raystonn

-- The center of your digital world --
 
*" This is because the memory bus, which is shared by both processors can't keep up with the memory demands of two simultaneous runs. The CPUs are data starved."*

Now this is where the AMD 760MP (EV6 Bus Architecture) comes in...
The AMD CPUs have a dedicated bus for syching their Caches, rather than taking bandwidth from the Memory bus like the P3 does... now if only they would hurry and get those 760MP boards out to the public... btw.. haven't heard much on this... Will the 760MP chipset support Dual Durons? Seems to me that unless AMD has disabled the functionality, the Duron is a close enough relative to the T-bird that it ought to be SMP-compliant also...

--Fltsimbuff
 
I've really seen RDRAM as a joke when paired with a P3 system... Especially when compared to DDR SDRAM. However, From what I've heard, the P4 taking full advantage of RDRAM's bandwidth can blow DDR SDRAM away...

--Fltsimbuff