SpecInt/SpecFP - Intel vs AMD

Page 4 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
To keep things in cronilogical order, which you choose to ignore, here is an excerp from Tom's review after the kt133a chipset came out in which he clearly makes reference to to earlier test in which you choose to question his conclusions:


This kind of phenomenon is not new at all. Whenever a new technology is introduced, it is susceptible to failures and criticism. DDR-memory is seen as the proper companion to AMD's latest Athlon-processor with 133 (266) MHz processor bus clock ('C'-types), which was officially announced in combination with the still unavailable AMD760 DDR-chipset two months ago. However, the recent article 'The Three Musketeers' already showed that the 133 (266) MHz FSB Athlons are able to outperform their 100 (200) MHz FSB ('B'-type) counterparts even with normal PC133 SDRAM already.

Instead of prematurely reporting on those DDR-motherboards that nobody is able to buy, we chose to introduce six new motherboards with VIA's latest Athlon chipset, the Apollo KT133A. The only difference to its predecessor Apollo KT133 is the added support for the above mentioned 133 (266) MHz processor bus clock used by the new 'C'-type Athlons. This beefed up Athlon-chipset from VIA might not offer support of DDR-memory, but it resembles mature technology, and provides reliability as well as good performance.


And even this test is now a bit dated as the amd 760 chipset are available. In the future please refrain from drawing conclusions until you have all the facts. There is nothing worse than someone questioning someones integrity on his own site if the accusations are unfounded.

A little bit of knowledge is a dangerous thing!
 
My 'beef' is not at all with anything you suggested, but merely with Tom's assertion that DDR is faster than SDRAM in that test because the 133MHz FSB outperformed the 100MHz FSB. Well what does that have to do with memory type? Absolutely nothing. The test using the same FSB showed identical scores. This indicates the performance increase was due to the FSB increase alone.

Therefore, I do question the conclusions in his benchmarks. The point we're debating now though is not DDR vs SDRAM but whether Spec is a reliable benchmark for the CPU and memory subsystem; and it is. Raw numbers are always better than a reviewers claims. Checking on Tom's raw numbers, his claims are a bit off.

-Raystonn

-- The center of your digital world --
 
Looks like nobody has contributed any technical facts to my dual cpu statement. Nobody has refuted any of information mentioned at the sites I stated prior to this post. Good.

"Cock-a-doodle-do" is what I say to my girl when I wake her UP in the morning!!
 
Actually we have choosen to ignore you but we are curious as to how your freind still works for a company that ceased to exist.

A little bit of knowledge is a dangerous thing!
 
Don't be so quick to dance around the issue. You made a statement, a statement which was false. You brought the issue to this forum right here and I quote you:


Tom's benchmarks aren't very reliable. Take, for example, the SDRAM vs DDR benchmark comparisons: http://www.tomshardware.com/mainboard/00q4/001030/athlon-15.html

Quite obviously you question the reliability as well as integrity of Tom's testing. You fail to mention that at the time of this review it was not possible to test a 266 fsb athlon with pc-133 SDRAM as the chipset to support it was not out. When one became available and it was possible to test the 266 fsb athlon with sdram and compare it to the ddr ram solutions he did so right here and still found there to be a significant increase in performance.
Seeing how at the time he was using the only available chipset which required a 100mhz fsb processor for comparisons his conclusions were accurate. As a chipset became available which would allow for the use of the 133 fsb athlons naturally these test had to be revisited and revised, that however does not mean that his prior conclusions were wrong as they were most definatly not, only that there was now a second option that did in fact allow an exact processor fsb bus comparison.
And guess what? A point you continue to ignore that even with the same fsb at 133 (or 266 double pumped) the ddr still did in fact give a significant performace increse over sdram.

The fact which you seem to be trying to make is that when DDR ram is run with a athlon that uses a 100 FSB it shows no siginifact performace increase over the pc-133 sdram solution.Ok. This I never questioned and I will agree with you on this one point. You do need a processor that is able to take advantage of the increased bandwith of the DDR memory. And The Athlon "c" is such a processor. To draw a conclusion about DDR RAM based soley on 100 FSB comparisons is rather weak to say the least.
Another quote:

"This indicates the performance increase was due to the FSB increase alone."

No, not exactly, please refer to these tests after the 1.2 gig athlon c was coupled with a chipset that supported 133 fsb and pc-133 SDRAM and compared to DDR memory

"http://www.tomshardware.com/mainboard/01q1/010314/amd760-08.html#mpeg_4_encoding_flask_mpeg

As you can plainly see your assumtions are incorrect. As I do undertand this was not the intented prpose of this thread to discuss the merits and or lack of DDR Ram perhpaps we should bring this debate to another forum. However, bear in mind that it was you that choose to open this can of worms by clearly stateing that Tom's testing was not reliable. I found it neccesary to discredit you on this count and this count alone and to do so it was necessare in detail to show you why you were wrong.

A little bit of knowledge is a dangerous thing!
 
Hmnn. Maybe we should be comparing benchmarks that ARE NOT SSE2 enabled. That would be more fair, no? Also more real world since there aren't many SSE2 enabled apps out there yet.
 
"The issue" has nothing to do with SDRAM and DDR. It has everything to do with the fact that Tom made an assertion that is not true. I could really care less about benchmarks he did not perform. I am discussing those he _did_ perform. None of the thigns you mentioned has any bearing on the fact that in the benchmark he did perform, his concluding claims after showing the data did not match with the data. He was wrong.

All I'm saying here is that the claims of reviewers do not outweigh numerical data/analysis. This goes for all reviewers, not just Tom. I don't really care whether DDR is faster or not as that's not the purpose of bringing up that benchmark. The purpose is to show that a reviewer can state a conclusion which is in direct contrast to the numbers he just showed. We're lucky he chose to show the numbers, else we'd have no way to double check him. Spec is a benchmark that shows a great deal of data for each benched system. It does not suffer from these editorial conclusions that may or may not be based in fact.

-Raystonn

-- The center of your digital world --
 
We're testing the performance of one CPU (and memory subsystem) against another CPU (and memory subsystem). This means testing the full capabilities of each CPU. We're not doing this relative to what apps are on the market. SSE2 apps are going to start springing up bigtime in the very near future. Even AMD will be supporting SSE2 in their next processors.

-Raystonn

-- The center of your digital world --
 
Ok. Then let's compare them using a SSE2 enabled benchmark when they both are SSE2 enabled. Or, if AMD is not enabled by the time SSE2 startes surfacing in apps then Intel is all the better. I recognize Intel is ahead of the game here by includeing SSE2 in the P4. All I'm saying is that to compare raw processing power we should play on a level playing field.
 
No my freind it is you that is wrong. Please explain how you can draw such a conclusion. The test Tom did with a 1.2 gig athlon using DDR clearly shows that it outperformed the SDR equiped athlon at 1.2 gig. see for yourself:

http://www.tomshardware.com/mainboard/00q4/001030/athlon-15.html

If you wish to argue that this was unfare do to the fact the SDR was limited to a 100 FSB that is fine. However at the time of the reveiw this was the only availble platform for the athlon 1.2 gig with SDRAM. This makes Tom's conclusion right and you wrong. Face it!

If you wish to compare the scores at 100 mhz and slower DDR ram for both platforms then so be it. However this point is addressed in Tom's reveiw as well and still makes his conclusions correct.
One would hope that the claims of reveiwers are in fact backed up by numerical data/analysis as to which I beleive Tom's are.


A little bit of knowledge is a dangerous thing!<P ID="edit"><FONT SIZE=-1><EM>Edited by Ncogneto on 04/15/01 02:59 PM.</EM></FONT></P>
 
Your Quote

"The 100MHz memory bus tests using SDRAM and DDR are identical. The speed increase occurs when he looks at the 133MHz bus. Those performance increases are due mostly to an increase in the memory bus, not the DDR memory. Tom states, "The AMD760-platform with Athlon 1200/133 MHz is performing extremely well, scoring almost 11% better than an Athlon 1200/100 on a Asus A7V motherboard." Well gee, you're giving it a 33% memory bus speed increase.

"SDRAM would give you the same numbers when performing at 133MHz."

Unfortunately numbers do not back up this claim that you present. Please look at these benchmarks just once. And yes They are done with the a7v133 using a 133 fsb. You can plainly see your claim is the claim that is incorrect.

http://www.tomshardware.com/mainboard/01q1/010314/amd760-08.html#mpeg_4_encoding_flask_mpeg

A little bit of knowledge is a dangerous thing!
 
"All I'm saying is that to compare raw processing power we should play on a level playing field."

That's not true at all. We should be measuring the maximum performance of each CPU with all the bells and whistles. The performance of one CPU has nothing to do with the capabilities or performance of another.

-Raystonn

-- The center of your digital world --
 
I've looked at the numbers over and over. You cannot compare a 133MHz frontside-bus DDR system with a 100MHz frontside-bus SDRAM system and claim that it's the memory that's faster when the 133MHz system perform better. I've got no doubt that the first system will perform better than the second system, but this doesn't mean it's the RAM that's causing it. This is especially evident when the data in the results clearly shows a 100MHz DDR system that's performing identically to a 100MHZ SDRAM system.

However, I don't want to continue debating Tom's benchmark forever. If we cannot agree on it then we should just drop it and move to some common ground. Otherwise this will be unproductive and soon the "monkey" people will start showing up, flinging bananas and turning this into another pointless flame war.

My original point was that Spec is a reliable benchmark. The massive amount of data that's provided is very valuable. Every set of results has a link on the right hand side where you can get very detailed reporting on the exact system specs and exactly how it performed on every sub-benchmark within the whole.

-Raystonn

-- The center of your digital world --
 
Unfortunately for you, Tom's review isn't the only one.

Check this out:
<A HREF="http://www.anandtech.com/showdoc.html?i=1448&p=5" target="_new">http://www.anandtech.com/showdoc.html?i=1448&p=5</A>

It compares PC2100, PC1600, PC100, PC133, mixed with Athlon C and B. Go look.

--------
Time is never time at all. You can never ever leave, without leaving a piece of you.
 
you quote:

"I've looked at the numbers over and over. You cannot compare a 133MHz frontside-bus DDR system with a 100MHz frontside-bus SDRAM system and claim that it's the memory that's faster when the 133MHz system perform better. "

Well, yes you can when at the time of the comparison there is no 133 fsb SDR solution available....get it? So what is your point? Yes NOW there are such systems with a 133 fsb and sdr ram to test... so lets do just that and see if your claim that the added performance was only do to the fsb increase in the DDR system as you keep saying.......

http://www.tomshardware.com/mainboard/01q1/010314/amd760-08.html#mpeg_4_encoding_flask_mpeg

OH OO! LOOKY MOMMY! The increase in performace was not only do to the fact of the increased FSB! DDR actually had alot to do with it! What do I do now? Hmmmm, well an apology addressed at Tom might be expected.

A little bit of knowledge is a dangerous thing!
 
Other reviews are pointless. I don't care if DDR is faster than SDRAM. I keep having to repeat this. I was merely pointing out a benchmark in which his conclusion, at the time he performed the test, did not match his data, at the time he performed his test. That's a flaw with his testing procedure, not anything at all to do with what he was testing.

-Raystonn

-- The center of your digital world --
 
"Well, yes you can when at the time of the comparison there is no 133 fsb SDR solution available"

No, you can't. You cannot compare the two and get any relevant information as to whether the RAM is truly giving you a performance benefit. It might be the RAM, it might be the frontside-bus. There's no way to tell when comparing those two situations. When you do a comparison measuring one component you must ensure all other components are identical, otherwise your results are meaningless.

Because we don't know which of the two changes to the system caused the increase in performance, we take a look at some of the other data and see a DDR and SDRAM system both with the same frontside bus (100MHz). The results are identical. This makes one lean toward the frontside-bus being a larger contributor to system performance than the type of RAM being used.

All of this is standard scientific method.

-Raystonn

-- The center of your digital world --
 
Raystonn, itss amazing how many times you have to repeat this stuff to the lemmings, and they are still in denial.

Don't give in to the lemmings. they will try to wear you down by splitting hairs and rewording your posts to fit their needs.

Lemmings think they are smarter than Intel engineers and do everything they can to dicredit and belittle you.
 
Well then Ray, aren't you yourself guilty of the same thing? You made the conclusion that it was not in fact the DDR memory that gave the increase but in fact the increased FSB of the processor. This conclusion you made without checking benchmark results that were at the time readily availble to you, a luxury that at the time of Tom's testing he himself did not have. After looking at these results you find that the 10% performance increase statement did hold true in Tom's conclusionary statement and made your statement, which I quote here , the one that was in error:

"SDRAM would give you the same numbers when performing at 133MHz."

A little bit of knowledge is a dangerous thing!
 
So now that I've proved you wrong, you're in denial? Give me a break. You're pathetic.

From now on when someone asks you how fast your computer is, tell them your benchmark scores.
 
"Well then Ray, aren't you yourself guilty of the same thing?"

Guilty of restating Tom's numbers in plain english? Again, I'm only analyzing this one benchmark test of Tom's. The point is not to show whether DDR is faster than SDRAM, but to show that Tom's statements at the time did not accurately reflect the data he had available to him at the time. Surely he didn't have access to a time machine to go check whether DDR was going to be shown faster when new buses came out in the future. Therefore his conclusions should have been restricted to the data he produced in his benchmarks. I've already shown that not to be the case.

-Raystonn

-- The center of your digital world --
 
You've proven nothing. The argument was that Tom's conclusion in that benchmark did not reflect his data in that benchmark. The given propositions were the data he showed in the tables and his concluding statements. The propositions logically support my argument, and hence my conclusion.

If you're arguing about DDR being faster than SDRAM, then you're arguing an entirely different thing. I agree with you that in fact DDR is somewhat faster than SDRAM. But again, this is an entirely different subject.

-Raystonn

-- The center of your digital world --
 
The arguement was that DDR memory did not give any more performance. Your claim was that the benchmarks between DDR and SDR were unfair, that SDR were 100MHz FSB and DDR were 133MHz FSB and that was the difference. I've shown you that DDR does infact increase performance, when compared against a SDR system of the same bus frequency.

From now on when someone asks you how fast your computer is, tell them your benchmark scores.
 
This goes to show that communication is key in a discussion. We are arguing two entirely different things. The trigger of this line of discussion in this thread is back one page. This is the response, given by BHC, that sparked my mention of why I trust raw data, such as that from Spec, more than I trust reviewers statements:

----------------
"Raystonn,
Go back to your original post on Spec. I have to disagree with you on the validity of the comparison. For example, the Athlons were only using PC133 SDRAM in those tests, but I am sure the P4s were using 800 MHz RDRAM. Moreover, there are little info on other components. That is why I would trust results from people like Tom (our host) more. He always makes sure all parties get the best MB and memory, and he always keeps other components (graphic card, hard drive, etc) the same. He also tests with or without SSE2 optimized, which is only fair since most users won't go out to buy all new software. All in all, the numbers on spec.org should not be taken as the last words when we compare processors."
----------------

My argument in showing a problem with that one benchmark by Tom was that these reviewers' benchmarking conclusions aren't necessarily all that reliable. The data must always be analyzed. I've never been arguing that DDR was not actually faster than SDRAM, merely that Spec was a viable and reliable benchmark for CPUs and their respective memory subsystems.

-Raystonn

-- The center of your digital world --
 
The 100MHz memory bus tests using SDRAM and DDR are identical. The speed increase occurs when he looks at the 133MHz bus. Those performance increases are due mostly to an increase in the memory bus, not the DDR memory.
Click to expand...
From now on when someone asks you how fast your computer is, tell them your benchmark scores.
 
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom