Conflicting CPU Benchmarks on THG?!?!?!

[WATYF]

Maybe someone can help me out with this... I'm trying to decide on which processor to use for my next upgrade. So I've been checking out THG's benchmarks from a number of different articles that have been up lately, and I found something very confusing.

My main application for this system is digital audio recording... so I've been looking at the Audio benchmarks mostly... I ran across a test THG uses titled "Syntrillium - CoolEdit Pro 2.1" described as "Amplitude Normalizing". Now, I know what this is, since it's a type of plug-in processing function that I would be performing on a very routine basis, so this benchmark was of particular interest to me. But then I noticed this disconnect.

In this article:

http://www6.tomshardware.com/cpu/20040322/prescott-17.html

... you can see that the P4's far surpass the AMD's in the ability to handle this type of function.

But in this article:

http://www.tomshardware.com/motherboard/20040619/socket_775-44.html#audio_benchmarks

... it's the EXACT opposite. The AMD's outperform the P4's hands down.

The same is true for this benchmark:

http://www6.tomshardware.com/cpu/20040601/socket_939-22.html


So my question is... why is there such a difference in the results of this benchmark, and more importantly... which one is RIGHT???


WATYF

 

[endyen]

I am going to guess that it is dependant on the format of the file. It would seem that P4 is faster if it's wav format. Does that make any sense?

 

[trooper11]

well ive been doing alot of audio processing as of late and ive noticed the following:

-athlon 64 outperforms the p4 chips when using dbpoweramp in what i have used it for, that being ripping cd tracks to mp3 and also to wav.

- athlon 64 also outperforms the p4 chips when using besweet for audio work such as converting to 32bit ac3 files for dolby digital 5.1 tracks

so what it comes down to is what apps you use and what processes you use

 

[WATYF]

Actually... I don't think it's either of those things guys. If you compare the first and third links I posted, you can see that the P4's have the exact same results (i.e. the 3.2Ghz is at ~100.5, etc) but it's the A64's that get faster all of a sudden. I just need to figure out why the A64's are so much faster from one benchmark to the next. It looks like they're using the same exact mobo in the tests, so I can't think what else would give them such a performance boost.


WATYF

 

[bandikoot]

<A HREF="http://www6.tomshardware.com/cpu/20040322/prescott-17.html" target="_new"> Click </A>
<A HREF="http://www.tomshardware.com/motherboard/20040619/socket_775-44.html#audio_benchmarks" target="_new"> Click </A>
<A HREF="http://www6.tomshardware.com/cpu/20040601/socket_939-22.html" target="_new"> Click </A>

 

[aaron2504]

is there anyone on this forum who isn't AMD baised!

 

[ChipDeath]

Most people here are un-biased. I buy whatever performs best for what I want at the time, at the right price.

The majority of enthusiasts play the latest games on their PCs. Because AMD chips are better for gaming applications than Intel ones, they tend to win out with most people. The fact they're generally cheaper too is just a bonus.

---
Epox 8RDA+ V1.1 w/ Custom NB HS
Summer's here! so ease off the overclock...
XP1700+ @166x12 (~2Ghz), 1.475 Vcore
2x256Mb Corsair PC3200LL 2-2-2-4
Sapphire 9800Pro 400/730

 

[bandikoot]

You mean people do things on their computers other than playing games and watching porn? 😱

<font color=blue> Failure is not an option. These days it comes packaged with the software. </font color=blue>

 

[trooper11]

i dont know why you owuld say that, most aorund here will choose either processor. some tasks favor one over the other though and certain apps favor one over another. i just pointed out my personal findings with the apps i use.

 

[WATYF]

I don't want to start yet another AMD vs Intel war... I just use whatever's faster... right now I run an XP2800 (T-Bred B) and I intend to use the P4 3.2 for this upgrade...

...that is, unless someone can explain to me why the A64's did so much better in those second set of benchmarks... I still haven't gotten an explanation on that one.


WATYF

 

[ChipDeath]

There does seem to be a discrepancy in the results for that particular test.

Looking just at the A64 3400+ (Socket 754, Single Channel DDR400), it goes from 178.0 seconds to 86 seconds. That's a huge difference.

Both systems use 1Gb of memory, but the slower one has 2-3-2-6 timings, Versus 2-2-2-6 for the faster system, but that should give <1% difference, if that!

They both use an Asus K8V mobo. In fact the faster one suggests It's using a S940 (SK8V) mobo, so I'll assume that's a typo, and it's the same mobo as the earlier review, although it's using a more recent BIOS. (1006 v 1004)

They both use the Same sound card, But the faster system Has 2 16Mb 250Gb Maxtors in RAID0, whereas the slower one has a single, 8Mb cache Maxtor drive.

Both systems are normalizing a 2.6Gb Wave file.

So put simply, the faster system has slightly tighter memory timings, but is running a considerably faster disk subsystem - considering the file being operated on is 2.6 times larger than the amount of memory (2.6Gb Vs 1Gb), I would expect a good boost in speed from that alone, possibly as much as is seen here.

I'll go have a look for some RAID benchies along similar lines.

---
Epox 8RDA+ V1.1 w/ Custom NB HS
Summer's here! so ease off the overclock...
XP1700+ @166x12 (~2Ghz), 1.475 Vcore
2x256Mb Corsair PC3200LL 2-2-2-4
Sapphire 9800Pro 400/730

 

[ChipDeath]

I'm sure there must be hundreds of Single Drive Vs. RAID 0 comparisons out there, but I don't seem to have found that many after a quick look. I found a review of an MSI 875p Neo mobo at <A HREF="http://www.xbitlabs.com/articles/mainboards/display/msi-875p-neo_5.html" target="_new">x-bit labs</A> that shows that Intel's RAID controller does actually deliver 100% improvement for sequential Reads, and very near that for sequential write operations, when using a RAID 0 array....

I knew RAID 0 could theoretically double disk performance, but I didn't think it was actually <i>that</i> much faster! not bad I guess. I've never really bothered looking into RAID 0 much - I don't like the idea that one dead disk loses all your data...

Anyway - I think that pretty much explains the discrepancies in the two reviews...

<pre>holy [peep]! I must be <i>really</i> bored at work today....</pre><p>
---
Epox 8RDA+ V1.1 w/ Custom NB HS
Summer's here! so ease off the overclock...
XP1700+ @166x12 (~2Ghz), 1.475 Vcore
2x256Mb Corsair PC3200LL 2-2-2-4
Sapphire 9800Pro 400/730

 

[WATYF]

Even though one of the "faster" tests uses 250GB drives, if you compare these two tests:

http://www6.tomshardware.com/cpu/20040322/prescott-04.html

http://www6.tomshardware.com/cpu/20040601/socket_939-14.html

...you can see that they're both using the same exact HDD setup... (80GB/RAID 0) but the second one shows the A64's as being a lot faster. I don't think it's the HDD change that's causing the variance in results. And if it was the HDD setup that was causing an increase, you would see it in the P4's as well... but they maintain the same performance throughout all the benchmarks. There has to be some kind of fundamental difference between what he's using for the first test, and what he's using for the second two.... or else the second tests are just plain wrong.

The other differences are so minor, there's no way that they'd cause a +100% increase in performance. I mean... I'd LIKE to think that a BIOS revision and some memory tweaking can do that much for your CPU, but that's just not realistic.


I emailed THG about this... maybe they'll have an explanation for it.


WATYF

 

[ChipDeath]

Ah yes - I mis-read the first one, I thought it was using a single 80Gb Drive.

Seems strange... 😱

---
Epox 8RDA+ V1.1 w/ Custom NB HS
Summer's here! so ease off the overclock...
XP1700+ @166x12 (~2Ghz), 1.475 Vcore
2x256Mb Corsair PC3200LL 2-2-2-4
Sapphire 9800Pro 400/730

 

[sonoran]

The bigger issue would be if hard drive changes cause a significant difference - they're not benching CPU performance so much as I/O performance.

 

[Guest]

Maybe AMD got tips from nVidia and they're able to release Magic bios!
Yay optimisation!

Asus P4P800DX, P4C 2.6ghz@3.25ghz, 2X512 OCZ PC4000 3-3-3-8, Leadtek FX5900 w/ FX5950U bios@500/1000, 2X30gig Raid0

 

[Piccoro]

That is a very good question.
Would a codec being updated cause such a result?
Maybe the settings were not the same?

The graphs between the <A HREF="http://www6.tomshardware.com/cpu/20040601/socket_939-22.html" target="_new">socket_939_22 page</A> and <A HREF="http://www.tomshardware.com/motherboard/20040619/socket_775-37.html" target="_new">socket_775-44.html#audio_benchmarks page</A> match among all the CPUs. The <A HREF="http://www6.tomshardware.com/cpu/20040322/prescott-17.html" target="_new">prescott-17 page</A> matchs the Pentium scores with the other, but your right...the Athlon scores are different on the older review.

The only things I noticed was the improved memory timings among all the platforms, Bios updates and they did use a different board for Socket 940.
But nothing that would explain across the board improvement for all of the Athlon's.
Interesting. I'd like to know what the difference is now too. hehe

 

[endyen]

It's probably something silly like "someone forgot that the ht bus was set to 400" It's not like Tom's would do anything to make Intel look good. Right Omid?