Sign in with
Sign up | Sign in
Your question

To everyone who thinks that THG is biased (repost)

Last response: in CPUs
Share
September 24, 2003 3:10:14 PM

Slvr_Phoenix stole my damn thunder, so now I'm reposting one of his posts from another thread here, it would be a good read for those of you who just read the conclusions of the articles without bothering to actually seriously examine the benchmarks and equipment.

I've said this before, LOOK AT THE HARDWARE VERY CAREFULLY, it makes a massive difference. Never believe a site that doesnt post detailed specs on what they are doing or uses hardware that isnt of a similar performance level.

Anyways, on with the show.

Shadus

--[Nothing below this line is mine]--

To everyone who thinks that THG is biased by slvr_phoenix
I went through the THG benchmarks and counted the number of times that the stock Intel CPUs beat the A64 FX. (Meaning that no 3.4/3.6GHz P4EEs were included in this count.) I counted stock Intel beating AMD 27 to 14. (I'm not sure where THG got their win-count numbers frankly.)

Here however is how it breaks down:
Gaming = Intel 9 / AMD 10
Encoding = Intel 8 / AMD 0
Real Use = Intel 6 / AMD 3
Synthetic = Intel 4 / AMD 1

For games it's a tie between Intel and AMD.
For encoding Intel totally rocks AMD's world.
For real-world applications Intel is stronger.

And what this shows is that even if you chop out all synthetic benchmarks and all encoding benchmarks so that you drop the Intel bias completely (some would even argue that this would heavily bias it towards AMD) you still end up with Intel on top, though in a much closer race.

Now, one could easily pick and choose certain benchmarks to make AMD look much better. Just as one could easily pick and choose certain hardware to make AMD look better. For example, THG went out of their way to up the performance of their hard drive system so that it couldn't possibly become a bottleneck for determining the CPU's performance by using a SATA RAID0 array. Is anyone else doing this?

For example, Anandtech was only able to squeeze a piddly 4.02 GB/s of bandwidth for their P4 system. THG on the other hand got 4.95GB/s of bandwidth on their P4 system. That's almost an entire 1GB/s of bandwidth lost in Anandtech's P4 setup! Yet Anandtech was able to squeeze 5.315GB/s for their Athlon64 FX system, which is much closer to the extreme that THG pushed their's to for 5.6GB/s.

Anandtech biased? Never!

I briefly glanced at HardOCP's setup as well and had a similar laugh. The P4 was paired on an Intel mobo (Intel boards are known for rock-stability but low performance) while the A64FX was thrown onto an Asus mobo (Asus boards are known for extreme performance).

HardOCP biased? Never!

I obviously don't have time to go through every single review one by one. But frankly it's usually very easy to determine how biased a review is likely to be. Obviously the more benchmarks that are run the better because this gives a more rounded view. Also anyone who gives very little information on their test platforms is immediately suspect. And as always anyone who configures one platform significantly better than another platform is biased.

When the scores between Intel and AMD are as close as they are right now, even a tiny bias can make a huge difference. So far I haven't seen a single review site go through as much effort as THG did to remove all possible bias. If other review sites hadn't been so lazy in their reviewing process the results would have differed greatly. As far as I'm concerned THG is the only review site that hasn't biased their A64 FX benchmarks. Yet if you read the comments made in THG's review, it's very clearly biased, towards the A64 FX.

---[Following name had nothing to do with the above post.]---

Shadus<P ID="edit"><FONT SIZE=-1><EM>Edited by shadus on 09/24/03 11:12 AM.</EM></FONT></P>
September 24, 2003 3:39:04 PM

Here is a <A HREF="http://www.hothardware.com/hh_files/CCAM/a64fx_51_launc..." target="_new"> review</A> that used excellent test rigs with comparable high performance mobos including 10k Raptor drives. And I still don't see your beef with Anands article, it was excellent. Maybe check the section on 64bit apps while you are there. I can imagine that UT2003 64bit is going to be wicked. I'll have to read the Hard review to comment.
I guess giving the<i> Crown of Performance</i> to a Paper Launched CPU isn't biased either now is it? You give a good point that everyone will put their spin on their review. Maybe reviews should just post Test Rigs and Benchmarks. Leave the commentary for the readers to invent, since reviewers can't be partially objective right? They must ALL be Intel or AMD fanatics. Riiiiiiiighhhhhhht.
September 24, 2003 3:41:50 PM

Quote:
I've said this before, LOOK AT THE HARDWARE VERY CAREFULLY, it makes a massive difference.

Yeah it makes a difference when the race is so damn close between these CPUs. As it stands, Intel still ownz in encoding, and AMD is owning on Gaming. For "Real World Performance" it seems pretty even. So buy whichever CPU suites your tastes.
Related resources
September 24, 2003 3:49:41 PM

I definitely agree with this post for the most part. However, I still strongly hold that I would much rather prefferred overclocked AMD & Intel processors in a totally seperate article.

There is the case that why did THG bother to OC the 3.2 to 3.6GHz and not even bother Overclocking the Athlon 64. That would have been a more fair comparison. Even at that, I'm not impressed with the Athlon 64's current 32-bit performance. Its just that the P4EE with such an overclock makes it look like its 20% faster than the Athlon 64. Currently, I am impressed that Intel was able to squeeze so much extra performance out of the Socket 478. I've said it before and I'll say it again, I don't like excessive socket changes and right now, Intel is working harder to provide a longer lasting platform than AMD. Don't get me wrong, socket changes are an absolutely GOOD thing if done sparingly.

My OS features preemptive multitasking, a fully interactive command line, & support for 640K of RAM!
September 24, 2003 4:09:38 PM

That review most certainly does not use "excellent" test rigs. The Intel rigs are crap. They used a DFI Lanparty 875 motherboard for Bob's-sake! It's notable for reviews with comments such as "<font color=red>The fact that the manufacturer still relies on a <i>two-phase controller</i> of the old design can probably only be attributed to lower costs. <i>Allowances have to be made for the memory timings</i>: the fastest setting in the case of the DDR400 is CL2.0-2-2.7, which causes the board to <i>lose some of its edge in terms of memory performance</i>. In addition, the <i>voltage has to be increased</i> to 2.7 Volts so that stable operation is guaranteed. <b>DFI ends up in the <i>midfield</i> in the performance test.</b></font color=red>"

Meanwhile the AMD systems rock along on kick-arse top-notch Asus motherboards. Yeah. That's fair.

<pre><A HREF="http://ars.userfriendly.org/cartoons/?id=20030905" target="_new"><font color=black>People don't understand how hard being a dark god can be. - Hastur</font color=black></A></pre><p>
September 24, 2003 4:14:15 PM

wwwwaaaaaaaaaaaaaaaaahhhhhhhhh my pussy hurts!
September 24, 2003 4:17:16 PM

please whine some more. I think Asus mobos suck in my opinion too.
September 24, 2003 4:18:53 PM

You do realize that "Golden Sample" Asus nForce3 is only running at 600Mhz right?
September 24, 2003 4:28:10 PM

hey silverp knows what hes talking about... as anyone whos talked with him KNOWS. like anyones gonna listen to some no-name just passing thru, and is only spamming during their short visit at that

-------

Quote:
i hate to admit but im a worthless lying idiot jackass

- scamtron
September 24, 2003 4:47:50 PM

we all appreciate your 2 cents

Added silver p and phial to the Intel pukers when things got rough list ; )
September 24, 2003 5:03:55 PM

Quote:
Never believe a site that doesnt post detailed specs on what they are doing

That would be THG leaving the reader to interpret whether the review had 3 different P4 EE CPUs or changed the multipliers and voltages on one engineering sample to make it look like these 3 CPUs existed.
Quote:
LOOK AT THE HARDWARE VERY CAREFULLY, it makes a massive difference

It makes a trully MASSIVE 1%-2% difference. I'm sorry but the esteemed Slvr_Phoenix may have let his emotions get to his head in this post. The DFI Lan Party Pro875 mobo lags between .5% and 1% behind the MSI 875P Neo, which is a rocking board IMHO. I agree with you that HardOCP is gay to have used an Intel board, but comparing the DFI to the Asus is by no means unfair, especially considering the current limitation of the NF3 150.
September 24, 2003 5:12:21 PM

Add me too then...Silver wrote down a well thought out argument...you said waaaahhhh. Things aren't rough for Intel at all. The processor they put out months ago, is showing well against the 'revolutionary' processor that AMD just came out with. After another couple of months, AMD will be playing catch-up again.
That wasn't the point of these threads though. The accusation was that Tom's was deflating the A64's benchmarks, and silver has clearly shown that there IS some data skewing going on, but its at the other sites that you hold so sacred. Get a clue...you've been marketed, sensationalized and you ate up hook, line, and sinker

Scamtron doesn't like my sig...
September 24, 2003 6:04:35 PM

I decided to read your review (http://www.hothardware.com/hh_files/CCAM/a64fx_51_launc...), I hate to talk about things that I honestly don't know about.

1) The FX also had a second hard drive for an unknown reason, but placing os on one drive and data on a second drive will increase performance as anything that accesses the vmem will not have to compete for disk time. We have no idea what they were doing with the second drive.

2) Of interest is the 64bit software tests they ran, in almost all cases the chip runs faster on the 32bit os than the 64bit os. That in itself is quite fishy.

3) The EE Did much better at this site than the FX... yet the conclusion said the FX was a faster chip EH?! Fscked. Lets look at the numbers... (remember some benchies are better with lower numbers... some better with higher, skimming graphs doesnt cut it unless you READ THEM.)
o PCMark2002 CPU - EE
o PCMark2002 Memory - EE
o 3DMark CPU - FX
o SpecView ugs-03 - Tie FX/EE
o SpecView proe-02 - EE
o SpecView light-06 - EE
o SpecView dx-08 - XP3200
o SpecView drv-09 - EE
o SpecView 3dsmax-02 - EE
o Kribibench Office - EE
o Kribibench Ultra - EE
o Cinebench - EE
o Novalogic Com - EE
o UT2003 Antalus - FX
o Q3 - No Score, Not Complete As of Yet. The FX only beat the standard edition p4 by 8.7%... the EE has performed better than 8.7% over the standard quite frequently. In fact in Antalus it got 12.6% better than the standard, if we apply that number to the standard's score the p4ee would be about 525.5 which would put the EE in first by a pretty comfortable margin... but we wont score this one till they post up a fixed benchie.
o 3dMark2001 - FX
o XMpeg/Divx - EE
o Magix - EE
o Business WinStone - FX
o Content Creation - FX
o Dhrystone 2.1 - FX
o Whetstone - EE
o Eight Queens - XP3200
o Matrix Ops - FX
o Number Crunch - EE
o Floating Point - FX
o Memory - Tie FX/EE (both dominated 3200 and maxed benchie)

Remove quake from the equation and you get: EE-16 (2 Ties), FX-10 (2 Ties), XP3200-2. (I counted the ties as a win for both- p3.2c got eaten alive.) The numbers are EE 57% Won, FX 36% Won, 3200 7% Won, P3.2C 0% Won...

They did 3 games, ee won 1, fx won 1, and the third was borked.

Post up another site and we'll go through their numbers, this one pretty much agrees with toms based on the numbers, its the conclusion that is off... This is a case of a site putting up the conclusion that people want to see.

Shadus<P ID="edit"><FONT SIZE=-1><EM>Edited by shadus on 09/24/03 02:12 PM.</EM></FONT></P>
Anonymous
a b à CPUs
September 24, 2003 7:28:59 PM

>o PCMark2002 CPU - EE
Irelevant (synthetic benchmark)

>o PCMark2002 Memory - EE
Irelevant (synthetic benchmark)

>o 3DMark CPU - FX
Irelevant (synthetic benchmark)

>o SpecView ugs-03 - Tie FX/EE
>o SpecView proe-02 - EE
>o SpecView light-06 - EE
>o SpecView dx-08 - XP3200
>o SpecView drv-09 - EE
>o SpecView 3dsmax-02 - EE
Interesting, I would consider SpecVIEW just one benchmark, won by the EE.

>o Kribibench Office - EE
>o Kribibench Ultra - EE
Again, 1 benchmark

>o Cinebench - EE
>o Novalogic Com - EE
>o UT2003 Antalus - FX
>o Q3 - No Score,
EE is faster on Q3, no doubt. Not that Q3 is still relevant though

>o 3dMark2001 - FX
>o XMpeg/Divx - EE
>o Magix - EE
>o Business WinStone - FX
Irelevant (seriously, office performance ?)

o Content Creation - FX
Irelevant (completely broken and biased benchmark, even if the FX wins it somehow)

o Dhrystone 2.1 - FX
Irelevant (synthetic lowlevel benchmark)

o Whetstone - EE
Irelevant (synthetic benchmark)

o Eight Queens - XP3200
o Matrix Ops - FX
o Number Crunch - EE
o Floating Point - FX

Above benches seem like real specialty/syntethic cases; interesting perhaps, but would you buy a cpu because of its 8 queen performance.. take them with a grain of salt.

o Memory - Tie FX/EE (both dominated 3200 and maxed benchie)
Irelevant (synthetic low level benchmark)

>Remove quake from the equation and you get: EE-16 (2 Ties),
>FX-10 (2 Ties), XP3200-2. (I counted the ties as a win for
>both- p3.2c got eaten alive.) The numbers are EE 57% Won,
>FX 36% Won, 3200 7% Won, P3.2C 0% Won...

This is so senseless.. Pick a representative group of benches, and make a conclusion based on that (eg gaming, or media encoding, or 3D/CAD,.. etc). Throwing everyhting on a pile, giving the same weight to a new DX9 game as to "Dhrystone" or Sandra memory performance tells you *nothing*.

Why is it that people need to simplify all this to one magical number to declare a "winner" ? Its not a football match, just pick the benches that matter to you, compare prices, and buy whatever suits your needs.

= The views stated herein are my personal views, and not necessarily the views of my wife. =
September 24, 2003 7:48:18 PM

Quote:
That would be THG leaving the reader to interpret whether the review had 3 different P4 EE CPUs or changed the multipliers and voltages on one engineering sample to make it look like these 3 CPUs existed.

In case you weren't aware that even THG can eat crow...
Quote:
<font color=green>Update Sept 24,2003: Unfortunately we have made a mistake in the original article: In addition to the official P4 EE 3.2GHz we had included benchmark scores of the P4 Extreme 3.4GHz and 3.6GHz. These values were planned for a future THG article and were not intended to be included here. We would like to apologize especially to those readers who misinterpreted our charts. The two bars of the P4 Extreme 3.4GHz and 3.6GHz have now been removed. However, this issue does not affect our conclusion as we have only compared the official P4 3.2GHz EE to all other test candidates in our original article. For your information: The press sample of the P4 Extreme provided by Intel does not have a multiplier lock and is already designed for higher clock speeds.</font color=green>


Quote:
It makes a trully MASSIVE 1%-2% difference. I'm sorry but the esteemed Slvr_Phoenix may have let his emotions get to his head in this post. The DFI Lan Party Pro875 mobo lags between .5% and 1% behind the MSI 875P Neo, which is a rocking board IMHO. I agree with you that HardOCP is gay to have used an Intel board, but comparing the DFI to the Asus is by no means unfair

If by "a trully MASSIVE 1%-2%" you mean the actually recorded 3% to 9% difference that using the DFI Lanparty board makes when compared to a top end mobo, counted against the 2% to 7% that the vast majority of the P4EE vs. A64FX benchmarks were won by then you're correct. Otherwise you're completely wrong.

Quote:
especially considering the current limitation of the NF3 150.

You mean the limitation that according to the benchmark scores out there was in fact only in a rare few products? You mean the limitation that caused such dramatic problems that whenever it did make an appearance it made such a dramatic effect on performance that even a 2500+ Barton could beat the A64-FX? <sarcasm><i>Oh yes. I can see how this might cause confusion or slightly biased benchmark scores.</i></sarcasm>

I'm sorry, but the use of a bad mobo really does make all the difference in how the balance of those benchmarks turn out. The proof is in the millions of benchmarks already out there. Look at the numbers and do the math. Your argument is totally flawed by the facts.

And since when did emotions come into this? That's got me more confused than anything. I'm talking about choices in hardware that result in statistical innacuracies and you're talking about emotions? I don't get it.

<pre><A HREF="http://ars.userfriendly.org/cartoons/?id=20030905" target="_new"><font color=black>People don't understand how hard being a dark god can be. - Hastur</font color=black></A></pre><p>
September 24, 2003 7:54:08 PM

Quote:
That would be THG leaving the reader to interpret whether the review had 3 different P4 EE CPUs or changed the multipliers and voltages on one engineering sample to make it look like these 3 CPUs existed.

That overclocked EE will perform the same as an intel released EE at 3.4/3.6ghz since they were only dealing with multiplyers, and not fsb. I did think it was good to include that actually because it shows how the FX may fare against the prescott. Yes i do think it was a fault of them not to say how they came about getting these cpus, but the performance is undeniable, if intel released a 3.6ghz EE it would perform the same as toms overclocked cpu. I beleive that tom overclocked to 3.6 because they thought/had information that intel would release an EE at that speed. They probably did not overclock the FX because it has a locked (or possibly no?) multiplyer, thus overclocking it would require increasing the cpu's bus speed and effect memory bandwith performance.



Proud owner of DOS 3.3 :smile:
September 24, 2003 8:24:21 PM

Did you bother to read why I was even looking at those benchies? Did you bother to read the post I was responding to? I was simply responding to someone elses post and looking at the actual numbers they were looking at when they were making their conclusions. Does any number matter to me except the actual games I play which NO ONE does benchies on? No.

When I see a site start coughing up MMP game benchies I'll pay alot of attention, until then... shrug. I'm argueing with a senseless moron to make a point... that he didn't bother to read the article, he only read the conclusion, and he didn't bother to really look at the benchies themselves except to note that the fx had longer bars than the ee.

Shadus
September 25, 2003 2:13:28 AM

Quote:
I am impressed that Intel was able to squeeze so much extra performance out of the Socket 478. I've said it before and I'll say it again, I don't like excessive socket changes and right now, Intel is working harder to provide a longer lasting platform than AMD.

You forget to mention 1 thing! The Sockect 478 platform came after the Athlon platform. It's not true that Intel put more life than AMD into there platform. I think that AMD still own the crown with the Athlon platform when we talk of long time compatibility/upgradability.

I wish the Socket 939/940 will last long!

--
Would you buy a GPS enabled soap bar?
September 25, 2003 2:23:34 AM

Um Socket 940 won't live to next year.

The whole point was that 939 will be the real FX platform. Already a socket change à la 423->478. Sucks to be those who will be suckered into paying for an overpriced system already, and then not being able to upgrade.

--
<A HREF="http://www.lochel.com/THGC/album.html" target="_new"><font color=blue><b>This just in, over 56 no-lifers have their pics up on THGC's Photo Album! </b></font color=blue></A> :lol: 
September 25, 2003 8:17:12 AM

Eh? I've seen on several sites that AMD is claiming they will continue to release 940 cpus simultaneously with 939 throughout 2004 so that people who buy now aren't completely left out in the cold. I admit that I wouldn't buy a 940 because of the memory, but that's still a little over a years worth of upgrade potential.
Anonymous
a b à CPUs
September 25, 2003 10:38:26 AM

Quote:
I've said this before, LOOK AT THE HARDWARE VERY CAREFULLY, it makes a massive difference. Never believe a site that doesnt post detailed specs on what they are doing or uses hardware that isnt of a similar performance level.

Anyways, on with the show.

To everyone who thinks that THG is biased by slvr_phoenix
I went through the THG benchmarks and counted the number of times that the stock Intel CPUs beat the A64 FX.
Here however is how it breaks down:
Gaming = Intel 9 / AMD 10

In this he includes 3 times quake3? When I count them I get at :

Amd wins: UT2003 , Spinter cell, WC3, Gun metal
Intel wins:Quake
Equal (less then 2% difference)commance4,X2

Thats 4 for AMD and 1 for intel

In synthetic benchmarks its 2 for AMD (3dmark) and none for intel (aquamark is again equal)


Quote:
Encoding = Intel 8 / AMD 0

Yep intel is superior here altough I wonder why he uses 3 different video and 3 MP3 encoders ? 1 is enough to me except of course if you just want to add benchmarks were intel wins.
Quote:

Real Use = Intel 6 / AMD 3

Really depends on what program you use :

DIEP, ScienceMark MolDyn , Plasma ,3DS Max 5.1 Studio PC, Realstorm Raytracing ,Lightwave 7.5 Raytrace , POV-Ray , picCOLOR NT all won by the athlon FX-51 and others won by intel so its kinda equal there intel has a small edge here in 32 bit.

Quote:

For games it's a tie between Intel and AMD.
For encoding Intel totally rocks AMD's world.
For real-world applications Intel is stronger.

In gaming "AMD totally rocks Intels's world is the reality in real time applications intel is slightly stronger.

Quote:

I obviously don't have time to go through every single review one by one. But frankly it's usually very easy to determine how biased a review is likely to be. Obviously the more benchmarks that are run the better because this gives a more rounded view. Also anyone who gives very little information on their test platforms is immediately suspect. And as always anyone who configures one platform significantly better than another platform is biased.

Or someone who uses the same kind of benchmark over and over just to slant the numbers or who uses overclocked CPU's from a CPU tha isnt even on the market without even giving a hint there overclcoked ,... is biased too .

Quote:
When the scores between Intel and AMD are as close as they are right now, even a tiny bias can make a huge difference. So far I haven't seen a single review site go through as much effort as THG did to remove all possible bias. If other review sites hadn't been so lazy in their reviewing process the results would have differed greatly. As far as I'm concerned THG is the only review site that hasn't biased their A64 FX benchmarks. Yet if you read the comments made in THG's review, it's very clearly biased, towards the A64 FX.

LOL sure all the rest are wrong only THG is right, he even was biased towards AMDso in fact the A64 is even worse LOL sure .
Anonymous
a b à CPUs
September 25, 2003 10:46:07 AM

Quote:
I decided to read your review (http://www.hothardware.com/hh_files/CCAM/a64fx_51_launc...), I hate to talk about things that I honestly don't know about.Remove quake from the equation and you get: EE-16 (2 Ties), FX-10 (2 Ties), XP3200-2. (I counted the ties as a win for both- p3.2c got eaten alive.) The numbers are EE 57% Won, FX 36% Won, 3200 7% Won, P3.2C 0% Won...

here http://www.tweakers.net/nieuws/28882 is a site that gives summary of some other review as you can clearly see AMD dominates the games , intel the rendering/encoding and its a tie in other applications . I can make a review in wich intel wins 10 out of 10 benchmarks and another in wich amd wins all 10 not that hard . Little numbers like that say nothing you have to look at the general lines.

all reviews looked at :

Naast Ace's Hardare, AnandTech, Tech Report en Lost Circuits hebben de volgende sites reviews van de Athlon 64 geproduceerd: Extreme Tech, Hot Hardware, Gamers Depot, Athlon XP, Tom's Hardware Guide, Hexus, HardOCP, Digit Life, AMD Zone en TecChannel.
September 25, 2003 12:24:50 PM

> LOL sure all the rest are wrong only THG is right

It wouldn't be the first time. Nor would it be the first time if toms had to eat crow.

Shadus
September 25, 2003 2:51:44 PM

For those who might own a Socket 940 mobo, they can still upgrade, but only to the Opteron Socket (from Anandtech site).

From Anandtech site, it mention that the arcitecture of 939 might get a "rework" (don't know the correct term), ie, AMD is "planning" to reduce the Hyper transport channel from 3 (940, and I think at least the 1st 939) to 1. I was wondering, beside not being able to have a dual AMD63FX, could their also be other disadvantage? Also, can't the chipset manufact. do something about the other 2 Hyper Tans? Eg. 1 deicated to South bridge, 2nd to AGP, 3rd to...whatever next most important hardware?

System Integration...yeah right, thanks to marketing, more confusion
September 25, 2003 4:02:04 PM

Quote:
In this he includes 3 times quake3?

Damn straight. Each of the benchmarks under Quake were of <i>extremely</i> different demos that each have their own unique characteristics. It's happened before where one Q3 demo sees one CPU win and another sees a different one win.

Further, the Q3 engine is used in a great many games out there. So even though each game using the Q3 engine has completely different performance characteristics, you'd just lump all of those games together because they use the same base engine?

Quote:
Equal (less then 2% difference)commance4,X2

When performance differences are this close as it is, ties don't exist unless the numbers are actually identical. Or if you want to actually do the math, a tie would be determined by a defined minimal distance (5% beign a nice round number) between the maximum statistical differences. So if the maximum performance difference in any benchmark is 10%, than a 'tie' would be if the difference in performance were less than 0.5%.

Quote:
altough I wonder why he uses 3 different video and 3 MP3 encoders ?

Because if just one was chosen it could easily just be the one out of a whole group that was optimized specifically for a certain product. By including numerous different codepaths you ensure a more rounded view of a processor's strengths and weaknesses. You wouldn't run just one game to determine gaming performance or just MS Word to determine office performance. So you don't run just one encoder to determine encoding performance.

Quote:
Really depends on what program you use

You couldn't have made a more blatantly obvious statement if you'd tried. And it's sad how many people just don't get the simplicity of this. If people actually looked for a CPU that meets their needs instead of a CPU that won n out of x benchmarks they'd be <i>much</i> better off. But <i>that'd</i> require intelligence.

Quote:
Or someone who uses the same kind of benchmark over and over just to slant the numbers

If it were the exact same benchmark and was used just to slant the numbers then you'd have a point. <i>Neither</i> is true however.

Quote:
or who uses overclocked CPU's from a CPU tha isnt even on the market without even giving a hint there overclcoked ,... is biased too .

And again <i>neither</i> of your points is valid. 1) The coloring of the overclocked CPUs <i>was</i> a different shade than the stock P4EE and the graphics before the benches <i>showed</i> that it was a 3.2GHz CPU, so there clearly <i>was</i> "a hint" there for anyone to see if they'd bothered to look. 2) None of the OCed CPUs were used in THG's determination of who won and why. In fact most of the article was even written as though the whole P4EE didn't even exist. 3) THG also later ate crow and retracted their OCed P4EE scores from the benchmarks.

So do you see just how little ground you have to stand on now?

Quote:
LOL sure all the rest are wrong only THG is right, he even was biased towards AMDso in fact the A64 is even worse LOL sure .

It certainly wouldn't be the first time that THG was one of the very few that told the complete and honest truth instead of caving in to AMD's marketing demands to bias results.

And I never said that the A64FX was even worse. <i>Opteron</i> is a good chip. AMD has really done a good job there, especially with the use of the hypertransport layer for inter-CPU communications and the use of the ondie memory controller bypassing a northbridge-style architecture for high-speed inter-CPU memory transfers. It's a brilliant design <i>for a server</i>.

Why AMD felt that they had to rename one of their <i>server</i> chips as a desktop chip <i>just</i> to try and grab that ever-elusive performance crown instead of being content with the actual A64's performance upon release (which consequently was <i>also</i> released under a lower clock than the Opteron/FX) is however another rant all together.

<pre><A HREF="http://ars.userfriendly.org/cartoons/?id=20030905" target="_new"><font color=black>People don't understand how hard being a dark god can be. - Hastur</font color=black></A></pre><p>
September 25, 2003 4:10:11 PM

Quote:
For those who might own a Socket 940 mobo, they can still upgrade, but only to the Opteron Socket (from Anandtech site).

Exactly. And why? Because Socket940 isn't even a desktop socket. It's a <i>server</i> socket. AMD knew it. Intel knew it. We know it. Yet no one seems to pay attention to the blatantly obvious, that the A64FX currently is <i>just</i> an Opteron sold under a different name. Had people included Opterons in their benchmarks there would have been foul cries of cheating. However just the simple task of renaming the exact same chip somehow placates everyone? :\ I don't get it.

Quote:
I was wondering, beside not being able to have a dual AMD63FX, could their also be other disadvantage?

Nope. Those hypertransport layers that will be reduced are specifically for inter-CPU communication for memory transfers since each CPU has it's own unique memory bank. It won't affect performance in any way to get rid of them if you only allow single CPU setups.

Quote:
Also, can't the chipset manufact. do something about the other 2 Hyper Tans? Eg. 1 deicated to South bridge, 2nd to AGP, 3rd to...whatever next most important hardware?

Nope. Those hypertransport layers are specifically for inter-CPU communications. There is a completely different (and considerably slower) hypertransport layer already in use for communication between the CPU and everything else.

<pre><A HREF="http://ars.userfriendly.org/cartoons/?id=20030905" target="_new"><font color=black>People don't understand how hard being a dark god can be. - Hastur</font color=black></A></pre><p>
September 25, 2003 4:35:08 PM

Quote:
The FX also had a second hard drive for an unknown reason, but placing os on one drive and data on a second drive will increase performance as anything that accesses the vmem will not have to compete for disk time. We have no idea what they were doing with the second drive.


The system specs say the second hard drive was for the 64-bit OS. There may have been a driver issue with the SATA and the 64-but Windows. It wasn't a RAID configuration, so there would be no performance advantage, and in fact that HD wasn't even used in the comparison tests with the P4.

-------------------------------------------
<font color=blue> "Trying is the first step towards failure." </font color=blue>
September 25, 2003 7:23:41 PM

Quote:
Damn straight. Each of the benchmarks under Quake were of extremely different demos that each have their own unique characteristics. It's happened before where one Q3 demo sees one CPU win and another sees a different one win.

Further, the Q3 engine is used in a great many games out there. So even though each game using the Q3 engine has completely different performance characteristics, you'd just lump all of those games together because they use the same base engine?

Then why not UT? Why from all avaible engines (and there are plenty) do they happen to pick the quake 3 in wich intel almost always won? If they use jedi knight no problem but 3 times quake 3? ANd using all them scores ? Nope cant understand that except if you think he's biased towards intel quake 3 doesnt equal UT, WC3 AND splinter cell .

Tell me do you honestly believe quake 3 is equal to 3 modern games?

Quote:
than a 'tie' would be if the difference in performance were less than 0.5%.

Dont agree no benchmark is accurate a 0.5% it depends on many factors wich can vary 1 or 2%

Quote:
Because if just one was chosen it could easily just be the one out of a whole group that was optimized specifically for a certain product. By including numerous different codepaths you ensure a more rounded view of a processor's strengths and weaknesses. You wouldn't run just one game to determine gaming performance or just MS Word to determine office performance. So you don't run just one encoder to determine encoding performance.

But encoding is just a small part and now he uses 8 benchmarks about as much as he used games sorry but encoding isnt that important 1 perhaps 2 would have made it clear intel is the winner there.

Quote:

You couldn't have made a more blatantly obvious statement if you'd tried. And it's sad how many people just don't get the simplicity of this. If people actually looked for a CPU that meets their needs instead of a CPU that won n out of x benchmarks they'd be much better off. But that'd require intelligence.

Thats just what I said the conclusion from THG to add all the benchmarks together and then prenounce the EE as a winner is wrong.

Quote:
If it were the exact same benchmark and was used just to slant the numbers then you'd have a point. Neither is true however.

Quake 3 is used 3 times, 3 mp3 encoders, 3 ... Yes there different programs but thats still no reason to say 3 mp3 encoding programs carry as much wait as 3 of the top selling games. As I said anyone can find numerous small programs were intel or amd wins and thus slant the score if you just add everything up at the end.

Quote:
And again neither of your points is valid. 1) The coloring of the overclocked CPUs was a different shade than the stock P4EE and the graphics before the benches showed that it was a 3.2GHz CPU, so there clearly was "a hint" there for anyone to see if they'd bothered to look.

For that you have to know THG some people dont, showed that it was a 3.2? nope the graphics just said 3.4&3.6 perhaps you mean on th configuration? Not many people read that you know.

2) None of the OCed CPUs were used in THG's determination of who won and why. In fact most of the article was even written as though the whole P4EE didn't even exist. 3) THG also later ate crow and retracted their OCed P4EE scores from the benchmarks.

Quote:
None of the OCed CPUs were used in THG's determination of who won and why. In fact most of the article was even written as though the whole P4EE didn't even exist.

Euh not quit the used the 3.2EE to say the A64 isnt all that great .

Quote:

3) THG also later ate crow and retracted their OCed P4EE scores from the benchmarks.
So do you see just how little ground you have to stand on now?

Seeing the reaction you think he had much choice? Now he can claim it was a mistake, seeing it takes a while to make such a review kinda hard to believe .

Quote:

It certainly wouldn't be the first time that THG was one of the very few that told the complete and honest truth instead of caving in to AMD's marketing demands to bias results.

I am sorry but your implying out of 15 mayor review sites I have seen 13 are biased towards AMD? Kinda like a big conspiracy to make sure people buy AMD CPU's? Marsian perhaps? Sorry not quit buying that .

Quote:
Why AMD felt that they had to rename one of their server chips as a desktop chip just to try and grab that ever-elusive performance crown instead of being content with the actual A64's performance upon release (which consequently was also released under a lower clock than the Opteron/FX) is however another rant all together.

Kinda like intel with there EE? Who cares what AMD/intel does as long as it gives us good proccesors? If they throw in server CPU because there better who cares? You dont see the difference once there in your computer you know, just perhaps programs that run a bit faster.
September 25, 2003 8:15:57 PM

Quote:
For example, Anandtech was only able to squeeze a piddly 4.02 GB/s of bandwidth for their P4 system. THG on the other hand got 4.95GB/s of bandwidth on their P4 system. That's almost an entire 1GB/s of bandwidth lost in Anandtech's P4 setup! Yet Anandtech was able to squeeze 5.315GB/s for their Athlon64 FX system, which is much closer to the extreme that THG pushed their's to for 5.6GB/s.

You can't compare the results of ScienceMark (Anandtech) to the results of Sandra Max3 (THG).

SiSoft has this to say in their FAQ about bandwidth results.

<i><b>Q: Why is Sandra (2002 and later) memory index so high compared to other benchmarks?
A: This is due to using the latest instruction sets and techniques (see above) for obtaining the highest possible efficiency and thus performance out of the system. This should show what the system is able to do.</b></i>

Sound like a dig at the other benchmark programs?
Quote:
When the scores between Intel and AMD are as close as they are right now, even a tiny bias can make a huge difference. So far I haven't seen a single review site go through as much effort as THG did to remove all possible bias. If other review sites hadn't been so lazy in their reviewing process the results would have differed greatly. As far as I'm concerned THG is the only review site that hasn't biased their A64 FX benchmarks. Yet if you read the comments made in THG's review, it's very clearly biased, towards the A64 FX.

Hasn't biased their benchmarks? Are you sure?

I'm trying to figure out why THG would used such varied memory timing between platforms. OK, I'd accept the inavailability of elite memory for FX-51 as an explanation but why on the other platforms (see below. Sorry about the formatting).

Why go to the trouble of all this work and use varied memory timings between platforms.

Mistake or a "tiny bias"?.

- from the <A HREF="http://www17.tomshardware.com/cpu/20030923/athlon_64-20..." target="_new">THG's Test Configuration Table</A>

<i><b>Intel 875 (Socket 748) Asus P4C800-E Deluxe Rev: 1.02
Bios: 1011 BETA 006

4 x 256 MB CL 2.0-2-2-5 (200 MHz)
4 x 256 MB CL 2.0-2-2-5 (133 MHz)

Nvidia nForce (Socket 462) 2 Ultra Asus A7N8X Rev: 2.00
Bios: 1006

2 x 512 MB CL 2.0-3-3-6 (200 MHz)
2 x 512 MB CL 2.0-2-2-5 (166 MHz)
2 x 512 MB CL 2.0-2-2-5 (133 MHz)

Nvidia nForce 3 150 (Socket 940) Asus SK8N Rev: 1.03
Bios: 1003 BETA 002
2 x 512 MB CL 2.5-3-3-6 (200 MHz)

VIA K8T800 (Socket 940) MSI K8T Master 1-FAR (MS-9130)
Bios: 1.0Bd Rev: 1

2 x 512 MB CL 2.5-3-3-7 (200 MHz)

VIA K8T800 (Socket 754) MSI 8KT Neo (MS-6702)
Bios: 1.0 Rev: 1.0

2 x 512 MB CL2.0-4-4-8 (200 MHz)</b></i>

<b>56K, slow and steady does not win the race on internet!</b><P ID="edit"><FONT SIZE=-1><EM>Edited by phsstpok on 09/25/03 04:37 PM.</EM></FONT></P>
!