PDA

View Full Version : I'm not quite sure I understand Tom.


ChrisRay
06-19-04, 06:22 PM
This review really seems to make me Wonder.

Are AMD CPUs "really" this bad at Quake 3? Not from my experience. I mean I know Intel has always dominated the Quake 3 benchmark. But was it this much of a domination?

http://www.tomshardware.com/cpu/20030513/athlon_xp-12.html

I mean 3200+ is barely faster than a p4 2.4 Ghz @ 133 FSB.


Prior to the actual launch of the processor, we at THG received a benchmark guide for our test sample that explains to the press the best environment for testing the CPU. Among other things, it contains recommendations for many benchmarks and BIOS settings. It also advises testers to exchange a DLL before starting Sysmark 2002 in order to attain better results in the Media Encoder. We, however, did not make any changes to the benchmarks and stuck to accepted standards

Also this kinda surprised me. Whats the difference between the .dll? Is it SSE? 3dnow? Why is one standard and one not? Curious Curious. I mean I have read alot of tom reviews and I keep questioning some of them because I dont understand his reasonings for benchmarking this way.

And one other thing: before installing the operating system, AMD recommends deactivating the APIC mode (Advanced Programmable Interrupt Controller) in BIOS in order to boost performance. How does that work? When APIC is deactivated, you only have 15 available IRQs, instead of the standard 23. This puts the brakes on the system, making it go down to the level of Windows 98 and causing hardware conflicts under Windows 2000/ XP.


How do I do this? I didnt know it boosted performance ;)


http://www.tomshardware.com/cpu/20030513/athlon_xp-05.html#amds_recommendations

Sazar
06-19-04, 06:44 PM
its a bigger gap than I have seen @ other sites but its reasonably accurate.... HT and the 200mhz fsb help out...

if its any consolation athlon64 processors slap the p4's around in the same bench :)

http://anandtech.com/chipsets/showdoc.html?i=2088&p=10

corse thats with nv40 ultra afaik...

ChrisRay
06-19-04, 06:45 PM
its a bigger gap than I have seen @ other sites but its reasonably accurate.... HT and the 200mhz fsb help out...

if its any consolation athlon64 processors slap the p4's around in the same bench :)

http://anandtech.com/chipsets/showdoc.html?i=2088&p=10

corse thats with nv40 ultra afaik...

Not according to tom..

http://www.tomshardware.com/cpu/20040601/socket_939-16.html#opengl

Sazar
06-19-04, 06:58 PM
I am sure the card he used was different... also the resolution is different...

ChrisRay
06-19-04, 07:04 PM
I am sure the card he used was different... also the resolution is different...


Btw the 3200+ Barton is on a 200 FSB ;)

He used both a 640x480 and a 1024x768 reso, I'm trying to understand toms results is all. What demo does he use, Why does it seem to different than "most" results I have seen.

(yes this is in account for my 3200+ I have and I'm trying to get an idea of what its capable, Its unfortunate but seems only toms uses alot of different CPUS to do his benchmarks.

Chris

ect ect.

Sazar
06-19-04, 07:17 PM
yah I know but the fact that you have a better implementation of dual channel coupled with synched mem/fsb does more for socket 478 800mhz qdr fsb cpu's than it does for socket a's... there is a reason why amd was struggling to barely keep up with intel before the athlon64's came out...

I also don't understand tom's result.. they have a bigger intel bias than I have seen on many other sites...

:shrug:

I guess its the way they operate perhaps :)

GlowStick
06-19-04, 07:35 PM
Quake 3 is really optmized for Intel's sse2 stuff i belive, however if someone could force quake3 to run in intel mode with a a64 i bet you will see a BIG preformance increase.

Sazar
06-19-04, 07:37 PM
a64's have the sse2 instruction set as well... or are you saying there is a specifc intel path :confused:

/me was never in q3

Woodelf
06-19-04, 09:31 PM
I don't know why anyone with winxp would want to disable APIC. You need more than 15 IRQ's to cope with all of the new (serial ata, usb, 1394, etc...) hardware. It's been My understanding that you cant properly just disable it in the bios, and that it need's to be enabled prior to installing windows for higher than 15 IRQ support. Correct me if I'm wrong. :)

ChrisRay
06-19-04, 10:52 PM
I don't know why anyone with winxp would want to disable APIC. You need more than 15 IRQ's to cope with all of the new (serial ata, usb, 1394, etc...) hardware. It's been My understanding that you cant properly just disable it in the bios, and that it need's to be enabled prior to installing windows for higher than 15 IRQ support. Correct me if I'm wrong. :)


You're right, I tried to disable it and well. Windows XP took a nice craaaaash :)

Son Goku
06-19-04, 10:58 PM
yah I know but the fact that you have a better implementation of dual channel coupled with synched mem/fsb does more for socket 478 800mhz qdr fsb cpu's than it does for socket a's... there is a reason why amd was struggling to barely keep up with intel before the athlon64's came out...

I also don't understand tom's result.. they have a bigger intel bias than I have seen on many other sites...

:shrug:

I guess its the way they operate perhaps :)

tbh, I used to go to Tom's Hardware more. Read used to. He might have gotten better, but I take what he says with a grain of salt, unless it can be confirmed elsewhere.

Things might have become better since then, can't say for certain.

Edit: Here's a link

http://slashdot.org/articles/99/10/11/1545205.shtml

Here is what happend: Tom posted Quake3test benchmarks, this was at a time when the current test did not correctly calculate the fps scores from timedemo, this was a fact stated by John Carmack and Brian Hook. Hook stated in his .plan that Tom's benchmarks were inaccurate. Tom responded by saying Hook didn't know what he was talking about (keep in mind Hook was the #2 programmer on the quake3 project up untill he moved over to verant) and Tom refused to release any facts about how he ran the timedemo's other then that he did something or other to make it work. I could have made up #'s and claimed the same thing, and I have no reason to beleive that this was not what Tom did. Eventualy it boiled down to a pissing contest Tom vs Hook as to who knew how the Quake3 engine worked. What Tom was doing was the equivalant of me going up to Linus and telling him he doesnt know jack sh*t about linux. This had nothing to do with "Tom's opinions", this had to do with the fact that he was posting lie's that favoured a certain company using unverifiable techniques that the developers said were invalid.

This is how I remember it too. The first post that sparked this thread was from Tom himself, though this one was not. I seem to remember 8 versions of an appology on Tom's site also, the first several not sounding like appologies.

Well it's past history now, and his reviews might have gotten better. tbh, this has given me certain doubts concerning him though...

GlowStick
06-19-04, 11:50 PM
a64's have the sse2 instruction set as well... or are you saying there is a specifc intel path :confused:

/me was never in q3
bingo, when a quake3 based game starts on my intel machine it lists what sets it uses its the P3 set. When it starts on my amd machine its a diffrent set, less sse2 or whatever and all that.

Viral
06-19-04, 11:58 PM
It's to do with SSE, not SSE2 which Q3 doesn't support.

There is a patch for Athlon XP/64 processors that enables the use of SSE in Q3. The problem was Q3 was made to search for Pentium!!! or greater (while still pentium) as they were the opnly processors out at the time that had SSE.

The performance on AXP/A64 processors is much better when this patch is used, but no, the XP's still don't deserve there ratings when compared to the P4'C's or moreso, the A64's.

Dazz
06-20-04, 06:06 AM
Then again the Athlon XP's where out before the Pentium4 C.