Why does the 9700 perform better on the AMD platform?

[CurtCold]

Ok, I was looking at Toms charts, and the 9700 performs better on the 2700+ platform, while the rest of the cards seem to just be in a jumble performance wise (4600 performs better than the 9500pro at some things, vice verse) If the P3 is the most powerful processor, then why doesn't the 9700 perform better under this platform?

Memory bandwidth?

Linkage

 

[IBdaMac]

I think you meant to say that the p4 is the most powerful processor, but AMD Athlon processors work better with Directx applications. It's just the architecture of the processor. AMD Processors are better built processors which is why they perform better with lower clock speeds.

 

[CurtCold]

Originally posted by: IBdaMac
I think you meant to say that the p4 is the most powerful processor, but AMD Athlon processors work better with Directx applications. It's just the architecture of the processor. AMD Processors are better built processors which is why they perform better with lower clock speeds.

I understand that AMD processors do more processes per clock cycle, hence the whole "rating" system. I just thought it was intersting that the card performed more fps under the athlon regime.

 

[majewski9]

thats actually a good question! Other than the directX maybe ATI has optimized drivers more for AMD lately or maybe that 333 fsb shines with a 333fsb Radeon.

 

[chizow]

Originally posted by: majewski9
thats actually a good question! Other than the directX maybe ATI has optimized drivers more for AMD lately or maybe that 333 fsb shines with a 333fsb Radeon.

Umm, this is the whole point that Imtim86 brings up every other day and Intel and ATI fanboi's flame in response.

P4 + 9700pro = low minimum frame rates.

Dunno the exact science behind it, it just happens. Yah, your average fps might be higher overrall, but again, its all about peaks and valleys. Also, the 9700pro scales much better with faster CPUs, which indicates it has plenty of bandwidth, but its starved for data from the CPU. OTOH, the GF4 flatlines around 1.8ghz, with little noticeable difference in frames, indicating it lacks the bandwidth to benefit from extra CPU power. I'm sure it has something to do with Intel's quad-pumped architecture not being fully utilized the majority of the time, whereas the Athlon pumps more through its pipeline per clock, although it has less bandwidth. I guess you could compare it to 8x AGP with 1 vertex shader vs. 4x AGP with 2 vertex shaders.

Chiz

Edit: Probably also explains why my "lowly" 2.4ghz Athlon with a totally unoptimized Radeon 9700pro (stock drivers) outperforms a P4 @ 3ghz in 3dmark2k1

 

[Y23KC]

I'm actually thinking of going with an athlon rig. Here is more score with a overclocked 2.53 with stock 9700pro.

 

[Spicedaddy]

Ouch, I need a CPU upgrade...

 

[nRollo]

In that particular link, the difference is most likely due to the higher memory bandwidth of the nForce2 motherboard.

 

[Noid]

check out my AMDbox

I had 16,149 on my AMDbox when i was at 191 FSB... but I had to see if I could break my oc... I did 2 days ago.. so now ...

Im at 189.4 x 11.5 = 2177 gives 15,831

Unbreakable now

My 85 dollar Via 333 keeps up with the N2's pretty good ... egh?

Just cuz someone got a 3dMark like that ... doesn't mean he can do it 24/7

I just finished running Prime95 with 3dMark DEMO looped on NON-DEFAULT with sound ON,, at 1024 @ 32 bit with APP pref ON and the Hardware manager running ... for a 12 hour period. (on air cooling)

 

[glugglug]

My guess is it has something to do with the P4 caching micro-ops rather than caching x86 instructions, to have the advantage of not needing to re-decode stuff already in cache..... but still has to do it if your code gets purged from the cache (or maybe even if it hits a new branch target). Maybe the ATI drivers decode to close to 12K microops causing stuff to get purged.

Hope that's not it... I read the Hammer converts x86-code to micro-ops before storing it in L2, let alone L1.

 

[bunnyfubbles]

21623 **AMD XP2700 & Epox 8RDA**

Date: 2003-1-17
Res: 1024x768 32bit
OS: Microsoft Windows 2000
User: oppainter@xtremesystems.org
CPU: Unknown 2885 MHz
GPU: ATI RADEON 9700/9500 Series

That's the second highest 3D mark Score showing now... verdict? nForce2 + Tbred B = win It'll be very interesting to see what Bartons can deliever. The Northwoods sure got a performance boost from the extra cache...

 

[Y23KC]

That test looks very strange. The reason I say that, is because in Tom's review of the 3.06ghz proc, it does much better than what he shows in the vga test. Of course, I take Tom's conclusions like a grain of salt so it doesn't really matter. Both platforms are really good at the moment and the barton should see a good increase.

 

[chizow]

Originally posted by: Y23KC
That test looks very strange. The reason I say that, is because in Tom's review of the 3.06ghz proc, it does much better than what he shows in the vga test. Of course, I take Tom's conclusions like a grain of salt so it doesn't really matter. Both platforms are really good at the moment and the barton should see a good increase.

The results are the same on any hardware site. That's a lot of salt, you should watch your cholesterol.

Chiz

 

[Y23KC]

The results are the same on any hardware site. That's a lot of salt, you should watch your cholesterol.

Chiz

Probably, thought I saw something else I guess.

 

[nRollo]

Chizow:
"The results are the same on any hardware site. That's a lot of salt, you should watch your cholesterol."
Just not the one we're on.

http://www.anandtech.com/cpu/showdoc.html?i=1746&p=14
The P4 is faster than the Athlon 2.8 on an nForce2 in the UT2003 bench that stresses cpu, and 800 points higher in 3dMark.

http://www.anandtech.com/cpu/showdoc.html?i=1746&p=15
It's faster at 2/3 games here as well.

 

[Y23KC]

Originally posted by: Rollo
Chizow:
"The results are the same on any hardware site. That's a lot of salt, you should watch your cholesterol."
Just not the one we're on.

http://www.anandtech.com/cpu/showdoc.html?i=1746&p=14
The P4 is faster than the Athlon 2.8 on an nForce2 in the UT2003 bench that stresses cpu, and 800 points higher in 3dMark.

http://www.anandtech.com/cpu/showdoc.html?i=1746&p=15
It's faster at 2/3 games here as well.

Thanks, knew they were different, just didn't have the energy to argue