Low 3DMark06 score AMD help

[ibleet]

Sorry I am not an AMD guy so this one baffles me.

I have 4 Intel rigs with similar specs and they all score over 11,000 3DMark06.

This AMD rig scores 6,600 (SM2.0 2665, SM3.0 2760, CPU 2340). The CPU score is a little low, but the video scores are like half what they should be.

AMD X2 6000+ cpu 3.0Ghz, running at 3014 mhz, multi 15, bus speed 200.9 mhz
2x1GB Patriot PC26400, DDR2-800 dual channel, 5-5-5-16 running at ~750mhz (cpu/8)
Gigabyte M57SLI-S4 AM2 board, bios Award rev. FHe, 2-26-09 (nforce 570SLI chipset)
BFG 9800GTX+ 1024Mb, gpu clock 750, mem clock 1000, shader 1836 (gpuz) (PCI-Express link width x16)
Hiper type R, HPU-4K580, 580W modular PSU
WDC WD1500AHFD-00RAR4 10,000rpm raptor 150GB Hard drive SATA150 16MB cache
Dual opticals, 1 SATA, 1 IDE

OS - Windows XP Home SP3
Using the latest Nvidia driver

The last AMD rig I had (5200+, Crosshair) also scored very low 3DMark06 as compared to similar Intel rigs.

This PC should score at least 12,000 and I am stumped why the score is so low. Any ideas?

[GI-JOE]

CPU bottleneck would be my guess.

[ibleet]

I'm confused. How can the CPU bottleneck this rig?

I have an Intel E6750 @2.66Ghz, P35 DS3L, 8800GT, and it scores over 11,000 3DMark06, with a slower CPU. Something has to be severely wrong. Either that or AMD's have inferior benchmarks.

[GI-JOE]

I'm confused. How can the CPU bottleneck this rig?

I have an Intel E6750 @2.66Ghz, P35 DS3L, 8800GT, and it scores over 11,000 3DMark06, with a slower CPU. Something has to be severely wrong. Either that or AMD's have inferior benchmarks.

For one thing, an E6750 is faster than an AMD x2 6000+. The core 2 architecture can do more instructions per clock than AMD's x2 offerings. Also, I do believe that benchmarks tend to favor intel a bit more, but that is my opinion.

[ibleet]

[GI-JOE]

I would say for certain that it is a CPU bottleneck. Take the 9800gtx and put it in an intel rig and see what score you get then.

[ibleet]

Ya I already did that and it scores double in my E6850 rig.

Take a look at the -12v and -5v readings in Hardware Monitor once...I started up HWMonitor, ran 3DMark06 and those are the results.

Also note the 10-12c difference in the cores. My guess is that either the CPU or the HS is not flat enough because I pulled it and applied new AS5.

[Major_A]

Why does the score of a generic benchmark bother you? Try a real game at a real resolution and see what the results are then see if you think you have an issue. 3DMark06 is a poor indicator of actual gameplay and performance.

I wouldn't rely on HWMonitor too much. While it is giving you the correct motherboard readings and showing you what they are being reported as they still may be incorrect. For some reason my temps are being reported cooler than ambient conditions on this AMD rig and it's outside and air cooled.

Ambient temp as of writing, 81 degrees Fahrenheit.

[ibleet]

I never said it bothered me. This is a customer's rig. I no longer own anything AMD and this is exactly why. I am just curious about the low performance is all.

I have always used 3DMark06 as a benchmark and find it to be very useful in determining gaming ability. Besides that, my Intel rigs can all run Crysis on high settings while the AMD rig gets auto selected to medium, and then it runs at lower FPS (FRAPS) than the Intel rigs running on high settings.

The AMD rig also gets a substantially lower score (~50% less) when running Lightsmark. So regardless of the validity of benchmarks, the AMD rig is left in the dust in actual gaming FPS as well.

I guess the problem is that I didn't realize the limitations of AMD cpu's. I figured a 3Ghz dual core AMD cpu would be roughly equivalent to a 3Ghz Intel dual core cpu. I never expected the performance to be nearly HALF. I figured at the most, it would be 15-20% lower, but definitely not 50%. Oh well, it is what it is.

So what you are all saying is that this is NORMAL for AMD X2 cpu's? If that is the case, then my advice to the owner of this "so called" gaming PC, is to swap out the board and cpu for Intel counterparts. Maybe an E7400 or E8400 and either a P43 or P45 board.

It's a total waste to have that video card in that rig! Earlier today I put the 9800GTX+ in my Intel rig and the 3Dmark score was nearly 13,000.

By the way, I don't use HWMonitor exclusively. I also use CoreTemp, RealTemp, speedfan and Everest to monitor temps / voltages. Most times, they are within a few degrees / volts of each other once the proper TJMax is set.

Thanks for the input...anyone else have a theory? I figured every AMD fan would come to the rescue.

[Major_A]

That Windsor is an older CPU so it is what it is. Most people who are casual gamers aren't willing to spend more on Intel builds for more FPS. Most of the benchmarks you are running are also CPU limited (i.e. low resolution benchmarks 1024x768 or 1280x1024). With widescreen LCDs pretty much the norm I'd say that those resolutions are almost extinct (please don't refer to STEAM statistics). If you have the resolution and the bandwidth try this benchmark Cryostasis Tech Demo. It loads the games AI and PhsyX so you are actually getting a more indicative representation of what the gaming experience will be like. Maybe that AMD CPU still can't keep up but if you have time it's worth a shot.

For **its and grins have you tried flashing that card to a GTS 250?

[ibleet]

Thanks for the great input Major. I'll try that benchmark tomorrow and let you know how I make out.

No, I haven't considered flashing the GPU, but maybe once it gets into a suitable rig I might change my mind.

[ibleet]

Any AMD fan wanna take a stab at this?

Well anyway here is the update:

Did a fresh install of the OS to eliminate software issues, and updated to XP Home SP3. Swapped out the PSU with no significant change in voltages, temps or performance. Ran memtest86 with no errors on either stick. There are no device conflicts and the event viewer was error and warning free the past few days. I even flashed the bios to the latest revision.

Swapped out the 9800GTX 1Gb, for an 8800GT 256 and the score remained nearly identical, thus enforcing the bottleneck theory. I placed the 9800GTX in my e6850, 3Ghz rig and my score was 12,875. The AMD cpu score was 2317 while the Intel cpu score was 2662, which is reasonable (~15%). The SM2 and SM3 scores jumped dramatically from 2665 to 6138, and from 2760 to 6171 respectively, which is an increase of over 130%. Overall 6,612 to 12,857 is an increase of over 90%. The same performance increases were also realized using Lightsmark 2007 and while playing COD4 and Crysis using FRAPS.

The only thing remaining to do is to swap out the board and cpu with Intel counterparts and then retest everything. The last time I did that with a crosshair and 5200+ windsor, the 3DMark06 score when up by nearly 75%, from 5,800 to 10,100. I know its an older benchmark, but it still has some merit, at least to me.

I expected a 15-20% increase, but not 90%+.

[Major_A]

That doesn't make any sense. What those scores are telling me is this...

A) The CPU scores are wrong.
B) Because if they were that close there is no way there is that much difference in the SM3 tests. The only difference there is the CPU.

Try a benchmark based on an actual game.

[ibleet]

I know it doesn't seem to make any sense...that's why I'm having a hard time with it myself.

Here is the data. Hopefully this makes it more understandable: (no overclocking on any systems)

Mfg. --- Speed --- Ram (all DDR2800) --- Video --- HD speed --- 3D Score

#1 - Intel E7400 --- 2.8ghz --- 2Gb --- 8800GT 256 --- 7200 --- 11,370
#2 - Intel E6750 --- 2.66ghz --- 2Gb --- 8800GT 512 --- 7,200 --- 11,568
#3 - Intel E6850 --- 3.0Ghz --- 4Gb --- 8800GTS 512 --- 7,200 ---12,528
#4 - AMD 6000+ --- 3.0Ghz --- 2Gb --- 9800GTX 1024 --- 10,000 --- 6,612
#5 - Intel E6850 --- 3.0Ghz --- 4Gb --- 9800GTX 1024 --- 7,200 --- 12,857 (rig #3 with rig #4's video card)

From #4 to #1 = +72% increase
From #4 to #2 = +75% increase
From #4 to #3 = +89% increase
From #4 to #5 = +94% increase

I'm willing to bet that if I took the board and cpu out of system #3 and replace the AMD board and cpu from #4 that it would score over 13,000.

I'll do some actual gaming benchmarks (probably crysis) and post those results here later today.

[ibleet]

Here are some more results:

Lightsmark07
- 1280x1024
Intel E6850 (# 3 above) - 447.2 fps, AMD 6000+ (#4) - 199.1 fps
- 1680x1050
Intel E6850 380.1 fps, AMD 6000+ 186.9 fps
From AMD to Intel rig = +95% increased fps

Crysis Benchmark Tool - High Spec, 1680x1050
- GPU Test
Intel - 32.89 avg fps, AMD - 21.95 avg fps
- CPU Test
Intel - 30.71 avg fps, AMD - 22.02 avg fps
From AMD rig to Intel rig = +50% increased fps

The Intel rig is smashing the AMD rig every time with increased performance of no less than 50%.

[Major_A]

Just looking at the Crysis benchmark alone that says a lot in itself. It just shows that Ghz isn't everything. CPU architecture has a lot to do with it. Crysis is a still very GPU bound and still scales very well as the processors get faster. Do you have another game that hammers a GPU like that you can test, like Far Cry 2, Fallout 3 (outdoors of course), STALKER? If you do get FRAPS pick a point and run from one point to the other and try to be consistent. Then see what the benchmarks tell you there.I have a feeling though that the AMD CPU just won't be able to keep up.

Is this a friend's computer or something? If so you might want to point them in the direction of a cheap dual core Intel (like a E5300 or something) with a decent motherboard and overclock the snot out of it. Just an option.

[ibleet]

Ya it's a customer's PC. He picked it up tonight and has decided to upgrade the board and cpu to Intel. Most likely an E7400 / P43 or P45.

Thanks for all your input. It was great bouncing all my ideas off of you and you kept me moving in the right direction.

[ibleet]

Final Update:

I swapped out the AMD board and cpu for Intel: E7400, Gigabyte P43-UD3L. No other components were changed and no overclocking!

3DMark06 score increased from 6,612 to 12,484
Lightsmark score increased from 186.9 to 496.5
Crysis Benchmark increased from 21.95 avg fps to 35.95 avg fps

The AMD board and CPU cost ~$180. The Intel board and CPU cost ~$200. That's a considerable performance increase for $20.

This is the 2nd time that I have personally seen this type of performance boost by merely changing the board and cpu from AMD to Intel counterparts.

Draw your own conclusion from these facts.