Poor 3DMark06 scores in crossfire

[Gintok]

Greetings I'll try to keep it short, sweet and informative.

I have a Gigabyte MA790FX-UD5P motherboard paired with a AMD Phenom 9950 Quad core while using a ATI 3870 for graphics. Using this set up I achieved a score around about 12000. Having another 3870, I also added this to my machine only to achieve a score of 1000 more (total 13000). Is it me or is something not right.

Card seems to function fine. Just want some opinions on whether a score improvement like this is normal. I'm of the idea it should be 16000+.

 

[cjl]

It's strongly processor dependent. Overclock the CPU and you should see significant improvement.

 

[Gintok]

It's strongly processor dependent. Overclock the CPU and you should see significant improvement.

I thought PCMark05 was processor dependant and that 3DMark06 was graphical based. Either way still confusing to see why I got such a poor score. Makes me wonder what's wrong.

 

[cjl]

3dmark is supposed to be a graphics test, but it is extremely processor dependent. My machine (i7+4870 quadfire) scores 17k with my CPU at 3.2GHz, and 27k at 4.2GHz. Both of those scores have the GPUs at stock clocks, the only change is CPU clock.

Games tend to be more graphically dependent than 3dmark, and as a result, they should show more improvement with the second card. Nothing is wrong with your setup.

 

[Gintok]

3dmark is supposed to be a graphics test, but it is extremely processor dependent. My machine (i7+4870 quadfire) scores 17k with my CPU at 3.2GHz, and 27k at 4.2GHz. Both of those scores have the GPUs at stock clocks, the only change is CPU clock.

Games tend to be more graphically dependent than 3dmark, and as a result, they should show more improvement with the second card. Nothing is wrong with your setup.

Holy crap, I thought you would have hit way higher than that with an i7. Are you using 3DMark06 or a newer version of software?

 

[cjl]

That's in 06. I score around X15k in Vantage.

 

[Gintok]

Personally I'm still of the opinion that I (and you should be getting a better result somehow)

 

[cjl]

Well, as I said, the result is about right for your hardware, and the benchmark isn't really that representative of the load that games put on your system.

 

[ZipperSnake]

It seems like the whole 3dMark06 score is more CPU dependent (since it contains a CPU test (graphics run only by CPU)! ) then the Shader Modeling (SM) 2 and 3 scores are. And the Tomshardware Graphics charts shows the same I think. They almost double from 4500 to 8800 something for SM 3 for a 3870 VS two 3870 Crossfired.

http://www.tomshardware.co.uk/chart...ity/3DMark06-v1.1.0-HDR-SM3.0-Score,1631.html

I can not find anywhere what hardware (CPU and stuff) and drivers they used for this test. Anybody else?

I also have two 3870 and I have the same problem as many else seem to have: 3dmark06 scores 9700 without CF and 11300 with. No way near doubling there! But the strange thing is that it is working some... I' think I'll investigate how much the CPU will impact the SM scores. I'm aware that my CPU score in 3dmark06 is low.

 

[Rustyy117]

If your running the test at default res. of 1280x1024 then the CPU may be bottlenecking the graphics cards, Thats why when cj1 OC'ed his i7 his score jumped up alot!.


I had the same thing:

Core i7 920 at stock + HD5850 = 19K Score
Core i7 920 at 4.1ghz + HD5850 = 25K Score

Just OC that CPU and you will see some big improvements. Just don't expect double the score because you doubled the graphics power. I'm pretty sure 16K is achievable if you OC that CPU.

 

[Gintok]

But still an AMD 9950 Phenom Quad with 2 3870s in CrossFire should have had a higher score than the one I achieved.

 

[Rustyy117]

Your not understanding me.

You have a CPU bottleneck at that resolution. Meaning that your CPU is limiting the 3870's performance.

You could put 2 x 5970's in crossfire and still get the same score because your CPU would be holding them back. 3DMark06 is VERY! CPU hungry.

 

[Gintok]

Your not understanding me.

You have a CPU bottleneck at that resolution. Meaning that your CPU is limiting the 3870's performance.

You could put 2 x 5970's in crossfire and still get the same score because your CPU would be holding them back. 3DMark06 is VERY! CPU hungry.

So in your estimation, replacing the CPU with say a Phenom II 955, how much benefit would I get??

 

[Gintok]

Your score is not poor its what it SHOULD be.

Remember the two 3870s in crossfire mode is the same speed as the ATI 4850. The 4850 on your rig would get about 12k-13k on your spec.

You need to be humble because your score is spot on. Yeah your score would be higher if you OC'd your CPU, but your score is accruate as it stands

Wait I thought the 4850 was equivalent of a 3870, not two 3870s were equivalent of a 4850?

 

[Gintok]

You will only get about 1k more. Yeah the Phenom II 955 is alot faster than your current Phenom 9550 but 3Dmarks bulk scores are derived from the SM2.0 and SM3.0 scores which are GPU dependant. Also upgrading just for a higher 3Dmarks score makes no sense if your PC is actually running the games you want.

9950 (the one you speak of is Intel). But I appreciate your comments and responses. For the most part I'm very happy with my rig but was saddened by the 3dMark06 scores. I'd still say it's still quite a good machine over all.

 

[Rustyy117]

Agreed, Anyway why not get some "free" performance by OCing your current CPU, theres no point in upgrading your CPU just for the sake of 3dmark06.

Your 3870's are 2 generations old now. If your that worried about 3dmark then sell the 2 3870's save up for awhile then buy a 58xx card and OC your CPU.

 

[Gintok]

Who told you that, do you not read reviews before you purchase?

The 4850 is equivalent to a Nvidia 9800 GTX+ or GTS250

the 3870 is equivalent an 8800 GTS 320MB or 9600 GT.

I read a lot of reviews actually. In regards to the 4850 and the 3870. The 3870 I bought apparently outperformed the 4850 in some performance categories as well as price. Plus the 3870 I had was GDDR4 which was quite rare at the time as the.

I was told the 2900XT played against the 880GTS while my card being the next gen played against a 9800GT. Previous generations pitted a 1900series against the 7900 line IIRC.

Anyways thanks for the help

Agreed, Anyway why not get some "free" performance by OCing your current CPU, theres no point in upgrading your CPU just for the sake of 3dmark06.

Your 3870's are 2 generations old now. If your that worried about 3dmark then sell the 2 3870's save up for awhile then buy a 58xx card and OC your CPU.

I'm very happy with performance even though the benchmarks don't seem to show it. I can overclock but I don't think I need to as I can run everything I need comfortably with really good frame rates. Power consumption seems to go up when OCing too so I'll save it for when I need it.

 

[Gintok]

+1 omgitzfatal

The Phenom 9950 is an "OK" processor I wouldn’t replace it, even the 3870 in crossfire yields decent performance, I would be content with your rig for another 1-2 years. If you can manage to OC the processor its worth a shot though you'll only get about 3GHz-3.3GHz stable though because the Agena Phenom Is were not good overclockers but it should give about 1k extra to your score. Also if you can find a stable overclock for your video card's mem/clock and flash the firmware you might be able to add another 1k. That is an easy 2k extra in 3dmarks.

It wasn't that great of an overclocker but neither were the chips in my price range available so for me, it actually was the best overclocker and the best choice

Bang for buck, been really happy because at the time, this was the highest thing AMD had, (Spider platform). I've tried a bit of overclocking to see how far it can go and yeah it can hit 3.1 easy but power consumption goes much much higher. Cards are fairly good because they outperformed many other cards at the time in it's field so was really pleased at the performance, just mystified at the 3dMark result.

Finally can put this one to rest, thanks for the help./

 

[Rustyy117]

Yeah No problem, what matters is performance in games, and im glad your happy with the performance you get in games. F**k 3dmark