Unigine Valley 1.0 Benchmark Thread ** Post your scores**

[Elfear]

Two cards at 1300/1650.




One card at 1360/1675

 

[Face2Face]

^^ Looking good. Very weird on the 4 card crossfire scaling...

 

[Kenmitch]

Seems to be scaling a bit better although not nearly as well as the guys at OCN. Maybe X79 vs P67?

My guess is it's a driver issue. If you rename the .exe on valley to the same as heaven's you may get better scaling if so....Worth a try at least.

The 13.4 CAP 1 notes state the following: Unigine Heaven: Improves CrossFire performance in 3 and 4 GPU's

^^ Looking good. Very weird on the 4 card crossfire scaling...

Yep....Gotta be a driver issue.

 

[Elfear]

My guess is it's a driver issue. If you rename the .exe on valley to the same as heaven's you may get better scaling if so....Worth a try at least.

The 13.4 CAP 1 notes state the following: Unigine Heaven: Improves CrossFire performance in 3 and 4 GPU's



Yep....Gotta be a driver issue.

I'll try renaming the .exe and see if that helps. I'll give trifire a go too and see how it scales.

Thanks for the tip with the Crossfire profile. That helped a lot.

 

[stahlhart]

Updated score -- GTX 670 SLI 1293/1825

 

[stahlhart]

Results chart is missing Elfear's impressive X2 score.

 

[MTDEW]

Results chart is missing Elfear's impressive X2 score.

Agreed, it should be listed.
His single GPU score isn't too shabby either.

 

[Face2Face]

Agreed, it should be listed.
His single GPU score isn't too shabby either.

Results chart is missing Elfear's impressive X2 score.

Done and Done

 

[Sable]

Specs in sig. Card is a reference EVGA 680.
According to Valley - GPU 1342mhz, Memory 3429mhz
Temperature went up to a toasty 82 degrees C.

 

[stahlhart]

Single card score: GTX 670 1293/1825

 

[stahlhart]

Updated score -- GTX 670 SLI 1293/1830

 

[Face2Face]

Updated score -- GTX 670 SLI 1293/1830

How did your score increase so much without changing your GPU clocks???
I don't think 5Mhz to the memory is going to add that much....

 

[stahlhart]

How did your score increase so much without changing your GPU clocks???
I don't think 5Mhz to the memory is going to add that much....

I switched the PCIe slot assignment for the two cards (one has a higher factory overclock and ASIC score), and reduced throttling by adjusting the power limit upward in AB.

I had earlier them reversed prior -- the slower card in the top slot driving the monitor -- in an effort to keep the temperatures more closely balanced and lower overall, but just for grins I decided to switch them around and live with the top card running hotter. Prior to this, I had been thinking that some stability issues I've been having trying to push my Heaven score higher were temperature-related. When I realized it was actually more of an upper clock limit issue, I went back to having the faster card in the top slot and running hotter, and got a little more out of it. I'll more than likely switch them back for normal usage.

4094 to 4117 is only 23 points, though...

 

[Face2Face]

4094 to 4117 is only 23 points, though...

Ok, no problem. 23 points for no clock speed increase is enough to ask questions :whiste: I will add your SLI setup! Thanks

 

[mojothehut]

Okay so I'm disappointed. After seeing folks with very similar setups, I think my score is about 100 or so lower than where it should be.
AMD 975@4.0
Sapphire 7950 (900/1250)
1400 score =(

Okay jpeg isn't loading, here's the link =\

https://www.dropbox.com/s/9dsztgxd40f1g69/Capture.JPG

 

[stahlhart]

Okay -- here's how it played out in more detail:

I have two MSI PE 670s; one has a factory overclock on 1254, and the other 1241. The 1254 has an ASIC of 90.0%, and the 1241 86.3%. The 1254 runs marginally hotter than the 1241; if I run the 1254 in the top slot and the 1241 in the bottom slot, I get about a 4C difference top to bottom. If I reverse them, the 1254 in the lower slot only runs about 1C hotter. Since I know temperature causes these cards to throttle back, I was initially thinking I'd get my best results with the cards in that arrangement. And I ended up getting to 4094 that way (switching the TIM to Gelid GC-3 made a huge difference in load temperatures; prior to that I was never able to get much higher than 3900.

So I then decided to see what I could do with just one of my cards in Valley -- so, natually, I'm going to go with the 1254 card here, so that gets moved to the top slot. Once I got what I could out of that (the 2221 score), I noticed that the 1254 was running pretty decent, and when I put the 1241 back in, I just dropped it in the lower slot. Ran another SLI run, and saw that it went up over 4100. I bumped memory up +10 from there, and got the score. I crashed with a TDR about five seconds after I took the screen capture, so for all intents and purposes it was a suicide run.

 

[stahlhart]

Ok, no problem. 23 points for no clock speed increase is enough to ask questions :whiste: I will add your SLI setup! Thanks

Thank you -- no worries; Elfear can and will top it, I'm certain.

 

[Face2Face]

I crashed with a TDR about five seconds after I took the screen capture, so for all intents and purposes it was a suicide run.

This is what it's all about.. Thanks for explanation.
GTX 670 Power Editions - Nuff said..

Okay so I'm disappointed. After seeing folks with very similar setups, I think my score is about 100 or so lower than where it should be.
AMD 975@4.0
Sapphire 7950 (900/1250)
1400 score =(

Okay jpeg isn't loading, here's the link =\

https://www.dropbox.com/s/9dsztgxd40f1g69/Capture.JPG

Try to clock her up some more - You have plenty of head room.

 

[Elfear]

Thank you -- no worries; Elfear can and will top it, I'm certain.

Oh man. Gauntlet has been thrown. :biggrin:

Nice scores by the way. Your memory runs much higher than mine.

 

[stahlhart]

Oh man. Gauntlet has been thrown. :biggrin:

Nice scores by the way. Your memory runs much higher than mine.

Thank you -- I'm pretty much bouncing around at the ceiling here, and just got one good bounce last night. Funny thing is that I can push memory up higher from here, and it probably won't improve the score -- I had some runs earlier where the score dropped a little. So I'm more or less at the upper limitation of this build now, I think (though there are new beta GeForce drivers today, so you never know).

I honestly didn't think I'd top 4094. One guy at OCN reached 100fps with his 670s -- I think he's got FTWs.

Anyway, good luck... :thumbsup:

 

[Sable]

Small correction in your table. My gpu speed wasn't the 1198 shown on nvidia inspector, it was OCed to 1342MHz that I put in my post.

 

[Face2Face]

Small correction in your table. My gpu speed wasn't the 1198 shown on nvidia inspector, it was OCed to 1342MHz that I put in my post.

I don't think your GPU was running at those speeds. If so, your score would be higher. If you look at Stahlhart's pics he is showing higher clocks on valley but it's not correct. You may want to ask him how to find the correct clock speeds - It should report correctly in GPU-Z

 

[Dankk]

Try to clock her up some more - You have plenty of head room.

Something definitely seems wrong with his performance. My 7870 (with very modest overclock) should not be beating his 7950... should it?

 

[Face2Face]

Something definitely seems wrong with his performance. My 7870 (with very modest overclock) should not be beating his 7950... should it?

Your right, it shouldn't be...

 

[stahlhart]

I don't think your GPU was running at those speeds. If so, your score would be higher. If you look at Stahlhart's pics he is showing higher clocks on valley but it's not correct. You may want to ask him how to find the correct clock speeds - It should report correctly in GPU-Z

I read mine from the Afterburner OSD -- take the core speed as is (for each card, it's the core offset in AB added to the factory clock), and divide the displayed memory clock in half. I'm not sure how Unigine arrives at the numbers it does.