GPUPI challenge! This time, the gimp beats you! Or not.

[Bubbleawsome]

Do we know anyone on this forum that still has a TITAN/black? I wonder if their enhanced DP helps. It seems like it does, considering my 280x is only a little behind a 1600Mhz 970.

 

[DrMrLordX]

Thanks, biostud! It's interesting to see how close your score was to Makaveli's i7-970.

Bubbleawsome, I'm not sure who has Titan cards, but you could probably find someone looking through sigs in the video forum.

 

[WiseUp216]

Did a 32m run - .502s

And improved my 1b time (if a 2nd try is allowed ) - 38.917s

 

[DrMrLordX]

WiseUp216 has taken the lead! And yeah, you can take as many cracks at it if you like, I don't mind. I'll post the fastest score reported in this thread for any particular CPU/GPU.

edit: woops, totally missed .vodka's post. .vodka has taken the lead with R9 290 in triple Xfire!

 

[cmdrdredd]

Full System with GTX970 overclocked to 1560/8000 and running this in OpenCL. Cuda gave slightly slower results.

1B

32M

 

[Lepton87]

Thanks, biostud! It's interesting to see how close your score was to Makaveli's i7-970.

Bubbleawsome, I'm not sure who has Titan cards, but you could probably find someone looking through sigs in the video forum.

Titan should be just a bit faster than 780, about 15% clock for clock, I would have confirmed it but it shows me that MSVCP120.dll is missing error even though I installed that C++ package.
UPDATE:

Got it working, the results are:

Standard:
Device time for pi calculation: 46.998 s
Device time for memory reduction: 0.903 s

DP enabled:

Device time for pi calculation: 47.020 s
Device time for memory reduction: 0.391 s

What's the meaning of that second result that changed?

 

[DrMrLordX]

Titan should be just a bit faster than 780, about 15% clock for clock, I would have confirmed it but it shows me that MSVCP120.dll is missing error even though I installed that C++ package.
UPDATE:

Got it working, the results are:

Standard:
Device time for pi calculation: 46.998 s
Device time for memory reduction: 0.903 s

DP enabled:

Device time for pi calculation: 47.020 s
Device time for memory reduction: 0.391 s

What's the meaning of that second result that changed?

Interesting results! I am not sure exactly what the memory reduction time is, to be perfectly honest. You could always ask _mat_ at XS, or I could.

Btw, sorry I missed .vodka's post. He is now in the lead with triple Xfire!

 

[biostud]

Interesting results! I am not sure exactly what the memory reduction time is, to be perfectly honest. You could always ask _mat_ at XS, or I could.

Btw, sorry I missed .vodka's post. He is now in the lead with triple Xfire!

I'm pretty sure it is limited to a single GPU, no CF/SLI scaling.

 

[DrMrLordX]

I'm pretty sure it is limited to a single GPU, no CF/SLI scaling.

Seems to work for .vodka, though the scaling is obviously not perfect.

 

[.vodka]

Triple xfire, lol, it's a SINGLE 290. I wish I had three 290s. Check the screenshots in my post, GPUz clearly says it's one GPU.

XS original GPUPI thread

post 3 here has a single 290x with similar results. Hawaii seems to love this benchmark in particular.

 

[biostud]

Seems to work for .vodka, though the scaling is obviously not perfect.

Just seems that the r9 290(X) is superior to nvidia in this scenario.

 

[DrMrLordX]

Too true, .vodka has clarified that he only using one card for this bench. My mistake!

Deders, sorry I missed your post about biostud's time etc. I'm trying to make sure that the total time turned in by the bench (pi calculation + memory reduction) is the one entered into the leaderboard. Right now I have a time of 47.056s listed for biostud's 7990 in the 1b category (full system). I can't tell if he listed the pi calculation time only (omitting the memory reduction) or if he gave me the sum total time.

biostud, is the memory reduction time included in the scores you've turned in?

 

[biostud]

7990

32M
Device time for pi calculation: 0.668 s
Device time for memory reduction: 0.063 s

1B
Device time for pi calculation: 45.997 s
Device time for memory reduction: 0.617 s

 

[Deders]

Ahh that could be my bad, I must have misread the post before having a coffee, thinking that the 32M and 1B where the calculation and memory reduction values.

 

[DrMrLordX]

Thanks again biostud. The times you posted above are actually faster than the ones I had on file, so I'll put them up as your new times.

Not a problem, Deders. I find myself succumbing to varying degrees of confusion lately. Many edits have been necessary to clean up my (admittedly somewhat minor) mistakes.

 

[DrMrLordX]

GPUPI has finally been updated to v2.0 which introduces hwbot submissions and multi-GPU support! Check the link in the OP, it should take you where you need to go to get it.