• We should now be fully online following an overnight outage. Apologies for any inconvenience, we do not expect there to be any further issues.

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

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Bubbleawsome

Bubbleawsome

Diamond Member
Apr 14, 2013
4,834
1,204
146
Oh, just to clarify 3.8Ghz was the turbo. Poor wording on my part. I think the default 1 core boost was 3.9Ghz and 4-core was 3.6. Was clearing up that is is somewhere in the middle.

Took it back to 4.3Ghz and I'll test later today.

Also, the GPU was at 1100Mhz.
 
Last edited:
Yuriman

Yuriman

Diamond Member
Jun 25, 2004
5,530
141
106
HD7850 OC results:

VQzqQjL.png


HD7850 1050: 1m 40.599s
Device time for pi calculation: 99.495 s
Device time for memory reduction: 1.104 s

HD7850 1100: 1m 36.056s
Device time for pi calculation: 94.996 s
Device time for memory reduction: 1.059 s

HD7850 1125: 1m 33.919s

Device time for pi calculation: 92.902 s
Device time for memory reduction: 1.018 s

1150: 1m 31.845s
Device time for pi calculation: 90.859 s
Device time for memory reduction: 0.985 s
- video driver restarted

1175: 1m 30.347s
Device time for pi calculation: 89.266 s
Device time for memory reduction: 1.081 s

1175 w/ 1400 ram: 1m 30.305s
Device time for pi calculation: 89.227 s
Device time for memory reduction: 1.078 s

1200: 1m 28.413s
Device time for pi calculation: 87.370 s
Device time for memory reduction: 1.042 s

1225: 1m 26.624
Device time for pi calculation: 85.590 s
Device time for memory reduction: 1.034 s

I tried various memory overclocks ranging from stock (1200) to 1700mhz and found improvements to be within margin of error.
 
Last edited:
R

Ramses

Platinum Member
Apr 26, 2000
2,871
4
81
DrMrLordX said:
Is that the proper clockspeed? And are you using turbo?
Click to expand...

Plain Jane stockety stock, turbo's on but not going to do anything with all the cores active. If it did there might be a fire and alarms would go off at the nuke plant down the road. :)
 
Yuriman

Yuriman

Diamond Member
Jun 25, 2004
5,530
141
106
Here are my socket-only results:

KmbdtrA.png


Stock (3600mhz 4 cores)
20m 13.235s
Device time for pi calculation: 1185.106 s
Device time for memory reduction: 28.129 s

4000
18m 16.624s
Device time for pi calculation: 1069.509 s
Device time for memory reduction: 27.116 s

4400
16m 34.737s
Device time for pi calculation: 968.094 s
Device time for memory reduction: 26.643 s

4600
15m 51.754s
Device time for pi calculation: 925.782 s
Device time for memory reduction: 25.972 s

4800
15m 13.210s
Device time for pi calculation: 887.096 s
Device time for memory reduction: 26.114 s


Looks like an HD7850 is roughly 10x faster than an Ivy Bridge i5 for this.
 
ShintaiDK

ShintaiDK

Lifer
Apr 22, 2012
20,378
146
106
GTX980 1240Mhz.

Default:
Device time for pi calculation: 37.829 s
Device time for memory reduction: 1.115 s

32M:
Device time for pi calculation: 0.441 s
Device time for memory reduction: 0.063 s
 
DrMrLordX

DrMrLordX

Lifer
Apr 27, 2000
22,945
13,031
136
Ramses said:
Plain Jane stockety stock, turbo's on but not going to do anything with all the cores active. If it did there might be a fire and alarms would go off at the nuke plant down the road. :)
Click to expand...

heehee

Well if you think it's not kicking in, I can omit it I suppose.

Yuriman said:
Here are my socket-only results:

4800
15m 13.210s
Device time for pi calculation: 887.096 s
Device time for memory reduction: 26.114 s


Looks like an HD7850 is roughly 10x faster than an Ivy Bridge i5 for this.
Click to expand...

Not surprising. I'd still like to see a 5820k or 5960x do the CPU version of the test. Would be interesting to see.

More scores added. ShintaiDK is now the one to beat!
 
Last edited:
Bubbleawsome

Bubbleawsome

Diamond Member
Apr 14, 2013
4,834
1,204
146
If I can get my 280x to keep a 1250Mhz clock I might have a chance. It's time to RMA it anyway. It's been unstable since I got it.

Drooling over that 980 though.
 
R

Ramses

Platinum Member
Apr 26, 2000
2,871
4
81
DrMrLordX said:
heehee

Well if you think it's not kicking in, I can omit it I suppose.
Click to expand...


I think that's just the way it is with these, my ambient is low and temp are fine, but the higher clocked FX just aren't programed to do much turbo-ing unless it's under load on one or maybe two modules. I had an 8650 that I disabled it and just clocked it up, wasn't a significant enough difference to matter other than in benches so I just left it alone. And bought a 9590 that was "overclocked" out of the box. The lower power FX chips are supposed to have more aggressive Turbo since they have more margin for heat/voltage I've read.
 
Bubbleawsome

Bubbleawsome

Diamond Member
Apr 14, 2013
4,834
1,204
146
Put it back to 4.3
32M
00h 00m 11.051s PI value output -> 4286DFB04

Device time for pi calculation: 10.626 s
Device time for memory reduction: 0.425 s

1B
00h 17m 34.126s PI value output -> 5895585A0

Device time for pi calculation: 1044.027 s
Device time for memory reduction: 10.099 s
 
Deders

Deders

Platinum Member
Oct 14, 2012
2,401
1
91
GPU is actually running at 1215/7000 during the test.

OpenCL GPU: NVIDIA GeForce GTX 780 (12 CUs, 941 MHz)
OpenCL 1.1 CUDA 7.0.18 is ready.

Compiling OpenCL kernels ... done.

Calculating 1.000.000.000th digit of PI. 20 iterations.

Allocated device memory : 335546368 Bytes
Batch Size : 20M
Reduction Size : 64

00h 00m 00.255s Batch 1 finished.
00h 00m 01.330s Batch 2 finished.
00h 00m 02.334s Batch 3 finished.
00h 00m 04.368s Batch 4 finished.
00h 00m 08.938s Batch 5 finished.
00h 00m 13.020s Batch 6 finished.
00h 00m 14.033s Batch 7 finished.
00h 00m 15.034s Batch 8 finished.
00h 00m 17.059s Batch 9 finished.
00h 00m 21.595s Batch 10 finished.
00h 00m 25.649s Batch 11 finished.
00h 00m 26.665s Batch 12 finished.
00h 00m 27.666s Batch 13 finished.
00h 00m 29.699s Batch 14 finished.
00h 00m 34.271s Batch 15 finished.
00h 00m 38.354s Batch 16 finished.
00h 00m 39.367s Batch 17 finished.
00h 00m 40.368s Batch 18 finished.
00h 00m 42.392s Batch 19 finished.
00h 00m 46.928s Batch 20 finished.
00h 00m 50.876s PI value output -> 5895585A0

Device time for pi calculation: 49.949 s
Device time for memory reduction: 0.927 s

For some reason it won't allow me to do the CPU test. Nothing shows up in the drop menu.
 
Last edited:
sm625

sm625

Diamond Member
May 6, 2011
8,172
137
106
I have an i7-4700EQ system I'd like to run this on. But it refuses to run. The openCL and the standard exe both will not even start. It is win 8.1-64.
 
DrMrLordX

DrMrLordX

Lifer
Apr 27, 2000
22,945
13,031
136
If you guys have any technical issues running the program, please report them in the author's thread. I do not understand why the program would refuse to run on functional CPUs.

sm625, do you get any errors when attempting to run the program?
 
Deders

Deders

Platinum Member
Oct 14, 2012
2,401
1
91
DrMrLordX, My GPU is a 780, not a 970. (although it might have been if i'd been able to wait 2 or 3 months)
 
DrMrLordX

DrMrLordX

Lifer
Apr 27, 2000
22,945
13,031
136
Update: looks like _mat_ over at XS has weighed in on the issue of GPUPI behaving strangely (or flat out not working) under certain scenarios.

Looks like it might be an OpenCL driver issue. Maybe?

Anyway, he has a handy support thread that may help. Anyone that is unable to get the program to run on a particular part of their hardware (Intel iGPUs, CPUs, etc) is advised to check it out.
 
DrMrLordX

DrMrLordX

Lifer
Apr 27, 2000
22,945
13,031
136
Thanks for the continued input. biostud, any chance you could run GPUPI on your CPU? I am interested in seeing Haswell-E results. It shouldn't take but a few minutes to finish 1b.
 
Deders

Deders

Platinum Member
Oct 14, 2012
2,401
1
91
Hi again, just wondering about the consistency of the table, for instance you put my time as 50.876s which is (Device time for pi calculation: 49.949s) +
(Device time for memory reduction: 0.927s) but for biostud's 7990 result they weren't added together.
 
V

.vodka

Golden Member
Dec 5, 2014
1,203
1,538
136
CPU: 2500k @ 4.5 GHz ∼1.3v, 1866 10-10-10-24 1T

32M
Device time for pi calculation: 19.260 s
Device time for memory reduction: 0.795 s

1B
Device time for pi calculation: 1025.663 s
Device time for memory reduction: 24.612 s

GPU: R9 290 Tri-X

32M
@1075/1425
Device time for pi calculation: 0.356 s
Device time for memory reduction: 0.079 s

@1180/1600
Device time for pi calculation: 0.338 s
Device time for memory reduction: 0.065 s

1B
@1075/1425 (daily)
Device time for pi calculation: 23.473 s
Device time for memory reduction: 1.065 s

@1180/1600 (just to see how much faster it'll get)
Device time for pi calculation: 21.414 s
Device time for memory reduction: 0.969 s



I'd like the faster GPU results to be added to the table.
 
biostud

biostud

Lifer
Feb 27, 2003
19,934
7,041
136
DrMrLordX said:
Thanks for the continued input. biostud, any chance you could run GPUPI on your CPU? I am interested in seeing Haswell-E results. It shouldn't take but a few minutes to finish 1b.
Click to expand...

1B 5820K @ 4.3Ghz - 09m 25.828s
32M 5820K @ 4.3Ghz - 10.500s
 
Last edited:
You must log in or register to reply here.

TRENDING THREADS

COMPANY
RESOURCES
FOLLOW
Top Bottom