CUDA vs nVidia OpenCL User Poll
log in

Advanced search

Message boards : News : CUDA vs nVidia OpenCL User Poll

Previous · 1 · 2 · 3 · 4 · Next
Author Message
Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2528
Credit: 740,580,099
RAC: 0
Message 18592 - Posted: 15 Feb 2014, 15:43:54 UTC - in response to Message 18584.

a - I have four macs with old NVIDIA cards (GeForce 9400 and GeForce 320M) and since the project restart they have become very sluggish running C@H GPU tasks (much more sluggish than they were prior to the project restart). I don't care if the change causes a dip in my C@H RAC, as these units collectively account for approximately 5% of my total C@H RAC. In fact, unless something changes soon, I am going to cease crunching C@H GPU tasks on these units - I have already done so for two other macs with old NVIDIA cards (GeForce 9400M and GeForce 320M).


The CUDA an OpenCL apps have the exact same config and optimization really does work both ways. If you machine is too sluggish, reduce the items_per_kernel and the kernels_per_reduction to improve video response.

lvanst
Send message
Joined: 21 Mar 13
Posts: 1
Credit: 209,915,582
RAC: 0
Message 18593 - Posted: 15 Feb 2014, 16:46:52 UTC

My Radeon/7750/OpenCL stomps my Nvidia/650Ti/Cuda in performance testing...

From a numbers perspective (SP/DP GFlops) the cards are very close in performance.

So, for me, an OpenCL app running on a native OpenCL architecture is #1

Now, if chose to not write a native OpenCL app for my Radeon architecture, I'll run whatever you send me.

However, that would rob you of significant productivity in a mixed marketplace.

jjwhalen
Avatar
Send message
Joined: 17 Apr 10
Posts: 20
Credit: 202,526,055
RAC: 0
Message 18594 - Posted: 15 Feb 2014, 16:57:58 UTC - in response to Message 18534.

Option a) for me.
____________
Best wishes:)

DXT
Send message
Joined: 2 Aug 12
Posts: 1
Credit: 21,207,833
RAC: 0
Message 18595 - Posted: 15 Feb 2014, 17:00:32 UTC - in response to Message 18563.
Last modified: 15 Feb 2014, 17:03:11 UTC

I'd prefer CUDA !! vote for A

greetinx, DXT
____________

jayBee666
Send message
Joined: 21 Sep 09
Posts: 1
Credit: 1,712,249,430
RAC: 0
Message 18596 - Posted: 15 Feb 2014, 18:04:27 UTC

I would prefer CUDA. I'm using my CPUs for another BOINC project. So less CPU usage is mandatory. I would like the same for AMD GPUs. Exactly the way it was before the crash. Maximum GPU and minimum CPU usage. I don't like the solo workunits at all.

Profile Handicap [SG-FC]
Avatar
Send message
Joined: 1 May 12
Posts: 2
Credit: 2,626,465,616
RAC: 361,249
Message 18598 - Posted: 15 Feb 2014, 21:41:54 UTC - in response to Message 18596.

I would prefer CUDA. I'm using my CPUs for another BOINC project. So less CPU usage is mandatory. I would like the same for AMD GPUs. Exactly the way it was before the crash. Maximum GPU and minimum CPU usage. I don't like the solo workunits at all.



#2

mark
Send message
Joined: 12 Aug 11
Posts: 1
Credit: 1,049,798,178
RAC: 1,490,705
Message 18599 - Posted: 15 Feb 2014, 23:04:00 UTC - in response to Message 18534.

a a

Bill Eckard
Send message
Joined: 19 Jul 11
Posts: 1
Credit: 8,435,646
RAC: 0
Message 18600 - Posted: 16 Feb 2014, 1:40:07 UTC

CUDA is definitely faster and I prefer it with my 660 GPU cards. Using less CPU resource is also a big plus as my CPU can idle during its processing off periods, while the GPU processing continues at high speed.
____________

Profile fthibaud0001
Send message
Joined: 26 Dec 09
Posts: 2
Credit: 422,684,863
RAC: 195,071
Message 18603 - Posted: 16 Feb 2014, 8:13:15 UTC

Manual select in "Preferences for this project Collatz Conjecture preferences";
if not possible or too difficult to implement, then option "a)"

Regards

Profile [AF>Amis des Lapins] Phil1966
Send message
Joined: 25 Dec 11
Posts: 17
Credit: 296,621,902
RAC: 345,194
Message 18605 - Posted: 16 Feb 2014, 9:33:56 UTC

Source of the below info : PrimeGrid Preferences :

CUDA is usually faster on pre-Kepler GPUs (GTX 5xx and older).
OpenCL is usually faster on Kepler GPUs (GTX 6xx and newer). Use whichever is faster.

=> Possible to choose between OpenCl and CUDA + adapt the "block size"
This can be used to vary the kernel size within GeneferCUDA

If possible, this would be the ideal solution.

Every cruncher will be able to try CUDA and OpenCL for nVidia and use whichever is faster / optimal.

If not possible to implement, than solution A.

Again, Thank You Very Much, and please accept my appologies for this "second vote".

Philippe

pelpolaris
Send message
Joined: 14 Jul 09
Posts: 4
Credit: 1,307,382,547
RAC: 2,948,103
Message 18606 - Posted: 16 Feb 2014, 11:51:02 UTC

a)

gomeyer
Send message
Joined: 1 Mar 12
Posts: 4
Credit: 62,112,081
RAC: 0
Message 18607 - Posted: 16 Feb 2014, 11:59:43 UTC

a)

I'd be happy with a choice that says "CUDA or nothing". OpenCL is a waste on my equipment.

Mark
Send message
Joined: 17 Aug 11
Posts: 1
Credit: 342,827
RAC: 0
Message 18608 - Posted: 16 Feb 2014, 12:47:57 UTC - in response to Message 18534.

A)
____________

opr
Send message
Joined: 23 Apr 12
Posts: 1
Credit: 8,469,809
RAC: 0
Message 18609 - Posted: 16 Feb 2014, 15:28:25 UTC

I have ~10 years old Computer but still after downloading and installing intels cl-driver these latest workunits seemed to go through hour or two faster. Should it be available in options to choose wether use it or not? This obviously speeds up things, at least a little. C-option is interesting too. I can practically run only 1 cpu-wu and 1 gpu-wu at the same time , processor usage set to 75%.

opr

Leon Majal
Send message
Joined: 18 Feb 11
Posts: 1
Credit: 4,548,230
RAC: 0
Message 18611 - Posted: 16 Feb 2014, 16:16:16 UTC

I say A

Fire$torm [BlackOps]
Avatar
Send message
Joined: 13 Mar 10
Posts: 8
Credit: 416,166,690
RAC: 0
Message 18612 - Posted: 16 Feb 2014, 17:43:25 UTC - in response to Message 18611.

I say A


Ditto Here!!!!!!
____________

Jozef J
Send message
Joined: 2 Aug 13
Posts: 17
Credit: 1,343,479,849
RAC: 3,824
Message 18613 - Posted: 16 Feb 2014, 19:12:48 UTC

CUDA FOREVER! Nvidia.

A.)

"My God—it's full of stars!

Bill Brumbaugh
Send message
Joined: 10 Nov 09
Posts: 1
Credit: 655,371,316
RAC: 430,321
Message 18614 - Posted: 17 Feb 2014, 1:15:38 UTC

Cuda all the way.

Profile heffalumpen
Send message
Joined: 2 Mar 11
Posts: 15
Credit: 1,068,795,532
RAC: 1,247,752
Message 18615 - Posted: 17 Feb 2014, 13:38:19 UTC

A !!

mathew
Send message
Joined: 23 Oct 12
Posts: 3
Credit: 50,064,546
RAC: 0
Message 18619 - Posted: 17 Feb 2014, 14:24:42 UTC - in response to Message 18615.

A!

Previous · 1 · 2 · 3 · 4 · Next

Message boards : News : CUDA vs nVidia OpenCL User Poll


Main page · Your account · Message boards


Copyright © 2018 Jon Sonntag; All rights reserved.