Posts by Ninos Y

1) Message boards : Number crunching : 2CPU per GPU Task? (Message 1909830)
Posted 31 Dec 2017 by Profile Ninos Y
Post:
In multicore systems, is there any advantage in dedicating 2CPUs or 3 CPUs to every GPU task?

Thanks.
2) Message boards : Number crunching : GPU Lockups (Message 1859273)
Posted 3 Apr 2017 by Profile Ninos Y
Post:
Hi Shaggie,

This maybe a shot in the dark, but try turning off hyper-threading and dedicate 1cpu/gpu task if you haven't done so already. I've also experimented with 2cpu/task with my workstations that have high core counts.
Good luck!
3) Message boards : Number crunching : how to avoid cuda50 in favour of _SoG (Message 1832093)
Posted 23 Nov 2016 by Profile Ninos Y
Post:
great suggestions, thank you all.

For me, I had _SoGs running nicely with 3 tasks per GPU, then noticed some triplicate errors (something like that), and I believe this is what triggered the servers to abandon the distribution of _SoGs for me in favour of the CUDA50s. Then I scaled back to 2 tasks per GPU, now it is back to _SoGs.

There seems to be checks and balances to make sure the Berkley team isn't getting nonsense results.
4) Message boards : Number crunching : how to avoid cuda50 in favour of _SoG (Message 1831811)
Posted 21 Nov 2016 by Profile Ninos Y
Post:
I would like to know how to exclude the server from sending me cuda50 tasks.

Currently using Boinc Manager with 7.6.22.

If I delete the setiathome_8.00_windows_intelx86__cuda50.exe, it eventually just reuploads it (along with all the previous cuda iterations)

I am NOT interested in installing Lunatics.

I have 2 GTX 950s, running 2 tasks per GPU, they should be adequate to use _SoG.

How do we tell the servers: 'no thank you' to cuda50?
5) Message boards : Number crunching : Wow! Unexpected Result From SoG Command Line in Stock (Message 1823445)
Posted 11 Oct 2016 by Profile Ninos Y
Post:
you probably should copy over the command lines to the .sah program as well.
Its seems that the server will deliver a cluster of .sahs and a then alternate with a cluster of .SoGs.
6) Message boards : Number crunching : Receiving Arecibo 2009 Data? (Message 1821276)
Posted 2 Oct 2016 by Profile Ninos Y
Post:
How did you know you were getting old data?
7) Message boards : Number crunching : Does cuda core count still matter (GTX 780Ti)? (Message 1819805)
Posted 26 Sep 2016 by Profile Ninos Y
Post:
The GTX 780Ti has incredible 2880 cuda cores, more than the GTX 980Ti and GTX 1080.

The only other card, The Titan X has more @3072.
The Titan Z has more b/c there are 2 embedded GPUs @5760.

Would this still be a factor since now the platform is mainly openCL?

Have a look at the openCL benchmarks link:
http://www.tomshardware.com/answers/id-2268443/opencl-video-card.html

Its seems the Titan Black (2880 cuda cores) does well on the folding section.
8) Message boards : Number crunching : A Windows 10 update that blocks OpenCL use (Message 1819514)
Posted 25 Sep 2016 by Profile Ninos Y
Post:
reinstalling your video card driver alone is not enough, the problem will return once windows 10 decides to do another driver update.

reinstall your video card manufacturer driver
click start, type 'devices and printers', choose your computer, right click and go to 'device installation settings' and tick off: 'no, let me choose what to do' and then tick off: 'never install driver software from windows update'

your done.
9) Message boards : News : "Baffling" "signal" "from HD 164595" is probably none of the above. (Message 1814017)
Posted 31 Aug 2016 by Profile Ninos Y
Post:
If this truly has an alien origin, a signal that strong from a star 95 light years away would likely come from a Kardashev Type II civilization.
10) Message boards : Number crunching : Cruncher Upgraded - Penta-Nano (Message 1813684)
Posted 30 Aug 2016 by Profile Ninos Y
Post:
Very professionally done and you have a great work area with the proper precautions taken; The 4U case looks terrific too.

There is a lot of compute power here in those AMD cards; This will break top 10 easily soon.
11) Message boards : Number crunching : New Computer Build (Message 1798868)
Posted 26 Jun 2016 by Profile Ninos Y
Post:
and remember, Quad-memory channels for both of these CPU types, so always buy RAM in 4's, and install 4 at once.

The E5-2670 are socket r, not v2, v3, or v4 those are still DDR3


DDR3/DDR4 status doesn't matter:

the CPU is capable of running QUAD memory channels. These CPUs runs so fast, if you try to use triple channel or 2 channel, your memory control will bottleneck.

The V3/V4 are DDR4, but again - quad channel.
12) Message boards : Number crunching : New Computer Build (Message 1798832)
Posted 26 Jun 2016 by Profile Ninos Y
Post:
and remember, Quad-memory channels for both of these CPU types, so always buy RAM in 4's, and install 4 at once.
13) Message boards : Number crunching : Spicing UP the RAC a little... (Message 1798759)
Posted 26 Jun 2016 by Profile Ninos Y
Post:
I agree with Zalster, its ALL about the CPU cores now. How things have come full circle eh?

I'm sorry, but you need a real CPU core to drive your GPU instances, not a make-believe one.

As for the physical-core vs. HT-core issue, it's true you are sharing cache with one physical core. But the later model CPUs have "smarter" HT technology and have less cache thrashing. Its just a way of programming to run 2 threads on one core, so yes I could see how the HT-cores won't perform as well.

more physical cores = more WU done = higher RAC no doubt about it.
14) Message boards : News : SETI@home Version 8 for Android and GPU to be released January, 21st (Message 1758722)
Posted 23 Jan 2016 by Profile Ninos Y
Post:
It seems that once the GPU percentage reaches 99%, the remainder to 100% is verrrrry slow. happening with Cuda42 and Cuda50. I have multiple machines, all showing same symptoms.
15) Questions and Answers : GPU applications : Allocate different # of WUs on different devices (Message 1751957)
Posted 28 Dec 2015 by Profile Ninos Y
Post:
I have 2 NVIDIA GPUs, I want device0 to run 2 WUs, and I want device1 to run only 1 WU.

If we use the basic script that changes the gpu_usage to 0.5, this is "spread" across all connected devices.

How do you allocate different numbers of WUs to one GPU to the next.
There must be a script command that can do this.





 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.