Advice

Warren Hardy
Warren Hardy
Joined: 6 Mar 19
Posts: 2
Credit: 359,092
RAC: 2,149
Topic 220889

Hello,

I am coming from SETI@HOME. I have a Linux based AMD Ryzen 5 2600 with 16GB RAM and dual GTX 1070TI GPU's. Currently I am running a SETI version which is customized for Nvidia GPU cudas.

With SETI ramping down I wish to switch to Einstein@home. I am running Einstain while my GPU's are still knocking out SETI. 

 

My Questions

What is the best optimization for this system? 

Is there optimized code for NVidia? Or is it 1 Nvidia code for all NVidia GPU's

Will more RAM helpful, I noticed I am using using 95% memory, and have tasks waiting for RAM, how much more RAM would be useful? 

All I want to do is optimized my system to get the maximum out of it 

Thanks

Munocat

Keith Myers
Keith Myers
Joined: 11 Feb 11
Posts: 649
Credit: 562,747,287
RAC: 1,013,362

Your cards are useful here. 

Your cards are useful here.  But E@H does not employ any CUDA code, they use OpenCL applications.

You won't be seeing the kind of crunch times you're used to seeing with Seti.

 

Warren Hardy
Warren Hardy
Joined: 6 Mar 19
Posts: 2
Credit: 359,092
RAC: 2,149

Thanks Keith,  You helped me

Thanks Keith, 

You helped me a lot on seti, shame they do not use CUDA. Does other project use CUDA? 

Betreger
Betreger
Joined: 25 Feb 05
Posts: 893
Credit: 480,043,564
RAC: 488,618

GPUGrid uses CUDA

GPUGrid uses CUDA and it really works out my GTX1060s

Keith Myers
Keith Myers
Joined: 11 Feb 11
Posts: 649
Credit: 562,747,287
RAC: 1,013,362

Very few.  You can look at

Very few.  You can look at this project list and look for the Nvidia logo to see which projects can use Nvidia.

https://boinc.berkeley.edu/projects.php

Doesn't indicate CUDA applications though, just that they have Nvidia compute apps which more than likely will be OpenCL based.

Asteroids has an old CUDA55 app but it is not compatible with the newer cards like RTX.

The best match to what we used to run at Seti is the acemd3 CUDA10 app at GPUGrid.net.

There are math projects that use Nvidia too apparently but I have no knowledge of those since those type of projects disinterest me greatly.

 

mikey
mikey
Joined: 22 Jan 05
Posts: 5,673
Credit: 540,862,098
RAC: 80,470

Collatz for example uses both

Collatz for example uses both Nvidia and ATI gpu cards.

steffen_moeller
steffen_moeller
Joined: 9 Feb 05
Posts: 77
Credit: 534,338,318
RAC: 2,301,636

If considering to run some

If considering to run some project other than E@H, with a short look at global curcumstances these days, you may decide to run Folding@Home to keep your GPUs busy and add Rosetta@Home to it for your CPUs.

mikey
mikey
Joined: 22 Jan 05
Posts: 5,673
Credit: 540,862,098
RAC: 80,470

steffen_moeller wrote:If

steffen_moeller wrote:
If considering to run some project other than E@H, with a short look at global curcumstances these days, you may decide to run Folding@Home to keep your GPUs busy and add Rosetta@Home to it for your CPUs.

Does Folding use AMD/ATIgpu's or just Nvidia gpu's? And I believe Folding can use cpu's too.

MarkJ
MarkJ
Joined: 28 Feb 08
Posts: 407
Credit: 66,065,894
RAC: 293

Warren Hardy wrote:My

Warren Hardy wrote:

My Questions

Will more RAM helpful, I noticed I am using using 95% memory, and have tasks waiting for RAM, how much more RAM would be useful?

I am running Ryzen 3600’s with 16Gb of memory. The CPU gravity wave tasks can use up to 2Gb each. I use an app_config file to limit the machine to 6 at a time. Not only does it reduce the run time (at the expense of half the throughput) but also the memory needed. Given the tasks are quicker you get through slightly more per day than maxing out the CPU cores.

Jim1348
Jim1348
Joined: 19 Jan 06
Posts: 364
Credit: 200,992,089
RAC: 7,410

mikey wrote:Does Folding use

mikey wrote:
Does Folding use AMD/ATIgpu's or just Nvidia gpu's? And I believe Folding can use cpu's too.

It uses OpenCL.  I run it mainly on Nvidia cards, but AMD works too (e.g., my RX 570 does fine).

The new Core 22 work units are very efficient, but they can run the card slightly harder than the Core 21 ones, so check your temps (and don't overclock).

PS - it works on CPUs too, but since the output of the GPUs is much greater, there is not much point to it.

Save the CPUs for Rosetta.

 

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.