Hi all. After a long time, I am resuming my SAH crunching. I have a few questions regarding GPUs and Win10 machine advice in general:

@ejthedj
I took a look at your raft of systems. Almost any older intel cpu also processing on a 3 digit iGPU is likely to run very slowly on the cpu side. If you set one of your 3 work groups (home, school, work) on the Seti website to cpu only. And switch all your older/slower/3 digit version iGPU systems to cpu only (into that group), I expect the RAC on each system to go up. And probably go up faster.

The only one this trick won't quite work on is the box you have with both a Quardro gpu and a iGPU on. That one you need in a group that only allows AMD and/or Nvidia gpu tasks.

You have a bunch of machines that have 4 digit iGPU versions. Those may run the gpu faster without slowing down the cpu processing. Since you have 3 groups I would put them in a group by themselves with both cpu and iGPU proceesing.

And then wait a while and see if it does make a difference on your machines.

Not running an older iGPU on one of my boxes made a major difference in the cpu processing speed.

Respectfully,
Tom

---

A proud member of the OFA (Old Farts Association).

At present, the Windows platform is the slowest at SETI. Both Linux and macOS can use the CUDA Special App which is close to 3 times as fast as the OpenCL App.

Is that set in stone, or is there any way of compiling a special app for Windows to double or treble the output? All my cards use both Cuda and OpenCL.

It's just a matter of someone producing an application for Windows based on the code that Petri has developed.

---

Grant
Darwin NT

The next development in the "official" applications will be the incorporation of the data from Parkes, plus a few tweaks here and there. Such releases are often based on the most stable optimised applications.
So it all depends on.....

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

Delayed, not killed - just read on a few posts:
https://setiathome.berkeley.edu/forum_thread.php?id=83569&postid=1976780

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

At present, the Windows platform is the slowest at SETI. Both Linux and macOS can use the CUDA Special App which is close to 3 times as fast as the OpenCL App.

Is that set in stone, or is there any way of compiling a special app for Windows to double or treble the output? All my cards use both Cuda and OpenCL.

It's just a matter of someone producing an application for Windows based on the code that Petri has developed.

Raistmer, our Windows app developer recently started playing with Petri's codebase. So would make an assumption he is trying to wrangle it into a Windows application.

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

Hi all. After a long time, I am resuming my SAH crunching. I have a few questions regarding GPUs and Win10 machine advice in general:

@ejthedj
I took a look at your raft of systems.

I apologize. After a while I finally remembered that many older generation Intel iGPU's were labeled "4000" and "5000" etc. So my advice should have been put those in the CPU-only group.

Some of the 3 digit iGPU's may be significantly faster. Especially anything with "Iris graphics". I don't remember about the rest.

The older the design of the Intel cpu chip the more likely that crunching with the iGPU will slow the entire system down and lower production.

Tom

---

A proud member of the OFA (Old Farts Association).