Nothing beats the power savings of just running a host with a lot of CPU threads and some dinky graphics card to display the desktop. The power vultures are the GPUs.

So you are arguing against processing tasks on GPUs? Or for lower power bills :)

Tom

---

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

Nothing beats the power savings of just running a host with a lot of CPU threads and some dinky graphics card to display the desktop. The power vultures are the GPUs.

So you are arguing against processing tasks on GPUs? Or for lower power bills :)

Tom

No, well known that gpus outperform productivity on the equivalent power over cpus. But put more than one into a host and the power bills skyrocket.

If I am most concerned about the power bill, I would just run one multi-core cpu. Always aware what the actual power draw is on a host when you drop the gpus out of service and just run on the cpu and watch the power usage plummet on the UPS front panel display.

---

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

A proud member of the OFA (Old Farts Association)

I was reading this review of was a 3990x "the best choice" for a rendering cpu.

And it appears that depending on how many variables you crank into his analysis the 3950x was actually winning in many cases :)

But he was also thinking of rendering farms.

Oh, well. I think the 3950x is still the best high thread count cpu for the buck that I can afford.

Tom

---

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

They did tests of power draw at various thread count engagement at Anandtech and the 3950X was more efficient than the 3900X. Mainly due the 3950X runs at less Vcore per core than the 3900X due its better binned dies.

The charts on this page show the difference.
https://www.anandtech.com/show/15043/the-amd-ryzen-9-3950x-review-16-cores-on-7nm-with-pcie-40/2

---

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

A proud member of the OFA (Old Farts Association)

Now here is an EYPC result. A 7 slot MB with a e-atx form factor but lots of I/O and memory anyway.

https://www.anandtech.com/show/15628/the-gigabyte-mz31ar0-motherboard-review-single-socket-epyc

Tom M

---

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

Now here is an EYPC result. A 7 slot MB with a e-atx form factor but lots of I/O and memory anyway.

https://www.anandtech.com/show/15628/the-gigabyte-mz31ar0-motherboard-review-single-socket-epyc

That does look rather good and tempting...

Happy fast crunchin'!
Martin

---

See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)

Now here is an EYPC result. A 7 slot MB with a e-atx form factor but lots of I/O and memory anyway.

https://www.anandtech.com/show/15628/the-gigabyte-mz31ar0-motherboard-review-single-socket-epyc

That does look rather good and tempting...

Happy fast crunchin'!
Martin

Here is another temptation. Take advantage of the high Pcie lane count on a lower cost EPYC that can run even cheaper ver1 as well as ver2 (8c/16t) EPYC cpus.
[url]https://www.phoronix.com/scan.php?page=article&item=asrock-rack-rome&num=1 [/url]
This probably would be for the "high gpu count" experimenter.

We have no first hand reports that this combo or the previous one would allow 8+ gpus on a single system even with 1 to 4 or 1 to 8 riser expansion cards.

Tom

---

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

Well I made it! Top 1000, team Isle of Man top 500, position 1 and 50 million credits. All before the shutdown/hibernation/whatever. I wasn't sure I could manage before the apparent deadline.
I have been scouting around looking for a suitable alternative to seti and signed one computer up to foldingathome but they seem to be having expansion problems as only one WU so far.
Any offers for the many cpu threads approach? I am not a terrific fan of gpus as I have enough trouble keeping temperatures within bounds even with three fan radiators and a big pump.
jsm

Well I made it! Top 1000, team Isle of Man top 500, position 1 and 50 million credits. All before the shutdown/hibernation/whatever. I wasn't sure I could manage before the apparent deadline.
I have been scouting around looking for a suitable alternative to seti and signed one computer up to foldingathome but they seem to be having expansion problems as only one WU so far.
Any offers for the many cpu threads approach? I am not a terrific fan of gpus as I have enough trouble keeping temperatures within bounds even with three fan radiators and a big pump.
jsm

Both Rosetta@Home and World Community Grid are 100% cpu only projects. Einstein@Home also runs cpu tasks. PrimeGrid also runs cpu tasks.

In fact many of the projects are CPU-based. It is possible that the projects that run GPU-based processing are a minority?

Rosetti@Home is running Coronavid19 tasks. You can set the length of time target for your cpu tasks in Rosetti. The higher you set it the more ram it takes. 8 hours seems to take 2GB, 1 to 2 hours seems to take 1.6GB or less.

WCG's largest ram user appears to be the Africa Rainfall project. It also has 2 cancer projects, 1 Zika project, 1 TB project, 1 Hiv/Aides project and 1 "Microbiome Immunity" project.

---

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

It is possible that the projects that run GPU-based processing are a minority?

Yes. Not all algorithms works fine on GPU's and develop GPU apps is not an easy task due the way they works. Actually i know few programmers who can effectively do that. A lot could tell yes i know how to do and even could make it work, and probably will run faster than in the CPU's due the huge amount of parallel processing power the GPU's has, but rely make it efficient and optimized is not for the most of them. This is why try to find guy's like petri, jason, raistmer, etc. is not an easy task.

---

Here's a spectacular review and performance test for the latest mobile from AMD in an Asus laptop:


Linus Tech: Ryzen9 4900HS



Specs-tacular!!

Enjoy!
Martin

---

See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)

Here's a spectacular review and performance test for the latest mobile from AMD in an Asus laptop:


Linus Tech: Ryzen9 4900HS



Specs-tacular!!

And here's a very positive review from Tom's Hardware:

Asus ROG Zephyrus G14 Review: AMD Ryzen 9 4900HS Tested
wrote:

One of the best AMD-based laptops we’ve ever seen...

Enjoy!
Martin

---

See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)

Any offers for the many cpu threads approach? I am not a terrific fan of gpus as I have enough trouble keeping temperatures within bounds even with three fan radiators and a big pump.
jsm

Your AMD boxes have enough ram to run up to 50% of your threads with Rosetti@Home/Eienstein@Home. Your intel boxes mostly have 8GB of ram so they should be able to run World Community Grid.

Are you running some of your boxes "headless"?

Tom

---

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

Thanks for the suggestions. I will evaluate when seti runs dry. The AMD boxes are on a 4 way KVM switch but all the other m/cs are stand alone.

Any offers for the many cpu threads approach? I am not a terrific fan of gpus as I have enough trouble keeping temperatures within bounds even with three fan radiators and a big pump.
jsm

With Rosetta allow 1.3GB RAM per thread, and plenty of HDD space depending on how large you set your cache. With a 1 day cache Rosetta is using just under 7GB running the default 8hr Tasks. Their CPU application doesn't come close the the Seti Windows AVX application for working the CPU hard.
With Seti my CPU generally sat around 70°c (peaks over 75°c), on Rosetta it's around 60°c (peaks to 65°c) and that's with lower fan & pump speeds (the lack of heat from the GPUs would be helping with the lower CPU temperatures too).

---

Grant
Darwin NT

Will the tasks be allowed to run out of your stored cache? My 2990 ran out of cpu some time ago although there are gpus aplenty for the one gpu I am running.
jsm

Will the tasks be allowed to run out of your stored cache? My 2990 ran out of cpu some time ago although there are gpus aplenty for the one gpu I am running.
jsm

Yes. Rosetti@Home is standard BOINC. You can run both at the same time.

My Rosetti set up is also running World Community Grid and PrimeGrid side by side with Rosetti.
I expect the usual rules for limiting processing threads on a 2990wx still apply.

And I have been using the max_concurrent_tasks on each project to allow them to "play together". I have every reason to believe that Rosetta will take as many threads as you want to give it.

Tom

---

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

Yes. Rosetti@Home is standard BOINC. You can run both at the same time.

Note that's Rosetta with an 'a'.

The actual joining url is http://boinc.bakerlab.org/rosetta/

Yes. Rosetti@Home is standard BOINC. You can run both at the same time.

Note that's Rosetta with an 'a'.

The actual joining url is http://boinc.bakerlab.org/rosetta/

Opps, my flying fingers spellcheck apparently went on strike (again) or else its trying to keep it's "social distance".

Tom M

---

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

https://compfuncmacsbudd.ml/features/how-to-overclock-amd-ryzen-cpus

---

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