Folding Forum

Hello,

I am new to folding and I would like to contribute.

First experiments with my laptop showed that the utilization of my working machine for folding is not very practical for some reasons.

Hence I decided to setup a box especially for FaH with a low cost CPU and low power consumption - the latter is very important here in Germany due to high energy cost - on a 24/7 schedule under Linux Ubuntu. I still have to obtain the hardware so my hardware decisions are still open.

For this setup I figured out two reasonable CPU's:
- Intel Celeron J3455 with a 10 Watt TDP at base frequency, but the on chip integrated CPU is not utilizable in FaH.
- AMD Athlon 5350 , Kabini core, on chip integrated GPU Radeon HD 8400 with a 25 Watt TDP. This 15 W extra consumption makes already over 30 EUR extra cost per year.
Apart from the iGPU the two CPUs show with respect to the benchmarks almost the same computing power.

Both CPUs could be used with boxes in miniITX and microATX form factors, the primer with the advantage of a much smaller box size, the latter with the possibility of the integration of an aditional descrete GPU in the box.
In addition Intel sells for the Celeron a cheap nice tiny Barebone box which would be the easiest setup at all.

Ok, with this informations I would like to ask you for some additional information which I could not figure out to make the most efficient provisioning decision:
- Would the iGPU of the Athlon utilizable for FaH to justify the 15 W extra power consumption?
I downdloaded the latest GPU.txt but I could not interpret it in this respect.
- How could I ensure that a still especially for FaH to be provisioned descrete GPU is utilizable for FaH? Works this automatically or what would I need to configure to get this running? Could there be a conflict with the iGPU?
For that I could not interpret the GPU.txt as well.

Because of rapid changes of available hardware offers I would appreciate it if you would response somewhat quickly.

Thank you very much in advance for any kind of help.

Cheers,
Peter

Hi & Welcome to the forum.

how much power are you thinking of using, if power costs are your primary concerns even a minimalist system is going to use 50-60 watts at idle, the consumption from the cpu/gpu is only going to add to that. Personally I would not fold on the cpu and would look to fold on a low power gpu instead, anything from NVidia's 10xx line of gpu's would be the most power efficient, for the lowest power consumption I would look at a 1030,1050 or 1050ti.

FAH is compute-intensive which means low power = low production and high power = high production. (Those statements are not "linear" to suggest that twice the production will require twice the power, but they're still reasonable statements.)

Modern hardware is more efficient than older hardware and GPUs are more efficient than CPUs, but Nathan_P's comment is certainly reasonable. First, though, EVERY WU completed before the deadline is useful to science, so a lot depends on how much you choose to contribute. If the hardware has a slot that will accept a high-/moderate-/low-cost NVidia GPU, that is a really good plan.

THE GOOD: The HD 8400 is a GCN based GPU, so is good for all graphics cores currently in use.
THE BAD: It has no private RAM, and shares system RAM with the CPU, so running both CPU cores and GPU cores seems like a bad idea.
THE UGLY: You may find that while it can run current cores, it cannot complete them before the deadline.

As a general rule, ALL iGPUs are under-powered. Some work and some don't but if they can't meet the deadlines, they might as well not work. Take a look at the Single Precision GFLOPs and compare to most consumer-grade discrete GPUs.

Hello,

thank you all for your guidance. This brings me really one step further. I will choose the Celeron J3455 or the somewhat weaker J1900 on a mainboard for a micro ATX box and one of the NVIDIA GPUs. Presumably I will let run the CPU idle and not use for FaH but I will test that.

Yet I have just two short questions:

@Nathan_P: Do you mean with your acronyms GPUs with the exact designations "NVIDIA GeForce GT 1030", "NVIDIA GeForce GTX 1050" and "NVIDIA GeForce GTX 1050 ti"?

And my concern as a FaH newbie: These GPUs are used by graphic cards by different vendors. Can I rely on that all these GPUs by all these vendors are accepted by FaH or which measures do I have to take to ensure that my graphics card selection works with FaH?

Thank you very much in advance for your answers.

Cheers,
Peter

FAH requires GPUs that use either a chip from NVidia or from ATI / AMD. It requires the OpenCL drivers from the chip manufacturer. I've never seen a situation where the company that packages the board/fan/VRAM mattered.

Yes, that's what he means. Many people prefer the NV 10* series GPUs.

About the only place where manufacturer matters for GPU usage is with laptops. Some have custom drivers to handle switching from an integrated, low power GPU to a higher power discrete GPU. The custom drivers often do not include all software components needed for folding.

Sometimes for folding on a laptop the drivers from nVidia or AMD can be substituted, but this is not always the case.

Thank you very much. Now I'm ready to get started!

contrib854 wrote:Hello,



@Nathan_P: Do you mean with your acronyms GPUs with the exact designations "NVIDIA GeForce GT 1030", "NVIDIA GeForce GTX 1050" and "NVIDIA GeForce GTX 1050 ti"?

And my concern as a FaH newbie: These GPUs are used by graphic cards by different vendors. Can I rely on that all these GPUs by all these vendors are accepted by FaH or which measures do I have to take to ensure that my graphics card selection works with FaH?

Thank you very much in advance for your answers.

Cheers,
Peter

Yes I do mean those gpu's - its just less to type , and the brand that you buy shouldn't matter.