Folding Forum

Mxyzptlk wrote: ↑Mon Feb 17, 2025 4:50 pm That's pretty darn fast... At most I'll only have a 5080, but I'll wait to see some of those numbers before I pull the trigger.

You aren't fooling us. You'll end up with a 5090 in each rig!


They are impressive, but just like the last series, it might be hard to find work units that will really push them some.

A lot of projects leave the 5090 underutilized. I have two questions. As more people get the 5000 series, will new projects start fully utilizing the 5090? And if it's a matter of not all CUDA cores being utilized, would unlocking vGPU support in the driver and then splitting it into two (or more) vGPUs allow them to be fully used?

arisu wrote: ↑Tue Mar 18, 2025 3:28 am A lot of projects leave the 5090 underutilized. I have two questions. As more people get the 5000 series, will new projects start fully utilizing the 5090? And if it's a matter of not all CUDA cores being utilized, would unlocking vGPU support in the driver and then splitting it into two (or more) vGPUs allow them to be fully used?

There no known ways to increase utilisation of wide high end GPUs. Occasionally projects with large amount of atoms come along and they tax those GPU decently, other than that, majority of projects are medium or small sized.
VGPU will not be supported on desktop Nvidia GPUs, while AMD seem to have it, but there is no massive gains from it, though it is not slower

The newer consumer Nvidia GPUs can be unlocked to support vGPU but if there aren't massive performance gains then it doesn't really matter.

Would three 5080s be more helpful for the project than two 5090s? It's a lot cheaper and has approximately the same PPD (~80M). Or is the 5080 also difficult to fully utilize?

Efficiency usually peaks with x60 model of each generation ... it sometimes shifts a little bit toward x70 depending on the costs and power draw.

There are only so many cards that can be fit in a machine before the efficiency savings are offset by the more expensive motherboard.

enroscado wrote: ↑Sat Feb 01, 2025 1:04 pm Wow, that's some power...

I wonder what it will crunch on Ubuntu/Linux.

GeForce RTX 5090 (Palit GameRock) under Ubuntu 24.04 with driver 570.124
Max PPD (project 16525) 61 209 402
Min PPD (project 16525) 59 049 102
Power consumption of the card is ~550W
GPU temperature is 76°C, fans at 52% (1950 RPM)

Retvari Zoltan wrote: ↑Fri Mar 21, 2025 6:52 pm

enroscado wrote: ↑Sat Feb 01, 2025 1:04 pm Wow, that's some power...

I wonder what it will crunch on Ubuntu/Linux.

GeForce RTX 5090 (Palit GameRock) under Ubuntu 24.04 with driver 570.124
Max PPD (project 16525) 61 209 402
Min PPD (project 16525) 59 049 102
Power consumption of the card is ~550W
GPU temperature is 76°C, fans at 52% (1950 RPM)

How often are you receiving WUs? Do you find that your card is waiting for a WU most of the time, or is it folding most of the time?