Folding Forum

Is the version 8.3.5 folding faster than 7.6 on Ubuntu?

No.

The client doesn't do any of the actual folding work, just handles WU downloads and uploads and provides monitoring and control functions. The actual work is done by the downloaded folding cores and v8 uses the same ones as v7.

V8 is faster on Apple Silicon, because the client and cores are compiled native.

Are there GPU cores for Apple Silicon now, or is Mac folding still CPU only?

Still no GPU folding on macOS.

OpenMM now has a Metal3 plugin, but I don't think anyone has tried to build a mac fahcore, let alone tested it.
The client would need code to look for and whitelist Metal gpus.
There might also be different args to pass to such a core.

Don't hold your breath. It may never happen.

Thanks for your answers. I thought it was an update of the client and the core.

calxalot wrote: ↑Fri Feb 23, 2024 9:03 pm V8 is faster on Apple Silicon, because the client and cores are compiled native.

To clarify, the v7 client doe not have the necessary options to identify an Apple Silicon processor to the Assignment and Work Servers to download the native core. So the Intel core ends up being downloaded and used when a CPU WU is processed with the v7 client running on macOS. It does work well with a large fraction of the performance of the native core while running under Rosetta 2.

With bonus points, the difference in PPD can be significant.

I’m not running v7 anymore, and projects vary wildly. But I believe I’m getting +25% PPD vs intel cores under Rosetta.

calxalot wrote: ↑Sun Feb 25, 2024 2:25 am With bonus points, the difference in PPD can be significant.

I’m not running v7 anymore, and projects vary wildly. But I believe I’m getting +25% PPD vs intel cores under Rosetta.

I can confirm the difference between similar projects from V7 to V8 on a M2 Mac mini is 48% increased PPD so nothing to laugh at. I will track it a bit more but my last V7 and first V8 project were the same so it is likely as close to 1:1 as possible.