Is there an app for AMD gpus? I am considering getting an AMD gpu; however, I looked at the current applications for this project, and I the only GPU apps that I see are for nVidia.

Am I missing something?

Thanks.

Unfortunatelly not. We have applications only for nVidia and not resources to make AMD gpu app. Some people wanted to rewrite it for AMD gpu but they never finished it.

Too bad. Their FP64 is superior to NV. :/

One good news - there is a work on an openCL application so if everything goes well we will have also application for AMD GPU.

Oh! Do these asteroid gpu tasks rely on 64 bit compute power?

I've been wondering which ( or do all) boinc projects rely on 64 bit precision gpu computing.

Like I heard milkyway is all about the 64 bits on gpu.

Are there currently Nvidia tasks available? Because looking at the leaderboard, I don't see many folk with more computing power than myself. And I'm only running CPUs. I would have thought anyone with a Nvidia running would have massive amounts more points than me.

Any task can be sent to either CPU or NVIDIA GPU. However, GPUs are not that efficient on this project. For example, a medium-class GTX 1060 needs about 30 minutes for one task, while a medium-class i5-8600K can do a set of 6 tasks in less than an hour thanks to the AVX-optimized application. So, in contrast to other projects with GPU applications, this project is pretty much CPU-dominated.

Any updates?

I tried to find A@H app source on forum.

Do anybody know about gpu or cpu app source code?

---

Hi guys,

As I said in one of my posts here:

Creating OpenCL application to support AMD GPUs was always at our RoadMap for years. Even now there is ongoing development on that. No mater there were some turbulence during the years, postponing and getting back on, we are doing our best to make it happen. Once we have solid PoC application we will be glad to provide it to our contributors.

Cool things takes time.

Thanks.
Georgi

OpenCL 3.0 is on the way.
Supposedly OpenCL 1.2 with selective optimizations, rather than the full package like OpenCL 2.0

source


I would also not try to run 2 WUs at a time.
It seems a 2080 Ti (with mild power reduction to 180W) does a WU below 10 minutes.
I've ran them from 125 to 290W, and Asteroids seems to be able to tax these GPUs pretty much completely.

Excited to get some AMD GPU WU in future! I got 4 hungry RDNA2 cards.

Hi folks,

We also would like to have it already done. Unfortunately the past two years messed up pretty well with our plans, as many others.
So at this momment I can't give any prognosis.

One thing is sure though, we'll find a way to have that OpenCL work finished at some point.

Thanks for understanding,
Georgi

---

“The good thing about science is that it's true whether or not you believe in it.” ― Neil deGrasse Tyson

Probably because CUDA is the highest performing development platform and is better known with much more developer support.

CUDA predates OpenCL by two years. That is a lot of time passed for developers to get comfortable with the CUDA SDK's

CUDA only runs on Nvidia cards and is optimized specifically for that hardware.

OpenCL has to make itself compatible with all architectures and is not optimized for any specific architecture.
It has to be a generalist and never will have the performance of CUDA.

---

A proud member of the OFA (Old Farts Association)

It has to be a generalist and never will have the performance of CUDA.

This is an old myth.
Now OpenCl, with LITTLE optimization, has the SAME performance of Cuda (and, sometimes, better).
You can also convert, almost automaticaly, the Cuda code to OpenCl with a lot of tools.

Hi folks,

We also would like to have it already done. Unfortunately the past two years messed up pretty well with our plans, as many others.
So at this momment I can't give any prognosis.

One thing is sure though, we'll find a way to have that OpenCL work finished at some point.

Thanks for understanding,
Georgi

No problem.
P.S. Is the code open source? Do you need help from community to develop it?