Early this year it frustrated me that I could not get cuda WU's.
Now it frustrates me that I get cuda31 WU's which are slower than opencl.
It is commendable that moowrap tries to accommodate all of us.
I cannot expect moowrap to optimise every GPU variant - so please let us decide.
Yours very faithfully,
JohnMD

You might be able to disable the cuda application with the <exclude_gpu> tag in cc_config.xml

<app> specifies the short name of an application (i.e. the <name> element within the <app> element in client_state.xml). If specified, only tasks for that app are excluded.

So that might work, if the both versions have different names, but I don't think it's the case.


What definitely works is the Anonymous platform mechanism, but than you might need to update the application yourself one day if a new one is released. There is an app_info.xml, which I use, you'll need to change the file names in it, but than it should work. Check what's in your client_state.xml and replace it in the app_info.xml.

EDIT: OTOH I see that most tasks you get are OpenCL anyway, so not sure it's worth the effort. The server seems to know by now, that the cuda app is slower on your system, so it shouldn't send you many cuda tasks anymore.

---

Strange to find out that the problem with these inefficient CUDA3 WUs is still not solved. :(

@project: Please send out the old WUs only on old PCs.

i would be good if the Users can decide which Version should run.

Got the ugliest version on the fastest cards (RTX20xx).
Now the GTX1080 runs faster then the RTX2080 with lower Power consumption because the RTX Cards get only Cuda31.

Make a selection in the preferences or put the app_info.xml and app_config.xml in one thread.

All other ist wasting of calculation power.

delete me

Just found this in the SETI.Germany Wiki.

It's in german, but the cc_config example you should understand. They use the file name of the executable instead of BOINC's appname in the <exclude_gpu> tag. With that you should be able to exclude a specific executable, but no idea if it actually works that way.

---

Does this project utilise Intel GPU such UHD 630 that ar integrated into the new I7 processors ? I note at present we can only tick Nvidia and ATI cards.

According to the applications page: nope, no apps with "opencl" and "intel" in the name. In fact, there are no Intel OpenCl available from distributed.net, so this isn't just a Moo! Wrapper issue. No idea if it's possible (and a good idea) to run the AMD/nVidia OpenCL application on Intel GPU via anonymous platform.

---

Yes, this is a horrible implementation of supplying slow arse CUDA31 tasks to fast GPUs on newly attached machines. Very inefficient and are watse of energy and time.

Please consider changing your system to have the server make the switch sooner. It shouldn't take as long as it does.

wee still got cunda31 task on nv 3090.
how is bout project? uupdates?

wee still got cunda31 task on nv 3090.
how is bout project? uupdates?

Used to only take a couple hours or so for it to figure out the CUDA31 sucks and OpenCL is better. Now it runs days and days with never switching to the better app.

Not sure what happened to this project, but seems admins are MIA and don't give a shite.

Pretty disappointing!