I think you missed the point. The question was whether IBM controlled the software, and whether they know how to code for GPUs. It is clear that the project scientists do know, at least to a point, and that it doesn't work here.

For the 123.456.789 time:
- IBM workers (in WCG) only code special versions of BOINC & work with server applications (input databases, manages WU, running server BOINC, manage OS on servers)
- scientists work with development & alignment (every app has a version, which is specially coded just for that science) GPU / CPU apps, that use BOINC, but also work with databases (inputs & evaluate outputs).
WCG can only help them in giving them some suggestions on which apps some other projects have run great.

So in 123.456.789 time: you're barking at the wrong tree!

Now that we've reached batch 200,000, I thought I'd take a look back to when this project was young.

We've submitted 82,851 years worth of work and returned one billion, nine hundred seventy-one million results since the project began ten years ago in 2013. Back then it was normal for projects to last a few months or years. Who could have predicted that MCM would become the backbone and eventual home of WCG?

The last estimate said we had 32.5 days of work unit availability left, which roughly correlated with our 200,000th batch.

https://www.worldcommunitygrid.org/forums/wcg/viewthread_thread,45360_offset,0

Perhaps we have begun a new phase of this project.
Perhaps we will begin a new phase shortly.

Either way, thanks to all for your outstanding contributions and enthusiasm, and I hope the contributions we have made in the prior decade pave the way for discoveries to come.

MCM1_0200035_1912_0 Linux 4.8.0-59-generic Valid 2023-06-09 08:12:50 UTC 2023-06-11 13:17:56 UTC 51.5 / 52.16 90.3 / 90.4

Just curious as to what kind of machine would take over two days to finish a unit.

Cheers