Finally, a gpu app on Rosetta@Home

Message boards : Number crunching : Finally, a gpu app on Rosetta@Home

To post messages, you must log in.

1 · 2 · Next

AuthorMessage
Profile [VENETO] boboviz

Send message
Joined: 1 Dec 05
Posts: 1991
Credit: 9,520,400
RAC: 12,860
Message 109296 - Posted: 27 May 2024, 5:19:25 UTC

From Ralph@home:

Starting this week, we will begin testing RoseTTAFold All-Atom, our state-of-the-art application designed to predict the structures of proteins in complex with ligands such as DNA and other biological molecules. It also models post-translational modifications. We are initially targeting only Windows machines with or without Nvidia GPUs for this test


I would prefer and opencl/rocm/oneApi app that run over all platforms (i'm Amd boy), but it's, despite all, an historical moment for this project!!
ID: 109296 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Grant (SSSF)

Send message
Joined: 28 Mar 20
Posts: 1670
Credit: 17,494,103
RAC: 24,547
Message 109297 - Posted: 27 May 2024, 6:08:24 UTC

Looks like a new CPU application, with, maybe, possibly, a GPU application at some stage (maybe...).
Grant
Darwin NT
ID: 109297 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile [VENETO] boboviz

Send message
Joined: 1 Dec 05
Posts: 1991
Credit: 9,520,400
RAC: 12,860
Message 109298 - Posted: 27 May 2024, 6:40:41 UTC - in response to Message 109297.  
Last modified: 27 May 2024, 6:49:56 UTC

Looks like a new CPU application, with, maybe, possibly, a GPU application at some stage (maybe...).


Yes.
But it's the first time that R@H admins speaks explicitly about work for gpu on our computers
ID: 109298 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile [VENETO] boboviz

Send message
Joined: 1 Dec 05
Posts: 1991
Credit: 9,520,400
RAC: 12,860
Message 109321 - Posted: 30 May 2024, 16:18:16 UTC

The first batch of this app on Ralph: all errors!!
Waiting for a better app
ID: 109321 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Grant (SSSF)

Send message
Joined: 28 Mar 20
Posts: 1670
Credit: 17,494,103
RAC: 24,547
Message 109323 - Posted: 31 May 2024, 4:49:42 UTC - in response to Message 109321.  

Looking at the error messages, the Tasks sent out were faulty.

A quick search gives this result
the reason you typically see the “End-of-central-directory signature not found” error when working with a zip file is because the file is either corrupt, the file download is incomplete, or it’s a multi-part archive file and the other components are not found, or the zip file is not actually a zip archive file. The most probable reason for most users to encounter this error when trying to unzip a zip archive though is that the zip file download is either incomplete, or the zip archive is corrupt.
With everyone getting the same error, i can't see it being a download corruption issue (unless there is some really odd configuration issue with the download server).
I'm guessing it was archived incorrectly- wrong flags set or there are files within that archive that are missing from the archive that should be there.
Grant
Darwin NT
ID: 109323 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
hadron

Send message
Joined: 4 Sep 22
Posts: 68
Credit: 1,556,165
RAC: 556
Message 109325 - Posted: 31 May 2024, 10:43:52 UTC - in response to Message 109298.  

Looks like a new CPU application, with, maybe, possibly, a GPU application at some stage (maybe...).


Yes.
But it's the first time that R@H admins speaks explicitly about work for gpu on our computers

On Windows, with an NVidia card -- whoopee. Call me when they come up with something that works.
ID: 109325 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile [VENETO] boboviz

Send message
Joined: 1 Dec 05
Posts: 1991
Credit: 9,520,400
RAC: 12,860
Message 109326 - Posted: 31 May 2024, 13:56:24 UTC - in response to Message 109325.  
Last modified: 31 May 2024, 13:57:22 UTC

On Windows, with an NVidia card -- whoopee. Call me when they come up with something that works.


:-))

After YEARS and YEARS of waiting for a gpu app, i think we have to be satisfied (a little bit)
Patience, we must have patience :-P
ID: 109326 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Grant (SSSF)

Send message
Joined: 28 Mar 20
Posts: 1670
Credit: 17,494,103
RAC: 24,547
Message 109328 - Posted: 31 May 2024, 22:22:59 UTC - in response to Message 109326.  
Last modified: 31 May 2024, 22:23:12 UTC

After YEARS and YEARS of waiting for a gpu app, i think we have to be satisfied (a little bit)
If they had released a GPU maybe, maybe.
So far, it's only a CPU application.
Grant
Darwin NT
ID: 109328 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile [VENETO] boboviz

Send message
Joined: 1 Dec 05
Posts: 1991
Credit: 9,520,400
RAC: 12,860
Message 109330 - Posted: 1 Jun 2024, 16:35:11 UTC - in response to Message 109328.  

If they had released a GPU maybe, maybe.
So far, it's only a CPU application.


In Italy we say "Se son rose, fioriranno" (letteraly "if they're roses they'll bloom", something like "time will tell").
ID: 109330 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile [VENETO] boboviz

Send message
Joined: 1 Dec 05
Posts: 1991
Credit: 9,520,400
RAC: 12,860
Message 109343 - Posted: 5 Jun 2024, 12:50:40 UTC

After some days of down, the servers of Ralph restarted and are releasing some wus.
All errors, as a week ago :-(
ID: 109343 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile [VENETO] boboviz

Send message
Joined: 1 Dec 05
Posts: 1991
Credit: 9,520,400
RAC: 12,860
Message 109351 - Posted: 7 Jun 2024, 9:59:06 UTC - in response to Message 109343.  

After some days of down, the servers of Ralph restarted and are releasing some wus.
All errors, as a week ago :-(


Another batch with the same app and the same error...
Why?
ID: 109351 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile [VENETO] boboviz

Send message
Joined: 1 Dec 05
Posts: 1991
Credit: 9,520,400
RAC: 12,860
Message 109381 - Posted: 14 Jun 2024, 6:45:52 UTC

The new app (0.02) has a lot of problems....
ID: 109381 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Grant (SSSF)

Send message
Joined: 28 Mar 20
Posts: 1670
Credit: 17,494,103
RAC: 24,547
Message 109382 - Posted: 14 Jun 2024, 7:14:11 UTC - in response to Message 109381.  

The new app (0.02) has a lot of problems....
But at least they are running now.
They just need to fix up the CPU usage & the lack of any checkpointing, and the inability to suspend them, and the fact that they never seem to complete...
Grant
Darwin NT
ID: 109382 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile [VENETO] boboviz

Send message
Joined: 1 Dec 05
Posts: 1991
Credit: 9,520,400
RAC: 12,860
Message 109384 - Posted: 15 Jun 2024, 6:49:30 UTC - in response to Message 109382.  

They just need to fix up the CPU usage & the lack of any checkpointing, and the inability to suspend them, and the fact that they never seem to complete...


Only little detalis... :-))
ID: 109384 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile [VENETO] boboviz

Send message
Joined: 1 Dec 05
Posts: 1991
Credit: 9,520,400
RAC: 12,860
Message 109392 - Posted: 20 Jun 2024, 9:33:00 UTC

Seems that the hw request for this app is an Nvidia gpu with, at least, 6gb of ram...
ID: 109392 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Grant (SSSF)

Send message
Joined: 28 Mar 20
Posts: 1670
Credit: 17,494,103
RAC: 24,547
Message 109393 - Posted: 20 Jun 2024, 10:00:32 UTC - in response to Message 109392.  

Seems that the hw request for this app is an Nvidia gpu with, at least, 6gb of ram...
Actually, with the latest version, it's (effectively) an 8GB minimum (i can't think of any cards with more than 6GB and less than 8GB). With the previous version my 6GB card was able to process work with no issues. With the new version, it is 1MB short of the minimum needed to process work.
1MB

6144 MB is the minimum, i've got 6143 MB on the GPU in this system.
1MB!
Grant
Darwin NT
ID: 109393 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile [VENETO] boboviz

Send message
Joined: 1 Dec 05
Posts: 1991
Credit: 9,520,400
RAC: 12,860
Message 109394 - Posted: 20 Jun 2024, 15:00:14 UTC - in response to Message 109393.  
Last modified: 20 Jun 2024, 15:01:01 UTC

Seems that the hw request for this app is an Nvidia gpu with, at least, 6gb of ram...
Actually, with the latest version, it's (effectively) an 8GB minimum


I hope that, after the debug of problems, they will start to think to "optimize" the code.
Not every R@H volunteer has a gpu with 8gb of memory...


Meantime, i continue to run Folding@home
ID: 109394 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Grant (SSSF)

Send message
Joined: 28 Mar 20
Posts: 1670
Credit: 17,494,103
RAC: 24,547
Message 109398 - Posted: 21 Jun 2024, 4:22:40 UTC - in response to Message 109394.  

I hope that, after the debug of problems, they will start to think to "optimize" the code.
Not every R@H volunteer has a gpu with 8gb of memory...
As i mentioned, it was running in 6GB of VRAM before. So hopefully they can bring it back under 6GB again.
Grant
Darwin NT
ID: 109398 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Sid Celery

Send message
Joined: 11 Feb 08
Posts: 2114
Credit: 41,098,908
RAC: 22,244
Message 109399 - Posted: 21 Jun 2024, 9:08:40 UTC - in response to Message 109393.  

Seems that the hw request for this app is an Nvidia gpu with, at least, 6gb of ram...
Actually, with the latest version, it's (effectively) an 8GB minimum (i can't think of any cards with more than 6GB and less than 8GB). With the previous version my 6GB card was able to process work with no issues. With the new version, it is 1MB short of the minimum needed to process work.
1MB

6144 MB is the minimum, i've got 6143 MB on the GPU in this system.
1MB!

Writing this in complete ignorance, so ignore if not relevant.

Is it like the situation we had with Raspberry Pi devices that time(?) when an amount of RAM needed was hard-coded in but free RAM was fractionally below, so only 4Gb devices could run tasks, even though actual RAM called was way less?

Similarly now 6144Mb VRAM hard-coded, but only 6143Mb free on those cards?

Or doesn't it work that way with VRAM? Disappointing if so.
Not so relevant for me as I think I've gone over the top having 2Gb & 4Gb cards and I doubt that will change anytime soon. Or ever
ID: 109399 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Grant (SSSF)

Send message
Joined: 28 Mar 20
Posts: 1670
Credit: 17,494,103
RAC: 24,547
Message 109402 - Posted: 21 Jun 2024, 10:15:56 UTC - in response to Message 109399.  

Is it like the situation we had with Raspberry Pi devices that time(?) when an amount of RAM needed was hard-coded in but free RAM was fractionally below, so only 4Gb devices could run tasks, even though actual RAM called was way less?
At this stage we don't know.
The previous version was able to run on cards with 6GB of VRAM without issue. Updated version came out and with it the message about the 6144 MB minimum if you had less than that.
But we don't know if the new version can't run on 6GB of VRAM, or if the minimum they've set is just that little bit higher than it needs to be.

To my way of thinking, if 8GB was the minimum amount then you'd set it that (or even 7.9GB), but to pick a value as specific as 6144MB? (to me that would be 6.2GB, or 6.1GB if it needed just that little bit more than 6GB).
It's just an odd value to use.



And from an installed base perspective, the vast majority would still be 4GB or so, but ruling out even those with 6GB would take another pretty big chunk out of the available compute resources.
It wasn't until the RTX 20 series that all models of card (including the bottom end models) had more than 4GB of RAM.
Grant
Darwin NT
ID: 109402 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
1 · 2 · Next

Message boards : Number crunching : Finally, a gpu app on Rosetta@Home



©2024 University of Washington
https://www.bakerlab.org