No Progress on Einstein
1 Sep 2023 19:31:42 UTC
Topic 230037
(moderation:
I have five Einstein tasks, all with more than 2 days remaining. I have suspended all but one.
That one shows 02:16:26 elapsed, 1day:14:52:47 remaining. But only 19.500% progress, with no progress being made for two days.
Is this normal? I don't like it.
I'm thinking of aborting them all. Any reason why I should not do that?
Application
Gamma-ray pulsar search #5 1.08 (FGRPSSE)
Name
LATeah1090F_1208.0_8237340_0.0
State
Suspended - user request
Received
8/25/2023 3:45:12 AM
Report deadline
9/8/2023 3:45:11 AM
Estimated computation size
105,000 GFLOPs
CPU time
01:27:22
CPU time since checkpoint
00:03:49
Elapsed time
02:16:26
Estimated time remaining
1d 14:52:47
Fraction done
19.500%
Virtual memory size
273.85 MB
Working set size
283.11 MB
Directory
slots/23
Process ID
12800
Executable
hsgamma_FGRP5_1.08_windows_intelx86__FGRPSSE.exe
Steven Gaber
Oldsmar, FL
S. Gaber
I think I solved the
)
I think I solved the problem.
I just reset the project.
There are now five tasks each with 5 hours to go. Plus one with 0.9 CPUs + 1 AMD/AT GPU.
That is acceptable.
Thank you.
Steven Gaber
Oldsmar, FL
S. Gaber
Steven Gaber wrote:I think I
)
Don't be surprised if it comes back.
If there is an underlying issue, that will not have solved it.
I looked at your tasks list. It shows the new tasks you received as a result of the project reset and none of the ones you previously had. It shows two aborted tasks you received in early August and nothing else completed for the balance of the month. This suggests an ongoing problem for some weeks.
Resetting the project throws away all current tasks and data, and forces a fresh download of everything. It doesn't remove any previously completed work so the lack of such tasks in your list on the website is a big red flag.
I think you may find it becomes 'unacceptable' once more.
Your machine has a "Ryzen 7 5700G with Radeon Graphics" 8 core, 16 thread processor. You are running CPU tasks as well as a GW GPU task for a new search, all using that 8 core CPU.
There is an announcement thread for this new search. People running searches like this one should keep an eye on the posted messages for any issues/problems that might arise. There were indeed some.
The problem of most importance to you was posted in quite a few places. Here is one of them, which was confirmed by the message that immediately followed.
These GW GPU tasks are both compute and memory intensive and are really only suitable for a mid to high end modern discrete GPU with at least 8GB VRAM. You are trying to crunch such a task on a GPU built in to the CPU as well as crunch additional CPU tasks. It's not surprising that your CPU is struggling with the combination.
My advice would be to temporarily suspend ALL CPU tasks and allow the solitary GPU task to try to complete without all that competition. If it does so in an acceptable amount of time, resume the CPU tasks for just long enough to report the completed GPU task and download one or more replacements. Immediately again suspend 4 CPU tasks, leaving 1 running, and see how long both the CPU task and the GPU task take to complete. If these two seem to have unduly slowed in their rate of progress, your machine is already overloaded. If not, maybe you could try 2 CPU tasks with the 1 GPU task.
When you find a combination that doesn't cause a slowdown, change your preferences to make that the default arrangement. You really need to experiment to find this arrangement where the machine is NOT overloaded. It may involve choosing a less demanding GPU search.
Cheers,
Gary.
The Op was having an issue
)
The Op was having an issue with the CPU Gamma-ray Pulsar Search #5, which I am as well. Not the GPU all-sky search.
Tom McFarland wrote:The Op
)
Actually, no he wasn't. He was having an issue with a GW GPU task running on an internal GPU and fighting for CPU resources that the GRP #5 tasks were also trying to use.
The OP's post was a full month ago and none of the tasks that were current then are still on display now. That being the case, how can you possibly state that there was a problem with GRP #5 tasks at that time?
I analysed the problem before writing my response so I know for sure that over-committing of CPU resources by running CPU tasks on all threads, as well as trying to support a GPU task on that same CPU, was causing a dramatic slowdown for CPU task run times. As far as I'm aware, there is no problem with the #5 search. There will just be overly long run times (and possibly other undesirable side effects) if you over-commit resources.
So, what is your problem? Your computers are hidden so unless you fully describe the issue, or provide a link to the host in question, or change your settings to 'un-hide' your computers, nobody will be able to diagnose the situation and offer advice.
Cheers,
Gary.