LHC@home

For a while I couldn't get any Theory tasks (Message 37971). That problem eventually resolved itself. Now I do get tasks, but it doesn't matter how I configure project preferences...they all just use one CPU of my quad core.

At one time I used all four CPUs for a couple of days. Now it doesn't matter if I configure preferences to use 2, 3, or no limit CPUs.

I also run Einstein@Home GPU tasks, which use one CPU per task to feed the GPU. So I could run an E@H task and a two CPU LHC task concurrently. Now after starting a Theory task using one CPU and it's established at 25% of processor potential, if I start one E@H task, the LHC task immediately goes into Waiting to run.

I only have 6 GB of memory (8 GB is on order), and now I commonly see just a few MB free. But if I use a browser, then close it, there is more than enough free memory to run the E@H task. Their working set is no more than ~158 MB. But even forgoing E@H for now, there is no way I can download a LHC task that uses more than one CPU.

I have had a few computation errors and some tasks that self-aborted due to memory issues. The memory issues will be eliminated when I have 12 GB, But I don't understand what changed.

I can also mention that BOINC CPU benchmarks used to test two CPUs followed by a single CPU, but recently changed to using just one CPU for the duration of the task At that time BOINC Event Log reported just one CPU. But that has changed too. Now CPU benchmark uses all four CPUs followed by one CPU, and Event Log reports four CPUs. I also can't explain that, and unfortunately I don't remember if I ran the benchmark before or after I started another LHC task. I'm just 2:27:xx into the new task, so I won't be able to download another task until this evening.

Given that CPU benchmark now reports four CPUs instead of one, it will be interesting to see if I now can get a multiple CPU task. But I wanted to report this unusual behavior, and get the opinions of people more knowledgeable than me. I welcome all replies and I will keep you posted.

There was no way I could edit my first post, so I had to reply to it to add information.

As I said, never before when I ran BOINC CPU benchmarks did all four CPUs get tested, for a total of 100% of processor capability. And when I was only able to get 1 CPU tasks the benchmark test reported just one CPU.

The one thing that I did differently last night was to run Memtest64 for ~3.5 hours. Maybe the processor memory manager wasn't working correctly, and somehow Memtest64, which used every bit of processor power retrained it. Don't ask me how...it's just a guess anyway.

It will be interesting to see what happens next time I download a task. I'll keep everybody posted.

What are you Max # jobs & Max # CPUs, settings in the preferences?

What are you Max # jobs & Max # CPUs, settings in the preferences?

The maximum # jobs was 1 and the maximum # CPUs didn't matter. I could config it for 2,3, or unlimited, but I still just got one CPU tasks.

But apparently that has changed. I won't be able to download another LHC task until this evening. But this morning when I resumed an E@H 1 CPU task the CPU Usage reported by Task Manager jumped up to 50% and the LHC task didn't go to Waiting to run as it had been doing. So something clearly has changed. I've configured LHC project preferences to use 3 CPUs. I'll keep all but one E@H task in a paused state, and I'll pause the one that is running when I need to use a browser or do anything else.

The extra memory I ordered should be here by 2/20/2019. That along with two 2 GB DIMMs I'll leave in the other channel will eliminate all memory problems.