Posts by ProDigit

1) Message boards : Number crunching : Quchem multithreading (Message 939)
Posted 16 Jul 2020 by ProDigit
Post:
I'm wondering if those multi thread WUs are the ones giving validation errors?
I've got 900 of those, where I presumably not get any points for.
2) Message boards : Number crunching : Validation errors (Message 938)
Posted 16 Jul 2020 by ProDigit
Post:
Seems like somehow I've managed 900 long WUs with validation errors. They're all mostly processed on Ryzen 3000 CPUs, that are running pretty much independently for months now.

https://quchempedia.univ-angers.fr/athome/results.php?userid=243&offset=0&show_names=0&state=5&appid=

I'm wondering why this happens?
The system runs stock with motherboard overclocking.
Most projects don't show any errors..
3) Questions and Answers : Web site : Forum issue on mobile (Message 923)
Posted 3 Jul 2020 by ProDigit
Post:
On the mobile site, the left bar (where the username is located) is too wide, not allowing me to see much of the text.

Below is an image on how the website looks like on a ~2100 by 1080p screen (minus the titlebar):



In contrast, this is how the same forum looks like (ignore the dark theme) of the boinc forums:



As you can see, I can read much more over there.

Maybe the badges are the cause of this?
4) Message boards : Number crunching : missing computers (Message 922)
Posted 3 Jul 2020 by ProDigit
Post:
I can only presume that in the infant stages, getting the data is more important, than to get the data validated.
Like damotbe wrote, they have an ai system that scans for the results, so they can determine if a result is erroneous or not, by a certain percentage range.

Once the project gains more computational resources, they can be assigned to verifying or duplicating WUs.
5) Message boards : Number crunching : Not getting small WUs anymore (Message 921)
Posted 3 Jul 2020 by ProDigit
Post:
We analyze short workunits, called OD9_athome. Many calculations did not give valid results and we managed to produce an artificial intelligence that can predict reasonably well the calculations that give bad results. This is a very good result that we should be able to report at the end of 2020 or the beginning of 2021. We're already preparing the next step and we're generating new molecules as we speak. Soon, we should have between one and two million of small molecules.

Keep us updated!
I'm keeping this project on the small servers for now.
6) Message boards : Number crunching : Quchem multithreading (Message 920)
Posted 3 Jul 2020 by ProDigit
Post:
First 4 core WUs also processing well!
My PPD count for the day went up by 2x (I do share projects)

I believe 4 core WUs are a big improvement, especially on AMD Ryzen 3000 series CPUs, over the single threaded WUs!
Shared l-cache might be helping out a lot!

Good job Quchempedia team!
7) Message boards : Number crunching : Quchem multithreading (Message 915)
Posted 30 Jun 2020 by ProDigit
Post:
Good news!
My first two, real multithreaded, 8 CPU QChem Long WUs arrived and are crunching with success!
Instead of 3 days for 1CPU, they seem to crunch at around 7.5 to 10 hours est...

The boinc site, "Max # of CPUs for this project" wasn't set, but has a maximum setting of #8 cores.
8 is a good setting, perhaps 12 would also be possible, and slightly better.
24 is too much, as it'll hog up an entire 3900x, and most of a 3950x.
The good thing about such WUs is on Ryzens, they should have shared L-3 cache, allowing data to be shared between cores...
Would be nice if Quchempedia could use that advantage, though current 3000 series chipsets have 2 CPU blocks of 6 or 8 cores.
4000 chipsets should be able to share that data between all cores.
8) Message boards : Number crunching : "Multithreading" in prefs (Message 914)
Posted 27 Jun 2020 by ProDigit
Post:
Why would you prefer a x-thread WU, over running x- amount of single threaded WUs?
In Linux it doesn't make sense, but I get what you're saying if you're running virualbox, and each WU loads in it's own virtual environment.
9) Message boards : Number crunching : Long work units. (Message 913)
Posted 27 Jun 2020 by ProDigit
Post:
As far as I know, Quchempedia doesn't send out multi cpu WUs.
They're all single thread WUs.
If you just want your PC to crunch on 1 WU, set your CPU count to 1.
If you're sharing your CPU with other projects, adjust the app_config.xml file, to have a max of x-amount of WUs at a time.
10) Message boards : Number crunching : Not getting small WUs anymore (Message 912)
Posted 27 Jun 2020 by ProDigit
Post:
Some of my computers are a bit slower.
They have only 1,5 to 2GB of RAM (with 1,8GB available).
They can't run the Long WUs.
Yet they never seem to get any work.

Is there a shortage in normal WUs?
11) Message boards : Number crunching : Credits for t8=t4=t2=5.000 seriously? (Message 847)
Posted 21 May 2020 by ProDigit
Post:
Again, to keep the conversation in line with the original OP's thread,
The LONG WUs should get higher credit.
Especially now that I see more and more units hitting 3 and even 4 or 5 days!
in that timespan I could finish a whole lot of small WUs, and get a much more consistent score.
I think QChemp needs to adjust the PPD scoring on these units.
And perhaps should start thinking in pairing them in cpu core pairs instead (find a way to do parallel computing on those WUs.
They take way too long!

a 3,5 days task can easily be given to systems having 8CPU threads. It should be done in about 10 hours that way, which is way more reasonable!
So find a way to create those WUs using parallel computing of 4 threads or more.
I think 8 threads is easy to get, as most PCs nowadays have 8 threads.
And leave the short WUs for PCs with less than 8 threads, doing WUs on a single thread like is now.

I'm sure you could learn something from other projects, how they do it.
Lots of projects serve 4, 6, 8, 10, 12, and even 16 core WUs.
And they might do 24 soon, as it'll be more common, to find such amount of cores on a PC in the near future!
12) Message boards : Number crunching : 2,5 days long and counting... (Message 846)
Posted 21 May 2020 by ProDigit
Post:
I got about 7 WUs of over 3.5 days on a Ryzen 3900x, and several on my 3950x.
I would plea to make such long runners higher in PPD.
They're worth more than the 5000 credit assigned to them!
13) Message boards : Number crunching : Credits for t8=t4=t2=5.000 seriously? (Message 831)
Posted 4 May 2020 by ProDigit
Post:
Your CPu isn't overcommitted.
Why do you use 2 CPUs on a WU?
Is this a setting you have, or are you running beta projects?
I only get single core WUs (1 CPU per WU).
If you're not running the CPU at 100% yet, I'd increase the setting so 1 more core would be used.
It appears Goofy, and Prime isn't using a lot of CPU resources.
14) Questions and Answers : Windows : "Postponed: vm job unmanageable restarting later" status in quchempedia (Message 818)
Posted 27 Apr 2020 by ProDigit
Post:
I set my RAM usage to 95% when not used and 90% when in use.
Yet of the 32GB I have, only 14-16 GB are used.
15) Questions and Answers : Windows : "Postponed: vm job unmanageable restarting later" status in quchempedia (Message 814)
Posted 26 Apr 2020 by ProDigit
Post:
I think with that much memory, boinc may be thinking there's virtual memory (swap) getting used.
I'm running 32 threads right now, and I can tell you, you don't need more than 500MB per thread for QuChemPi.
On a system with 64 cores, I would assign between 32GB and the next value up (40GB or 48GB or so).
More is just wasting memory needlessly.
16) Message boards : Number crunching : Credits for t8=t4=t2=5.000 seriously? (Message 813)
Posted 26 Apr 2020 by ProDigit
Post:
If you set your CPU to 99%, and can see that one or more threads are running at below 50-75%, setting CPU to 100% is NOT over-committing the CPU!
If on the other hand, some GPU projects report incorrect CPU usage (eg: they report 0.29CPU per GPU, and you have 3 GPUs, but in reality they're using 0.997CPU, then you're over committing the CPU. Setting CPU to 99% will still show 100% CPU utilization in this scenario.

In most scenarios, an over committed CPU will not result in a task taking 3x longer to finish, since threads are always shuffled around.
17) Message boards : Number crunching : Checkpoint? (Message 795)
Posted 19 Apr 2020 by ProDigit
Post:
That's a concern... VM makes checkpoints, but has several other drawbacks. Native App (Linux) has not, but works better.
One hope, is to get a funding grant or an enthusiastic volunteer.

Why do you say native linux does not have a checkpoint?
Other projects do have a checkpoint on linux?
Yes, in a remote area where I live, the electricity is not always stable, and we have an outage anywhere from once a month to a few times a week.
This is really problematic for tasks taking 3+ days, without checkpoint!

FAH used to set checkpoints every 10 minutes, but I find it excessive.
A checkpoint can be set every hour, and is a much better interval.
An hour lost of crunching is about $0.01-0.02 lost. It's negligible, or for long projects, I'm ok with every 3 to 5 hours.

Do you see weeks of long WU processing on a modern processor (eg: 4Ghz), or an older one (eg:2,xGhz)?
18) Message boards : Number crunching : 2,5 days long and counting... (Message 794)
Posted 19 Apr 2020 by ProDigit
Post:
Thanks, but the percentage bar is not working correctly on those.
it works like normal long units, and starts slowing down near to 80-90%, ever getting slower the closer to 100%. At 99% or over, it doesn't appear to move at all.
I'm running 2 units for 3+ days now!
More and more of these are hogging up my CPU queue, meaning, they're processing not allowing other WUs to be processed.
You must find a solution to when this is happening, that the result will be uploaded before these wus stretch themselves to infinity time.
My electricity is valuable, and I don't want to spend a week running a task that in the end gets aborted!
19) Message boards : Number crunching : 2,5 days long and counting... (Message 763)
Posted 13 Apr 2020 by ProDigit
Post:
I am going to let my "long" long tasks run, they are completing and validating.

Yes, but at 2,5 days, you could have crunched and validated 3 long WUs, and got triple the score.
20) Message boards : Number crunching : Short Tasks run for 14 hours and counting. (Message 752)
Posted 12 Apr 2020 by ProDigit
Post:
After facing a few of my own, (not virtualbox, just native Linux)
It appears that some WUs created by QCP (including long ones) happen to slow down near to the last 10%, to where they run 'days' to complete.

Without any official word from QCP,
If a short WU runs for more than 2 hour on a 3,5Ghz, or more than 4 hours on a 2GHZ CPU,
I personally would say to cancel the WU.
The runtimes for short WUs are 1 hour on 3,5-4Ghz, and 2 hours on a 2Ghz CPU.
I'd give them the benefit of no more than twice the runtime.

If any long WU runs for more than 1 day on a 3,5Ghz, or 1,5days on a 2Ghz CPU, I would do the same.
Standard WUs run 17 hours on a 3,5-4Ghz CPU and should run about 20 hours on a 2Ghz CPU.


Next 20

©2024 Benoit DA MOTA - LERIA, University of Angers, France