Message boards : Number crunching : Level 3 cache requirements?
Author | Message |
---|---|
Drago75 Send message Joined: 13 Mar 20 Posts: 4 Credit: 521,989 RAC: 0 |
Does anybody now if Rosetta 4.20 wu's need a lot of Level 3 cache of the CPU? I tried to run 8 Rosetta together with 4 QuChem on my Linux Laptop 8 core CPU but it slows down QuChem considerably. Ram doesn't seem to be the problem as this config uses 15.2 of my 16 GB. |
Grant (SSSF) Send message Joined: 28 Mar 20 Posts: 1679 Credit: 17,801,285 RAC: 22,726 |
Cache has little if any impact. Overcommitting your CPU is what kills processing times. Check what else is using the CPU. Grant Darwin NT |
Drago75 Send message Joined: 13 Mar 20 Posts: 4 Credit: 521,989 RAC: 0 |
Hey Grant. In total I ran 12 wu's on 16 possible threats without GPU work or anything else. Thanks for the info... |
Grant (SSSF) Send message Joined: 28 Mar 20 Posts: 1679 Credit: 17,801,285 RAC: 22,726 |
If QuCHem work is hugely cache dependant, then running a lot of anything else at the same time will impact them considerably, even if the other running Tasks aren't affected by cache size or utilisation. Even though their performance isn't impacted, they will still be making use of that cache, which will impact those Tasks that are extremely cache sensitive. Grant Darwin NT |
Grant (SSSF) Send message Joined: 28 Mar 20 Posts: 1679 Credit: 17,801,285 RAC: 22,726 |
If QuCHem work is hugely cache dependant, then running a lot of anything else at the same time will impact them considerably, even if the other running Tasks aren't affected by cache size or utilisation. Even though their performance isn't impacted, they will still be making use of that cache, which will impact those Tasks that are extremely cache sensitive. Grant Darwin NT |
Drago75 Send message Joined: 13 Mar 20 Posts: 4 Credit: 521,989 RAC: 0 |
Allright, then I will not pair QuChem with anything else anymore... |
Mad_Max Send message Joined: 31 Dec 09 Posts: 209 Credit: 25,936,992 RAC: 12,835 |
Use of additional ("virtual") thread always decrease performance of other running threads. Its normal - because this share same compute unit in same physical cores. If this slowdown of a single thread performance not very big (like in 10-30% range) then its normal and there is nothing to do about it. Only if slowdown of a single thread is very big like twice slower (so total throughput of all threads combined decrease) then it indicates that something went wrong. Like not enough cache size or other issues. But a low to moderate slowdown when using virtual threads is both normal and inevitable. |
Message boards :
Number crunching :
Level 3 cache requirements?
©2024 University of Washington
https://www.bakerlab.org