I just finished the analysis of concurrent running of these tasks. I cannot say there is anything landmark within the results but maybe some interesting insights. Also, when (inevitably) users ask what the ideal amount of concurrent tasks are for these work units, this link can just be referred to. This is intended to be an informal look at the concurrent running of these tasks and not an actual study that would require tighter controls.
Of course, argument over the validity of the statistical analyses chosen for this is welcomed but I cannot say I am in any way an expert in statistics.
I am just going to post the link to the report since there are figures, graphs, etc.
This was a lot of fun to do and I hope it helps someone out there.
I hope the kids had alot of fun doing the testing as well, especially analyzing the results and coming up with the conclusions. It should help them later on when they have to work for a lving and when they do tasks around the home and their general lives as well!! Thank you for teaching the kids critical thinking and testing to come up with the best answers.
So for my purposes 3x,4x or 5x are possibly my best choices.
I believe I am memory constrained at 5x.
Only testing on the rtx 3080 ti and the Titan V will tell me which.
Apparently 4x is my sweet spot on my EVGA Rtx 3080 ti. That and Overclocks of +21/+1500.
I FINALLY made it onto the bottom of the Top 50 list. I still need to determine if 5x produces more production, less production or the same (within a reasonable variation of processing time).
Happy Friday!
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
So for my purposes 3x,4x or 5x are possibly my best choices.
I believe I am memory constrained at 5x.
Only testing on the rtx 3080 ti and the Titan V will tell me which.
Apparently 4x is my sweet spot on my EVGA Rtx 3080 ti. That and Overclocks of +21/+1500.
I FINALLY made it onto the bottom of the Top 50 list. I still need to determine if 5x produces more production, less production or the same (within a reasonable variation of processing time).
Happy Friday!
Tom M
Based on what I saw, don't try to go above about 75-80% VRAM usage (in the context of total available VRAM).
Apparently 4x is my sweet spot on my EVGA Rtx 3080 ti. That and Overclocks of +21/+1500.
I FINALLY made it onto the bottom of the Top 50 list. I still need to determine if 5x produces more production, less production or the same (within a reasonable variation of processing time).
I now suspect one major speed up would be to try to Multi-thread the CPU processing at 49.5% and 99.5%.
Limited time sub-project. Apparently will run out of data in a couple of months? Not worth the time.
Does the data structure make it possible to process the data in parallel? If the processing/scan is purely linear then I would guess not. Unless you had each CPU thread run a subset of the data. Post its results into a smaller data structure and then do a final CPU process of that on a single CPU?
I am, of course, an amateur at this.
Respectfully,
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
In my experience, multithreading has no effect on the total runtime of the work unit.
I have enabled two cpu threads per gravitational wave WU on my computers where the CPU is not running any other projects in parallel. I can't see any effect on the total runtime. Also, according to the system monitoring in Linux, only one thread seems to be used although two are enabled.
In my experience, multithreading has no effect on the total runtime of the work unit.
I have enabled two cpu threads per gravitational wave WU on my computers where the CPU is not running any other projects in parallel. I can't see any effect on the total runtime. Also, according to the system monitoring in Linux, only one thread seems to be used although two are enabled.
Granted. If the application does not have CPU multithreading programmed for and enabled it will make no difference how many cpu threads you make available.
For an example of cpu multi-threading in action you can go to PrimeGrid. I have one task from there using 16 cpu threads.
Respectfully,
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
In my experience, multithreading has no effect on the total runtime of the work unit.
Granted. If the application does not have CPU multithreading programmed for and enabled it will make no difference how many cpu threads you make available.
For an example of cpu multi-threading in action you can go to PrimeGrid. I have one task from there using 16 cpu threads.
Well, I gave it a whirl. It black screens my box fairly quickly. I did babysit it through one WU. Same box was rock solid stable on FGRPB1G. Oddly, F@H stresses the hardware A LOT more and its generally stable on most projects. I'll chalk it up to AMD and their awful drivers. I wish I hadn't bought this card, but it was during Newegg lottery days.
Great work! Thanks for
)
Great work! Thanks for sharing.
Boca Raton Community HS
)
I hope the kids had alot of fun doing the testing as well, especially analyzing the results and coming up with the conclusions. It should help them later on when they have to work for a lving and when they do tasks around the home and their general lives as well!! Thank you for teaching the kids critical thinking and testing to come up with the best answers.
Tom M wrote: Thank
)
Apparently 4x is my sweet spot on my EVGA Rtx 3080 ti. That and Overclocks of +21/+1500.
I FINALLY made it onto the bottom of the Top 50 list. I still need to determine if 5x produces more production, less production or the same (within a reasonable variation of processing time).
Happy Friday!
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
Tom M wrote:Tom M
)
Based on what I saw, don't try to go above about 75-80% VRAM usage (in the context of total available VRAM).
Tom M wrote:Apparently 4x
)
I now suspect one major speed
)
I now suspect one major speed up would be to try to Multi-thread the CPU processing at 49.5% and 99.5%.
I am, of course, an amateur at this.
Respectfully,
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
In my experience,
)
In my experience, multithreading has no effect on the total runtime of the work unit.
I have enabled two cpu threads per gravitational wave WU on my computers where the CPU is not running any other projects in parallel. I can't see any effect on the total runtime. Also, according to the system monitoring in Linux, only one thread seems to be used although two are enabled.
Ereignishorizont wrote: In
)
Granted. If the application does not have CPU multithreading programmed for and enabled it will make no difference how many cpu threads you make available.
For an example of cpu multi-threading in action you can go to PrimeGrid. I have one task from there using 16 cpu threads.
Respectfully,
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
Tom M
)
The same goes for Milkyway also.
Proud member of the Old Farts Association
Well, I gave it a whirl. It
)
Well, I gave it a whirl. It black screens my box fairly quickly. I did babysit it through one WU. Same box was rock solid stable on FGRPB1G. Oddly, F@H stresses the hardware A LOT more and its generally stable on most projects. I'll chalk it up to AMD and their awful drivers. I wish I hadn't bought this card, but it was during Newegg lottery days.