Have you rebooted since using the coolbits command?
Have you checked in NVIDIA X Server Settings panel that those changes have in affect occurred?
Is your memory clock shown for Level 3 19500Mhz?
Rebooted.
Server Panel is exposing the ability to offset -/+
With a memory offset of 3200+ the Panel is claiming: 21702
I got below 2 minutes but I had gpu speed 100+ when it crashed.
Trying 3200+ with no gpu speed off set.
Tom M
3200 seems wayyyy too high, what is the rationale for this number? or was it an arbitrary choice? you only get a 500MHz penaly for P2 compute mode. for comparison I run mine at +1000 getting the memory rate to 19500.
with a memory OC this high, I can only imagine that the memory is generating all kinds of errors and the EDR (GDDR6X has EDR built in, similar to ECC but not exactly the same) is working overtime to resend failed data transfers. at a point, more OC will just stop giving you more performance and it will plateau as more and more errors are detected and resent. have you shown that +3200 is any faster than +1500 for example?
3200 seems wayyyy too high, what is the rationale for this number? or was it an arbitrary choice? you only get a 500MHz penaly for P2 compute mode. for comparison I run mine at +1000 getting the memory rate to 19500.
I believe it was a mental error. I have just dropped it to +1000 with a +100 on the gpu speed. I am hoping the task processing time stays below two minutes.
Your rtx 3080 ti is running 110 seconds, mine is running 118 - 124 seconds.
Besides running the memory faster what else are you doing that speeds it up that high?
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!
3200 seems wayyyy too high, what is the rationale for this number? or was it an arbitrary choice? you only get a 500MHz penaly for P2 compute mode. for comparison I run mine at +1000 getting the memory rate to 19500.
I believe it was a mental error. I have just dropped it to +1000 with a +100 on the gpu speed. I am hoping the task processing time stays below two minutes.
Your rtx 3080 ti is running 110 seconds, mine is running 118 - 124 seconds.
Besides running the memory faster what else are you doing that speeds it up that high?
Tom M
I'm running it watercooled (cooler temps get you higher core clocks, observed clocks ~1750-1815 MHz), clocked +100core/+1000mem capped at 300W power limit. as well as with Petri's latest libsleep v3.1
I'm running it watercooled (cooler temps get you higher core clocks, observed clocks ~1750-1815 MHz), clocked +100core/+1000mem capped at 300W power limit. as well as with Petri's latest libsleep v3.1
Observed clocks in your range. Power not limited. +100core/+1000mem
I just removed my gtx 1080 and the rtx 3080 ti is processing faster. Hmmm.....
I guess I had better go find my copy of libsleep v.3.1
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!
Yes, Boinc screensaver is one of Windows screen savers. So if you select there 'None' Boinc screen saver is also disabled. From Power options you can then select if you want your screen to go blank after some time.
I'm running it watercooled (cooler temps get you higher core clocks, observed clocks ~1750-1815 MHz), clocked +100core/+1000mem capped at 300W power limit. as well as with Petri's latest libsleep v3.1
Observed clocks in your range. Power not limited. +100core/+1000mem
I just removed my gtx 1080 and the rtx 3080 ti is processing faster. Hmmm.....
I guess I had better go find my copy of libsleep v.3.1
Tom M
looks like you're now down to about 109/110 second run time so i guess you found the right package :)
how much power is your card using while running Einstein FGRPB1 GPU? you can see power use by running "watch -n 1 nvidia-smi" (this will loop the output every second so you can get a near-real-time reading)
how much power is your card using while running Einstein FGRPB1 GPU? you can see power use by running "watch -n 1 nvidia-smi" (this will loop the output every second so you can get a near-real-time reading)
. . My ears just perked up. Where would one run that command? And where would one see the result?
. . OK.. found it! running it in terminal window and results are in same. :) Thanks :)
I have dropped to 0,0,0 to see if GPU utilization will go back up And because of power bill prediction.
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:Keith Myers
)
3200 seems wayyyy too high, what is the rationale for this number? or was it an arbitrary choice? you only get a 500MHz penaly for P2 compute mode. for comparison I run mine at +1000 getting the memory rate to 19500.
with a memory OC this high, I can only imagine that the memory is generating all kinds of errors and the EDR (GDDR6X has EDR built in, similar to ECC but not exactly the same) is working overtime to resend failed data transfers. at a point, more OC will just stop giving you more performance and it will plateau as more and more errors are detected and resent. have you shown that +3200 is any faster than +1500 for example?
_________________________________________________________________________
Ian&Steve C. wrote:3200
)
I believe it was a mental error. I have just dropped it to +1000 with a +100 on the gpu speed. I am hoping the task processing time stays below two minutes.
Your rtx 3080 ti is running 110 seconds, mine is running 118 - 124 seconds.
Besides running the memory faster what else are you doing that speeds it up that high?
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:Ian&Steve C.
)
I'm running it watercooled (cooler temps get you higher core clocks, observed clocks ~1750-1815 MHz), clocked +100core/+1000mem capped at 300W power limit. as well as with Petri's latest libsleep v3.1
_________________________________________________________________________
Ian&Steve C. wrote: I'm
)
Observed clocks in your range. Power not limited. +100core/+1000mem
I just removed my gtx 1080 and the rtx 3080 ti is processing faster. Hmmm.....
I guess I had better go find my copy of libsleep v.3.1
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!
Harri Liljeroos wrote: If
)
Hi Harri,
Does that include the BOINC screensaver?
Stephen
Yes, Boinc screensaver is one
)
Yes, Boinc screensaver is one of Windows screen savers. So if you select there 'None' Boinc screen saver is also disabled. From Power options you can then select if you want your screen to go blank after some time.
Tom M wrote: Ian&Steve C.
)
looks like you're now down to about 109/110 second run time so i guess you found the right package :)
how much power is your card using while running Einstein FGRPB1 GPU? you can see power use by running "watch -n 1 nvidia-smi" (this will loop the output every second so you can get a near-real-time reading)
_________________________________________________________________________
Thanks again. Your rigs are
)
Thanks again. Your rigs are doing well, keep on crunching! :)
Stephen
Ian&Steve C. wrote:how much
)
. . My ears just perked up.
Where would one run that command? And where would one see the result?. . OK.. found it! running it in terminal window and results are in same. :) Thanks :)
Stephen
???
165 / 3000 with power limit
)
165 / 3000 with power limit at 350 watts.
Still need to chase up the other solution.
I have dropped to 0,0,0 to see if GPU utilization will go back up And because of power bill prediction.
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!