OTOY Forums

i got some troubles here and i am not sure if it's my hardware or somehow software related so maybe you can give me a hint.
i got my hands on a cheap secondhand tesla m2070 which is passively cooled but i modded it with something like a heat duct and a fan so it should be ok in my opinion.. the problem is since it is a not a workstation product i can't get out the temperature readings from the sensor with nvidia-smi.exe commands or other harware monitor programs:
from http://microway.com/hpc-tech-tips/2011/ ... your-gpus/

You’ll notice that unfortunately the M-series passively-cooled Tesla GPUs do not report temperatures to nvidia-smi. This is because the temperature data is reported directly to IPMI, allowing the system’s BMC to properly control chassis cooling.

the tesla card is used as secondary card next to a quadro 4800 which i did not use for rendering yet.
the other specs should be just fine:
mobo: ASUS P6T7 WS SuperComputer
24gb RAM
1500W power supply (this machine is meant to hold another 2 teslas in the future...)
win7x64
ok now my troubles:
octane preview window works just fine but when i hit render the at 2500x2500 the rendering will stop after 5 or 10 minutes. i checked gpu usage via nvidia.smi a couple of times and noticed that the preview window always stays at 99% while after hitting the render button the cpu usage mostly is 99 but sometimes falls down to 85 but a second later it is 99% again. after the stop i immediately launched the octane preview window and that one worked (the gpu had nearly no time to cool down while doing that).
i tried both path tracing and direct light with diffuse bounces.
so my question: does the preview window something other than the render button? does a higher resolution draw more power out of my gpu than the preview window?
any help is appreciated (also regarding reading out that gpu temperature)
tnx.
boris

p.s. if i am sure it is a temperature problem i am thinking about a water cooling solution especially when there will be 4 gpu's inside that WS...

boris wrote:i got some troubles here and i am not sure if it's my hardware or somehow software related so maybe you can give me a hint.
i got my hands on a cheap secondhand tesla m2070 which is passively cooled but i modded it with something like a heat duct and a fan so it should be ok in my opinion.. the problem is since it is a not a workstation product i can't get out the temperature readings from the sensor with nvidia-smi.exe commands or other harware monitor programs:
from http://microway.com/hpc-tech-tips/2011/ ... your-gpus/

You’ll notice that unfortunately the M-series passively-cooled Tesla GPUs do not report temperatures to nvidia-smi. This is because the temperature data is reported directly to IPMI, allowing the system’s BMC to properly control chassis cooling.

the tesla card is used as secondary card next to a quadro 4800 which i did not use for rendering yet.
the other specs should be just fine:
mobo: ASUS P6T7 WS SuperComputer
24gb RAM
1500W power supply (this machine is meant to hold another 2 teslas in the future...)
win7x64
ok now my troubles:
octane preview window works just fine but when i hit render the at 2500x2500 the rendering will stop after 5 or 10 minutes. i checked gpu usage via nvidia.smi a couple of times and noticed that the preview window always stays at 99% while after hitting the render button the cpu usage mostly is 99 but sometimes falls down to 85 but a second later it is 99% again. after the stop i immediately launched the octane preview window and that one worked (the gpu had nearly no time to cool down while doing that).
i tried both path tracing and direct light with diffuse bounces.
so my question: does the preview window something other than the render button? does a higher resolution draw more power out of my gpu than the preview window?
any help is appreciated (also regarding reading out that gpu temperature)
tnx.
boris

p.s. if i am sure it is a temperature problem i am thinking about a water cooling solution especially when there will be 4 gpu's inside that WS...

it's normal that launching renders at high resolutions the power will float from 85 to 99%. This happens for all gpu based rendering engines.
Octane for Max has a bug launching the render at high resolution (eg: 3500x2500), when stopped Max will crash (using all gpus)
You checked the gpu temperature? Install Geforce drivers and use MSI Afterburner.

tnx for the answer.

You checked the gpu temperature? Install Geforce drivers and use MSI Afterburner.

that's the problem. since my tesla is a m-class product, i just can't get that core temperature. for that i think i would need a server board supporting IPMI and a win server edition... i just tried msi afterburner. it just reports stuff about the quadro.
you mean install Geforce Driver instead of the quadro/tesla drivers? hmm. not sure if that will work. and all the good things about my workstation cards will be gone then i think. also i am not sure if a geforce driver will support 6gb ram since there is no geforce card with 6gb...
anyway i will try some improved cooling and stress testing and report later..