Under the hood of a M2090
- Thread starter Psi*
- Start date
Psi*
Tech Monkey
Assuming that is degree C, that is pretty hot but I have run my i7-920 like that for days. It was in the system for a couple of years. There is an old thread here where Rob suggested i go with that initially then upgrade later. Later was finally this morning.
I now have a i7-990X as the host for C2070 with similar speeds as the system hosting the M2090. I use Indigo Extreme as the TIM which required running Prime95 on all cores with the water pump turned off. The CPU maxxed for several seconds ... impressive.
That is 100 deg C. I used it on the i7-920 just pulled & on the other i7-990X. Pretty sure water was starting to boil in the CPU water block. Because when I turned the water pump back on a burst of water with fine bubbles came out. We need a funny face with sweat. Like I sad tho, I have done this twice before.The M2090 + system is faster tho. Still looking into that. These systems are quite similar. Same m/b, same CPUs, same RAM, etc. ... so manufacturing differences could be root cause of the difference performance, but still checking things. Both boxes are running now & that is all that counts at the moment.
Last, Prime95 is now available as 64 bit & can automagically spawn threads on all available cores. Humbling seeing 12 threads fully pegged in Task Manager.
Last edited:
Regarding monitoring the temperatures on these cards, NVIDIA told me that the "nvsmi" tool has existed for a little bit, and I don't remember that being mentioned here. I was told that the tool was only for Linux but it does seem to be available for Windows as well:
http://www.nvidia.ca/object/software-for-tesla-products.html
http://www.nvidia.ca/object/software-for-tesla-products.html
Psi*
Tech Monkey
nvidia-smi.exe, the actual application, was installed with the Windows Nvidia driver. I have used it on both the C2070 and the M2090 systems to compare what can be reported. As a result, I am pretty confident that the M2090 does not report temperature. Percent of memory usage is reported which is especially important to know when the project exceeds the memory of the GPGPU.
That's a bit strange, on account that it explicitly states "GPU Monitoring" on that page. It came about in the discussion of the K-series... I asked if the new cards could have their temperatures monitored, since the older ones could not. Then I was told about smi, and that yes, in fact, you could monitor the temperatures on previous cards.
Now, as I mentioned above, I was told that smi was for Linux-only, but possibly I misunderstood. I'm wondering if temperature monitoring, for some reason, is Linux-only, and that's what he meant.
But that makes little sense.
Now, as I mentioned above, I was told that smi was for Linux-only, but possibly I misunderstood. I'm wondering if temperature monitoring, for some reason, is Linux-only, and that's what he meant.
But that makes little sense.