I've replaced the pair of 980Ti's with a single 1080Ti and single-tasked it's doing about 1900 CPH with the same command-line below. It's only been cooking for a few days now so I'll have to give it time to see if it stays stable.

The new Ti is an amazing performer, isn't it.
How are your temp/fan% ?

1060 https://setiathome.berkeley.edu/workunit.php?wuid=2505759191
1080 https://setiathome.berkeley.edu/workunit.php?wuid=2505759161
and another pair
1080 http://setiathome.berkeley.edu/workunit.php?wuid=2505776592

---

To overcome Heisenbergs:
"You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones

After a night of cooking it was hovering around 65C and the fans were about 40%-50% I believe -- single-tasked though so you can see it cooling off between tasks so I bet it'll run a bit louder when I switch to double-tasks. NVSMI was also suggesting it was nowhere near TDP so I suspect there's still room to push it harder.

@Shaggie76:

I also have a 5960X, and was having BSOD problems once in 2-3 days, or twice a day. Turns out it was Nvidia driver 378.88. After updating to 378.92, I haven't had a single BSOD (knocking on wood) for 10 days now. Worth checking out?

After a night of cooking it was hovering around 65C and the fans were about 40%-50% I believe -- single-tasked though so you can see it cooling off between tasks so I bet it'll run a bit louder when I switch to double-tasks. NVSMI was also suggesting it was nowhere near TDP so I suspect there's still room to push it harder.

I had to rise the power CAP from 250W to 285W so that power draw is not a limiting factor. It hits sometimes 272-280W. To keep temp below 67C I had to push the fan to 98%. The GPU load hovers at 93-99%. P2 state has 2088 MHz set and 2040MHz running as GPU clock and mem clock is 10690MHz. I can not get the GPU to do CUDA at P0.

---

To overcome Heisenbergs:
"You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones

I found data from my first night of crunching:



So the temps were better than I recall but I'm guessing double-tasking will push it harder -- I'll give it another week or so before I do that.

I was doing a bunch of reading about Tesla Compute Cluster (TCC) mode -- you can enable this if your primary GFX is something else like IGP. In your setup you might be able to enable it since you have multi-GPU; it might get you to P0. If you try let me know -- I'm designing my next build and considering getting a chip with IGP if it means I can use TCC mode for the discrete GPUs.

Running a GUPPI vlar right now:

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 381.65                 Driver Version: 381.65                    |
|-------------------------------+----------------------+----------------------+
| GPU  Name            TCC/WDDM | Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce GTX 108... WDDM  | 0000:01:00.0      On |                  N/A |
| 40%   69C    P2   151W / 250W |   1534MiB / 11264MiB |     92%      Default |
+-------------------------------+----------------------+----------------------+

Definitely some power-headroom there; hopefully double-tasking eats that up.

40% 69C

Almost 70°c and the fan running at only 40%? Those fan profiles really do go for quiet over cool.
On my Gigabyte GTX 1070s the fan speed is 48% with the temperatures at 64°c (there are 3 fans per GPU- so they shift a lot of air). Power consumption is about 60% of maximum. Ambient temperature is 30°c.
65°c seems to be the target temperature for these cards. When the air conditioner is running the fan speed drops right back, and the GPU temps stay around the 62-64°c region.

---

Grant
Darwin NT

40% 69C

Almost 70°c and the fan running at only 40%? Those fan profiles really do go for quiet over cool.

I don't know if the temp displayed by Open HW Monitor is the same as NVSMI -- as you can see in the graph lower down it flirted with 60C during a night of crunching. So are you measuring with NVSMI or something else?

I don't know if the temp displayed by Open HW Monitor is the same as NVSMI -- as you can see in the graph lower down it flirted with 60C during a night of crunching.

True, hadn't noticed the difference between the graph & text readings.

So are you measuring with NVSMI or something else?

I'm on Windows and use GPUz

---

Grant
Darwin NT

...
I was doing a bunch of reading about Tesla Compute Cluster (TCC) mode -- you can enable this if your primary GFX is something else like IGP. In your setup you might be able to enable it since you have multi-GPU; it might get you to P0. If you try let me know -- I'm designing my next build and considering getting a chip with IGP if it means I can use TCC mode for the discrete GPUs.

I'll google the TCC mode. Thanks.

EDIT: Did some googling

root@Linux1:~/KWSN-Bench-Linux-MBv7_v2.01.08# nvidia-smi -g 2 -dm 1
Changing driver models is not supported for GPU 0000:09:00.0 on this platform.
Treating as warning and moving on.
All done.
root@Linux1:~/KWSN-Bench-Linux-MBv7_v2.01.08# nvidia-smi -g 3 -dm 1
Changing driver models is not supported for GPU 0000:0A:00.0 on this platform.
Treating as warning and moving on.
All done.
root@Linux1:~/KWSN-Bench-Linux-MBv7_v2.01.08# nvidia-smi -g 1 -dm 1
Changing driver models is not supported for GPU 0000:06:00.0 on this platform.
Treating as warning and moving on.
All done.

---

To overcome Heisenbergs:
"You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones