HWinfo "struggling" after adding 4th GPU

[candre23]

I use HWinfo in conjunction with rainmeter to monitor my system. It's been working great for years, but a couple weeks ago I added a 4th (or 5th, depending on how you count them) GPU to my ML rig and HWinfo has been really struggling ever since. Now, upon starting HWinfo, it takes a solid 4-5 minutes before the sensors window appears. If I exit the application and restart it, again there will be a several-minute delay before the sensors are visible. Once running, HWinfo is extremely sluggish to display GPU metrics. Though I have the update period set to 2 seconds, the VRAM, GPU capacity, and temp values only update every 90 seconds or so. Previously, with three GPUs, values updated every tick without fail.

My hardware and software setup is not exactly normal. I'm running win10 on a gigabyte MD70-HB0 dual xeon board, which doesn't officially support win10. I'm using four nvidia P40 GPUs (plus the built-in AST2400 BMC video adapter, if that counts). With three P40s, all was well. The problems only started when I added the 4th card.

If these slowdown issues are an inevitable result of running a goofy hardware/software combo, then I can live with it. But if there's something I can do to improve matters on my end, I'd love to hear it.

Thanks.

 

[Martin]

Such huge delays should not be there. Please attach the HWiNFO Debug File so I can analyze in detail what's going on there.

 

[candre23]

Debug attached. I enabled debugging, shut down HWinfo, started HWinfo, waited for it to start reporting values, let it record for a few min, and then shut it down again before zipping up the .dbg file. Hopefully this gives you enough data.

Here's a screenshot of my rainmeter panels showing the delayed value updates for the GPUs. When in use, the GPU utilization changes rapidly and previously the trendline would be jagged like the other trendlines. But since adding the 4th P40 the values update very intermittently, and the result is those wide square-wave-style trendlines.

 

[Martin]

Looks like the problem is not HWiNFO but some general problem in your system.
I can see that you are getting lots of WHEA errors which should be visible also in HWiNFO. These are caused by a hardware/driver failure and cause excessive delays in the entire system.
You should be also able to see those errors in the Windows Event Viewer.
Those WHEA errors are generated by your NVIDIA GPUs.

 

[candre23]

Hmmm. All 4 GPUs are using riser cables. I'm guessing that 4th cable was just too much for the PCIe bus to handle. I could probably get away with dropping down to PCIe 2.0 speeds, but I can't find a way to do that on my board, so I might be SoL.