Hi, Welcome to the forum! I’ll send you a PM.
After many hours of running what I did to make it work is not without problems. It ran all the benchmarks for hours without problems. I left the rig running folding @ home and everything ran up until I just got back on this rig 10 mins. after I get on the rig I got a driver fail. I did a reboot and what do you know W10 decides to do a upgrade and bam I’m right back where I started. The thing is all driver updates is suppose to be turned off from MS. I set it to disable myself in 3 different ways in the OS. The K2200 & M40 are not supported in this driver, what I did was let the driver setup the P40 first, then manually install the same driver for the other cards, picking M60 and then M10 for those, yea yea I know now they are not a close match to what I’m actually using. I will do more research and find better matching cards to use and test with.
The one thing I did notice is it sets AdapterType for the P40 to =1 where it used =2 in other driver installs.
The AdapterType is not removed or # changed adding EnableMsHybrid=2 to the display out card and EnableMsHybrid=1 to the P40. no changes were made to the 3rd card M40 in Registry. I need to find a better match for the M40 I think using M60 is causing problems.
I used 3Dmarks Time spy as well and it put full load on the card.
Edit: But you have to be very quick to take a look or SS after it is over for GPU-Z to show the results.
Edit: or place max to all you are interested in seeing.
I’ve tried the 511.65 GRID driver and the card started into WDDM mode without any changes to the Registry. The P40 also showed up as the high performance option straight away
If you don’t add the EnableMsHybrid to, in this case:
computer\HKEY_LOCAL_MACHINE\SYSTEM\ControlSet001\Control\Class{4d36e968-e325-11ce-bfc1-08002be10318}\0000
then you get a virtual display, but I haven’t figured out if or how to change the resolution to something above 1280x1024. If you add it, the virtual display disappears, however the TimeSpy benchmark fails.
I haven’t tried doing any changes to the AdapterType which is normally set to 1.
I also did a quick Time Spy benchmark, with stock settings, where the GPU utilization reached 99% most of the time, according to GPU-Z.
Yes pretty much the same here with my setup with installing the driver the normal way.
I forgot to mention it set the P40 auto to WDDM mode and showed high performance option straight away for the card.
It was adding in the other two cards that became a little tricky and not following the OP way to get everything working.
I have now change the Quadro K2200 to tesla P4 and left the M40 using the M60 and after more than 18 hours “hoping I’m not speaking too soon” everything is still running stable with full load on both the P40 & M40 doing F@H.
Edit: BTW I have a Folding @ Home Team ABIT Folding Wolves 13285 the old but not forgotten ABIT Motherboard team. That anyone is welcome to join, I’m not trying to steal members as I don’t know if this forum has a F@H team. But if anyone would like to really test there video cards out your more than welcome to throw as many WU= work unit our way as you’d like.
Edit: Its for a good cause, trying to find cures for many different human diseases.
Generally the “solution” for the virtual display created by the Tesla cards is to just set your desktop to “only show on X” ; X being your main monitor output.
@Abit_Wolf jealous you got that P40. And welcome to the “Titan X because they’re cheap and you can” gang 
I bought an EVGA GEFORCE GTX 1660 SC ULTRA GAMING for 140 the Titan X Pascal cost me 105 $ more.
Most I have paid for a video card in years, let’s just say I got a little carried away with my bid. lol
The last time I spent that kind of money on video cards is when I got a pair of GTX 480 that I still have.
That I should have sold them a long time ago. I’m going to have to downsize soon it’s already getting hot here on the south coast. No AC in my shop.
Edit: I really wanted to pair the Titan X and Tesla P40 together in one rig and did, but with the time I spent on that I could never find a driver that would support both so I moved on.
I started moving cards around yesterday in an attempt to downsize and sell off some of the cards. Plains didn’t work out as planned, now I have 2x K40m 2x K20m, and an M40 unused sitting on the bench. Had problems with the MB’s in use excepting some of the cards or pairs.
I’ll try to work more with that later tonight…
Tried the GRID 511.65 and originally thought it wasn’t working as running 3d mark demo didn’t display the benchmark, it just ran in the background. Tried windows key and tab and in Nvidia control panel but couldn’t get it to show up. The benchmark completed though and games seem to display fine so will keep using this driver for my setup of P40 and 12500 iGPU as output. Thanks Abit_Wolf for the driver suggestion.
Edit: Turns out with the 511.65 driver I have a second display in System>Display I just changed the setting there from default extend these display to show only on 1 and 3d mark works now as well.
Your welcome. Yeah that’s one thing I don’t really like about 3DMark is how it doesn’t alllow other 3rd party software to use its OSD while the benchmark is running. It has its own OSD but not with any real info like temps. clock speeds. You can monitor it some what with GPU-Z, but 3dm runs the CPU test last, so by the time that test ends and it exits to show scores, most of the info in GPU-Z about the GPU/load/clocks/temps has left the screen. This is why I like running Heaven benchmark it will allow 3rd party like Afterbunner OSD, plus it has a much better OSD of its own that shows clock speed and you can tell what cards are in use by those and the temptures of the cards. Too bad it doesn’t put or keep a full load on these more up to date cards.
Could do temp monitoring over the network and watch it on your phone or capture it, tho I don’t know the software off the top of my head without research xD
Edit: It’s late so I asked the machine spirit for an answer:
Code just needs nvidia-smi and python
import subprocess
from flask import Flask, render_template
app = Flask(__name__)
@app.route('/')
def index():
temperature = get_gpu_temperature()
return render_template('index.html', temperature=temperature)
def get_gpu_temperature():
output = subprocess.check_output(['nvidia-smi', '--query-gpu=temperature.gpu', '--format=csv'])
temperature = int(output.decode().strip().split('\n')[1])
return temperature
if __name__ == '__main__':
app.run()
Html page
<!DOCTYPE html>
<html>
<head>
<title>GPU Temperature Monitor</title>
</head>
<body>
<h1>GPU Temperature: {{ temperature }}°C</h1>
</body>
</html>
1 Like
You can put a check in GPU-Z to log the info, but for me its kind of hard to read I prefer a graph layout.
Could get some sweet graphs with Prometheus and Grafana.
Thanks for the info, I may look into this at a later date. I have much going on ATM and benchmarks results are not the most important.
1 Like
If someone wants to take a look, this maybe be a good config to capture temps GitHub - nickbabcock/OhmGraphite: Expose hardware sensor data to Graphite / InfluxDB / Prometheus / Postgres / Timescaledb
1 Like
Hi guys, what’s up?
Just to let you know my last driver setup has been running stable since my last post about it.
P4, M60, and P40 have no issues.
Off-topic: I was wondering if anyone has tried Stream/ShatterLine game with a P4/P40 setup?
I played 6+ hours on the P40 with K2200 output. The missions play very well, but the PvP is not as good, but as you know ping plays a big part there.
1 Like
What drivers did you settle on?
1 Like
Keep in mind P4=K2200 M60=M40.
Edit: Strange the driver doesn’t support K2200 or M40 but yet NCP sees them correctly.
Not that it matters so long as things are working correctly.
2 Likes
Hi Gang, Tesla P100 is in the house.
Just got it up and running only a small test run so far. The temperatures look okay for now.
But under a heavy load, I may need a better fan/higher RPM. There were some problems
with the setup but could be only in my case. The Asus X99-A USB 3.1 would not boot after
installing the P100 and removing the card it boots normally. Could have been from my overclock settings.
Note: If this happens it could be Above 4G setting is disabled.
Seems after installing the card it reset my bios back to stock settings
whereas in stock settings it was disabled.
Used driver 528.49 on the first test but so far am unable to set P100 as High-performance.
Registry tweaks haven’t helped. Will test a new driver soon I’ll post back if and when I get it working.
Edit: Almost forgot I was unable to change the mode of the card in CMD to WDDM from TTC.
I think this is driver related as before with the P40. 511 your up. lol
1 Like
After a DDU and driver change to 511.65 the P100 was set to high performance automatically.
But the Quadro K620 was not setup.
Rather it was set to MS display adapter and auto set P100 to high. WDDM was also auto set.
After manully installing the same driver to the K620=P4 Reboot: Then the Registry was edited.
No adapter type were changed or deleted. Add EnableMsHybrid=2 to K620=P4 folder, =1 to P100 folder.
Running benchmarks now and will post the results later.
Edit: I wasn’t very happy with the 3D benchmark results they were lower than the P40.
So I moved on for now to testing F@H and that looks very good as for time per frames & PPD.
I think if possible we would need a modified Geforce driver to see better performance in 3D gaming/benchmarks for the P100/P40.
As I did even see the lower performance with the P40 vs M40
or Titan Xp with the 511.65 drivers used here for the P100/P40.
Other Data Center drivers & Geforce drivers gave better numbers in benchmarks & gaming.
Could just be me and different hardware used as per the results, but I have tried to keep everything close as possible with OS and Hardware as possible.
I will do more 3D testing once the F@H work is completed. Maybe some tweaking will help.
So at this point, I would not recommend picking up a P100 or P40 as an upgrade for gaming.
1 Like