It's fine to drop or increase the power, the only issue is you'll potentially see a fair bit less performance from the card and it will run cooler if you drop it. That would be unacceptable to me though, as you've baught a card with XYZ specification and there is the expectation that it will perform to the advertised standard, not 500Mhz less or on half power.
Also, this isn't Windows 10 specific. I'm on Win 7 64.
So I reinstalled 17.9.3 Wattman, and did the other suggestions above:
- -50%, 0 and +50% power limit
- disabling GPU scaling
- force to Full Panel
- was unable to change color depth,8 was the only option available
Basically doing the above, still gave me the black screen issue after roughly about 10-20 minutes of Forza Horizon 3. Typically it happens after the temps reach about 75-77 deg, and I'm not sure if this is considered overheating.
So far the most stable change is to underclock the card in Wattman. I tried dropping the GPU frequency % to 52%, and I was unable to get the black screen issue even after 45 minutes of FH3. But of course, FPS took a hit, with the average FPS on Ultra dropping from a stable 60 to about 40-45 fps, with the max temps staying at 64 deg. Even though its stable, as mentioned above, this isnt really acceptable, as the card is essentially running at half the advertised capacity.
On another note, I've been reading that the GTX 1060 is more or less on par with the 580 in terms of performance, and also with less power consumption. I'm contemplating bringing the 580 back to the store and potentially replacing it with the 1060, but since I've never used an Nvidia card before, I'm unsure if the lower power draw on the 1060 might help to eliminate this black screen issue.
Starting to think the card is drawing too much power and crashing the motherboard in some fashion.
Just shutting off like it has is either overheating (which i'm fairly sure it's not, wattman says its not) or some power issue, we can't all have failing PSU's and we both have 800+watts available, that leaves the card overwhelming the motherboard or a duff card crashing the PCIE interface.
I'm assuming not everyone who has bought an RX card is having the issue, so I presume that makes is not totally a driver problem.
It would also explain why the motherboard hard reset doesn't work after a crash, why all the fans start spinning at max chat (There is still PSU power available, only the controller software, which requires the motherboard has crashed), why the card is more stable when undervolting and when setting the power draw to -50% or reducing the gpu frequency. (although there are other possible reasons for these latter points)
I haven't tested for this, I wish I could but I'm running out of other ideas right now.
I took my 560 out and i've put my old 6790 back in (which has 2x 6 pin PCI-E for 150watts, then another 75w from the mobo, totaling 225watts total. The RX560 or 580 in your case, is more powerful, only has one 6 pin and the PCIE available, totalling 150w available), switched the drivers over and there are no problems.
My motherboard is 2011 spec, so it's getting on a bit, maybe it can't handle the power draw thought the PCI slot and falls over or trips some protection?
I looked up your motherboard and see it was also released in 2011.
EDIT: Looked up the 580+ Nitro and I see it has a 8pin and 6pin PCIE power connectors, that should allow it to draw up to 300watts altogether. I assume you have them both connected? If so, sinks my theory a little bit for your case.
Please, poke holes in my logic.
Yes both 6 and 8pins are connected. The weird thing also is that the dead 7970 had a higher recommended psu of 550w GV-R797OC-3GD | Graphics Card - GIGABYTE while the 580 has 500w. Yet I've never gotten the black screen issue with the 7970.
I'm also out of ideas but I'm going to try a longer session of FH3 tonight at 50% gpu frequency to try get the black screen issue. It's still the only solution so far that's yet to give the issue.
I think I might end up replacing the 580 with a 1060 at the shop. I'm hoping the lower power draw of the 1060 can help alleviate the black screen issue.
Also another thing, I notice that there is actually a BIOS switch on the 580 card.
Might anyone know what the switch is for? I didn't alter it since getting the card, but currently the switch seems to be positioned away from where the HDMI slot would be located.
580 has dual UEFI bios.Some UEFI cards won't wokr with legacy BIOS systems although they generally don't get past the boot stage.
This is useful if you want to flash one of them. I'm not sure if there are any performance differences.
You could power down switch it then power back up, load a game and see it it solves the problem.
UEFI is 100% backwards compatible, if there is a problem with an old machine, check to see if there is an updated BIOS
It seems that I may have stumbled upon a solution for the black screen issue.
I came across this My sapphire rx580 nitro+ 8gb is hitting 80c during gameplay, is that normal? : Amd , and some people there suggested dropping the target temperature in Wattman. I dropped mine from the default 75deg to 70deg, and was then able to play FH3 on Ultra settings for an hour without encountering the black screen issue. The FPS was at a stable 60, while the temp hit a max of 72deg. I believe this is in exchange for higher fan speed, where I saw the speed go as high as about 1700 rpm.
This was also at 0% for the power limit, and with the BIOS switch at the position I changed yesterday (closer to the HDMI slot). It seems the BIOS switch is to toggle between Silent and OC mode, and it seems the current switch position is for Silent mode, where the clock speed is 1340mhz. The card was able to maintain this speed during the 1 hour session.
I'll probably change the switch back to OC mode tomorrow, where the clock speed is 1430mhz, and try again at 70deg target temp.