Sapphire R9 390 Nitro overheating

[8bit]

I have a Sapphire R9 390 Nitro (with backing plate) in my gaming machine, which until recently never had a problem. I got No Man's Sky (another story) recently and have been finding that after anything between 15 and 60 minutes of play, the screen would go off, I'd get the "no signal detected, check display cable" message from the monitor and some fans in the machine would spin right up to maximum speed.

Initially I thought it must just be an issue with the game because it wasn't happening with anything else I play, mostly Elite Dangerous and Mirror's Edge Catalyst just now but the other night it happened while playing Elite too.

The card is not overclocked (beyond the factory OC for that edition) and it's in a Corsair Carbide 540 Air case with two 140mm Noctua fans right in front of it. The fact that it's been fine until the past few weeks makes me wonder if whatever thermal transfer material they use from the factory has gone bad maybe?

Any suggestions etc. welcome...

[neonplanet40]

Any warranty left?

[kalniel]

What are your CPU temps? Case air intakes clear of dust?

[gupsterg]

GPU temp @ idle/loaded would be handy to know OP plus say room ambient temp.

[spolsh]

Have you tried a frh install of drivers for the card ? Maybe it got an update with a problem ?

[8bit]

Warranty - should be some yeah, I bought the card in November last year. Hoping it doesn't come to that though.

Temps - so I just played No Man's Sky for a little while with MSI Afterburner logging temps. The last recorded GPU temp was 77C and graphics card fan speed was 63%. CPU temp at that time was 52C (i7-6700K @ stock 4GHz under a Corsair H90 AIO water cooler). Idle temps right now are 40C for the CPU and 45C for the GPU. I cleaned the intake air filter on the case just recently so no issue there.

I have the latest AMD drivers (16.8.3) installed but I've had this problem with the last beta driver (16.8.2 I think, the first one that stated support for No Man's Sky) and the one previous to that, 16.7.something. Each has been installed by first cleanly removing the previous version with Display Driver Uninstaller. I always get an error message when installing new AMD drivers about a package install failure but I think that's just the control panel failing first time, the drivers do seem to install.

I did look on the Sapphire website to see if there was a new BIOS or custom driver for this particular card but didn't see anything very helpful.

[spolsh]

You've reseated the card and disconnected and reconnecte d the power wires ? (just in case someone accientally knocked it)

77 degress is 5 higher than Hexus said review card reached, but I doublt that'd be enough to cause the crashing - (just from comments on the net).

[gupsterg]

@OP

You're GPU temp I would regard as a non issue IMO, I've had 4 Hawaii cards in the past (2x Tri-X 290, 1x Vapor-X 290X, 1x DCUII 290X). My uses were gaming/folding@home (at times 70hrs continuous runs).

Install HWiNFO, besides GPU temp it will give you VRM temp/Amps/Watts. Here's an example screenie of data I monitored for a 70hr+ run of my Vapor-X 290X OC'd to 1100/1525 via custom ROM with stock cooler.

[CAT-THE-FIFTH]

The temperatures are fine - they are well under 90C which I believe is the point AMD Hawaii based cards start downclocking.

If you have overclocked your CPU,I would run it at stock clockspeeds.

[8bit]

OK, so if it's not actually overheating then I guess the question is why does it suddenly shut itself down and spin the fans up to max speed? I can still hear the game sounds through the speakers when this happens so pretty sure it's just the GPU being shut down.

@CAT, CPU is at stock speed.

[gupsterg]

Comes across as "black screen" issue, there are varying reasons / fixes for it. Usually tends to be lack of GPU voltage or RAM clock to high, try adding a bit of GPU voltage via MSI AB/TriXX and/or lowering RAM clock a bit. Try these 2 things even if you are not OC'ing card.

[8bit]

OK thanks, I'll give that a shot later.

[gupsterg]

When I first got my Sapphire Tri-X 290 I started viewing the Sapphire forum, my reason was to see if I could find an official OC edition bios to flash to my STD edition card. I found 2 sets ROMs members were sharing via PM, first was say "out of box" ROM which had no additional voltage offset being applied to all DPM states and second had +25mV (supplied by Sapphire support to an owner). Basically the one with the +25mV was what members shared in the "black screen issue" thread.

Once you've done testing if GPU core voltage addition fixes it you contact Sapphire support via this link to get permanent. Explain the issue you had and fix, they will ask for your original ROM and send you one with extra voltage.

[8bit]

OK so I just tried this with No Man's Sky while logging with GPU-Z. The display black--screen crashed after less than two minutes' play and the figures at the time of the crash are at the bottom of the following:

        Date        , GPU Core Clock [MHz] , GPU Memory Clock [MHz] , GPU Temperature [°C] , Fan Speed (%) [%] , Fan Speed (RPM) [RPM] , GPU Load [%] , Memory Controller Load [%] , Memory Usage (Dedicated) [MB] , Memory Usage (Dynamic) [MB] , 12V [V] , VDDC [V] , VDDCI [V] , VDDC Current In [A] , VDDC Current Out [A] , VDDC Power In [W] , VDDC Power Out [W] , VRM Temperature 1 [°C] , VRM Temperature 2 [°C] ,
2016-09-12 22:11:28 ,             1040.0   ,               1500.0   ,               62.0   ,              37   ,                1581   ,        100   ,                       19   ,                        1417   ,                        52   , 11.63   ,  1.180   ,   1.047   ,              19.8   ,              162.5   ,           233.8   ,            190.0   ,                   50   ,                   60   ,
2016-09-12 22:11:29 ,             1039.2   ,               1500.0   ,               62.0   ,              37   ,                1586   ,        100   ,                       20   ,                        1417   ,                        52   , 11.63   ,  1.180   ,   1.047   ,              20.3   ,              161.0   ,           241.5   ,            197.3   ,                   51   ,                   60   ,
2016-09-12 22:11:30 ,             1040.0   ,               1500.0   ,               62.0   ,              39   ,                1616   ,        100   ,                        4   ,                        1417   ,                        52   , 11.63   ,  1.195   ,   1.047   ,              16.2   ,              140.5   ,           233.8   ,            186.8   ,                   51   ,                   60   ,
2016-09-12 22:11:31 ,             1040.0   ,               1500.0   ,               62.0   ,              39   ,                1635   ,        100   ,                       15   ,                        1418   ,                        52   , 11.63   ,  1.164   ,   1.047   ,              19.6   ,              166.0   ,           191.5   ,            155.3   ,                   51   ,                   60   ,
2016-09-12 22:11:32 ,             1018.8   ,               1500.0   ,               63.0   ,              39   ,                1645   ,        100   ,                       45   ,                        1418   ,                        52   , 11.63   ,  1.172   ,   1.047   ,              20.3   ,              169.0   ,           224.3   ,            166.5   ,                   51   ,                   60   ,
2016-09-12 22:11:33 ,             1040.0   ,               1500.0   ,               63.0   ,              39   ,                1648   ,         99   ,                        3   ,                        1418   ,                        52   , 11.63   ,  1.180   ,   1.047   ,              17.2   ,              147.0   ,           229.3   ,            189.3   ,                   52   ,                   61   ,
2016-09-12 22:11:34 ,             1040.0   ,               1500.0   ,               63.0   ,              39   ,                1645   ,        100   ,                       20   ,                        1418   ,                        52   , 11.63   ,  1.180   ,   1.047   ,              18.6   ,              159.0   ,           242.8   ,            190.0   ,                   52   ,                   61   ,
2016-09-12 22:11:35 ,             1040.0   ,               1500.0   ,               63.0   ,              39   ,                1648   ,         99   ,                       12   ,                        1420   ,                        52   , 11.63   ,  1.180   ,   1.047   ,              18.5   ,              159.0   ,           240.0   ,            190.3   ,                   52   ,                   61   ,
2016-09-12 22:11:36 ,             1040.0   ,               1500.0   ,               63.0   ,              39   ,                1661   ,        100   ,                       20   ,                        1420   ,                        52   , 11.63   ,  1.188   ,   1.047   ,              19.8   ,              166.0   ,           252.8   ,            223.0   ,                   52   ,                   61   ,

The GPU and VRM temps look OK to me? Not sure what the VDDC values should look like, I can't post much more data than that because of the 15,000 character limit in Hexus forums...

As regards GPU voltage, Sapphire Trixx seems to default to +19mV. Obviously because the card is <1year old I don't want to risk damage, are you saying I should increase it by another 25mV or up to 25mV?

EDIT to add - I tried increasing from +19mV to +25mV and the game lasted a couple of minutes longer but still crashed.

[gupsterg]

VRM temps are sweet , they tend have a very high rating ie 100C+. As temp increases output capacity/efficiency is effected, at your logged temps they are very cool and a non issue.

GPU temp non issue.

Voltage depends to a few things (listing as simply/short as I can):-

i) the ROM is never set to manual voltage it is more like "auto calculated", so depending on GPU properties each may differ as to what VID/VDDC per DPM is set.
ii) Electronic Variable Voltage (EVV), even when card is in same DPM state (ie DPM 7 1040MHz) loading GPU with differing apps you will see differing VID/VDDC.

Upto +75mV is not an issue IMO, if you don't feel comfortable doing that then try upto +50mV. You could always lower RAM clock and just test.

[8bit]

So after recovering from that crash last night I fired up Mirror's Edge: Catalyst which played perfectly for about an hour until I'd had enough and saved and quit, no crashes in sight. I was running the GPU-z logs, I'll have a proper look at home this evening but I think the VRM temps were sitting around the mid 80's for most of that time with the GPU in the mid 70s.

If it wasn't for that single Elite Dangerous crash the other night I'd be thinking this was purely a software issue with No Man's Sky. I'll play about a bit with power and memory clocks on the card this evening under No Man's Sky and see if I can get it running stable, but it really should cope fine under the factory OC (which is pretty minimal above normal clocks IIRC) so am tempted to try and RMA the card while there's still some warranty.