So first off, my specs/config, (I'm going to go through all the crap I had to deal with in order to reach stability but if you want to go straight to what my FIX was, it is at the bottom).
System:
CPU: Core i5 2500K OC to 4.5 (H55 with 2x 120SP Quiet Fans)
RAM: 16GB (4x4) DDR3 Patriot Gamer Series 2
MOBO:Asrock Z77 Pro3 -> Intel DZ68BC v28 BIOS
Hard Drives:
60GB Kingston V300 (For Games)
90GB Kingston V200+ (For OS)
2x320GB Western Digital Blue
1TB Western Digital Blue
GPU: Sapphire R9-290 (BF4 Edition) Reference Card (290X BIOS)
GPU Cooling: Arctic Accelero Extreme IV + Gelid VRM Heatsinks + AKUST Aluminum VRAM Heatsinks
AUDIO: USB DAC (ELE-D02)
PSU:Corsair CX750M -> Rosewill CAPSTONE 750W
CASE: Corsair 330R
OS: Windows 8.1 (64bit)
2X: DELL 1080p Monitors; Tried both HDMI + DVI-D, and 2x DVI-D Configs
Addition GPU Info:
ASIC: 68.0%
Memory Type: EPILDA
Memory Name: EDW2032BBBG_DEBUG2
Clock Settings: 947/1250
PCI-Express 2.0 (Since the 2500K doesn't have the 3.0 Controller)
Cooling: While the aftermarket solutions brought my temperatures down by a lot, they did not resolve the black screen issues.
Other info:
Came from a GTX 670 which worked flawlessly. Nothing but headaches since rejoining Red Team (was rocking 6850 CF prior to GTX 670).
I only blackscreen (generally) while playing games . The only non-gaming blackscreen occured after booting to desktop and the screen would sometimes freeze around the time Sapphire Trixx would be loading and then get a blackscreen. After a reboot, it is okay. I have since uninstalled Sapphire Trixx and have yet to experience this blackscreen not while gaming without Sapphire Trixx installed (though I still get blackscreens in game). IMO this sucks because I don't like Afterburner and Catalyst lacks fan curve settings.
Blackscreen Issue:
Things I've Tried (and didn't work):
Things I have not tried (but solved it for other people):
What worked and how did I figure it out:
So after I installed 13.12 and tested it until I black screened, I decided there was no point staying with old drivers that still share the same issue so I decided to uninstall the drivers, clean it up with DDU (for those who don't know what DDU is, it is Display Driver Uninstaller which is a nice utility that doesn't blow like Driver Fusion [once the great Driver Sweeper]).
After installing 14.7 I tried to test Dota 2 when I realized at upon locking a character I would always black screen upon pressing the button! I had found a way to trigger the black screen on command. I abused this and tweaked settings for my GPU one a time, little by little only to realize that the only thing that stopped the black screen from occurring upon locking in a character was dropping the Core clock. This baffled me considering my GPU temperature wasn't anywhere close to overheating so why would a core clock change stop the black screens? I can only assume that it is due to a HARDWARE based issue and this definitely calls for an RMA if you can (I cannot in my case). I initialially dropped the GPU clock by 5% in Overdrive which stopped the triggering of instant black screen upon character lock in. However I still black screened in game so I ended up dropping the clock further by another 5% and as a result of this was able to finish my first game of Dota 2 ever with 290..
-------------TL;DR
TL;DR: I fixed my black screen issue by reducing core clock by 100MHz OR use the Overdrive setting in CCC, with -10% GPU Clock. Since my 290 is stock at 947/1250, at full load it would now run at ~850/1250. I was then able to boost my memory to 1300 without any issues. No more black screens! (though at 100MHz cost).
Edit: I had to drop my core clock even further to 845 for full stability after a lot more testing. The performance hit is noticeable, when benchmarking (Unigine Valley/FireStrike [3D MARK]).
Edit #2: After longer sessions of stress testing, I had to drop the core clock even further to 830 MHz. I managed to off-set some of my performance losses by flashing my 290 to a 290X. At this core clock (830) I get about 2300-2400 in Unigine Valley. I ran 947/1320 with a 290 BIOS and benchmarked a score at roughly the same ballpark(as a 290 at 947/1250, my card is stable enough for benchmarks but crashes randomly while in game). So basically 290 with 947/1320 performs near identical with a 290x at 830/1320. Ultimately in my case, using the 290x BIOS is a superior choice since the card is stable at 830/1320 as a 290x but not at 947/1320 as 290.
Note: I can run most benchmarks (at stock settings as a 290--as a 290x I HAVE to underclock to 980MhZ minimum to avoid instant blackscreen though still unstable) without black screens so benchmarks like UNIGINE, 3DMARK, FURMARK are POOR ways to test for stability of an AMD card suffering from blackscreens. Testing with actual games (graphically intensive ones of course) are a good start A decent stress test however I found was to test games while using software like OpenBroadcaster to stream (capture video) while playing. This appears to stress the system much harder (this was how I revealed that 845-850MHz core was still not fully stable).
Note#2: If you ARE able to adjust VDDC, adjust that instead of core clock as I suspect this will be able to resolve the blackscreen issue related to core clock. In my case the I can't access the I2C device on my card meaning I can't touch the voltage at ALL. It's also the reason why my card does not have sensors for VDDC etc on GPU-Z (based on chats/research on Guru3D and HWINFO forums, the IR3567B appears to provide the telemetry for those information/data). I tried using AMD's ADL SDK to manually access the I2C interface to no avail. Maybe the IOR 3567B IC is dead? Faulty? Counterfeit? There was a post about the specific model my (IR) PWM controller has but there were no responses, the user also had an issue with the IC.
-------------TL;DR
Although its working for the most part, it sucks that my card is unable to run at advertised spec.
Why is it so hard to identify a Fix?:
However I don't know if this truly a pure hardware issue or rather even worse-- both a HARDWARE AND SOFTWARE problem. If this was purely a hardware issue why am I able to trigger the black screen with 100% chance of occurrence using the 14.7 drivers but not 13.12 or 14.4? Does this imply a driver issue? Or do the latest drivers cause the hardware to behave in a manner that reveals non-readily apparent hardware issues?
I feel that without a doubt this is a HARDWARE issue so if you can RMA, do it. If you can't you are SOL as no fixes apply to everyone. You will have to attempt to trigger your black screens manually and test various configurations of settings to find the cause.
System:
CPU: Core i5 2500K OC to 4.5 (H55 with 2x 120SP Quiet Fans)
RAM: 16GB (4x4) DDR3 Patriot Gamer Series 2
MOBO:
Hard Drives:
60GB Kingston V300 (For Games)
90GB Kingston V200+ (For OS)
2x320GB Western Digital Blue
1TB Western Digital Blue
GPU: Sapphire R9-290 (BF4 Edition) Reference Card (290X BIOS)
GPU Cooling: Arctic Accelero Extreme IV + Gelid VRM Heatsinks + AKUST Aluminum VRAM Heatsinks
AUDIO: USB DAC (ELE-D02)
PSU:
CASE: Corsair 330R
OS: Windows 8.1 (64bit)
2X: DELL 1080p Monitors; Tried both HDMI + DVI-D, and 2x DVI-D Configs
Addition GPU Info:
ASIC: 68.0%
Memory Type: EPILDA
Memory Name: EDW2032BBBG_DEBUG2
Clock Settings: 947/1250
PCI-Express 2.0 (Since the 2500K doesn't have the 3.0 Controller)
Cooling: While the aftermarket solutions brought my temperatures down by a lot, they did not resolve the black screen issues.
Other info:
Came from a GTX 670 which worked flawlessly. Nothing but headaches since rejoining Red Team (was rocking 6850 CF prior to GTX 670).
I only blackscreen (generally) while playing games . The only non-gaming blackscreen occured after booting to desktop and the screen would sometimes freeze around the time Sapphire Trixx would be loading and then get a blackscreen. After a reboot, it is okay. I have since uninstalled Sapphire Trixx and have yet to experience this blackscreen not while gaming without Sapphire Trixx installed (though I still get blackscreens in game). IMO this sucks because I don't like Afterburner and Catalyst lacks fan curve settings.
Blackscreen Issue:
Things I've Tried (and didn't work):
- Different drivers (14.4, 14.7, 13.12 all Black Screen) cleaned with DDU
- Installing drivers, and making absolutley no changes on any driver settings, e.g. No overdrive, Trixx, MSI afterburner, GPU Tweak
- Testing with numerous setting tweaks with Trixx, Afterburner and GPU Tweak (one at a time, different software--not at the same time)
- Aftermarket GPU Cooling
- Black screen in my case is NOT a temperature issue--core temp caps about 65C on full load, 68C with FurMark,
- VRM(1) heatsinks are not super hot to the touch though I can't confirm numerical because this GPU lacks VRM sensors
- Off-topic: If you are thinking of getting an Arctic Accelero after market cooler, I think you should totally do it if you are willing to get your hands dirty. It was a pain to install the first time with the 670, but it was easy the second time around with the Extreme IV on the 290.
- I took forever with the Twin Turbo III on the GTX 670 and I had to MacGyver the VRM Heatsinks (they were tall, but you definitely need these heatsinks). It dropped my temperatures significantly (from core temp ~80 to 64 on full load with the reference EVGA Blower cooler) AND the thing was so quiet I thought it was broken!
- Flash GPU with numerous BIOS ranging from BIOS on 290->290X thread, to numerous on TechPowerUp that support EDW2032BBBG_DEBUG2 memory
- Keep 3D clocks permanently
- (PT1 Bios from 290 to 290X thread, although I have yet to try a BIOS for 290 where clocks are not dynamically changing)
- Split PCI-E from different cables from PSU (though the CX750M is single rail so this shouldn't be a problem)
- sfc /scannow (though it does say it has things it cannot repair)
- Underclocked GPU Memory (Blackscreened with 500mhz memory freq..)
- Reseat GPU in PCI-E Slot/ Clean
- Reapply thermal paste to GPU (Arctic Silver)
- Disable ULPS
- Bumping up power target or reducing it
- Overclocking slightly (someone said it helped)
- Keep GPU fan at full blast
- Full system scan with Malwarebytes and Windows Defender, found nothing.
- I know Windows Defender heuristics aren't very good but I only surf Netflix/Youtube or safe websites and avoid shady looking urls I have Adblock Plus + Ghostery and keep Javascript disabled on Chrome aside from exceptions added for trusted sites.
- Forced constant voltage in MSI Afterburner
- AHCI Drivers (Someone mentioned this in an anandtech thread)
- Performance setting for Power settings on Windows
- Reset CMOS
- Adjust VTT Voltage on motherboard, also known as VCCSA or voltage for "Processor System Agent" on Intel Motherboards. Someone claimed that raising this to 1.1V significantly improved stability however I tried all possible voltages from .92V to 1.3V for this and it had absolutely no effect on stability.
- Different Power Supply; CX750M and Rosewill Capstone 750W
- Try another PCI-E Slot
Things I have not tried (but solved it for other people):
- RMA (Not an option for me, don't have POP and I threw out the reference cooler before realizing I would be plagued with this issue, plus using aftermarket coolers generally void warranty)
- If I am unable to fix this issue without RMA I'm just going to go get an NVIDIA card.
- Tweak VDDC Offset (I can't do this, option doesn't show up in Afterburner/Trixx even with a custom BIOS). I think the reference Sapphire R9 290 doesn't have the hardware capable of controlling the voltage offsets through software..
- Complete system Reformat (Troublesome as of now but I may try later when I have time)
- Dropping monitor refreshrate from 120 to 60 Hz (I don't even have a 120Hz monitor...)
What worked and how did I figure it out:
So after I installed 13.12 and tested it until I black screened, I decided there was no point staying with old drivers that still share the same issue so I decided to uninstall the drivers, clean it up with DDU (for those who don't know what DDU is, it is Display Driver Uninstaller which is a nice utility that doesn't blow like Driver Fusion [once the great Driver Sweeper]).
After installing 14.7 I tried to test Dota 2 when I realized at upon locking a character I would always black screen upon pressing the button! I had found a way to trigger the black screen on command. I abused this and tweaked settings for my GPU one a time, little by little only to realize that the only thing that stopped the black screen from occurring upon locking in a character was dropping the Core clock. This baffled me considering my GPU temperature wasn't anywhere close to overheating so why would a core clock change stop the black screens? I can only assume that it is due to a HARDWARE based issue and this definitely calls for an RMA if you can (I cannot in my case). I initialially dropped the GPU clock by 5% in Overdrive which stopped the triggering of instant black screen upon character lock in. However I still black screened in game so I ended up dropping the clock further by another 5% and as a result of this was able to finish my first game of Dota 2 ever with 290..
-------------TL;DR
Edit: I had to drop my core clock even further to 845 for full stability after a lot more testing. The performance hit is noticeable, when benchmarking (Unigine Valley/FireStrike [3D MARK]).
Edit #2: After longer sessions of stress testing, I had to drop the core clock even further to 830 MHz. I managed to off-set some of my performance losses by flashing my 290 to a 290X. At this core clock (830) I get about 2300-2400 in Unigine Valley. I ran 947/1320 with a 290 BIOS and benchmarked a score at roughly the same ballpark(as a 290 at 947/1250, my card is stable enough for benchmarks but crashes randomly while in game). So basically 290 with 947/1320 performs near identical with a 290x at 830/1320. Ultimately in my case, using the 290x BIOS is a superior choice since the card is stable at 830/1320 as a 290x but not at 947/1320 as 290.
Note: I can run most benchmarks (at stock settings as a 290--as a 290x I HAVE to underclock to 980MhZ minimum to avoid instant blackscreen though still unstable) without black screens so benchmarks like UNIGINE, 3DMARK, FURMARK are POOR ways to test for stability of an AMD card suffering from blackscreens. Testing with actual games (graphically intensive ones of course) are a good start A decent stress test however I found was to test games while using software like OpenBroadcaster to stream (capture video) while playing. This appears to stress the system much harder (this was how I revealed that 845-850MHz core was still not fully stable).
Note#2: If you ARE able to adjust VDDC, adjust that instead of core clock as I suspect this will be able to resolve the blackscreen issue related to core clock. In my case the I can't access the I2C device on my card meaning I can't touch the voltage at ALL. It's also the reason why my card does not have sensors for VDDC etc on GPU-Z (based on chats/research on Guru3D and HWINFO forums, the IR3567B appears to provide the telemetry for those information/data). I tried using AMD's ADL SDK to manually access the I2C interface to no avail. Maybe the IOR 3567B IC is dead? Faulty? Counterfeit? There was a post about the specific model my (IR) PWM controller has but there were no responses, the user also had an issue with the IC.
IC Name: IR (international rectifier) 3567B N330P J971
-------------TL;DR
Although its working for the most part, it sucks that my card is unable to run at advertised spec.
Why is it so hard to identify a Fix?:
- The black screening is Random and it is difficult to identify the source of the problem when you can't trigger it on the spot (though I found a way in my set up but it may not occur on other setups).
- In my case, my Crapphire card has virtually no sensors! I couldn't tell if my voltage levels were bad, nor could I see if my VRM temperatures were too high. For those wondering why my idle temps are so high, I live in Hawaii and don't use A/C.
- No BSOD = No Event Viewer log entry to get a start on the source of of the problem. Howeever, I did manage to get a BSOD once when I fell asleep when a blackscreen occured. I woke up when a blue flash ignited my dark room. It was a VIDEO_TDR_FAILURE. Once again this BSOD is not very helpful as all it means is the GPU failed to respond within a time interval which generally only indicates GPU issues.
- All the solutions being posted on forums all over the place (my Google first 3 Pages of search results on the Blackscreen issue are all purple links) imply that the issue is all hardware related. This is due to the fact that people have fixed their issues with locking voltages, reducing core/memory clock values--all hardware related issues and not to mention the people who posted with 100% fixes were those who RMA'd their card.
I feel that without a doubt this is a HARDWARE issue so if you can RMA, do it. If you can't you are SOL as no fixes apply to everyone. You will have to attempt to trigger your black screens manually and test various configurations of settings to find the cause.
