Hey, first thread here so it's going to be a mess by itself but the situation is quite messy too.
I bought a used Sapphire RX 470D Platinum, which I then flashed into a full RX 470 using the Sapphire Platinum equivalent. Had some issues flashing, first flash failed because of Windows being weird, then I flashed the original bios back using freeDOS and later successfully flashed the RX 470 bios using Ubuntu. Since then I've been experimenting with mild overclocking and undervolting, and I noticed that voltages would spike way high, sometimes up to 1.29v(for comparison, default voltage is 1.038v and I am currently undervolting to 1.01v, and max voltage is supposed to be 1.15v) though they were occurring sparringly so I initially wrote it off as LLC. Later on I discovered that those spikes can be induced by using the GPU encoder, and seem to occur more consistently using AVC/H265 specifically, but of course they still occur randomly throughout the day.
I have already tried underclocking both GPU and VRAM, even turning all the power limit all the way down to -50%, lowest it will allow, in conjuction with the underclock, but it seems to make no difference. I have already flashed it back to the original RX 470D bios, to no effect. I recently tried to record footage while I played an intensive game and the GPU temps went way higher than usual, which I feel confirms it isn't just the voltage sensors not working. I have also already tried a different, compatible VBIOS, once again to no effect. I have reinstalled the drivers at least twice, and one time with DDU. DDUing the drivers seemed to fix it initially, but then the random voltage spikes began again shortly after. Drivers before DDU were stable, only installed the beta ones now.
Full system specs, for those interested:
Sapphire RX 470D Platinum 4gb flashed to Sapphire RX470D Platinum 4gb, ASIC quality 81.2% and it was the same before the flash
Ryzen 3 1200 overclocked to 3.8ghz, already tried lowering the LLC on the motherboard just in case it was by chance causing the GPU spikes, to no effect. Have not yet tried setting it back to stock as I think CPU and MOBO don't have anything to do with it, this did not happen at all on my previous GPU(Powercolor RX 560D)
Asus Prime B350M-K, bios version 5222 AGESA 1.0.0.3, dated 10/14/2019
Corsair CX450 CP-9020120-BR
G.Skill Ripjaws 1x8GB 2666mhz CL15
2 Samsung HDDs, one's got a SMART error and the other has a SMART warning(these are really old and somehow still work so I have yet to get around to replacing them)
TP-LINK 150mbps wireless card, PCI-E 1x
Has anyone here had similar issues with your GCN4 cards? Or have any idea of what could fix, or at least mitigate the issue?
Lowering the GPU clocks and voltage does not seem to lower the maximum voltage spike, so I'm fresh out of ideas of what to do to attempt to fix this. I have run OCCT several times both for 3D and VRAM, and it never showed any problems there, no artifacts in games either. Videos also come out just fine when encoded by the GPU both on H264 and H265. It really is just the voltage spiking seemingly randomly and when using the GPU video encoder.
GPUZ image and example of voltage spike(random one, not GPU encoder enduced) on HWInfo attached. Voltage spikes can also be seen on GPUZ, so it's not just HWInfo misreporting either. If I missed anything, please tell me and I will answer as soon as I can.
Oh, another thing to note, GPU performance seems to be completely fine. I'll run and post results of some benchmarks if anyone thinks that'll help. Clocks on the stock VBIOS and the one I flashed are a bit different, with the flash going a bit higher, but it never showed any issues besides the voltage spikes, and like I said I have tried both underclocking GPU and VRAM and flashing the original VBIOS back to no effect.
I have also tried to overvolt the GPU(but not the VRAM) to see if it was a stability issue, once again to no effect. If I recall correctly, during this I was using the RX 470D VBIOS and I lost the aux current sensor from the card, simply vanished.
I apologize for the long and messy first thread, but I'm really concerned about the cards longevity and I already did a lot of messing around myself to try to fix it.
I bought a used Sapphire RX 470D Platinum, which I then flashed into a full RX 470 using the Sapphire Platinum equivalent. Had some issues flashing, first flash failed because of Windows being weird, then I flashed the original bios back using freeDOS and later successfully flashed the RX 470 bios using Ubuntu. Since then I've been experimenting with mild overclocking and undervolting, and I noticed that voltages would spike way high, sometimes up to 1.29v(for comparison, default voltage is 1.038v and I am currently undervolting to 1.01v, and max voltage is supposed to be 1.15v) though they were occurring sparringly so I initially wrote it off as LLC. Later on I discovered that those spikes can be induced by using the GPU encoder, and seem to occur more consistently using AVC/H265 specifically, but of course they still occur randomly throughout the day.
I have already tried underclocking both GPU and VRAM, even turning all the power limit all the way down to -50%, lowest it will allow, in conjuction with the underclock, but it seems to make no difference. I have already flashed it back to the original RX 470D bios, to no effect. I recently tried to record footage while I played an intensive game and the GPU temps went way higher than usual, which I feel confirms it isn't just the voltage sensors not working. I have also already tried a different, compatible VBIOS, once again to no effect. I have reinstalled the drivers at least twice, and one time with DDU. DDUing the drivers seemed to fix it initially, but then the random voltage spikes began again shortly after. Drivers before DDU were stable, only installed the beta ones now.
Full system specs, for those interested:
Sapphire RX 470D Platinum 4gb flashed to Sapphire RX470D Platinum 4gb, ASIC quality 81.2% and it was the same before the flash
Ryzen 3 1200 overclocked to 3.8ghz, already tried lowering the LLC on the motherboard just in case it was by chance causing the GPU spikes, to no effect. Have not yet tried setting it back to stock as I think CPU and MOBO don't have anything to do with it, this did not happen at all on my previous GPU(Powercolor RX 560D)
Asus Prime B350M-K, bios version 5222 AGESA 1.0.0.3, dated 10/14/2019
Corsair CX450 CP-9020120-BR
G.Skill Ripjaws 1x8GB 2666mhz CL15
2 Samsung HDDs, one's got a SMART error and the other has a SMART warning(these are really old and somehow still work so I have yet to get around to replacing them)
TP-LINK 150mbps wireless card, PCI-E 1x
Has anyone here had similar issues with your GCN4 cards? Or have any idea of what could fix, or at least mitigate the issue?
Lowering the GPU clocks and voltage does not seem to lower the maximum voltage spike, so I'm fresh out of ideas of what to do to attempt to fix this. I have run OCCT several times both for 3D and VRAM, and it never showed any problems there, no artifacts in games either. Videos also come out just fine when encoded by the GPU both on H264 and H265. It really is just the voltage spiking seemingly randomly and when using the GPU video encoder.
GPUZ image and example of voltage spike(random one, not GPU encoder enduced) on HWInfo attached. Voltage spikes can also be seen on GPUZ, so it's not just HWInfo misreporting either. If I missed anything, please tell me and I will answer as soon as I can.
Oh, another thing to note, GPU performance seems to be completely fine. I'll run and post results of some benchmarks if anyone thinks that'll help. Clocks on the stock VBIOS and the one I flashed are a bit different, with the flash going a bit higher, but it never showed any issues besides the voltage spikes, and like I said I have tried both underclocking GPU and VRAM and flashing the original VBIOS back to no effect.
I have also tried to overvolt the GPU(but not the VRAM) to see if it was a stability issue, once again to no effect. If I recall correctly, during this I was using the RX 470D VBIOS and I lost the aux current sensor from the card, simply vanished.
I apologize for the long and messy first thread, but I'm really concerned about the cards longevity and I already did a lot of messing around myself to try to fix it.
Attachments
Last edited:
