Question Strange issue with XFX R9 380

[Jmi20]

Im having a strange issue with an XFX R9 380 4GB.

wouldn't post with a Z77 board, wouldn’t post on a G41 board, but would surprisingly post on an old dell optiplex 745. Have yet to try it on a B450 board.

the Z77 board is most likely defective as it also doesn’t post with a 1050 ti installed.

when the gpu is installed in the optiplex, going into bios settings is impossible. Video mode 103 not supported by gpu. Bios splash screen is also different. But boots into windows 10…

installed drivers and ran msi kombustor to stress test, first used the furmark 1700 mb and all was well. But when i used the 3200mb test it crashes instantly.

tried using xmrig (not my card, not a miner. Owner is) and also fails. Randomx memory allocation failed plus another error (forgot). SRBminer crashes instantly on cngpu algo and all others.

on ubuntu xmrig gives error cl memory allocation failure.

Note: nvidia drivers are installed alongside the amd drivers on both OSes

The owner (friend) bought it used yesterday hoping to play a few games cause he couldn't find a new card, his old one was artifact-ing cuz of mining. no system has any game to test it with so i thought mining software would be a good alternative.

i want to believe these errors are because im running it on a dell optiplex which might not be fully compatible.

the card can be returned to the seller for 4 more days, but I wouldn’t be able to test it on a newer b450 board as its in my parents house, their city is locked down for the next week.

Could the vram of this card be defective? is there any definitive test we could use to determine if this card is defective using these old motherboards? It’s posting on the 745, it boots into windows. Non memory intensive stress test works, tried watching 4k on youtube no issues, no garbage on display… could this be one of those cards that don’t work on a non uefi system? (But it does post, on a dell optiplex of all things)

we want to cling on to the hope that this gpu isn’t defective and just incompatible with the dell. its very hard to find even a used one at a reasonable price atm…

any help appreciated, thanks.

 

[Jmi20]

Problem solved!

used ddu to uninstall nvidia drivers on windows. No issue since, so we are confident it would work in either of our main pc’s.

this is a very strange occurrence as in my friends main pc, he had drivers for his old gt 710 and his now dead rx 570 and posed no problem whatsoever.

 

[werberman]

Ya, this sounds like a VRAM error (either the controller or the memory itself, though I strongly suspect the latter). It is getting up there in terms of age - chips can fail for all kinds of reasons. Could be that one of the memory chips is physically coming apart from the PCB due to heat stress. It COULD be something else, just seems unlikely.

 

[werberman]

Problem solved!

used ddu to uninstall nvidia drivers on windows. No issue since, so we are confident it would work in either of our main pc’s.

this is a very strange occurrence as in my friends main pc, he had drivers for his old gt 710 and his now dead rx 570 and posed no problem whatsoever.

Oof, that's the kind of issue that would thoroughly piss me off. It's also puzzling, given your tests. I'd keep an eye on that one for a while, but I'm glad it resolved!

 

[Jmi20]

Oof, that's the kind of issue that would thoroughly piss me off. It's also puzzling, given your tests. I'd keep an eye on that one for a while, but I'm glad it resolved!

it seems i was mistaken, it isnt over yet.

its even more inconsistent now.
mining with xmrig now no longer produces an error, and is also able to allocate 4gb of memory, but sometimes the drivers crash other times i would be left with a black screen.

heaven benchmark works well, no artifacts or anything of the sort.

furmark in msi kombustor is inconsistent.

it seems the older the drivers i install the more stable it gets, up to a certain point…

i did some searches and it seems the r9 380 had black screen issues when new, so there must be a fault in the design, the only question is: is the black screen im experiencing now due to the “original” fault or some other defect.

There most be a fault somewhere in the software side of things. god, thank goodness i never get amd gpus. I personally stick to nvidia for desktop gpus… i love the new cpus though.

 

[Jmi20]

update:

i forgot to mention that the psu being used is a corsair ax850 as the stock optiplex one doesnt have 6 pin…

we suspected that the reason why it kept crashing is because of the psu. We exchanged it with a spare generic “topower” btc brand psu and it stabilized even further. We have yet to determine if the corsair is bad or just coincidence.

furthermore stability would improve even more after the driver crashes and the system doesnt auto restart or stay blank. This is true whether anybody would believe me or not, i cant believe it myself…

i could only surmise from this experience that amd gpus are garbage, or at least the old ones. Maybe also their drivers. Never had an issue with nvidia. Whether its an ancient gts 250, useless gt 710, or rtx 2080.

 

[Zerk2012]

update:

i forgot to mention that the psu being used is a corsair ax850 as the stock optiplex one doesnt have 6 pin…

we suspected that the reason why it kept crashing is because of the psu. We exchanged it with a spare generic “topower” btc brand psu and it stabilized even further. We have yet to determine if the corsair is bad or just coincidence.

furthermore stability would improve even more after the driver crashes and the system doesnt auto restart or stay blank. This is true whether anybody would believe me or not, i cant believe it myself…

i could only surmise from this experience that amd gpus are garbage, or at least the old ones. Maybe also their drivers. Never had an issue with nvidia. Whether its an ancient gts 250, useless gt 710, or rtx 2080.

XFX was a budget brand on most models the card is probably giving up because of the mining.

 

[Jmi20]

XFX was a budget brand on most models the card is probably giving up because of the mining.

no, sorry let me clarify.

my friend bought a used r9 380 yesterday as a temporary replacement for his gpu which broke due to mining.

we couldnt find a new one, specially at a reasonable price so we looked for a used one online. We found the xfx for about $100 which is about the same price as a new gt 710 (here). So we took it.

that seller also has a physical store and bought new components from him before so i kinda trust him, he offers a five day replacement for the used products he sells.

the card arrived yesterday but unfortunately me and my friend are both stuck at my apartment cuz of temporary lockdown measures here and cant get to our main rigs till next week. So were stuck with testing it with uhh, sketchy equipment.

none of the pcs in my apartment have any games so we used benchmarks and mining software to test it out.

at the moment it seems that the main issue is actually on the software side, whether the bios of the old motherboard is not “fully accepting” the gpu or the drivers are just trash is a mystery to me. But i cant say for certain unless i have a known good pc to test it with… all i could say is i might never buy an amd gpu. What a horrible experience. i would take their cpus any time though.

 

[Jmi20]

Final update!!!
Thermal paste (or pads)

there is a small heatsink underneath the heatsink of the card that appear to be for the vrms, we took it off and found whats left of the paste (or pads, couldn’t know). We put on some of those cheap pink thermal pads i had and reinstalled.

works without error.

 

[werberman]

Your saga is making way more sense now - that may be the end of it, but depending on how long it was dealing with thermal issues the card may have a permanently shortened lifespan. Personally, I would take it back if I still could, but given the market, I can see how that may not be an easy decision.

 

[Jmi20]

Your saga is making way more sense now - that may be the end of it, but depending on how long it was dealing with thermal issues the card may have a permanently shortened lifespan. Personally, I would take it back if I still could, but given the market, I can see how that may not be an easy decision.

Well, taking it back isnt my decision to make lol. Plus the seller stuck a warranty seal on the heat sink screws so we cant take it back now. Thermal paste on the chip itself seems new, maybe the seller forgot to change the vrms. There isn’t much discoloration on the board so i guess we’re safe.

very peculiar that a thermals issue would manifest in such a way. But all is good now, and hey, it overclocks great too!

i redact my statement about not getting an amd gpu. Id definitely get one, just not used.