Mine still here... not in use anymore but still alive, along with her younger sister ROG STRIKER-GTX760-P-4GD5 also alive and a death older sister ROG Poseidon Platinum GTX 780.
Got the memory chips last week for the Asus Matrix gtx 980 ti 6gb Platinum. Going to take a few weeks before I send it away to get it repaired as I got to save up money first
I have been trying to get the BFG 8600 GTS to come back to life after doing varies of cleaning which was a abit of a challenge as first time round it refused to boot but causes problems with the motherboard resetting it on purpose but after giving it a good cleaning all over the pcb as it took awhile and it came back from the dead which was a surprise to me might have abit of an issue with it not sure till I get a operating system installed on another hard drive to load 32bit to see if it working correctly as trying to remember where did I put the windows xp hard drive as I know it an 80gb but cant see it with the soot covering the label
Made an account just to hop in this thread. Some posts a while back related to some Vega cards caught my eye and I and wanted to post one of them. I shot for the one that had a VBIOS reporting as a Vega 64 Frontier Edition, despite it clearly being otherwise.
It works well, though getting a power draw reading off it is not going well, so I can for sure claim it's not finished. It lacks the yellow sticker but I did take off its backplate to see what else is there. If anything stands out I'm all ears. The "untest" sticker has my attention, as does the rectangle under it being covered in sticker residue. Seller mentioned it was grabbed from e-waste which I'm inclined to believe given the light damage. Also I noticed there's a yellow film of some sort on the LED switch. I've opted to leave it be.
And then for the one that people might care for a bit more, the AMD Radeon Pro V320. The entry here has some points that stand out to me a little. More looking to give discussion about this one since they're admittedly hard to come by due to what machine they were for.
First, I see it mentions 2 8pins, but this is not correct, it has an 8 and 6 pin. Second it IS an 8GB card but what it reports can vary. I know multiple outlets reported this having 16GB or "our version has 8GB instead of 16GB", and this does appear to be possible, but as far as I'm aware theres only 1 SKU and it has 8GB for this specific card. Having encountered 3 of them personally, and 5 if you include other outlets reports I think that's a fairly good sample size. In linux, depending on where you look it either reports as an AMD Radeon Pro V320, an AMD Radeon Instinct MI25, or both at the same time. I've seen it report 16GB in some fields but as for why I am not sure. If additional details are desired I can setup a machine with linux to poke at it as I have the hardware to do so (more on that below).
It can work under windows using the generic drivers, but no official windows driver is around. There's an old beta driver package from AMD that have 2 very close results, both of which work, albeit just barely so I cant advise those. One being labeled "6860:C0" which has a higher base fan %, the other being "Radeon Pro V320MxGPU" which i find causes issues with sound and when sound comes into the picture it ramps the fan til you shut off windows. Some users have managed to make it work with a Vega 56 and 64 driver, and in one case somehow a 6700XT driver which I would not advise doing, but it does at least match Vega 56 pretty closely so that one might be in the safer territory.
Theres one last drawback to this GPU and it seems to be locked to Lenovo devices only. No other board will accept it, with varying results, but all of them failing to even get to bios splash screen. Maybe theres something I'm missing but as far as I can tell it does not work outside of Lenovo machines. Anyway here's some photos. and GPU-Z.
It's a card custom to Google, more specifically, Stadia. Even more specifically, a GPU tied to a thinkstation that is used for local-ish Stadia testing on the off chance the normal service was not usable/the time needed to upload was too great and started compromising the workflow. It's not even the one used in their datacenters which handled the normal load. It's a very unusual GPU to the point where even Lenovo doesn't guarantee it will work under linux (it does and quite well at that).
If anything, it was likely Google that locked it down weirdly. That or it requires specialized components not found in consumer grade boards, but what it would require is beyond me. Linus Tech Tips tried installing drivers for windows with the GPU in a testbench but never named the board in use, and it was clearly running so whatever the requirement is, it's either Lenovo specific boards, or some other non-consumer grade requirement. Same video also mentions the 16GB issue I was referring to originally if you're curious at 15:15. MI25 cards by default have 16GB and since its based off that it was originally 16GB on that page, but its 8GB now that we have them to verify.
Made an account just to hop in this thread. Some posts a while back related to some Vega cards caught my eye and I and wanted to post one of them. I shot for the one that had a VBIOS reporting as a Vega 64 Frontier Edition, despite it clearly being otherwise.
It works well, though getting a power draw reading off it is not going well, so I can for sure claim it's not finished. It lacks the yellow sticker but I did take off its backplate to see what else is there. If anything stands out I'm all ears. The "untest" sticker has my attention, as does the rectangle under it being covered in sticker residue. Seller mentioned it was grabbed from e-waste which I'm inclined to believe given the light damage. Also I noticed there's a yellow film of some sort on the LED switch. I've opted to leave it be. View attachment 289892
Sorry, my fault. I am a collector of AMD Engineering samples and my posting was to short.
Now the longer Version.
You reported power issues and than could cause the error (chip IR35217 in red circle) :
The white spot may be a burning hole, but I can't see it exacly. And I never saw a white burning hole before and no other member mentioned it. If it is white paint or thermal paste it is not the error.
So this info is important for a possible repair!
And next point is it worth to repair it? In daily use the card may run until death as Vega frontier, but a few collectors like me will pay extra money for an AMD Engineering sample like your AMD Radeon RX Vega 64. Possible Qualification Samples like this card are much more interesting, when the BIOS file is NOT available on internet platforms, so pre release BIOS is worth to mention.
White dot is usually an indicates pin one or alignment on the pcb nothing interesting about that but to diagnose a faulty ic is involving someone who knows repairs using specialist tools which I know nothing about if you are in the USA you might be able send it to NWR he is moving to Kentucky
Sorry, my fault. I am a collector of AMD Engineering samples and my posting was to short.
Now the longer Version.
You reported power issues and than could cause the error (chip IR35217 in red circle) :
The white spot may be a burning hole, but I can't see it exacly. And I never saw a white burning hole before and no other member mentioned it. If it is white paint or thermal paste it is not the error.
So this info is important for a possible repair!
Took the card back out (now that my 7900 XTX is here, first one was defective sadly), and took the backplate off and it's just white ink on the chip there, nothing else as far as I can tell.
As far as the bios is concerned, are you certain this one is undumped as of yet? If so I'll gladly just dump it.
White dot is usually an indicates pin one or alignment on the pcb nothing interesting about that but to diagnose a faulty ic is involving someone who knows repairs using specialist tools which I know nothing about if you are in the USA you might be able send it to NWR he is moving to Kentucky
We don't have any dump of a Vega card with the subsystem ID of 6B76, that appears to be completely unique. We do havethis VBIOS which has a version number numerically one behind yours. It doesn't mirror yours entirely.
We don't have any dump of a Vega card with the subsystem ID of 6B76, that appears to be completely unique. We do have this VBIOS which has a version number numerically one behind yours. It doesn't mirror yours entirely.
Huh neat. Here's the VBIOS, I opted not to use the online submission since I didn't want to make a mistake anywhere in the submission and given the card isn't a retail one it seems easier to just drop it here. Also throwing in the V320 in case its needed, I doubt it is but maybe to someone? If you find anything interesting in the VBIOS for the Vega 64 let me know.
I do have a secondary question for you. The VBIOS is 256kb from GPU-Z, but using amdvbflash dumping that V320 came out to like 1mb. What exactly is the difference since thats quite literally 4 times bigger. I recognize this isn't super important but it has me curious.
Huh neat. Here's the VBIOS, I opted not to use the online submission since I didn't want to make a mistake anywhere in the submission and given the card isn't a retail one it seems easier to just drop it here. Also throwing in the V320 in case its needed, I doubt it is but maybe to someone?
Any submission from ES cards is tagged as Unverified and isn't shown under any manufacturer tags, only the Unverified tag. It's not likely that the hosting will go down and these files would be lost on the forum, but the point of the database is to keep everything together. No worries about making mistakes.
I do have a secondary question for you. The VBIOS is 256kb from GPU-Z, but using amdvbflash dumping that V320 came out to like 1mb. What exactly is the difference since thats quite literally 4 times bigger. I recognize this isn't super important but it has me curious.
Possibly a bug with GPU-Z, it's likely seeing Vega and expecting a 256kb ROM, so that's as much as it reads. However around this time AMD started to shadow the VBIOS across a 1MB EEPROM; there are two exact copies of the VBIOS on the same physical chip, presumably for redundancy though AMD had already been offering redundancy with their old arrangement of dual EEPROMs, one write protected and one not. You may find that a large chunk of the VBIOS is simply zeroed out as well with no clear "end of BIOS data" flag anywhere which GPU-Z is looking for but amdvbflash isn't.
Any submission from ES cards is tagged as Unverified and isn't shown under any manufacturer tags, only the Unverified tag. It's not likely that the hosting will go down and these files would be lost on the forum, but the point of the database is to keep everything together. No worries about making mistakes.
Possibly a bug with GPU-Z, it's likely seeing Vega and expecting a 256kb ROM, so that's as much as it reads. However around this time AMD started to shadow the VBIOS across a 1MB EEPROM; there are two exact copies of the VBIOS on the same physical chip, presumably for redundancy though AMD had already been offering redundancy with their old arrangement of dual EEPROMs, one write protected and one not. You may find that a large chunk of the VBIOS is simply zeroed out as well with no clear "end of BIOS data" flag anywhere which GPU-Z is looking for but amdvbflash isn't.
Took a look at the VBIOS from amdvbflash and you seem to be right, at least at a glance. Theres multiple parts where it's all 00 and FF, with the bulk being after 0x40000.
Sad time for the Asus matrix gtx 980ti platinum it decided to erase it vbios memory to unknown then later diced with a massive short circuit. I could literally smell something slowly releasing the funny magic smoke. Then touched the pcb it literally cooking itself to death as it was getting so hot it didn't thermally shut down because the safe limit isn't there. This time it dead the computer refuses to post with this graphic card installed so don't know if the short is on the pcie plug side or the pcie slot side. Just have to do a investigation on this to see what happened to it.
Pretty sure the core is dead as it won't turn on the fans to cool itself down because it not suppose to be in 0db mode as it never came with that feature on this graphic card.
Found the fault one of the cylinder style cap smd had a exploded bottem possible top bulge on the left side of the pcb as before I checked around the board the three vrm looks like linked to memory were shorting to ground using the multi meter then I accidentally knocked off the suspected cap and the short disappeared all together and checked all over the pcb that short disappeared as well so might be saved once again but still have to find the replacement smd to replace that faulty one including getting the core replacement some day as I'm in no hurry
Jesus this graphic card is possessed in some way originally memory issues then later knocked off the low esr 270 16v cap near the memory vrm which I resoldered back on then later had a total shut down with pcie power plug shorted to ground then accidentally knocked off the same cap that I resoldered a replacement cap in a same spec back on earlier which cleared up the short but blew the vbios primary chip in the pin one trace gone so had to hard flash the secondary vbios chip with the original vbios belongs to it. Since the original cap had broken legs had to salvaged one in the same specs from a broken Asus p6x58d-e board which I had laying around and this time the graphic card woke up from the dead once again with the original issue with artifacts.
Any shorter on the cap that I replaced and there be fireworks popping out of that corner
AMD Radeon Pro V540. It is like a dual RX5700 but with HBM2 memory and lower clocks/tdp.
It has one minidp port. (i think it is working like a video card like M8 or M25 or any other enterprise card)
I ordered one just for fun.
