Taking on initial release iterations of hardware in a new generation is always a gamble. In this case, it’s a gamble I lost. After securing one of the apparently very few actually available RTX 3080 cards on launch day for my new PC build I’ve had a few dramas to work through.
November 1 Update: Turns out, after testing with a different RTX 3080 — the issue lay elsewhere. With the RAM frequency to be more specific. Even though it tested fine and ran stable-as-can-be with a 1080ti in the system; some combination of the 3080 and the high frequency caused issues.
The below might still be of some interest, but also check out the top of this post for the latest information.
Initially it seemed like everything might be resolved by the 465.55 driver release. And at first I actually thought it had for me too. The Division 2, for example, went from crashing in minutes on 465.38 to lasting significantly longer on the updated driver. But ‘longer’ turned out to have a limit, at an outer stretch I’d crash within an hour or two and I started finding some gaming sessions crashing within 30 minutes.
Horizon Zero Dawn was similar — could sometimes get a two hour session, but othertimes crash much, much sooner.
Funnily enough though, it was Assassin’s Creed: Odyssey which became my ‘go to’ crash-test game while I was still trying to sort through the issues and see if I could resolve them. 15 minutes was the typical crash time for this one.
After a lot of pottering around the internet looking at potential solutions and also mainly for any sign of other people continuing to have these sorts of issues with the new driver I had to acknowledge my issue was likely different from the common ones which in the main were solved by the driver update. For one, my crashes were not always a clean simple CTD (Crash to Desktop). Sometimes it would be a BSOD (Bluescreen of Death) with a variety of stop codes attached — but generally memory management or video scheduler internal error.
I ran through a battery of system stability checks ranging from simple memtest through to a more intensive run of Prime95 in torture test mode and these all indicated no issues with the CPU, RAM or system stability more generally.
Most interesting to me though was the fact I could also run FurMark with a high resolution and 8x MSAA pushing the GPU to ~99% utilisation and let that run for quite a while with no issues whatsoever. My theory is that the GPU silicon might’ve actually been just fine, but rather the issue existed in the VRAM which this doesn’t touch as intensively as true gaming scenarios do.
I’ve admitted defeat and taken the card back to my vendor to kick-off the RMA process. Even if they don’t give me grief and try and tell me there are no issues and just to update my drivers (despite the clear advice I already had) — actually getting a hold of a replacement is likely going to take quite some time.
Advised dates of delivery for 3080’s placed today from my vendor are looking to be early December.
In the meantime, I’ve put my trusty ol’ GTX 1080ti in here so that I can at least continue to game — if not as gloriously smooth and high quality as before — while I await to hear my fate around the RMA process.
Although I will say, that I am at least somewhat heartened by the fact that I can run a game in a stable fashion again now. More than just for the sake of gaming itself too — I always start second-guessing myself when I need to RMA hardware. Did I just miss something? Should I have just done one more pass of DDU + Driver reinstall? Would one more reformat and reinstall of Windows given it a better shot? Did I miss something in troubleshooting and it’s actually a different component entirely??
Seems at least this time — no. It really was the card. My only remaining worry is that it was a compatibility issue between the specific card and my mobo, which would mean it’s incredibly unlikely they’ll be able to recreate the issue on their testbench. I did install the latest available BIOS for the board, but that release was still a couple of months before the release of the 3080 cards. Watch as one comes in the next week or two addressing the issue… ;)
Anywho, that’s my story for the day. As I said in the opening — going for Rev1.0 of new generation hardware is always a gamble. I just hope that I don’t end up waiting too long to rectify the mistake this time around!