Damnit… RMA Required on my RTX 3080

Damnit… RMA Required on my RTX 3080

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.

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.

So!

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!

Share this Post:

Share on twitter
Twitter
Share on facebook
Facebook
Share on tumblr
Tumblr
Share on reddit
Reddit
Share on pinterest
Pinterest

4 thoughts on “Damnit… RMA Required on my RTX 3080”

  1. Well, that sucks. I hope you won’t have to wait that long.

    I’m still holding off, although my new system really longs for a new graphics card. My main issue is that pretty much all 3080 models (if you can get them, that is) are three slots thick, which means my mainboard’s chipset fan, almost half of which is already covered up by my old 980 GTX, would be smothered completely by any of those. So I guess I have no choice – I need a water cooler.

    Which would be fine – I had thought about going for a water system from the get-go, not least because they look really cool if done right – if it didn’t mean that I’ll need to find (and get) an exact match between card and cooler, and I’ll also lose the card’s guarantee if I do it myself.

    I guess I’ll have to wait, probably until after christmas, for cards with pre-installed water cooler to come out and be actually available, and then build the rest around that.

    Reply
    • Good plan waiting it out a bit longer, I think. With the rapid stock turnover at the moment, you can at least be sure that when you do pick-up a card, any revisions made should be in your hands rather than worrying about languishing stock putting a 1.0 into your system.

      Still not entirely clear how much of a red herring (or not, given EVGA’s initial comments) the SP-CAP vs. MLCC stuff was, but it seems at least there was enough substance to it to warrant official statements from a few vendors to confirm they’d already made similar changes for their commercial releases, or would soon do so.

      I likely would’ve waited a bit longer myself except for the pressure my monitor puts on the graphics card, it really does make the 1080ti feel like a low powered mid-range card rather than the beast it is for anything up to and including standard 1440p resolutions. xD

      We’ll see how it goes though. Trying my best not to fret too much about it in the meantime, although I will say I have had a rough experience with this crowd in the past over a video card… So; we’ll see.

      Reply
  2. “Liking” this feels wrong, but it is a well-written account of a bad situation!

    I was interested in seeing your experiences because I remembered you had the Gigabyte Eagle card which was one of the culprits people talked about when capacitorgate became a thing, but I was optimistic that no news was good news until this. The supply issues around the 3080 have been pretty crazy and it seems bad that vendors don’t have some measure of RMA stock available. On Furmark – I know some cards/drivers auto-detect it now as a power virus and limit it, so I wonder if perhaps that caused the seeming stability you had?

    Either way, December lines up with the rumors I’ve heard about the 20 GB version launching now, so maybe you can convince them to get you an upgrade?

    Reply
    • I have (or had) the Gigabyte Gaming OC 10GB card, but fundamentally aside from the factory OC it’s the same card as the Eagle. And aye, it runs a full battery of SP-CAPs with narry an MLCC array to be found.

      However it does run higher spec SP-CAPs (470s, where some of the others were running 330’s or even down as low as 220’s) and ultimately has a higher on-paper rating than some of the cards that did include an MLCC array or two. Of course, there is still the potential for the performance characteristics of the SP-CAPs being the slower option to have had an impact… But from what I can tell, for the most part, people with this card and the Eagle have found stability post the 456.55 drivers.

      re: Furmark — possible. Monitoring it in the Aorus tool, it did appear to be fully utilising the card at least as raw frequencies went, it was still boosting to 2010MHz, occasionally 2025MHz which is inline with the speeds I was seeing in gaming tests. Memory utilisation remained low though, thus my commentary on that.

      Even so, aye — I have heard that it has been somewhat common practice to limit Furmark for a number of years now, so it could well be that you’re right.

      If price on the 20GB model doesn’t end up being too insane a jump (which I wonder whether it might be given the cost of the GDDR6X memory at the moment), that’s not a bad idea if I do end up waiting that long.

      My card was sent back to the vendor’s headoffice for testing, the store I was dealing with didn’t know whether any RMA stock had been kept as a buffer or not. Although frankly I wouldn’t object if they switched brand on me, I’d happily take an ASUS TUF OC instead for instance even though it was slightly cheaper than this one.

      Reply

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: