Looking for Suggestions and help for a Gigabyte GPU RMA

[DarkRogue]

Hey guys,

I'm looking for some help or advice on an RMA to Gigabyte I'm currently going through. I heard some horror stories, but I had to cross my fingers and hope regardless, and it seems like it may actually be problematic after all.

I sent in a Gigabyte GTX670 a couple weeks ago because it would crash anytime I tried to play any real games with it. I tried many driver versions, and it would always produce the same issue - if I ever loaded the GPU, it would crash, my display would blank in and out, and nvidia's drivers would tell me that they crashed and restarted. It didn't matter what game it was. If I enabled shadows, or AA/AF, or PhysX, or anything beyond low texture quality in most games, it would consistently crash after a half hour or so (sometimes more quickly, sometimes more slowly.)

It got to the point where pretty much anything was unplayable without constant crashing and resetting of my display. Even Path of Exile, a game I don't consider all that demanding as, say, Payday 2 or Borderlands 2, would crash if I didn't limit its FPS to 60 via the vsync option. And even then it still crashed, just less often.

I cleaned out the fans, etc. and monitored its temps via GPU-Z-- it never went above 70C. FWIW my resolution is 1920x1200.

Last week I got fed up with it because basically nothing was playable. I sent it in for an RMA, and am currently back on my 7 year old 8800GT. It struggles to run Path of Exile (and as a result I haven't attempted anything more demanding,) but I have yet to see a single crash with it, which does tell me that there were problems with the GTX670.

Here's the problem... Gigabyte received the card, and I've been monitoring the RMA page for updates; Today, I see this:

No trouble found after tests with multiple configurations

...Seriously?

What do I do? If they send me that same card back without doing anything, I'd have wasted $22 out of pocket sending them a clearly defective card.

Any ideas on how to get them to cooperate?

 

[DigDog]

where did you buy it? try to return it to the store, they get the RMA done much easier than the official channels.

 

[DarkRogue]

I bought it off Newegg, but it's been almost three years, so I doubt Newegg will be much help in this regard, which is why I went to pull an RMA with Gigabyte instead.

 

[LTC8K6]

Could it have been your power supply?

 

[DarkRogue]

I doubt it, but it's a possibility that I can't test. The PSU is about 5 years old, and as much faith as I have in Seasonic making reliable products, that's still a decent few years. Unfortunately, I simply don't have any other unit of similar capacity available to test with. As well, having swapped back temporarily to my 8800GT, everything seems fine. I may only get 20 fps with this card in these games, but at least it's stable.

 

[boozzer]

I sent in a Gigabyte GTX670 a couple weeks ago because it would crash anytime I tried to play any real games with it. I tried many driver versions, and it would always produce the same issue - if I ever loaded the GPU, it would crash, my display would blank in and out, and nvidia's drivers would tell me that they crashed and restarted. It didn't matter what game it was. If I enabled shadows, or AA/AF, or PhysX, or anything beyond low texture quality in most games, it would consistently crash after a half hour or so (sometimes more quickly, sometimes more slowly.)

how did this happen though? did you install new drivers? new os? my old 460 1gb was doing the exact same thing to me because I updated drivers. everything worked fine when I reverted back to old drivers. from that point on, it will do exactly what you described with every new driver release.

find out the cause, maybe you can reverse it. need to know how your gpu started behaving like it is broken.

 

[DarkRogue]

I don't know how it happened - to be honest, it probably was always like this.

In the beginning, I probably thought it was something with driver instability, or game conflicts, or whatever. Each time new drivers were released, I updated to it to try and see if there was any improvement -- there were none. This went on for probably 5 or 6 driver versions, or however many they've released in the past three years. Over time I found out that lowering graphical settings in most games reduced (but did not eliminate) the symptoms.

Basically, as far as I can recall, it's always done this, but back then it wasn't as frequent, so I decided to live with it for a little while and see if I couldn't figure out a way to resolve it.

Unfortunately, I was never able to other than figuring out that putting load on the card was a reliable way to get it to crash. If anyone wonders why I didn't RMA it earlier, it's because I couldn't afford the downtime of my only PC until now, and it is still within the warranty period anyway, so I didn't think it mattered when I RMA'd the card, as long as it was within the warranty period.

In the last few weeks, though, it's started degrading and crashing far more frequently to the point where I just got fed up with it.

 

[boozzer]

I don't know how it happened - to be honest, it probably was always like this.

In the beginning, I probably thought it was something with driver instability, or game conflicts, or whatever. Each time new drivers were released, I updated to it to try and see if there was any improvement -- there were none. This went on for probably 5 or 6 driver versions, or however many they've released in the past three years. Over time I found out that lowering graphical settings in most games reduced (but did not eliminate) the symptoms.

Basically, as far as I can recall, it's always done this, but back then it wasn't as frequent, so I decided to live with it for a little while and see if I couldn't figure out a way to resolve it.

Unfortunately, I was never able to other than figuring out that putting load on the card was a reliable way to get it to crash. If anyone wonders why I didn't RMA it earlier, it's because I couldn't afford the downtime of my only PC until now, and it is still within the warranty period anyway, so I didn't think it mattered when I RMA'd the card, as long as it was within the warranty period.

In the last few weeks, though, it's started degrading and crashing far more frequently to the point where I just got fed up with it.

5 - 6 driver versions is about few years. it could very well be around the same time when my 460 1gb started having the exact problem almost 3 years ago :\ I am surprise you can suffer through the constant driver crashes.

figure out when you started crashing. get the driver before that. it is what solved the problem for me. If your 670 was crashing like this the moment you bought it and you have never bother to get a new one/rma/return, I dunno how to help you.

 

[DarkRogue]

Well, I really don't think drivers are the cause of my problems, but even if I wanted to check that again, I can't because the card is at Gigabyte's service center.

As well, there's quite a conflict here when I need to look through older drivers to find a solution, while the common suggestion is to always use the latest drivers to fix issues.

I dealt with the crashes because I couldn't afford the downtime -- I can't exactly put work on hold because I need to RMA a part, so it was pushed back until now when I have some time between projects.

I just sent an email over to the Gigabyte USA 'services' address, so hopefully they'll try to help me out. For the time being, I'm stuck.

 

[boozzer]

yea I know it is weird that the solution is to use an old ass driver which is why I gave up on my 460 1gb and pass it on after a while.

I have no idea if that is reason enough for a rma. giga can easily deflect it as a driver problem cause by NV

when you do get it back if giga refuse to do anything for you, try older drivers. ones before support for 700 series, something around 320.XX or even 314.xx.

 

[sam_816]

Don't have any experience in rma-ing gpus but u can ask gigabyte for drivers they r using to test it. Maybe ask for the whole system configuration. Though your problem seems to b related to drivers but there should have been some changes with different drivers.. In case gigabyte refuses to share any info I think u'd be better off testing other variables starting with different cables, psu, ram etc etc. Or run ur card in friend's pc.

 

[Mushkins]

Honestly, this sounds like a PSU issue to me. When the card is under load, it draws a whole lot more power. If the PSU cant supply it, you get exactly what you described. Especially if it got slowly worse over time.

 

[amenx]

Dont you have another PC to try it in? Thats the first thing I would do before sending anything for RMA.

 

[sam_816]

How about rma the psu?

 

[DarkRogue]

when you do get it back if giga refuse to do anything for you, try older drivers. ones before support for 700 series, something around 320.XX or even 314.xx.

Well, technically I will be running some older drivers whenever I get it back; My 8800GT is apparently too old and was no longer supported by the version of the drivers I had been using on the GTX670, so for now I am running it with the 341.14 drivers. I have no idea how much 'older' I am willing to go, though, especially when it feels to me like it isn't really a driver issue. I used to think it was, but having upgraded through so many versions without any resolution, I'm doubtful.

Don't have any experience in rma-ing gpus but u can ask gigabyte for drivers they r using to test it. Maybe ask for the whole system configuration. Though your problem seems to b related to drivers but there should have been some changes with different drivers.. In case gigabyte refuses to share any info I think u'd be better off testing other variables starting with different cables, psu, ram etc etc. Or run ur card in friend's pc.

As I noted above, I highly doubt it is a driver issue, but we'll see if the 341.14 drivers do anything when I get the card back. As far as cables and stuff go, there's not exactly very many cables running to the GPU to begin with; I'm not running SLI, or anything. I don't have another PSU nor set of RAM available to test with, either.

Dont you have another PC to try it in? Thats the first thing I would do before sending anything for RMA.

I do not. This was the primary reason that I dealt with the issue for a while until now, because I don't have another machine to use, and could not afford the downtime then.

How about rma the psu?

I technically could, since I believe Seasonic has a 7 year warranty period on these units, but in my eyes, that's even more of a stretch than RMA'ing the GPU - at least with the GPU I have some suspicion as to it being faulty. I have zero evidence right now that tells me it would be a PSU issue, and it would be pointless to send in a PSU for service "just because" when it doesn't actually have any visible issues.

Honestly, this sounds like a PSU issue to me. When the card is under load, it draws a whole lot more power. If the PSU cant supply it, you get exactly what you described. Especially if it got slowly worse over time.

I am aware of that, and is why I chose my power supply appropriately for this build. 650W is more than enough for any single-GPU setup. It should potentially even do SLI, but in my opinion that might start to push it, so I won't be doing so. I do not subscribe to the idea that we must always run outrageously oversized 1KW+ PSU's for any system build regardless of components used.

Please do explain the logic though, because while it sounds like I'm getting a few responses pointing their fingers at the PSU, I'm not seeing it. For one, if the system at load runs into insufficient power, how and why is the graphical subsystem the only system to be affected in this manner repeatedly? One would imagine that if there was a power shortage, every system would feel the effects simultaneously.

Secondly, even if we assume the above is the case, we have to speak about the power drawn at load. Load should theoretically then be at maximum / exceeding maximum at the point where it resets -- that is, whatever scene the GPU is currently drawing. However, as long as it continues to draw that scene, it should mean that the load stays the same, does it not? In this case, this doesn't explain why when it resets like this, even when I do not touch any controls in the game and leave the scene exactly the same, I can get it to temporarily stop resetting by shifting to a 2D workload for a few seconds (Windows desktop) before tabbing back to the same scene, which should bring the load right back to where it was before, yet it won't crash again until some time later.

Finally, in terms of capacity, I highly doubt that I am running into power limits imposed by the PSU.
Again, my system runs off of a 650W PSU, which should be more than enough for a single GPU system. However, my CPU is indeed OC'd, and should therefore consume more power.

However, compare with another system my friend uses - he runs an original i7-920, OC'd by 1GHz or so. The original Nehalem was a 130W TDP part, whereas my Ivy Bridge is 'only' a 77W TDP part. I cannot find direct comparisons because they are several generations apart, but as far as I can tell from reviews, Nehalem at those speeds consumes just as much, if not more, power than my Ivy Bridge. In addition to this, he is running an eVGA GTX670 (one that I sold him a while ago when I decided to get Gigabyte's version for its Windforce cooler.) He runs games at the same resolution as I do - 1920x1200 - and at maximum graphical quality. All of this... on a 550W PSU; 100 watts lower capacity than mine! It all works fine.

To add to that, he even tested his system by running Prime95 and Furmark simultaneously without any crashes or issues. This should theoretically bring his system to maximum power load, and even in this case, his 550W PSU was sufficient to handle that. Granted, he could have pushed the power consumption slightly higher by running linpack, but then again, what game even stresses the system to the point where it'd be comparable to linpack?

This leads me to believe that capacity is not my issue here. If we argue that the PSU is degrading and not providing as much capacity as it should be, well, I have to refer back to my earlier issues, specifically why only a single subsystem experiences issues, when the PSU needs to be supply that power to every system at the same time. Even if it was degrading, I can't figure out how to test this without a spare PSU or PC available. If it isn't, and is fine, then it would be an extreme waste to send a perfectly good PSU to Seasonic for service just for them to turn around and say that there's nothing wrong with it.

I should also add that swapping a PSU is not an easy thing to do. In my opinion, the PSU is probably second only to the motherboard in terms of how annoying it is to swap in and out. (I have an Antec P182, so working in it is...tight.)

Having said all that, I have arranged with my friend, whom I sold my eVGA GTX670 to, to let me borrow his card to test with since he has said he's experienced zero issues with that card. The only problem is that it will take a couple days for us to arrange the time to swap it over. As well, whenever I get my card back, I'll probably bring it over to his PC and test it there as well.

In either case, Gigabyte finally did just get back to me with this:

Dear Customer,

Our engineers have reflashed the firmware, reapplied thermal compound, then cleaned the gold fingers.
The card is currently being burn-in tested, should be ready for shipping soon.

Best Regards,

Sounds like they didn't really do all that much. I don't know if the 'burn-in' process they're doing is standard, or if they actually did read my email and throw it on there to re-test for issues.

Bleh. I'll be happy if it just works.