Ok, here's the wall of text you've all been waiting on tenterhooks for.
@EarthDog and
@thesmokingman thanks for your patience guys! Didn't have a chance to look at your posts until now, late into the wee hours of Saturday here in sunny Blighty. Luckily I'm in full night owl mode so can properly digest them and pay the price tomorrow for not having slept and a messed up body clock.
EarthDog: That's interesting, so it looks like if the GPU is fairly heavily loaded ie framerate not too high, then it doesn't make much difference. I'll be curious to have a look at this again with my games. I first noticed it when I wasn't even looking for it. When I was using the GTX 780 Ti, I was playing DX8 classic Unreal Tournament 2004 where the card / system could render anywhere from a minimum of around 250fps to around a stupidly high 800fps if looking at the floor or similar. This meant with my 144Hz monitor and vsync on (regular flavour, none of this fancy stuff NVIDIA brought out later) I would get a solid 144fps without frame drops at all for the vast majority of the time, with correspondingly jaw droppingly smooth animation. Proper nerdgasm stuff for someone like me, hehe. However, this was only true with Max Performance. Set it to Adaptive and the fps would often drop to around 90-100fps or so, which was quite visible and pretty annoying with the stutters and increased lag. I found this to be true with the 580 and 590 (with different performances due to model differences, of course. I think a 590 is roughly as fast as a 780 Ti where video memory isn't full and SLI is working well, but don't quote me lol).
I hope to get around to repeating your tests with Metro 2033's built in benchmark and perhaps another game or two and get back to you about it, since that game loads the cards quite significantly. Note that I don't have Dirt Rally, so I can't do a direct comparison. Also, I suspect that this setting will matter more for some games than others and perhaps with DX version.
I read that forum thread, which looks inconclusive to me.
thesmokingman: That video is from 2012 and I remember the AMD cards of the time had a big problem with clock switching - dealbreaking for me. I seem to remember that this was a design fault that was only finally completely cured in the next gen cards, despite a BIOS update or two to address the problem. Why am I not surprised that AMD and NVDIA are finger pointing each other?
I don't ever remember seeing my NVIDIA cards do this or have any problems with switching clocks frequently. Perhaps my card already had a weakness there and the switching caused it to go over the edge? If true, it's certainly consistent with what I saw - died on starting a YouTube video which triggers the 3D clocks. I think 3D clocks can trigger even when set to Adaptive in some cases, which I do for desktop use.
I do love the delicious irony of an AMD user posting in the NVIDIA forum because they can't sign up on the AMD forum. If I'd had problems like that (clock glitches) with my AMD card I might think twice about getting another one, too. He doesn't say anything about RMAing that card though, as it could just be one bad card, like mine. He really should have done.
Hope that water heater install went well.
@Vario No, it was a brand new card in a sealed antistatic bag. I think Scan gave it to me in good faith and it's just another bad card. So far, Scan haven't done anything to make me suspect foul play and believe me, it doesn't take much to make me suspect this when it comes to warranty given some of the experiences I've had. They can only be bettered by Amazon, who will just replace kit on the customer's say-so no questions asked. Scan did ask me to do a little further checking to try it on another PC or with another PSU, not unreasonably.
@eidairaman1 I'm very much beginning to think this, with a heavy heart as the performance of this model is simply amazing, like the TPU review says. It's very fast and simply never, ever makes any significant noise, which is so important to me. I even had the framerate on this replacement hit 1700fps yesterday on some old game and I could barely hear any coil whine at all, which is simply astounding. I didn't keep it there for long, of course. I might give it one more go, "third time lucky". If that's bad too, then it will be no question of buying a different brand, likely MSI. Thing is, the coil whine on my MSI Gaming 780 Ti (both of them) was quite noticeable and annoying, ironically made worse by the quiet fans which don't mask it.
@newtekie1 Yeah, it could well be that. There's certainly a bad connection somewhere, that's for sure.
EVERYONE I have another update about this replacement. I got the card "working fine" this evening. In particular, it seemed to sit that little bit better/tighter in the slot and this time didn't crap out when I removed the PCIe power cables and reinserted them. Played a quick game, all stable.
Went out for the evening and put the PC into Sleep mode like I do all the time, reliably, even with my 780 Ti SLI setup previously. Came home, pressed the space bar, PC started up, that damned infernal 4 beeps from the graphics card and a locked up Windows (numlock stuck on) with no picture. Had to do the "wiggling" to get it going, where it's now worked fine for about 3.5 hours as I work on the PC.
This now convinces me without a shadow of a doubt that it's not just poor registration in the PCIe connector, but a faulty card like I thought from the start. I will therefore be definitely raising another RMA with Scan, requesting a refund this time. I doubt I'll get it as it's been three months before I reported a fault with it, but I might get a credit note, which would be acceptable. I could sit on it too (use one 780 Ti) and see if the 1080 Ti is worth buying when it comes out in March, depending on its price premium and performance.
Only thing stopping me from doing it right now after writing this post is the recovery of the cost of shipping it to them, which was significant at £26 by DHL, including £12 insurance. If I remember correctly, under the new Consumer Contracts Regulations for products sold mail order or over the internet, I'm entitled to have my shipping costs reimbursed for a faulty product. I want to be sure of my rights and quote the relevant bit of the legislation when making my request, to discourage them from trying to wriggle out of their obligations. I tried Googling for it, but couldn't find it, losing myself in a sea of confusing text, so I'd be really grateful if someone could please help me out with this and post a link to the relevant bit.
Only pulling the power connectors and / or removing the card and refitting it does. Note that when I pull the power connectors, I tried holding the card with my other hand to put the least stress on the PCIe connector, but it still failed. Had to do that little push up to get it going.
