[UPDATED] Reassuringly expensive Palit GTX 1080 gloriously kicks the bucket!

[alucasa]

Is adaptive default? Because I've been seeing people complaining about why their GPU doesn't max out always (well, duh) during gaming.

 

[qubit]

This is Awesome news !!! Finger's crossed for a trouble free card of BaddAzz proportions

Thanks revin.

Is adaptive default? Because I've been seeing people complaining about why their GPU doesn't max out always (well, duh) during gaming.

It is, yes and is likely the reason. As I said to EarthDog a few posts back, the framerate difference was quite noticeable on my older cards, but doesn't seem as pronounced on the 1080. By noticeable I mean 20% or more in some situations. I suspect that the vasly superior power characteristics of Pascal have something to do with it.

You can change that default to Max Performance globally. However, the downside to it is that web pages with videos and animations will trigger 3D clocks, causing the card to consume more power and run hotter. Crucially for me, was all that extra noise for nothing when this happened, which is why I leave it at a global default of Adaptive, but set it to Max Performance for each game.

 

[alucasa]

Meh, 20% less frame rate shouldn't matter too much on Wash dogs.

/joke

 

[Ferrum Master]

Meh, 20% less frame rate shouldn't matter too much on Wash dogs.

/joke

But the shampoo?

 

[alucasa]

But the shampoo?

And the rapidly moving tongue as well. That hit me after I made the post. Hmmm ...

Maybe, it matters more than I initially thought. Darn, Wash Dogs.

 

[qubit]

Meh, 20% less frame rate shouldn't matter too much on Wash dogs.

/joke

Perfomance loss here is not a problem..

But the shampoo?

But I'm not so sure about here. Ferrum, you've got a point.

 

[EarthDog]

Thanks revin.


It is, yes and is likely the reason. As I said to EarthDog a few posts back, the framerate difference was quite noticeable on my older cards, but doesn't seem as pronounced on the 1080. By noticeable I mean 20% or more in some situations. I suspect that the vasly superior power characteristics of Pascal have something to do with it.

You can change that default to Max Performance globally. However, the downside to it is that web pages with videos and animations will trigger 3D clocks, causing the card to consume more power and run hotter. Crucially for me, was all that extra noise for nothing when this happened, which is why I leave it at a global default of Adaptive, but set it to Max Performance for each game.

It should max out regardless... when I review GPUs, I leave it all on default and I sure hit max boost...

 

[qubit]

It should max out regardless... when I review GPUs, I leave it all on default and I sure hit max boost...

Don't take my word for it, have a play with Max Performance and see the difference for yourself, you might be surprised and I'm curious for your feedback. Even the tooltip description describes it as giving more performance, especially with older generations. It was quite noticeable on my 580, 590 and 780 Ti.

 

[arbiter]

Yea my gtx1080 decided to kick the bucket on Saturday, and just shipped it out for RMA today ;/

 

[EarthDog]

Don't take my word for it, have a play with Max Performance and see the difference for yourself, you might be surprised and I'm curious for your feedback. Even the tooltip description describes it as giving more performance., especially with older generations. It was quite noticeable on my 580, 590 and 780 Ti.

The tooltip only mentions its in its maximum performance state when 3D apps are running... it doesn't say it gets more performance, just that it is in the highest performance state on the card. It doesn't mention a thing about old cards either...

Adaptive specifically says it determines the proper performance state (which could be MAXIMUM!!!) based on GPU use... it goes on to say, " This mode helps save power when 3D applications do not demand the card's maximum capabilities".. that doesn't say anything for when it needs the max capabilities. So, I can see this behaving as you said on minecraft type things. I honestly don't have an 'easy' game in the stable to test... maybe ME: Shadow of Mordor as it gets over 100 FPS at our settings. But that isn't installed on the daily driver, just the review machine which doesn't have an NVIDIA GPU in there (and won't.. not putting a r7 250 into my daily driver for this testing, LOLOL!).

Anyway, some quick 3D testing for you in 3D TimeSpy. Notice the scores are all the same (within margin of error EASILY)? Notice the clocks are the same? GPU use the same? I think I missed power limit, but, I don't doubt it saves power when it can...

Note, this is on 6950X @ 4.2GHz and GTX 1070 Quick Silver...

I ran Optimal (default), Max, and then Adaptive:







I ran a much, MUCH easier test (getting almost 300 FPS) and the story remains the same all around...





You can see in the Physics tests (where its using very little GPU) in adaptive mode, the clocks go down a lot further than with Max perf...notice the GPU use drops, then the clocks do. But so long as its high like that, it will stay at amx boost bin (controlled by the card).

I am wondering what games won't have it boost up there except for something like minecraft or something of the like.

I would like to see your empirical testing in games when you get your card back.

PS - Done editing.. going to bed!

PPS - My update was to correct typos and add a mention of the old cards in the NVCP passage....

 

[Kanan]

Confirming, "Adaptive" works like max. performance mode in games, I have a 780 Ti and had always maximum clocks and performance.

btw. good luck with your replacement card.

 

[mobiuus]

dude i feel u and wish u best of luck and getting a new card A.S.A.P.!!!!

 

[CAPSLOCKSTUCK]

I hope you get it back quicker than the feeble effort Samsung has made with my daughters missing phone ( which they lost).

Returned by me on 28th November.....still no replacement even though they claim to have sent it last week.........and no tracking number available from them........ W *****S !!!!

 

[EarthDog]

@qubit ... I finally found something that makes it behave as you said it does.

I tested Dirt: Rally at 1080p, Ultra Low settings and no AA. I was averaging over 470 FPS with GPU use hovering around the mid 30s to mid 40s... I saw a drop, eventually, then it came back up. So we are seemingly on the edge of when it clocks down around that usage point.




I can't really see this happening in many normal circumstances at all though. Perhaps if you "trog" your system and gimp the hell out of it by limiting FPS you can manage to see this in more normal settings?

And here we are at medium settings, no AA... no drops...




Last but not least. Here is D:R at 1080p Ultra + 4xMSAA both adaptive and max... Not even 1 FPS difference.. clocks never budged outside of the normal temp bin drop. Adaptive, then max:






Anyway, I digress... this thread isn't really about adaptive and its effect on clocks (or lack thereof to the contrary of what was said)... but I think I have at least provided something to think about with this, albeit limited, testing. After "seeing for myself", it seems to me in a 'normal' situation one will still see full clocks and no FPS drops in gaming so long as they are using reasonable settings. If you play abhorrently low settings on a higher end card causing GPU use to go WAY down, perhaps you may have a point. What happened back then, I have no idea, but you are likely on to something with how the 580/590 behaved considering I don't think they had boost until Fermi (6 series), it was just idle/2D/3D clock domains. But ever since boost came about, which includes your former 780Ti, they behaved this way for all intents and purposes.

EDIT: Here is a good read for you. Pay close attention to what Jacob from EVGA says in posts and FX84 or whatever: http://forums.evga.com/Power-Management-Mode-Adaptive-or-Prefer-Maximum-Performance-m1562329.aspx

 

[thesmokingman]

I suspect we may have had a possible misunderstanding.

Are you saying that the driver itself killed the card? Also, I always installed the latest WHQL driver (never a beta or hotfix) the moment it came out, so if the driver, then several versions would have had the same fault in that three month period.
I agree that switching to 3D mode / clocks appeared to be the trigger. To be honest, I can't remember all the symptoms I saw now and all the situations, which weren't all that often, but I don't think it ever bombed out when playing a game. This might just be because I run the 3D power setting in Adaptive mode by default which saves on power (and noise on older cards that weren't so quiet) and have it switch to Max Performance for each game.

It still seems to me like the core reason was faulty hardware though. Do you have a link to a news article or forum post about this? I'm quite curious to learn more about this now.

Everyone: and in other news, Scan have emailed me to say that the card has been tested and a replacement is on the way, so it shouldn't be long now. Seeing as they only received the card yesterday, this is a really fast response and great service so far.

Sorry, been busy installing a tankless water heater. Man, that was a serious learning curve.

Mmm, I don't know if I'd say the driver killed it but it certainly didn't help it twitching all over the place everytime hw acceleration was called on the desktop. What happens is when an app makes a call for hw acceleration, the gpu switches power states, and *sometimes* it happens too fast and the card is not able to adjust voltage for the power state. Think of it as accidentally redlining your car's engine. Now normally this isn't that big of a deal but sustained blips into the redline can cause damage on some cards. And some cards are more tolerant of this than others. The links I posted above link to Mozilla and AMD's take on the issue. They basically blame the other party lol. I would assume Nvidia would take the same or similar stance.

Here's an example of hw aceleration flicker. If you google it you'll see many more examples ranging from black screens, to completely garbled screens and it can really run the gamut.

https://forums.geforce.com/default/topic/768365/gpu-screen-corruption-dvi/

This used to plague AMD for years, that's why I called that guy a troll. Hell, I've even authored a guide to combat it. Everyone who knows stuff knows this.

 

[qubit]

@EarthDog & @thesmokingman thanks for the replies guys.

I'll get back to you tomorrow* as I've had a long day and it's now gone 3am and I'm knackered and I want to give you an intelligent** response.

*Later today to be technically accurate as the date won't change, lol.

**Sorry, "intelligent".

 

[eidairaman1]

I hope you get it back quicker than the feeble effort Samsung has made with my daughters missing phone ( which they lost).

Returned by me on 28th November.....still no replacement even though they claim to have sent it last week.........and no tracking number available from them........ W *****S !!!!

Call them

 

[CAPSLOCKSTUCK]

Call them

i have spoken to them 8 times on the phone.

Here is the webchats and emails.

 

[KrisCo]

I had a similar prob many years back with a brand new 9600 pro. 2D worked fine, second a 3d application started up it shit the bed. Sadly that was during the time I was not as enlightened with RMA policies , cussed myself out for not getting the 9800 pro and ended up getting that.

My 5870 was plagued with issues when I first got it, grey screens non-stop (learned my lesson about buying flagship on launch date).

Sucks to have a nice gpu like that kick the bucket , but alas mass produced items will inevitably have a few duds in them. I'm not sure if things are diff for the rest of the world , but imo MSI has the best RMA and cust service out there. Or at least my experiences have been nothing but great.

 

[CAPSLOCKSTUCK]

finally............55 days

 

[revin]

the downside to it is that web pages with videos and animations will trigger 3D clocks, causing the card to consume more power and run hotter

That's a deal breaker !! Had that happen to my 5870 1-2 times, I usually just used 10.x or 12.x since I don't have major AAA games. But you can imagine how sweet that blower sounded at 2AM watching YouTube !

iirc: many years ago you all helped me discover that a GTX275 [may have been a counterfeit], but for sure was totally whacked out on 3D clocks, but was similar to what happened with yours.

 

[BiggieShady]

I finally found something that makes it behave as you said it does.

It happens when the GPU is overkill for a game or game settings (bottleneck is elsewhere and gimping the GPU results in exactly the same frame rate).
It happens even during some demanding games while in game menus (graphics card gets cool and quiet both because of lower boost bin and half of the gpu being unused and power gated)

Dirt series serves as a good example in this case since it's not very graphically demanding so GPU has hefty headroom and when in adaptive mode it can choose: should it use less cores at high clocks or should it use all cores at reduced clock. Of course, it chooses a sweet spot to keep the maximum achievable frame rate at least power usage possible.
If GPU consistently has unused streaming modules during the game, best thing for GPU is to suspend those parts and boost the clocks to fill the power envelope.

Wait I had a point ... ah, adaptive mode is exactly the same as max mode for 99% of situations. Usually there is enough parallelism in a single frame to occupy a complete width of a GPU, so only clocks remain to be controlled (maxed until power limit).

So, why does Max performance setting even exist in control panel ... because of remaining 1% for two reasons, (1) adaptive mode is bugged for some games, resulting in combo that's never allowed - low fps at low power, and (2) sometimes game runs right on the edge of two GPU power states (or even worse has wild GPU usage variations) so the adaptive mode constantly switches states around causing micro stuttering and you just need the GPU to hold a fixed state regardless of usage.
Interestingly enough boost algorithm works the same in both modes. It's not very well documented, but it seems to me as if max perf. setting only disables power gating for streaming modules.

 

[qubit]

Ok, so I received a replacement card really quickly: Scan received my faulty one on Monday and on Thursday morning (yesterday 19th) I have the replacement - that's a 4 day end-to-end turnaround time. 10/10 fantastic service and happy to recommend Scan to anyone. I should say that this isn't the first time that I've had good customer service from them either, so it's all good from a support point of view.

I compared the serial number to my old one and thankfully they're not the same. Surprisingly though, it's just one higher than my old one, eg old 314, new 315 (not the real numbers). It's such an incredible coincidence that I wonder if it's really a coincidence at all. Anyway, DPD delivered it, where it was wrapped in lots of bubblewrap and still sealed in its original antistatic bag, looking brand new.

I have now installed it in my PC and after some faffing around my verdict is: flaky. Yup, not "fixed" or "great" or "all good", but fucking flaky.

Turns out the card is really sensitive to seating in the PCIe slot as it wouldn't start properly most of the time, no matter how carefully and gently I fitted it. Motherboard gave the power good beep, then 4 beeps which had a different tone and appeared to be coming from the card, like this: deeeeet-dit-dit-dit. This was followed by no picture. Not even a background signal like the other one, there's simply no video signal.

Tried it in the second PCIe slot and got the same thing.

Refitted it a few times in both slots (checked slots, no dust or visible damage) and sometimes it would start properly, giving a BIOS picture. Seemed kinda random, with it most of the time not working. Then, I tried gently pushing up the corner of the card near the PCIe power connectors (power still on) pressed reset and the thing booted properly. It was ok after that, even after removing my finger and wiggling it gently a bit (power off when wiggling). Removed the card, refitted it and got the same beeping until I did the lifting again.

I've never seen anything like it. I've seen cards that are a bit fussy (hello GTX 285) but not to this extent. I visually inspected the PCIe connector on the card for damage and all around the card, but there's nothing to see. It's even sensitive to just unplugging the PCIe cables and reinserting them, forcing me to lift it again. I tried the other set of PCIe cables and got the same thing, ruling out bad cables. I tried googling for the beeping, but didn't find anything conclusive.

At the same time, I tried refitting my 780 Ti a few times and it was perfect every time, in both slots, so I think this card isn't quite right. My old GTX 1080 wasn't at all sensitive either and I could install it with wild abandon just like the 780 Ti.

This isn't the kind of situation I want, as the shop would simply return it as no fault found if it got a bit worse. A hard fault like the previous one is where you wanna be if you want to guarantee that the shop will actually replace your faulty item.

I then ran a couple of games just fine and even Furmark (don't shoot me!) where it reached a modest 73c, running so gloriously silently. Note I'm not overclocking it and don't plan to in the near future, but I might one day.

I wrote this post with the GTX 1080 installed and working fine - for now. Frankly, I don't trust it.

 

[eidairaman1]

Ok, so I received a replacement card really quickly: Scan received my faulty one on Monday and on Thursday morning (yesterday 19th) I have the replacement - that's a 4 day end-to-end turnaround time. 10/10 fantastic service and happy to recommend Scan to anyone. I should say that this isn't the first time that I've had good customer service from them either, so it's all good from a support point of view.

I compared the serial number to my old one and thankfully they're not the same. Surprisingly though, it's just one higher than my old one, eg old 314, new 315 (not the real numbers). It's such an incredible coincidence that I wonder if it's really a coincidence at all. Anyway, DPD delivered it, where it was wrapped in lots of bubblewrap and still sealed in its original antistatic bag, looking brand new.

I have now installed it in my PC and after some faffing around my verdict is: flaky. Yup, not "fixed" or "great" or "all good", but fucking flaky.

Turns out the card is really sensitive to seating in the PCIe slot as it wouldn't start properly most of the time, no matter how carefully and gently I fitted it. Motherboard gave the power good beep, then 4 beeps which had a different tone and appeared to be coming from the card, like this: deeeeet-dit-dit-dit. This was followed by no picture. Not even a background signal like the other one, there's simply no video signal.

Tried it in the second PCIe slot and got the same thing.

Refitted it a few times in both slots (checked slots, no dust or visible damage) and sometimes it would start properly, giving a BIOS picture. Seemed kinda random, with it most of the time not working. Then, I tried gently pushing up the corner of the card near the PCIe power connectors (power still on) pressed reset and the thing booted properly. It was ok after that, even after removing my finger and wiggling it gently a bit (power off when wiggling). Removed the card, refitted it and got the same beeping until I did the lifting again.

I've never seen anything like it. I've seen cards that are a bit fussy (hello GTX 285) but not to this extent. I visually inspected the PCIe connector on the card for damage and all around the card, but there's nothing to see. It's even sensitive to just unplugging the PCIe cables and reinserting them, forcing me to lift it again. I tried the other set of PCIe cables and got the same thing, ruling out bad cables. I tried googling for the beeping, but didn't find anything conclusive.

At the same time, I tried refitting my 780 Ti a few times and it was perfect every time, in both slots, so I think this card isn't quite right. My old GTX 1080 wasn't at all sensitive either and I could install it with wild abandon just like the 780 Ti.

This isn't the kind of situation I want, as the shop would simply return it as no fault found if it got a bit worse. A hard fault like the previous one is where you wanna be if you want to guarantee that the shop will actually replace your faulty item.

I then ran a couple of games just fine and even Furmark (don't shoot me!) where it reached a modest 73c, running so gloriously silently. Note I'm not overclocking it and don't plan to in the near future, but I might one day.

I wrote this post with the GTX 1080 installed and working fine - for now. Frankly, I don't trust it.

do you think if the case shifts it could cause an issue.

I'd say next time switch to a different maker.

Perhaps bend the PCI bracket or have that motherboard reseat...

this is unbelievable honestly

 

[qubit]

do you think if the case shifts it could cause an issue.

I'd say next time switch to a different maker.

Perhaps bend the PCI bracket or have that motherboard reseat...

this is unbelievable honestly

No eidy, I don't believe you'd ever be honest.

There's nothing to bend, really. Unfortunately cards are only supported at one end, so when they get a bit long and heavy the sagging doesn't help. It does look like it's sitting straight though and it has a backplate.

I don't think the mobo is out of alignment and frankly I'd rather not touch it as everything works apart from this one card. It's been installed there since 2011 so it's getting on a bit and you know that phrase about if it ain't broke don't fix it, lol.