Pidge from Nvidia has asked that user experiencing problems with the 880m list them here..

Ok ive been talking to an employee from Nvidia, his name is ManuelG and he is mod at on the Nvidia forums as well as one of there employees for the last 14 years.
but here is what he is requesting..

From ManuelG:
Here is what I need to do to move this bug over to devs
1.Setup a laptop that can reproduce the behavior the users are seeing
2.I would need description of the problem, notebook make and model, device ID, OS, driver version, basic steps to reproduce (i.e. run game x and then quit and then run game y)
3.Also if user is not on most recent SBIOS for laptop, mention which SBIOS version they are on
4.and last but not least, what apps they are running in the background that might affect GPU

I asked him him if user can list what other issues they are having and he said that would be fine.. But please Do what he asked first then list any issue you are having..
please include the system type and hardware.
Thanks.

 

They're just sending us on a chase for no reason.

Take any 880M SLI machine, load up Watch Dogs and set the AA to 8xMSAA and watch both cards throttle under 900MHz long before the temps reach the thermal throttle point of 93C or even 80C for that matter. They have the power restrictions too tight on the GPUs for them to perform at their rated speeds with a high GPU load level.

Sent from my HTC One_M8 using Tapatalk

 

I'm being extremely cynical here but I really don't believe there's any sincerity at all. Looks to me like they're realizing they have a potential public relations disaster on their hands and can no longer sit there and do nothing, so they respond with the typical "bare minimum" post that proves they're "doing something", but at the same time make end users jump through ridiculous hoops in that hopes we'll either get fed up, or we'll provide something that they (nVidia) can use as an excuse for why the cards are behaving the way they do.

And if this does indeed turn out to be the case, well I told you so.

I do hope I'm completely wrong though for the sake of all 880M owners.

 

That was exactly my thinking. It's sad.

Sent from my HTC One_M8 using Tapatalk

 

I've also been speaking with Manuel DumbDumb. He's sent my reports over to the QA team. He's recently purchased a AW 17 with the 880m. He's currently trying to replicate the problems we're experiencing.

 

I'm glad he's trying to replicate. Hopefully he finds something and can get it fixed.

 

Alright, let's hope something actually comes out of this.

 

When does this actually happen, as in has it been determined what the actual cause is of the flag being flipped? I'm about to pull the trigger on an AW17 with the 880M, but I don't want to waste my money if it's just gonna throttle right out the gate. I'm not big into overclocking so i'd probably be using it at stock clocks for the most part.

 

That "flag" is a different thing altogether. Apparently if the driver crashes one too many times (and too frequently, as is often the case when OCing), a flag gets written to the GPU that makes it perform even worse than Intel's iGPU. And this flag persists through reformats and reinstalls, so it's definitely a hardware flag.

 

So what are the actual issues being experienced with the 880M then? Would you recommend I get the AW17 now or look for another laptop that doesn't experience these issues (say with an 870M which I assume doesn't have any of the same issues)?

 

Thanks. Let me see if I can quickly test Watchdogs and see if I observe the same behavior.

Is this how it looks for you in Watch Dogs? Tested using Alienware 17 in my signature (single GPU, not SLI).

 

Since you're in here, I can do things directly.

My machine specs are in my signature. I'm running a stock vbios, however it is an updated one from the one that came with my machine.



When gaming, the slave card is *always* throttled. Here's a perfect example.

This is a 1 hour Bioshock run, as you can see, both cards throttled but neither hit its thermal limit.





First one is the primary, second is the secondary.

Now from what I just learned today, with SLI both cards are running at the lowest clock rate which means that my games are being throttled well below 954MHz.

I have a bunch more of these graphs on the forums that I can dig up for you, they paint the clearest picture of what these cards are doing.

 

Yeah very similar to that. I've never used that tool to monitor it, I make my own GPU-Z graphs but yeah, the clock goes all over the place.

Its the cards' built in power limit. High AA settings will push past the default setting and will bring the core speed crashing down. If that was raised, it wouldn't be an issue since the modded vbios does not exhibit that behavior.

And thank you for investigating this. Sorry about my comments earlier in the thread, its just that nVidia has pretty much ignored all of us angry owners until now.

Wait... you're throttling because of thermals... Hold on a sec, I'll show you what mine is doing without hitting 94C. Give me a few minutes, I lost all my original log files so I have to make a new one.

 

Thanks. I really appreciate the info. Let me check with QA and see if we have enough to pass on to dev team.

 

EDIT: I'm officially an idiot. My laptop was in power efficient mode. Wow. And that's usually the first thing I tell people to check. Tried again and with the driver downgrade things work great in C3. Need to try BF4 next.

I just tried BF4 and Crysis 3 - my GPU is getting throttled to hell. I just bought this machine as well - this is hugely disappointing.

I'm supposed to be able to get 60FPS on both BF4 and C3 at Medium settings at 1080. But I get nowhere near that. GPU utilisation keeps dropping to 48% in both cases. I haven't even gone above 82 degree celsius and it started happening.

I tried playing both at higher settings inititally and then dropped them down. Even when selecting low, my GPU utilisation tanked, despite the fact that the game was hitting 99% earlier. The FPS is all over the place. I'm seriously considering returning this unit as this was not what I expected at all.

This is on Windows 7 64-bit, with nvidia driver 337.88. Specs below.


Sager NP8278
120GB Samsung 840 EVO SATA III Solid State Drive
nVIDIA GeForce GTX 880M 256bit w/8.0GB GDDR5
8GB (1x8GB) DDR3L/1600 Memory
Intel® Core™ i7-4810MQ (2.8~3.8GHz) w/6M L3 Cache - 4 Cores - 8 Threads - SPECIAL SALE! $100 OFF

 

BTW, that graph is showing Bioshock Infinite right?

 

Yes that is Bioshock Infinite.

Here is Watch Dogs for you, just ran it.

This is the primary card



Now this is the secondary card



These are just the last few minutes of game play where it started to throttle for no real reason, here's the full time stretches

Primary:



Slave:



That just doesn't look like normal behavior to me...

337.88, Ultra, Ultra textures, 8xMSAA

As you can see, my temps were averaging 80-81C the whole time with automatic fans so something else caused the throttle, not the temps.

 

Since 3DMark is reproducible, I went ahead and ran that too.

3DMark 11 Performance mode, stretched.

Master:



Slave:



Single-GPU:



Here's Catzilla 1080p, both of them overlaid on top of each other



Both cards were below their 954MHz stock speed at the end of the test with the slave in the 800s.

 

I'm just gonna pop in and apologize to Pidge for the harsh comments earlier as well. Hope everything will work out in the end.

 

Don't worry, I'm sure ManuelG has thick skin to be able to do his job.

 

Is anyone else concerned about the fact that his card was at 94C? Lol

Sent from my HTC One_M8 using Tapatalk

 

It was probably a rush job and Alienware forgot to put thermal paste on it LOL.

 

Or its case in point that these cards have thermal issues that the 780M doesn't have as we all have been saying.

Sent from my HTC One_M8 using Tapatalk

 

FWIW when I tried running Valley benchmark with the 780M @ 954 MHz it crapped out after 2.5 minutes and reverted back to 849MHz probably due to insuffient voltage. But at no point did the temperature go above 80C.

 

Was that at 1v?

Sent from my HTC One_M8 using Tapatalk

 

Yep, with a peak of 1.012V.

 

Max or auto fans?

Sent from my HTC One_M8 using Tapatalk

 

Here's Valley, forced max fans.



I think I'm starting to see a pattern here...

Since the temperature is staying the same between the two cards, the second card is being throttled to maintain the same temps... The vbios must be set to aggressively throttle for temperature by default and my slave card is significantly hotter so it gets throttled more.

 

So I just ran Valley again, both cards @ 954 MHz. The run lasted exactly 2:32 before the cards reverted back to stock boost clock (849MHz). The driver did not report a crash or anything whatsoever. The benchmark simply paused and the screen turned black for a few seconds (but music kept playing) before both cards went back to stock speeds.

During that 2:32 duration the master GPU reached 69C, while the slave GPU was 65C with auto fans. The fan vents were enlarged and the slave heatsink was lapped, so temperatures are probably about 5-7C lower than would normally be.

System specs in sig.

 

is this problem just with the 880's or other 8 series too? like the 870?

 

Well I went ahead and flashed the modded vbios for comparison purposes.

Here is the result with the modded vbios. I didn't stop the benchmark until one card peaked at 87C. Max fans for this.



I confirmed by setting single GPU that its the slave that peaked at 87C, the master was the cooler of the two. There is almost a 10C temperature difference between them at max load.

It appears to be just limited the 880 but feel free to test.

 

Ugh! I hope a fix can be found for this. The modded vbios doesn't throttle to its 2D clock which keeps my fans going all the time and flashing back and forth between the stock and the mod is stupid.

I set the modded vbios to run with a temperature target of 87C and when I ran Valley, it hit 88C then dropped back to 87C and it did this for like 20 minutes but it didn't drop the core or the memory so what gives? How did it manage to run 993MHz the whole time? That confuses me. I think the stock vbios isn't working properly. Of course the mod doesn't work right either so the cards remain broken I guess.

Sent from my HTC One_M8 using Tapatalk

 

If it's a vbios problem, I guess it would affect every card? I've been happy because my new 880m machine is running things like Crysis 3 at 60FPS on High settings (1080), and doing much the same with Bioshock Infinite (DLC, nearly maxed out minus DX11 specific post-processing).

However, I keep MSIafterburner on my second screen and the GPU utilization graph is all over the shop. Is that normal?

What program are you using to monitor the clock speeds? I wouldn't mind giving it a shot myself to confirm the issue on non-sli machines (if there is one).

 

No, that's not normal.

You can use RTSS 64-bit v6.1.2 and HWiNFO64 to output your CPU and GPU statistics on screen as you game.

Here is an example...

 

Why does the secondary GPU (that throttles) lower the voltage as well as the core clock?
Almost like its trying to lower the temperature, or it have hit a power limit so its reducing the voltage to cut down on power consumption.

It looks a bit strange because the core clock doesnt seem to drop that far down, just too many times across a small time frame, so I dont understand why it reduce the voltage so much as well.

 

Part of the reason for that ridiculous behavior is that idiotic invention called GPU Boost. Using a modded vBIOS that kills GPU Boost helps eliminate that spastic GPU behavior. It never has worked as well as expected, especially on MXM modules. If it worked correctly the clock speeds would never drop below the 3D performance level base clock speeds when playing games or running benchmarks. Having "dynamic" or "adaptive" GPU clock speeds is a really stupid idea when you think about it. If the base clock speed is not enough, just overclock the thing. If the base clock speed is enough for a good gaming experience, there is no point in overclocking or having a clock speed "boost" when that is not required. It is only a marketing gimmick that was never executed well.

 

I really think that slave card is just defective. I put liquid ultra on it and it now has the same temps as the master but the clocks still drop. I'm going to start an RMA with Sager on Monday

Sent from my HTC One_M8 using Tapatalk

 

There's a reason it's known in some circles as Cripple Boost LOL.

 

I can't say this is the cause in your case but it is possible the second card is running slightly lower to sync with the first card. In SLI mode, our GPUs attempt to output the FPS at the same time to reduce micro-stutter introduced by two cards outputting FPS at different rates.

 

The master card is running faster than the slave though. When I disable SLI, the master card stays at or above 954MHz but in SLI, the slave drops as low as 888MHz bringing both down that low and way before the thermal throttle would be imposed unless there is something else going on. Have you heard anything about what's happening with these cards? I'm waiting to see the outcome here before I contact Sager.

Here's a pic of Bioshock Infinite after just a few minutes.



Here it is with SLI disabled.



Sent from my HTC One_M8 using Tapatalk

 

my laptop has just been collected and being returned to scan UK. Scan have said they no nothing of the issues we are describing.

 

Hi Pidge,

I see from your signature you have a gtx 880m in your Alienware M17, what are your experiences of what is being described in all the forum posts i have read over the last few days.

My Alienware m17 with gtx 880m is due to arrive in 3 days time, and it will be going straight back if i find this sort of under performance with such an expensive laptop. This simply shouldn't be happening and will only damage the Nvidia brand.

Do you have an update yet from Q&A on progress to fix this... Im not in the same league as some of the people (Mr Fox, Numbnumb) and quite often look for there thoughts before i make a purchase... because it was Nvidia i was buying i automatically presumed it would just work out of the box.... Im a new customer coming from your competitor... and so far it doesn't look good.

Please can you update us all on progress... i dont want to have to return the laptop as i don't believe Alienware are at fault, its down to nvidia to resolve this time, and so far there has been no response from Nvidia to even acknowledge the issue's many people are experiencing.

Looking forward to hearing some good news soon

Thanks

Ian

 

The chip still blows away it's competition, its also still faster than a stock 780M. It's just got some vbios throttling (which the 780M has too, there's a modded vbios for that card) that's overly aggressive and it runs hot. Shouldn't be a huge deal in an alienware machine because they are cooled better than Clevo machines and a single 880M will run its stock clock all day, it's SLI that doesn't like to run at advertised speeds.

Sent from my HTC One_M8 using Tapatalk

 

Thanks - I'll do that later tonight. I've been moving house so haven't had much time to try and see if this issues was still present.

Still happy with the card performance but I want to make sure that the GPU utilization drops aren't occurring at bad times. Mostly it seems to be when the card hits 60FPS and can't go higher due to Vsync but it's best to make sure. I am running a single 880m on a Clevo though and, while the temps do get high (86), I've yet to reach above the 90 mark yet thankfully. Good re-paste job by PowerNoteBooks.

 

I'm doing regression testing now. For those who have reported no throttling issues with driver 327.33, can you email me the .inf file you used just to make sure I am testing under the same conditions as you? Email me the .inf to [email protected]

 

Aaint +80 temps dangerous in long run for card?

 

Apparently they are normal for the 880M

Sent from my HTC One_M8 using Tapatalk

 

Hardly. My 680m did the same and that went strong for over 2 years nearly. It's when it gets to the mid 90's that things get worrying.

 

so on end it died? :S

mine like most of time dont go over 70. max what i saw was 75 and it was for a few sec just,then fans started to kick in strong

 

Nope. I sold the laptop with the 680m to by this new one. It's still going strong as far as I know.