Just got my 880M twins!

Read this... [Review] Nvidia GTX 880m for more understanding.

You will not now, not ever, get a good result using a stock vBIOS. Changing the clock speeds won't help because the GPU will not run even stock speed under load. Using a stock vBIOS the GTX 780M performed really bad like what you are experiencing, and so did 680M. You've got a fantastic piece of hardware with craptastic firmware. Don't be frustrated, just be patient. There is no point is pushing things harder and harder at this time because the only thing likely to happen is performance will worsen or you'll potentially cause damage to the CPU by attempting to use power to circumvent a firmware issue.

 

If you say fox is law! : D, do not even know how to thank all the support that you gave me so far, I will wait, BF4 with everything on ULTRA, MSAA X4 and another graphical configuration I can not remember the name now (so this setting is at 200%) is 10-20 fPS with it on 100% runs on average 40-60 fps,

 

It is real shame that NVIDIA does such a marvelous job on making hardware and a very horrible job on their firmware. I don't think any of us understand why this seems to be the case with every Kepler product. But, we know it will be awesome with the vBIOS mod.

 

It is also this performance actually hope she alone with original clock is stronger, get better performance than my two overclocked HD6990M:]

 

I was messing around with my 880M to and decided I really wasn't doing much at great risk of damage to a rather expensive video card.

Sent with love from my Galaxy S4

 

Poor GPU cooling, that's the main reason behind this. AW GPU cooling is far better compared to its competitors, still user wrongdoing/stupidity and liability (warranty) are the problem. AW vBIOS with some of these limitations/throttling removed...? it will never happen for obvious reasons, therefore the unlocked vBIOS is a win-win situation. Cards perform as they should/marketed (superior) and they are not responsible for any 'negligence'. Furthermore since driver v337.50 the power efficiency was improved beyond expectations, proving driver support silently. 780M Unlocked vBIOS, previous drivers @1006 1.037V (+37.5mV) v337.61 stock voltage...

 

Hi J95, Fox has spent the last two days trying to help me, first was that the processor was not stable now got to 4.5GHz, the problem now that my GTX880M, does not keep the original clock during games (BF4 using to test) I I open the game it is at most 954/2500 (core / memory) after some time (a few minutes) the clock speed of 954 falls, in the range 130-330, which makes anything unplayable, do you have any solution? I'm already mad at this ! if you can help I thank you

 

I was helping out R4/880m user and he failed to inform that he didn't rearrange the thermal pads (660m) even posted thermal pad layout reference pic, total waste of time since I ruled this out (core temp ok after re-paste) and focused on TDRs.
Very similar problem you're experiencing, try checking if all important components are properly covered.

 

Good idea from j95 to double-check and confirm thermal pad placement. Even if the core is not getting hot, the other components could be about to fry is pads are not correct, and those other components have no temperature sensors. Also, you are using NVIDIA 100W heat sink (I think I remember that you bought the proper part) and not using 6990M heat sink, correct? As far as I know, 680M, 780M and 880M thermal pad placement will be the same.

@j95, the image (not in spoiler) is 680M, correct.

You must mean this one...

I think that image is the one Clevo (Eurocom and UpgradeMonkey) are providing as guidance for their modules. I have noticed that Alienware heat sinks also have thermal pads on the components I drew a blue box around (below). Should these component NOT HAVE pads (be omitted) for best results, or it is optional and the green shaded areas are obviously minimum required? Is there anything definitive on leaving the components in the blue box without thermal pads?

 

I found that from my experience of replacing the pads, that you don't really follow those thickness exactly. Some chips need thicker pads, and some can be thinner.

I believe those 6 smaller chips in 3 sets of 2 on the top and near the top right are the VRMs. Probably the mosfets? I am not sure. The top 2 are the VRMs for the GPU die, the other for the vRAM. You can see the SMD capacitor (those tiny square things) near the mosfets.

Deep in the thread about the 580M overheating, I once saw someone who took a thermal IR shot of the GPU under load, which showed those VRMs to be the hottest part of the GPU (esp the ones feeding the GPU die).

 

Yep, 680m...as you said, identical thermal pad layout.

That's why I prefer 0.5mm pads, easy to adjust if needed.

Exactly what I use as reference since 580M.

 

Is it your IR shots? Or someone else's?

 

Nope, @Vytis

 

I remember seeing that image, but don't have a clue what thread it was in now. It might have even been an M17xR3 580M throttling thread.

I have been using pads on the components in the blue boxes (in addition to the green shaded areas) because Alienware heat sinks always have pads on those components. I believe they may be VRMs, but I don't have a schematic to identify what they are. Anyway, it's interesting that Alienware puts pads on them and Clevo does not identify them as components that need thermal pads.

Edit: Here is the post with the thermal images, and it was the 580M Thermal Throttling Conference Call thread.

http://forum.notebookreview.com/ali...ttling-see-post-191-later-29.html#post8400372

<iframe class='imgur-album' width='100%' height="375" frameborder='0' src="http://imgur.com/a/KzxQm/embed"></iframe>

 

Mr. j95 kindly linked it up there. Too bad the person who posted it, Mr. Vytis is no longer actively posting, but he seemed to have logged in earlier this month.

 

Large temperature difference between components, those (Blue) capacitors don't require padding. Beyond @1150 I started covering them... GTX 780M temperature and benchmarks on M18x R1

 

The big square stuff is actually the inductor, the caps are the tiny square thing beside the mosfet, I think.

Sadly, we only have 2 phases for the die, and 1 phase for the vRAM. Puny power supply compared to desktop GPUs which typically have 12 phases (some 14 phases) nowadays.

 

LOL, we were both looking for it at the same time... +1 Rep

I copied those to my imgur account for safe keeping because imageshack has a very special way of losing images... I am actually shocked they have not already vanished since they are hosted on imageshack. They have lost dozens of some of my images... just mysteriously vanish... and sometimes reappear later on with a different URL.

 

As i have said VRMs often run hotter than the core.

 

The instructions of the foam came as a photo exactly like the fox posted, and I followed all directions to the letter, I thought I had solved the problem of'' fall'' clock exchanged 337.81 BETA drivers for the upgradeyourlaptop / Eurocom provide, but drop clock continues = /, I will re install the beta drivers and try to do what FOX oriented, with nspector nvidia if anyone has other suggestions, I'm all ears

 

I have not disassembled the notebook again, just looked at the sink I'm not using and he says AMD 100W automatically assume that the other is 100W AMD also has some difference between heatsink from AMD to NVIDIA? are not the same?

 

AMD and NVIDIA heat sinks are different

 

ok, finally started up and everything seems to be fine but, when I look in GPU-Z, the bios and driver version is different.

 

know what the real difference between them? , And am now monitoring the temperature with GPU-Z to see how much is going, do you have any suggestions?
is ridiculous to say this, but I solved the problem with a tablet underneath the notebook to lift it a little temperature max video card 81ºC, summarizing temperature should be, what the that made me almost crazy --''
without the table gpu go to 90ºC , than clocks down =/

 

Please dump both versions, give them unique names, and send them to svl7 and johnksss for testing. One might perform differently (better) than the other one. You need to decide which one to go with and flash the other so they match. I'd pick the one that doesn't have a messed up name in GPU-Z.

 

I know eurocom (my second card with funny name in gpu-z) updated the bios on the card to make it work on my alienware but I guess upgrademonkey did the same?

 

You know those are two different version of the same driver right?
So I would start there first.

 

I noticed the same thing re your blue box comment, which was the same for my 560M - when I did my 670MX I decided to pad up those blue box areas too, I'm sure there's no negative to doing so.

 

The most important part to get are the vrms (the little black squares around the edge of the card) they are doing the heavy lifting.

 

Cool (the operative word), got them covered!

 

Two different version of the same driver? Im sorry but I didnt get that..

 

Two different version of the same driver? Im sorry but I didnt get that..

 

There's 2 different named versions out in the wild, but both are actually exactly the same.

 

Yeah, I can see that but the BIOS version is different rigth? Will it have any effect while running SLi?

 

It can cause problems with SLI using two different vBIOS. I would think having two different versions of the same driver might as well. I'd try to clean things up to eliminate both problems.

 

Finally

 

Well, fans are bearly running during 3Dmark, making the cards (heatsinks) EXTREME hot, guess its throttling.

Also getting these errors.

 

3dmark11 cant even run due to low fps (1-2fps) Worse 3000'dollars I ever spent.

 

Valley is showing 100-110fps the first 5 seconds and then jumps down to 5-10 because the FANS are not running making the cards hot.

But Im only seeing 65-70 degree on the cards though.

 

But Im not running 8Gbps, only 2,5Gbps.

 

Tried to disable SLi and see if it got better but still 1-2 FPS in 3Dmark11with dead fans.

But my physic score is 10550 with 4.5Ghz OC and during the test the fan is running perfect.

 

My 880's do the same thing in 3dmark11. Starts off strong but gets killed in some spots.

 

it happens in every application I have tried, cant even play a single game.

 

If the fans aren't running automatically, maybe you could use a program to manually run the fans at an rpm of your choosing. HWInfo64 might let you do this, and who knows this might sort it. Have you thought about doing that?

 

BUT, if you look at the picture with the crappy FPS....the core and memory has MAX MHz and are around 70 degree, it should not throttle now!

 

Sure, that's weird, but why not try it, it wouldn't hurt. Sounds like you're running out of options anyway. Ramping GPU fans up to 100% during testing just negates the fan issue you mentioned, I think it's worth testing. Not saying it's going to work & solve your problem, but if I was you I'd at least try it, close to zero effort really.

If fans aren't ramping up, that IS a major issue any way you look at it, especially if you know that they normally ramp up at certain temperature. Like I know my GPU fan kicks up to max rpm between 65 & 70 degC.

 

Cant change fan speed... Was just like the 680M, needed a modified BIOS to change the fan speed.

CPU speeds up but not the GPU.

 

Sorry you're having a frustrating time. It's a shame how poorly the stock vBIOS performs. It will likely turn out to be money well spent, with stellar performance, once svl7 and johnksss release the vBIOS mod.

Did you get matching firmware flashed in place on both GPUs and the mismatched driver issue corrected?

Have you tried disabling SLI and benching each GPU individually in the primary PCIe slot to see if the behavior is the same as SLI?

 

Yeah, its very frustrating.

I did fixed the driver issue but didnt flash the GPU's to the same BIOS but I tried to disable SLi and run some benchmark but its just the same.....zero performance, like I dont have a GPU at all, 10-12 FPS on 3Dmark11? Should be far over 120 FPS.

 

Ah, ok. Weird to me that you need a modified vBIOS to change fan RPM. On my laptop fan rpm is controlled independant of the vBIOS, it's part of the 'Compal EC', is that not the same with yours? For instance, fan control profiles are identical for me, both when I had my 560M in this laptop & when I have my now 670MX - for mine it's nothing to do with card or vBIOS it seems.