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

I actually did run catzilla last night my score was just under 8600. However it still barely boosted. I know this may be a hard question to answer but what's the chance the new cards will have the same problem?? I want a card to perform the way it is supposed to

 

Maybe if you fight with Dell you can get 970M as a warranty replacement.

Since nVidia chose to pull chips from the high performance bin that would normally be reserved for the desktop because of high leakage, they have to be expecting a number of replacements, it wouldn't surprise me if they set aside some 970M chips to replace them with. It wouldn't cost nVidia anything to do this because if they didn't release a card refresh, it would have led to more M290X machines being sold and from the fact the 780M costs significantly more for OEMs to source than the 880M, it tells me they ran out of low leakage GK104 cores so they put cores that should never be in a laptop in the 880M.

Sent from my HTC One_M8 using Tapatalk

 

If the core is at or above 954MHz, the card is functioning as intended. As messed up as it is, nVidia knew these chips would not be able to boost.

If the core is dipping significantly under 954MHz, the core needs more power than the default vbios is giving it.

Sent from my HTC One_M8 using Tapatalk

 

Ok so let's say the card is running as intended... should I be worried about this being an issue going forward? Especially with the new card having a higher core and higher boost? Will we not even get the rated core or is this really just an issue with the 880m?

 

I fully believe that Nvidia never intended for us to be able to have boost. These were just left over high performance desktop class cores that they stuck on a card to get rid of and used boost as a way to dynamically throttle the cards just enough that it isn't immediately obvious without looking at a core readout that it's throttling.

If you look at the core average across a period of time there are many spikes and drops but the card generally averages 940-960. This was likely the only way they could keep the temperatures in the low 80s because as we see with the modded vbios, even 954MHz shoots into the 90s in no time. The dynamic switching means the core voltage also goes up and down, hence the heat is reigned in.

Sent from my HTC One_M8 using Tapatalk

 

Well that being said I think we will see a completely different result with the new cards. I just ran another catzilla test before work and got 8767. My core fluctuated between 954 and 1012 (1032 is my max but only see it 20% of the time) with my attempt at an overclock. Usually stayed between 954 and 993 but did see few spikes above that as I said. I think for a split second it hit 940 but that was only once as very fast.

 

My SLI machine has crazy throttling on the second card but when I run it single GPU it doesn't drop below 954MHz.

I can push them to 1006 boost on the stock vbios but it will crash after that.

Sent from my HTC One_M8 using Tapatalk

 

Interesting... The problems are definitely more apparent on sli. I haven't had a crash yet at 1032 core and memory +350. But I need to do more testing anyways. At least I'm running cooler now. I just don't know how much luck I'll have getting a new card under warranty.

 

I honestly believe if I turn off SLI, I can probably get the single GPU to go up quite a bit for max boost but the tweak for 1006 didn't lead to any higher average (as expected, the power the limit doesn't change)

Memory tweaks though, that's a different story. I haven't tweaked the memory of these but my old pair did +400. I get leary about messing with memory for anything but benchmarks since there is no way to see the temp that's running at and the inductors are no doubt dumping their heat onto the memory.

Sent from my HTC One_M8 using Tapatalk

 

So wait are you saying monitoring my temp with afterburner will not show an accurate temp with my memory overclocked?

 

Memory temp isn't monitored, it doesn't have thermal sensors. That's why it is always advised to go light with memory overclocks. What can really kill a card besides too much voltage is the RAM frying.

Sent from my HTC One_M8 using Tapatalk

 

This is important for people to keep notice of. Knowing which brand and specs RAM is on your GPU is a good way of knowing base limits, but definitely go light on those clocks.

 

Hmm you think I should back off the 350 for memory then? Or is that reasonable?

 

With how hot the inductors get, I doubt that 350 is going to be stable longterm. I personally would go for 250 to be on the safe side but my Clevo card has Samsung memory.

Sent from my HTC One_M8 using Tapatalk

 

Well this won't be long term just until the 980m comes out then bye bye 880m. Maybe I'll see how it does. Can't possibly kill the card in like a month... At least I don't think lol.

 

You'd be amazed...

Sent from my HTC One_M8 using Tapatalk

 

Really? Alright I'll jus lower it a bit why risk it. It's not a huge performance gain anyways

 

Yeah actually clocking it too high can hurt performance.

Sent from my HTC One_M8 using Tapatalk

 

Ill just bump it down back to 250 I only just changed it to 350 two days ago. This wait for new cards is killing me though.

 

Your new Title is: Ethrem Cardkiller

 

Nah my cards work now that I fixed Sager's pad screw up. But yeah, I don't have fantastic luck.

Sent from my HTC One_M8 using Tapatalk

 

So I know you said the lack of boost is due to TDP but I have a question. Since the card does fluctuate and hit max clock here and there... is there a better OC tool that can lock the core at a certain level as long as its a stable OC?

 

More like Arotished Jr.

sorry couldn't resist

 

No. The modded vbios is your only option. It increases the TDP and enables the card to run at its maximum frequency for as long as you can keep it cool.

You may be able to underclock the stock vbios to get a consistent clock rate but while that would eliminate the up and down random nature of the clock rate and result in smoother frame rates, you would have a lower average.

Tweaking the frequency of the mod to where it doesn't throttle is your best bet and even 954MHz will hit the 90s with it. The dynamic up and down clocking dramatically lowers the voltage (stock voltage table is 1.012v @ 993MHz but only 0.981v @ 954MHz, and 836 only sips on 0.875v which runs so cool that the temp with auto fans never even hits 85) .

The stock vbios leverages core throttling to achieve a higher average core at much lower heat than would be possible otherwise.

Sent from my HTC One_M8 using Tapatalk

 

I though nobody had luck with a modded vbios so far? With the current state of drivers and such I thought Nvidia killed the option to run a modded vbios. Unless I'm missing something.

 

Modded vbios has been available for quite some time now on Tech Inferno forum, it just is extremely hard to contain the heat. 880M runs a good 8-12C hotter than 780M at the same clock speed.

http://forum.techinferno.com/genera...-editions-modified-clocks-voltage-tweaks.html

Johnksss and svl7 made it.

It isn't without its issues though. For example when I use it, I can't use XTU or it causes my system not to want to POST if I restart it instead of shutting it down and I have to pull the power cord to stop the shut off and power on cycle. Remove XTU and the problem is gone. Seems to be a Clevo problem.

Then there's the heat. Full 993MHz runs up to 94C and then throttles in no time. Pretty much mandates liquid ultra if you're going to have any hope of keeping temps under control.

Sent from my HTC One_M8 using Tapatalk

 

Wow it runs that much hotter just changing the vbios?? I get between 78 and 83 now after my new card and repaste. At 993 it never goes above 83. Even when it pushes my oc it doesn't creep up past 84.

 

Yes it runs that much hotter. The reason is that the cards run straight 993MHz with 1v each and the TDP is practically unlimited.

You and I have pretty similar temps. A burn in with Valley will have the master at 79C and the slave at 82C.

As soon as the modded vbios is flashed, I'll hit 93C 10 minutes into Valley on the second card and it will thermally throttle after about 15 minutes.

Like I said, nVidia managed the heat with the stock vbios by setting a low TDP limit that forces constant up and down clocking. Voltage causes an exponential increase in heat so it makes sense that having voltage fluctuating up and down runs the cards cooler than a constant voltage speed, especially since these cores appear to have high leakage (nVidia binned them for high performance - they can do 993MHz @ 1v while the 780M needs 1.025v or so to maintain that same clock usually and that means these are likely GK104 cores they had left over from desktop card binning)

Sent from my HTC One_M8 using Tapatalk

 

What a damn mess these cards are. I expected so much more and got spit on essentially. I don't care what these new cards cost I'm buying one if they won't warranty it. I inquired about this before but on my alienware laptop 17 the new card should be plug and play correct?

 

Yes. Alienware has never locked down cards to the BIOS. When I borked a BIOS flash of my 880s the first time, I stuck one of them in my M17xR1 to reflash it after my Clevo wouldn't even post. That machine had 260M in it lmao

Sent from my HTC One_M8 using Tapatalk

 

Wow ok that's crazy lol. Good to know that eases my mind... Soon as it's available through dell or can I order from anywhere? Don't wanna void my warranty.

Quick side note... What are your thoughts on the i7 4910mq? Will it last me a while?

 

If you order it from Dell they will rip you off for an extra couple hundred bucks.

See if you can pick up a Clevo card on eBay. They're gonna probably be 1k the first few weeks though.

As for warranty, Dell doesn't care what you have in the machine so long as you don't break the machine installing it and if your machine ever goes in for Depot service, I would remove the video card because people steal stuff at the Depot. Then again if you have to send your machine to the Depot it's time to reach out to Mr. Fox or another Alienware guru first.

Just to be clear though, the 880M benchmarks that are posted on notebookcheck I was able to manage to score the same or better with my stock cards so nVidia did still technically release an upgrade to 780M when comparing stock performance, its the overclocking that sucks.

4910mq is a solid chip. I've actually thought about selling my MX and getting the MQ because it appears to be binned better. Shouldn't need to worry about upgrading that for a long time. Not for gaming anyway.

Sent from my HTC One_M8 using Tapatalk

 

Is there any difference whatsoever ordering from another provider? Won't I need a different heat sink from Dell?

Thanks for input on the cpu I'll stick with it for a while... Even though I never see it hit the 4.1ghz dell said it's overclocked to. Adding to that I can't seem to figure out how to oc my cpu it let's me change the multiplier I'm bios but immediately reverts back when I try and save it.

 

Heatsink designs will be the same from one MXM card to the other. It's actually likely the TDP will drop with the new cards so your old heat sink will be even cooler.

Overclocking, just grab Intels XTU. You raise the turbo multiplier, not the base multi as that is locked on mobile chips.

Sent from my HTC One_M8 using Tapatalk

 

Which is the latest version and does it require any modifications or only the program? Runs in the background like afterburner? Forgive my ignorance. Also how much do you think is safe to raise the cpu?

 

This guide is for the 18 but should work just fine for the 17 as well.

http://forum.notebookreview.com/showthread.php?t=734696
How to Overclock the Alienware 18 and Haswell CPU (or actually have it run full stock Turbo Speed)

Sent from my HTC One_M8 using Tapatalk

 

Changing the bios concerns me because it does not look exactly the same as the picture Fox posted nor does it let me set those power and time settings the same. I'd love to get a slightly higher clock but not sure how to proceed with those instructions not fully matching up.

Let me add I'm on a13 bios.

 

Might have to shoot a message to Mr Fox about it but on my Clevo, I can do everything from Windows with XTU. All you do is turn up the multiplier for the turbo and then test it how he shows in that thread until you find the max stable point.

I have my chip sustaining 3.6/3.7 on all four cores with a 47W TDP which gives me a max temp of 80C and keeps the fan noise to a minimum. The easiest way to get a free overclock is to undervolt the chip. Mine does -85mv comfortably which gives me an extra 200MHz over the stock.

XTU reverts if it crashes or anything so unlike the BIOS, you can't break things so it won't boot.

Sent from my HTC One_M8 using Tapatalk

 

Hmm ok ill start with trying with xtu and see if I get results. Otherwise I'll definitely message him for some answers.

 

I flashed back to the stock vbios in my 880M last night and ran the Unique Heaven bench in a loop for about 16 minutes. Core temp got to 78c on max fans which it reached after 10 minutes. With very few exceptions the gpu was boosting to 993 and it went down to 980 during those times when it dropped off of max boost.

 

Sounds about right for single GPU.

Sent from my HTC One_M8 using Tapatalk

 

I've been doing some tweaking of the modded vbios and results are promising so far.



I'll just drop this off here. This is with auto fans. Basically the same temp stock gives me but with zero throttle.

Out of respect for svl7 and Johnksss I will not release my mod of their mod and will instead work with them to see about getting an official tweaked version from them. I am not done yet, I will be ratcheting up the clock speed bit by bit to see if I can get full boost without hitting over 90 but for most of us, 954MHz is plenty wouldn't you all agree?

Sent from my HTC One_M8 using Tapatalk

 

If that is 99% load on both cards, then you've really come across something special. Especially auto fans.

Now, if you could try that with a game that uses a lotta vRAM like watch dogs or Titanfall, that'd be awesome. Just to make sure it keeps the temp range with higher vRAM usage since johnksss said (I THINK) that the vRAM chips were what got overly hot.

 

Valley always has burned up the cards pretty nicely. It's not the memory, its the inductors that are getting hot which transfer heat across the PCB to the memory which causes memory errors eventually but that's what screws up memory overclocks.

Sent from my HTC One_M8 using Tapatalk

 

Well, still try it with a myriad of games and see. Haha maybe you could try that shadows of mordor thing with 6GB vRAM... put those 8GB mem buffers to some use huh?

 

Well this is looking extremely promising as a fix...



Sent from my HTC One_M8 using Tapatalk

 

Scrapped. I'm done with these cards. I don't get how something goes from working perfectly to epic failure so quickly.

I'm ready to throw my laptop out of the window so I'm just going to step away. These cards are obviously rigged to not cooperate.

Sent from my HTC One_M8 using Tapatalk

 

What happened?

 

I was playing with power targets... stock 117W is too low but the mod has 175W which is a bit much. I found that 130W had no games throttling and 135W had no benchmarks throttling and the heat was lower all the way around until I ran Bioshock Infinite. I don't know what the hell it is with Unreal Engine 3 but it cooks the living hell out of nVidia cards. It was about 25 minutes when I hit 94C on the slave and it thermal throttled.

I cut the core clock down to 954MHz @ .975v and 120W TDP and it still was creeping up to the 90s in minutes.

These cards are completely and totally broken and there's just no point messing with them. I'll take stock throttling with ~80C temps and an average of 940MHz over 954MHz consistent with 90C or 993MHz consistent with 94C and throttling.

Its just not worth the risk of destroying the machine. Its pretty sad that I can't get what I paid for but I'm not going to risk running those kinds of temps and frying the cards. I'm going to (impatiently) wait for the new cards and raise hell if I have to in order to get what I paid for because this is just ridiculous. nVidia should really take some responsibility for this mess and make it right. I imagine that they looked at "fixing" the vbios and realized what I did - the cards are too defective to be fixed - so instead of admitting they can't fix it, they put the gag on Pidge so they don't have to own up to it at all.

 

Yep. It's pretty obvious that Nvidia's complete silence, especially after acknowledging a problem exists is an total omission of guilt.

Instead of getting behind Pidge and fixing things they gave him one of these to chew on