Alienware 17R4 1080 overclocked vbios issue

Happy for you man!

Having a 7820 or a 6820 should be the same vBIOS right? I mean, that should work for both CPU configs?

But I still wonder why this only happens on some configs... Dell is replacing my mobo next monday so I'll probably just wait for that. What's 3 more days at this point?

I'll report if I end up flashing it before then.

 

Did the stuttering happen during everything , or only specific titles? Did Firestrike cause it to happen?

 

I suspect vrm cooling, can you monitor the gpu power consumption?

 

It wasn't heat related. It would happen with the GPU and CPU at 65 degrees.

 

i totally agree

 

vrm temp isn't monitored

 

Did the vbios downgrade and all my problems are gone so far. I still have some more testing to do, but so far it's pretty promising. I was able to recreate the issue with Doom and Witcher 3 over a dozen times and took many readings with HWinfo. I did the exact same thing with the older vbios and have not experienced a single stutter/freeze.

I'm pretty convinced it was a power issue from having a bad setting in their vbios. Not just the clock speed but probably just a high voltage or something. The maximum GPU power draw with the stuttering vbios was 190W and the highest I'm seeing now is 160W. That 30W very well could have pushed everything over the limits of the PSU. Possibly thermal/power limits inside the PSU resulted in temporary power drops, or something to that extent. I'll probably never know for sure since I don't have a larger PSU to test with. I'm just glad this is over. And my thermals appear a little better as well.

 

That seems odd you had to do that. My gpu pulls 200w max and zero stutters during gaming.

 

Delta 330w psu is good enough to pull 425W from the wall. If there is a thermal/power limit... Either the MB or Firmware, not from the PSU. Power cap was a well known problem we saw with the latest AW 18. Dell engineers crippled either the MB or firmware on those models.

 

for those of you monitoring the stuttering issue, if you open up GPUz, whenever the stutter happens, the GPU load will drop to 0% and PerfCap Reason will register VRel.

So the stuttering has absolutely something to do with power delivery to GTX1080. from my own investigation, it seems like the power delivery circuit is the culprit, possibly overheat due to the extra 30W power required by the new vBIOS.

 

We have had our hands on several systems that it doesnt happen on.
On these systems we run 4.4Ghz OC on all four cores and a 120Mhz OC on the GPU and still no issues.
The gpu has the latest vBIOS and hits 180-200w.
Seems odd. I wonder if its a driver thing going on.
Or perhaps a combo of system BIOS with the vBIOS.
Ima try and make it happen!

 

I'll be trying my own overclock tonight when I get time. I really want to make sure this problem is over before my replacement system gets sent out.

Speaking of that, boy are they slow on sending out a replacement system...

Sent from my SM-G950U using Tapatalk

 

I literally repasted mine 3 times. With Conductonaut and Kryonaut. I used Arctic Cooling pads and checked every single contact point more than twice. Used thin paper sheets to find out where it wasn't making contact. Even tried one run with double the thermal pads as these Arctic ones can get quite smooshed. It made NO difference. Freezing started around the same times as with stock paste.
Also tried limiting the GPU to almost half of it's turbo clock speed, no difference.

If it's VRM then some units are sensitive AF.

 

I think some units has a power system defect.
The modified cooling pad and the removed bottom cover of the system helped.

 

Change PhsyX in Nvidia Control Panel to CPU only and see if that helps. Also, are you running the original VBIOS posted in this thread?

 

No I'm running the newer vBIOS still. It came like this from factory.
Haven't downgraded yet, maybe I'll try it later.
Will try the physx

 

Do it. I had the exact situation as you. I downgraded the VBIOS and changed the PhysX setting and it works perfectly now.

 

Thing is: Dell is finally replacing my mobo and probably heatsink next monday. If it's working by then Idk if theyll replace it

 

I'm now running bios 1.0.8 and the previous vbios to the most recent one. Still getting the issues

This only happened since i had my motherboard replaced two weeks ago. Still not sure if it's a hardware or software issue.

 

Change PhysX to CPU only in Nvidia Control Panel Tek.

 

Thanks, I forgot to mention though I'd tried that when first suggested, doesn't seem to have made any difference

Just been playing rocket league, same old stuttering. Very frustrating.

 

Wow....
Games run fine now.
GPU temps 10ºC lower.
Performance seems the same, in Overwatch at least, not that much of a benchmark.

 

Is this on your problematic mobo? After vbios downgrade?

 

Yes.
Maybe on sunday ill update the vbios again and see if the issues come back.

 

i can save you the time and let you know the result. it does come back. i have tried it. the only real fix is to plug the thermal pad gaps between the power delivery circuits. yes i know you have checked, i was in the same position as you so i totally understand how you felt. i was very confident with my earlier work and visually confirming all the power delivery circuits are making good contacts with the heatsink only to find that i was actually wrong. Gaps still exist, some as tiny as 0.1mm, some not even visible to the eye.

 

Ok I checked nvidia control panel and for some reason some settings had reverted back. Changed them and now I'm gaming without stutter! Thanks those who had made the suggestions!

The question is though, can i upgrade to the latest bios and vbios or will that bring it back? Or is it purely phys-x being on the gpu which causes it?

 

Will the vBIOS posted above work on an i7-7820 with the latest system BIOS or would a roll back of the system BIOS also be necessary?

 

Worked for me

 

Is there an image showing where these locations on the board are? Not sure where the power delivery circuits are on the board.

 

Worked on my system. I didn't have any stuttering issues but it did raise the temps a few degrees. Upon rolling back and retesting with FireStrike I got the exact same score as with the new vBios but with lower temps so I am sticking with the original vBios for now as I don't really see any point. Even if it gives 10% like some claim it's a freaking GTX 1080 lol and I am already killing anything I toss at it so I would rather have lower temps for now.

 

Unfortunately, flashing back to old vbios is just masking the underlying problem. Odds are those not properly cooled VRM and chokes are already operating at the edge or over its max temperature specification at 150W. Without proper cooling for the power delivery circuit (VRM, chokes) your laptop is not going to live very long and prone to premature failure. Electronics don't like high heat, so I strongly urge you to fix the problem at hardware level.

 

If you can refer to iunlock excellent repasting Guide, the power delivery circuit is right on top of the gtx 1080 gpu, at the bottom left of gtx 1080 and on top of the cpu.

They are easy to spot. Example, 4 tiny black square chips on top of the gtx 1080, followed by 4 larger square silver chokes

 

As I said, I don't have said problem, I don't have any throttling or skipping on the original or the new vBios. The GPU simply runs about 5c cooler on the old vBios then the new one and as I don't see much to any of a performance hit I would rather take the 5c cooler temps.

 

This is a new machine that is less than a week out of the box. If it is having thermal issues I'd like to let Alienware get it upto the point where it will at least run games. I'm torn how much I want to mess with a new machine instead of letting it be a warranty issue.

I'm sort of grasping at straws as to why this is system number 2 having the exact same issue.

 

I circled what I think you're talking about. Am I correct?

Does the cooling unit cover these with the pads? So I should add more padding to them? The guide said .5 thickness, but should it be 2?

Thanks for the help on this.

 

The places you circled are correct but you missed out the same thing under the gtx 1080, to the left side. There are another 2 pairs of the power delivery.

The bigger silver chokes above and below the gtx 1080 use 0.5mm. You can keep to that. But the smaller square black VRM might need thicker than recommended pads. I can't give you a fixed pad size, you need to physically test each contacts to make sure they are good.

Also, be aware that if you over used thermal pads on the VRM, it may lift the heatsink up a bit and cause the silver chokes to loose contact with the heatsink. Every heatsink is made with a slight tolerance difference. So you need to customize the pads to your system needs.

The ones on the of the cpu are problematic as well. If you don't get it right, you will face core temperature differential nightmare like many of us do, including myself.

 

I installed the beefier thermal pads last night on the power delivery chips and the PCH.

I cranked the BIOS up to OC level 3 which overclocks to 4.2ghz and loaded up Battlefield 1. Similarly to what you said - I was able to play flawlessly for about 25 minutes and then minor stuttering started to occur. This is a major improvement from before, where if it was set to OC3, it would begin stuttering within 5 minutes.

Some other data points - with the liquid metal, and upgraded pads, on Overclock level 3, which is maxed out, my GPU was at 70 degrees Celsius and CPU hit 90 degrees while playing Battlefield 1.

If I turn off CPU performance mode in the BIOS, which restores the CPU back to stock clock speeds, it is buttery smooth seemingly forever (tested up to 5 hours of BF1 straight) with temperatures for the CPU and GPU never going above 71 degrees Celsius.

The temperatures are fine when fully overclocked, but the system still has trouble when playing something as taxing as BF1 for prolonged periods of time. I have no doubt if I was playing something less intense or not gaming at all, the laptop would be perfectly stable fully overclocked.

I also ran 3Dmark stress tests TimeSpy and FireStrike and it passed both of them while fully overclocked. It seems Battlefield 1 pushes the system a bit more than those tests though.

 

first off, well done on the rechecking the thermal pads.

the fact that you are getting about the same result as me suggest that the heatsink has reached its thermal limit after about 30min playing BF1. unfortunately i can't check what was the temperature of the heatsink area which made contact with the VRM when the stuttering happened. those who have a laser temperature gauge or IR camera can maybe provide us some input on this.

I must stress again for those who are unaware, the CPU and GPU temps might be really good or within range, but that doesnt mean the power delivery circuits are doing well because there is simply no temperature sensor on those things.

 

Will this or anything work on a Best Buy verision Kaby Lake 7700Hq /1070?

 

you are better off getting a new motherboard while you can

 

I am so glad I found this thread. I downloaded the OC vbios and the compatible sbios and everything was cool until I started playing project cars. After about 10min of gameplay my system would bsod like crazy or crash to desktop with no over clocking of the gpu or cpu at all. CPU temps never went over 72c on all cores and the gpu would reach the low 70 celsius range during gameplay.

The times when my system did not bsod and I was able to check HWINFO64 I saw that the gpu had sometimes pulled up to 210w with this OC vbios. I was like WOW!!!!! I dont know what power numbers you guys have been getting with this OC vbios but It was an act of God to be able to hit 185w under the old vbios and sbios 1.0.8 on my system.

Just for FYI my system is a 6820HK repasted with Conductonaut and Fujipoly extreme pads. GPU used to never go over 65c on the old vbios.

Im downgrading to the old vbios as of today and see how that goes.

UPDATE:

Downgrading to the old vbios solved everything and I did not loose much of any performance in game. Benchmarks suffered a little but that is fine.