** 1070 laptop: GT73VR, GT62VR, GT72VR reboot/crash problem **

How are you able to control the voltage? All the vbios for this card I have found do not let you control the voltage.

 

There are 2 approaches. 1 is to make msi fix it with a new gpu. The other is to apply the 8A vbios. I did the vbios and overclocked my 1070 and its been perfect for me. Not everyone feels this is a satisfactory solution. I understand that position. I am happy with the vbios as I can run the card overclocked.

 

The offset is greyed out for me too so I just use the curve in MSI afterburner, you select the point at the voltage you want to use, move it up to the clock you want, hit L to lock it.

I have no idea why that curve goes all the way up to 1.2V since pascal is hard limited to 1.09V...

The 8A vBIOS killed my display outputs (as all gsync/1BE1 vBIOSes do) so I started with a non-gsync GT62VR/1070 vBIOS (1BA1) which I've modded using @Coolane's tool for increased power limits and power slider (and lowered temperature limit) - no change made to the voltages or voltage control AFAICT

 

What happens if you do not lock "L"? Would boost 3 still go beyond the voltage set for the max frequency? For example if you flat the curve at 1 mv at 1911 mhz, will boost3 still go beyond 1 mv if temps are low?

 

Thanks for the answer, I installed yesterday thee vbios 86.... I need to test, but in afterburner I didn't find how to clock the speed or which value to choose on gpu clock.
If you can say me or have a link which explain it, would be grateful.
Same for or the voltage? Thanks

 

I'm rather late to the Pascal party, I suggest you try youtube or an indepth review of how Turboboost 3.0 works rather than rely on my limited experience drawn from my comparatively ancient SLI frankenrig - where I'm balancing many other things, such as compatibility of MXM GPU boards that are harvested from a desktop, observing temperatures on cards drawing double what the original 75W heatsinks were originally designed for (Fermi cards from 2012), and a laptop I'm not sure was ever tested at the kind of load (close to 400W total system power) I'm putting it under... not issues you guys with OEM units have to worry about.

 

Tested, at 1v, I increased the core frequency to 1911mhz (marketed speed from MSI) and flat line it. Ran fans at full speed and let the gpu go into the 30Cs degrees so Boost 3 has plenty of leverage. Ran Heaven benchmark at Basic so the gpu can fully clock and noticed the gpu running at 1v. I guess this is one way to control the voltage of boost 3. I used the 62VR 0D vbios that has gsync.

 

Curve Editor, limiting boost 3 to 1v at 1911mhz.

 

Looks like it craps at 1.05v. As long as you cap it at 1.04v, should be stable.

Nvm, needs more testing.

Extensive testing shows 1.013v is stable for the rev 1.0 gtx 1070 cards.

 

How do you block the voltage and other in afterburner, i can't find to do it ? thanks

 

You have to use the curve editor ctrl + f. Say you want your max frequency maxed at 1v. Slide the point at 1v to your desire max frequency and do the same for every point that is on the right of that.

 

If you're using the 8A Bios, this will cause strange things to happen because boost 2 clocks are disabled on it. So the voltage winds up being locked at the lower speed tier instead of the point you selected (you'll have to experiment to see).

 

The CTRL+F is not working, is there a specific msi afterburner version to use ? are you using the 8a vbios ? is it working at 1v ?

Do you have some strange thing like falkentyne is saying ?

 

I am using 0D vbios from gt62vr 6re. I am using the latest version of msi afterburner.

 

Is this vbios 86.04.31.00.0D you are using, is better than the a vbios version ?

 

The 8A vbios disables boost 3.0 because 2016 gtx 1070 from msi crashed. It is mostly voltage related. I flashed the official 0D vbios to test the theory. Locking the voltage to 10.12v, the crash does not happen during boost 3.0. I am using a buddy laptop for this test because his laptop was crashing during boost 3.0. So now we have two choices for people that crash during boost 3.0. Flash the 8A vbios and overclock or use afterburner and lock the voltage to 1.012v.

 

Ah, so it's a simple power state then. Probably some resistor on the MXM card that is set to the wrong ohm value. So, you change the resistor, which changes the voltage signal, change the revision of the card, and presto.

BTW I'm using the 8A Bios on a non crashing 1070 because I TDP modded it and even though I can't exceed 1900 mhz without crashing, I can run 1886 mhz +250 mhz overclock (.875-.887v) 24/7 and the clocks remain fixed at 1886 mhz, with no throttling. The 0D Bios with +250 mhz overclock would boost to boost 3 clocks, meaning 2086 mhz maximum and signal VREL and be jumping all over the place because of the temperature limit (42C) which cannot be changed (42C temp limit affects only boost 3 clocks. The throttling temp affects boost 2 clocks, which can be changed in the Pascal Bios Editor).

 

Yup, this is why I have 8A in my laptop as well.

 

Ho okay, but i have an K1608N but mine dont crashed, but i only have stuttering problem, does those need to be done to resolve this issu too ?

 

Oh not sure about stuttering.

 

Btw a tip for the curve editor. Edit the curve when your gpu temps are below 40 because of boost 3. The curve looks different when you are above 40 and below 40. The absolute max frequency is accurate when you are setting curve values below 40C. If you set a frequency when you are above 40C what will happen is your frequency will get bumped by 12 or 13 multiples depending on when you set the temp.

 

Guys,

Here is my small contribution to this topic.
Overall, I have to say for now that I'm confused !

If somebody can explain me wth is going on, I'll be gladful.

Here is my situation :

- I own a MSI GT72VR-6RE XFR273 (French machine)
- It is a 2016 GTX 1070 revision
- Factory bios was .0D => like 99% of you, I got instant crashes when I launched a game or randomly after X minutes.
- I installed the chinese .8A that disabled the boost. I kept this vbios for around 6 months without OC

Here is what's happening today (very weird !) :

- I made a few tests with .8A and .0D (installing one, reinstalling the other etc...)
- I did that bios switch for around 5 times : each time, I used MSI Afterburner in order to monitor my frequencies.
- Being tired, I let .0D since yesterday and guess what ?
Not even a single crash for now...

I also OC my GTX 1070 with .0D (with coolboost ON) at 220/200 and launch a firestrike bench : no crash
I could also see that my GC went up to 2039 mhz
I haven't locked anything on my CTRL+F curve.

I have to say that I'm a bit confused : wth is going on ? I used to have around 13000 points on Firestrike normal with .8A. With that "unlimited" boost, I went up to 14000 points !

Here are my settings :

- Repaste : IC Diamond 24
- Notepal U3 under the laptop
- Nvidia drivers : 385.41
- Bios : E1785IMS.117
- Vbios : 86.04.31.00.0D
- EC Firmware : 1785EMS1.107
- Dragon Center disabled
- OC on MSI Afterburner : +220/200, no locking on the curves.

Here are a few screenshots :



My only last issue is that, as you can see, my GC is locked at 1442 mhz all the time... I don't understand why it does not regulate since it is not sollicitaded...

EDIT : ok so weird, when I'm on my windows screen without any chrome, or steam window opened, my frequencies go down. But as soon as I opend a chrome tab to watch some youtube videos, my frequencies go automatically to 1442mhz and 4201 mhz.

How is that possible ??

Feel free to ask me any questions so we can find an explanation.

 

1442 MHz is the base clock and will stay on it if high performance is chosen in nvidia control panel under manage 3d settings. If you chose prefer maximum performance under power management mode in global settings, the clock will lock at 1442 when not being used intensively (it is not allowed to idle) if any app is open. You can prevent this by choosing adaptive or optimal for global settings and set maximum performance under "program settings" for your games only. Then the GPU will be allowed to idle (clock down) at all times unless your set apps are open.

And as for no crashes on the .0D vbios I have similar experience. Simply by reinstalling some random stuff and reflashing vbios my laptop started behaving normally even on the .0D vbios - no one trusted me. It should be in some of my posts earlier in this thread. My theory is that windows forcibly installs drivers in the background and interferes with the vbios installation, corrupting something in the process. With enough luck or disabling the windows update, this should not happen. No one tested it for me though from those who were affected so I can't tell for sure and I don't want to test it on my laptop anymore seeing it works as intended now.

 

Thx for your quick answer bro !

I can confirm that I have for now 0 crash, with .0D with an OC of 230/200.
Again, no voltage lock on the curve and a great performance boost.

My laptop is a K16XXX, 2016 GTX 1070. No RMA to MSI since the chinese .8A was fine for me...

I've just tried GTA 5, Endless Space 2, Hitman... Everything runs smoothly.

EDIT : indeed, the faulty setting (1442 mhz locked) was my nvidia 3D settings to "Maximum Performances". Once I switch to "Optimised Performances", my core clocks go down to 139/405 mhz even on Chrome

Thx !

 

So as heliada said, himself + myself today seem to have no more issue for the moment with .0D.

Simply do the test : it'll sound weird but :
- try to install .8A then reinstall .0D several times in a row
- as I said, I did it (unconsciously) like 5 times in a row (8A - 0D - 8A - 0D - 8A -0D) for testing.
- As a result, it seems I have no more crashes for now.

I'll let you guys know if something new happens.

 

EDIT : ok, so I got a crash... on Total Warhammer.

Instant reboot. My oc might be too high because if I look at my curves, my boost can go up to 2126 mhz max for 1200 mV...

I'm a try with no OC (max 1900 mhz).

 

I believe the voltage is locked to a lower value and cannot go as high. To confirm how high your voltage goes you may want to download hwinfo which will show you the max voltage you reach. You could also log it with MSI afterburner but you will need to edit some files in order to enable voltage monitoring. https://rog.asus.com/forum/archive/index.php/t-81916.html
Lowering the max voltage could probably prevent some crashes.

 

Thx!

I'll monitor this tonight after work.
So far, by resetting my OC (factory frequencies under msi afterburner => 0/0) my curve is limited to a max core frequency at 1900 mhz (starting from 1000mV).

I've been playing total warhammer and no instant crash or crash.

I'll give you more testing results tonight.

 

Ok here are some updates.
Basically, I got crashes with .0D

Got to go back to 8A.
I overclocked to 205/200. Settings are fine for me like this.

Put your power to "optimised" on the Nvidia Panel config, and put max perf to your games so your GC won't be full speed all the times.
Since I don't need that much power, I'm ok with it.

Nvidia has ****ed up those 2016 revision but as long as it's working at 1645 mhz (the speed it was designed to work), and since I can play at 205/200 (after that, I start getting some random crashes or artifacts)... I'm ok.

I don't also wanna send my laptop to RMA...

 

The reason you didn't originally get crashes going from 8A to 0D is because of that 'fast power off' crap in windows 10. It's the exact same thing when you set an overclock with MSI Afterburner, turn off the laptop, turn it back on (or reboot it), and the overclock is still applied in AB, even though you do NOT have AB set to load profiles on startup! I remember when I tested 8A Bios after going from 3A, and 8A was doing automatic overclocking up to 2000+ mhz, which it was not supposed to do. I had to do a full shutdown to stop that from happening.

 

Even at non OC 8A vbios 1645mhz will not have issues with games at 1080p.

 

Are MSI's phone support lines borked? I've been on hold for 45 minutes waiting to talk to someone so I can get a manager of sufficient authority to fix their mess.

Ha. So wait on hold for 50+ minutes to be told not a single manager is in. Sounds like a line of BS to me. Supposedly not a single manager comes in until 900 AM PST. And as a cherry on top, the rep said that they dont replace laptops, just repair them. To which I responded that the repair depot has had the laptop in three times, and a total of four months, and hadn't fixed it yet - and all they had to do to verify the white screen/no video out was to let the laptop sit there while logged in. And when the screen goes white just reboot it, and POOF! No video out.

That's a good way to piss me off more, I tell you.

 

USA working hours don't usually start until 9 AM. Plus we just got off a 3 day major holiday weekend.
You need to speak to a supervisor about the problem. And even though it's going to be difficult, you need to keep an even, calm tone about the issue. I honestly don't know their case load or what they go through, but sometimes, finding the correct person is half the success. And who did you buy the laptop from anyway? MSI? Or a reseller?

If you can get someone to be convinced that you need replacement for a completely new laptop with a newer revision 1070, or better yet, ditch the SLI and ask them for a single card upgrade for the same price, even if you have to pay the difference (1070 SLI->1080), 6820hk->7820hk, that might be your best bet. And the next time you send it in for this, make sure you explain on paper, in the box, exactly what the issue is and what it takes to reproduce it, and all the different conditions required for triggering the bug. For all we know they might just run 3dmark and that's it.
Forgot the SLI attachment--you don't want nor need SLI. What you need is a fully working, dependable laptop. That's the priority.

I hope you can convince them to swap out for a new model. Make sure you are assertive without yelling; raging won't help anything, and good luck.

 

I actually did type up how to reproduce it the second time I sent it back and that didn't help - they just said, as with this last time, that it failed 3DMark and they replaced the GPUs (for the second time).

I spoke with a manager, hours after calling support 3 times, and it's off to a repair depot for the 4th time. The only difference is that the manager will be the go between if they have issues reproducing the laughably easy to reproduce issue.

The manager alluded to the fact that all they can do is try and repair it, not replace it. If they can't figure out how to leave a laptop powered on for a couple of hours and wake up the screen (not sleep mode, that's off), then I'm going to have to take them to small claims court for a lemon that they can't figure out how to fix.

Pretty ****ty that this laptop has a 1 year warranty, ALMOST HALF of which will be gone by the time I get it back again. And that assumes they fix it the 4th go around. I haven't been able to use this POS since May!

Oh, and I would have sent this back for a refund to Newegg, but they don't do refunds on that laptop. Go figure as to why there was a rebate on that POS model.

The singular saving grace is that they have been footing the shipping bill since the start (literally since day one of ownership is when this was obvious, I just couldn't RMA to Newegg again since I cut off that rebate tag). Now, if they could just fix the ****ing thing that'd be just great.

 

After that long in the shop I'd request a free warranty extension be added to your laptop. And a free Lobster and Steak Dinner

Now that you've got the manager's ear, after he thinks it's fixed will be a good time for a thank you and that request.

 

Just got an email from MSI's repair depot, and they can't figure out how to make the easily, highly repeatable white screen issue appear - even though I wrote it all out for them.

Sigh. I've got a feeling that I'm going to have to take them to small claims court to get this resolved.

 

Wild, it does happen though, being the repairing end "unable to reproduce" isn't uncommon, frustrating though.

Tell them you'd like them to try again, with a different technician.

Sometimes shipping will "fix" a problem, bouncing around and when they work on it it won't show, then you get it back and "poof" it starts up again.

Can you return it to the seller for another one? Maybe call newegg and tell the what is going on and ask for them to swap one or for the to tell MSI to swap on for you?

 

I have been trialling a -50MHz core offset for a few weeks now, i.e. letting it turboboost as normal, zero crashes.

 

It is not the clock speed that causes crashes. It is the voltage on boost 2. I tested this and the crashes start at anything greater than 1.013v. As long as you keep it max 1.013v, no crashes.

 

It's been 5 months, Newegg won't RMA it since they have a 30 day window AND I had removed the UPC for a rebate offer - I tried after the repair depot didn't fix the problem(s) the first time I sent it to MSI.

The supervisor at MSI's repair depot has already stated that they can't do anything but try and repair it again and again unless they can replicate the issue - then they might swap me a different laptop or something else.

Likely I'll be buying a video camera and tripod to gather unbroken video evidence of this occurrence happening if they really can't replicate it - Something that I'd prefer not wasting my money on unless I can recoup that cost.

And it's not like this problem isn't easily repeatable - it has happened every time that I have copied part of my games over (I've got about 4 TB worth) - and that's excluding the times that the white screen issue has popped up when looping the Valley benchmark.

 

Two different use cases there, one high IO through SSD + HDD? The other GPU / CPU load. Could be power draw, could be SATA / PCIE with IO, did they swap out the motherboard yet?

That's a pain problem to reproduce and resolve. If you gave them instructions on how to recreate and it isn't happening, that's not necessarily on them.

Someone else mentioned testing with your PSU vs their PSU, sometimes they say not to send the PSU due to shipping weight. If you didn't send it then next time send it, and/or this time ask them to send you the PSU they are testing with or a new replacement to see if that's it.

Good luck

 

Did you say 'copying your games over'?
You mean it happens during hard disk access????? (or SSD access?)

 

No, it really is easy to replicate the white screen - start copying game files over to the spinner (or SSHD, or m.2 SATA drive) with sleep mode turned off and walk away.

The motherboard was allegedly replaced once and the GPUs twice. I have replaced ALL of the drives as well as tried a whole new Delta PSU - still happens.

Either - m.2 SATA drive or a Hybrid SSHD or a standard spinner or the striped m.2 NVME drives.

 

This seems to be a PSU related issue or some other fault on the mainboard that no end user could hope to fix. How do you know for a fact that they actually replaced the motherboard?
Whenever you ask for a motherboard replacement, you should put an identifying mark somewhere so you can see if it was actually replaced.
I agree with HMscott, this sounds like a power /amp related issue, especially if drive access make this occur MUCH more often.

 

Lower core offset means voltage is in effect higher at a given clock. As insufficient voltage is one cause of instability, my suspicion is that there is insufficient voltage at some point/s on the boost curve.

Since -50Mhz seems to have remedied my crashes, I'll be doing more restricted testing through parts of the boost curve to test this theory but its bloody slow going as I haven't found anything that quickly reproduces my crashes for simple testing and no GPU data logs I've collected show anything different prior to the shutoff. I observe a couple of stuttered frames, then a fraction of a second later, black screen everything dies.

At one point a week or so ago I set it back to 0mhz, crashed within a couple hours

 

Greater than 1.013v will crash regardless of clock speed. I set a profile where 1.013v is max voltage at 1911mhz (stock max boost 2). No crashes.

 

How did you do this?

Is this possible with the 8A Vbios at all?
Is it even possible to go past 0.881v?
I don't want to open up my laptop to reflash 3A right now....

 

I did it with stock 0D vbios. I let my gpu go to 29C and run Heaven Basic to trigger boost 2 max clock and voltage. No crashes.

I used the curve editor and basically click at 1.013v and raised it to 1911 mhz and flat line everything on the right of that. On the left, i overclock each point with the same offset used at 1.013v.

 

Can you post a screenshot of your curve graph?

 

I found this thread, sadly enough, after I had already RMA'd my GT73VR 6re (1070 SLI) to MSI for repair. It arrived on Tuesday and I had it back on Friday. I was miffed when I looked at the log attached to the box that said all they did was reflash a vbios. They had flashed 8A. MSI could have saved me and themselves money if they had just put 8A on the website for download.

So far, I have yet to have it reboot during game play. The two games I suffered those issues with were Path of Exile, and Grim Dawn. Both of which worked fine after the laptop was flashed with 8A.