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

I have a batch file which loads RWEverything (has to be run as administrator) and which programs a value of "91" into EC RAM register E3.
Just have to click it after windows finishes loading.

I managed to pull 355W from the wall yesterday by having Heaven benchmark and prime95 running at the same time. Had 12% battery drain an hour.
I don't know if the CPU will throttle if 330W is exceeded. I'm unable to pull that much. There is a 460W EC allowance due to there being an SLI configuration, but I'm unable to activate that (I think it requires the slave video card to be detected in the second MXM slot).

I could bypass that battery drain by removing the battery then programming 2 other EC RAM registers (31 to "09" and 42 to "64") to make the EC think the battery is connected, avoiding the "NOS disabled" cpu throttle at 160W or 180W system power (for 1070 and 1080 systems).

 

What overclocking were you achieving to drain that much? What is your firestrike mark using the auto fans? and using coolerboost?

 

I don't use auto fans. Well I do but the fans boost to 100% speed for me because I set the curves that way. Usually I use 100% fans all the time.
and i don't remember. but graphics score 21800 to 22050. and the 7820HK was at 4.7 ghz @ 1.270v and GTX at +140 core +675 memory and 230W TDP and I was running Heaven and prime95 at the same time.

 

Ok, you have convinced me. I have a graphics score of 18600, yours is much better. I will keep reading the other post and maybe I will have some questions. Is it possible to buy the needed pieces in europe?

 

Been done.

I'm in USA so I can't help you
but you need these

https://www.amazon.com/gp/product/B00HHH65T4/ref=oh_aui_detailpage_o00_s00?ie=UTF8&psc=1
https://www.amazon.com/gp/product/B072KYK2DR/ref=oh_aui_detailpage_o02_s00?ie=UTF8&psc=1
https://www.amazon.com/gp/product/B01DZC36GY/ref=oh_aui_detailpage_o06_s01?ie=UTF8&psc=1
https://www.amazon.com/gp/product/B01EV70C78/ref=ox_sc_saved_title_9?smid=A2WWHQ25ENKVJ1&psc=1 (for hooking up the pomona clip to the 1.8v adapter).

 

Well great. I just ordered an MSI GT73VR Titan-427 with 7820HK and GTX 1070 from XoticPC. I won't know the serial number till it ships.

 

Since these laptops are end of life in most countries (impossible to get in the netherlands for more than 6 months now unless imported), I assume it is more likely to get one of the later models with a 1.2 card rather than 1.0. Nevertheless, test it thoroughly when it arrives and if xotic allows, just ask for a refund if you see any flaws. It is not worth it going through with rma seeing how persistent msi is with not wanting to replace the gpus. And usually an rma would get you right past your return window ^^ I have been there, having faith in the company and then being stuck with a gt72vr that gave me a real headache with a different issue.

 

Ok thank you for the advice. I don't know when it will ship, it hasn't even started the cooling mod stage, but I'll check it out real good whenever I get it.

Sent from my SM-N950U using Tapatalk

 

Will this ones work???? I think the first one is different, but probably will do the job...
https://www.amazon.es/WINGONEER-Pro...qid=1543235447&sr=1-1&keywords=wingoneer+bios
https://www.amazon.es/dp/B072KYK2DR...olid=19M8MS5YHDR8L&psc=0&ref_=lv_ov_lig_dp_it
https://www.amazon.es/dp/B07GFSX5S7...olid=19M8MS5YHDR8L&psc=0&ref_=lv_ov_lig_dp_it
https://www.amazon.es/dp/B01NGTXASZ...olid=19M8MS5YHDR8L&psc=0&ref_=lv_ov_lig_dp_it

 

I can only help with the skypro, sorry.

 

Just buy a cheapo 1.8v SPI programmer from eBay(less than 5€).

The TL866 is a massive overkill of a programmer to just program a soic flash once or twice in its lifetime.

 

but is there any difference??? will it work? This one is the only one that I can get relatively fast....

 

If it supports the chip it will work. As I said I don't own this device. I am not techsupport.

 

Looks like it would work, but best to double check with the reseller to confirm.

 

You can also use this:
https://www.amazon.es/WINGONEER-EEP...ie=UTF8&qid=1543342619&sr=1-1&keywords=ch341a

Lots of people have used those, use it with flashrom:
https://www.flashrom.org/Flashrom

and grab the 1.8v and the SOIC clip and you are good to go.

 

So i changed the voltage using afterburner, and the problem goes away for a while until i decided to play Destiny 2 which crashes the system immediately as i start playing :/
Has anyone tried this game ?
games like Witcher was crashing my system before i changed the voltage but after that the problem goes away,but not for Destiny 2

 

What did you change the voltage to? Have you checked that the curve stays completely flat after a reboot? I know my curve was not sticking totally 100% flat and I had to open afterburner and click apply again to make it flat again (as upon each reboot there was 1 spot where it developed a slight bump up in voltage for some reason)... I suspect dragon center or windows or sth was messing with it for me... If 0.9V does not work in all scenarios, you can try 0.8V as that was the sweet spot for my bd prochot... Otherwise if you are out of warranty, well, you are screwed.

PS. it may be unrelated and game specific, any errors in the log? BSOD?

 

LUL i know im screwed long time ago, even with the warranty that not gonna help because all they do is limit the clock speed and not solve the problem,
also i uninstalled Dragon Center because it nothing but trouble
im going to change the volt to 0.8V and see if things change

 

changed volt to 0.8v and it works perfectly! Thanks!

 

Don't celebrate just yet. Test it for a while... Also it is possible the issue will surface at some point again and you will be forced to scrap the laptop or replace the gpu completely. No one knows if undervolting even helps permanently. Imo what is faulty is faulty, which is why I am glad MSI took mine back. If you do have warranty, I would push for it. If you are nice and patient and explain it way tooooo many times, they will exchange the gpu, or give you a replacement. You have a right for a fully functioning laptop, you did not sign up for the need to undervolt after all when you bought it.

 

Contact Tom Ho ( tomh at msi dot com i think, if that's not it, search this forum as it has been posted), and explain to him VERY politely that you have a version 1.0 MSI 1070 card that has the crashing problem at higher than 0.9v, your laptop is out of warranty, and you would like to exchange the 1.0 card for a working 1.2 revision card. This will cost you money so make sure you explain fully, politely and carefully that you understand the problem and you know what you're talking about. You have nothing to lose except being embarrassed.

 

Sadly i purchased mine one year ago from Ebay seller as refurbished for 2000 usd (and i told my self 2000 usd for sli 1070 isn't bad)
months after testing and windows reinstalling i found about those corrupted 1070s, reached msi and they told me my warranty is no more
before i found about the Voltage thing i was playing games on 4K resolution using Nvidia DSR, it cause a lot of fps lose but it works fine and solve the problem
for now i think im gonna live with it and adjust to the voltage method and take this as a lesson for the future

 

My laptop is Sli and one mxm 1070 card costs up to 800 usd
so i will end up paying as much as new laptop

 

Mmm ye well technically the undervolt does not make you lose much performance... and you can OC at the same time to around 1800MHz if you want to for sure which will result in having pretty much no impact in the end. If anything, the undervolt will save you on the electricity bill haha. The question is if the "fix" is permanent cause no one knows. No one even knows what is exactly wrong with those cards

 

As I said last time, ask him.
He may arrange a swap of the old cards for a fee. Maybe request a cross shipment and explain you are competent enough to remove the old cards and install the new ones.
There is NO harm in asking. And it costs you nothing to ask.

Complaining here about price isn't going to fix your problems.

 

Mobile 1070 is utterly power limited so unless you lock it to something really low (like MaxQ clocks/volts or below) it'll turbo up and eat the full 115W no matter what. Not really saving power, but getting you extra performance.

I was able to maintain 0.975V for 1975Mhz, well below crash voltage on my v1.0 vortex 1070s. That was with a significantly raised power limit to be able to maintain those clocks.

I tried benchmarking for a little while locking them to the full 1.062V and clocks were somewhere in the region of 2150mhz. Unstable though, it'd hit even a 200W power limit and clock down below 1.013V and crash fairly often

 

He has an SLI system and plays in 1080p. The likelihood of both cards being maxed with a mobile cpu are low. If they are not being used to 100%, then they will be allowed to run at full voltage as the power drawn will not be high enough by either card. Undervolting will lower power consumption in those scenarios as it might bring the clocks down from 1800+ to around 1600MHz on stock (or whatever OC is stable for such low voltage). And 0.975V is not even attainable by his gpus as they would crash the entire laptop due to faulty MSI design/parts/whatever it is.

Plus, the power limit will take the cards down to around 1600MHz if both cards are used at 100% which is in the safe voltage zone for these faulty cards. Unfortunately for 1070 SLI systems, especially with mobile cpus this just won't happen at 1080p in any game I know or am interested in looking into. Hence for this person playing at 4k on his 1080p screen was the workaround: the cards would be used to 100% in many games and power locked to around 1600MHz, which is low and safe voltage. At least if the SLI systems do have the same power limits on the gpus...

 

I also had 1070SLI on 1080p, the 100% reliable fix is to flatten the boost curve so the cards never go above 1.013V. Loading the cards heavier so they hit power limit and clock down, is not a 100% fix as on low load they will clock back above that point which risks a crash

In the below example, since the MHz is no higher on any point above 0.95V, the card will not use those points on the curve and will never ever use a voltage above 0.95V

 

I guess you only read the last message... what do you think I was talking about? Plus 0.95V would not help the guy in question cause his cards still crashed at 0.9V, while my recommended value of 0.8V (which you would have known if you read at least 1 page back) so far seems to work I guess... Anyway what do you think this "flattening" is? It's an undervolt. Most likely you knew what to do cause you saw it somewhere and it might even come from Falkentyne or me anyway as we were one of the first few to have suggested an msi afterburner curve for this as it replicates what the .8A vbios does. At least I know I figured it out on my own because it was BEFORE this thread was created and the only fix available was the .8A vbios which was just recently discovered at that time. And well, if you also figured it out on your own, then gg. But you could have still read at least 1 page back before commenting unnecessarily.
PS. what you did (undervolting) still may not be a fix, but a temporary solution and the problem could return any time or the card could simply totally die.

 

Hello. I've been encountering the same problem as mentionned and i have the K1611N (flagged as a defective unit)
I've read some of the posts on the thread but i'm utterly bad at understanding all of this. I still have my warranty (a few months left) and i would like to know what should i do.
I've read about reducing voltage (?) with the afterburner, but as i said, i don't understand anything and i'm scared to do anything bad, i wouldn't want to break the computer even more ...

Let's hope this reply won't be unseen

 

Hey there! You get reboots etc? If so, then you can either: install vbios 8a, decrease voltage in afterburner to 0.9V or slightly lower (if 0.9V does not work) OR push msi to replace your gpu. If you decide to do that then I suggest you open the laptop to verify that you have the faulty gpu first - you will also be able to see whether it was exchanged then. The faulty ones were revision 1.0.
Now if you want to try the vbios: use msi official tutorial on installing it (there is a how to under the vbios official file for your laptop - but that vbios would still have the same issue so you would need the .8A which lowers the voltage).
OR you can do the same thing as the .8A vbios manually using msi afterburner: open it, press ctrl+f, then flatten the curve COMPLETELY beyond desired voltage (If you want to OC I would do that beforehand as you can drag the stock curve up pressing shift and dragging one point). Whether you OC or not, flattening the curve horizontally beyond the chosen voltage (aka the same frequency for all voltage points) will result in the gpu never going higher in voltage. Keep us posted!

PS. if this is something that only started happening recently it is possible you don't have the same issue. Run hwinfo and check for thermals and throttling flags being tripped. Also check the event viewer for errors around the time stuff happens.

 

you will need to use afterburner by msi and hit control key and f I think. to bring up the frequency eq. you can go stock or OC if you want but the idea is to under-volt the card by taking your highest frequency at say .950 mv. and flattening out the curve from that point to the right if that makes sense. this issue is that card can not go over 1.013mv. or it crashes. by doing the above you are under-volting to a 0.950mv. you can go even lower if you are having high temps as this will also reduce the heat.

 

Thanks,i'll try that.

It's not something new, i've been encountering full crashes (complete freeze, randomly colored pixels, and complete restart of my computer)
for quite a while now.
I'm kinda scared to push for the warranty.
Also, i'm very sorry but i didn't understand everything you told me, my native language isn't english, and i'm trying my best to understand / write.

 

Well, i'm such a newbie to this ... So i understand i should put the max frequency at 0,9mV.
What's the highest frequency value i should put it to ?
And what if i mess up ? willmy computer be broken ?

Sorry for all these questions ... :/

Edit : i can't flatten the curve, it always goes upwards at around 1V

 

Use MSI Afterburner and make sure all points past the 0.900v point are completely flat and read the exact same mhz as the 0.900v point. The speed will be shown in white on the left when you click. Just click and use the up and down arrow keys to move the speed manually with the keyboard, then click and verify. That always works. Don't forget about the farthest right edge either. Then apply and save a profile.

 

The trick is to make sure your gpu is cold. Run the fan boost for at least 10 minutes. Then edit the curve editor. The reason is, the curve editor will display a different curve depending on your gpu temp. You wanna edit it when it is showing the default values.

 

Hi, i am new here, and i need some help or advice, i try to read some about throttlestop and undervolting. But i am new it this. Please help, i have MSI GT73VR i7 7820, with 1070, and not really long time ago its seems like Dragon Center dont do hes job any more. When you shifting power options its dont look like something realy change, just temperature seems like, but not performance... Can some body help with that? Sorry i'm not in the right topic

Sent from my SM-G965U using Tapatalk

 

This is no simple task, there are some tutorials on the forum already which can help you to get to average point. However, it does require time and patience to reach sweet spot for your system if you want more.

 

when i use custom fan on dragon center, sometime fan switch to to 0% , that can kill my computer if i do not notice it.
does someone get same problem ?
do there is a new vbios for gtx1070 ?? , I am still using the chinese vbios , that work fine. I can overclock +100 mhz clock, +200 memory

 

uninstall dragon center and download silent option. http://download.msi.com/uti_exe/SilentOption_1.0.1510.2301.zip

 

there are 2 versions of this card. ver. 1.0 ver. 1.2 I think or 2.0. anyway vbios is contingent on what version. I have the 1.0 version that came out of a MSI trash can. I have the OC bios version and can hit 2048 clocks at 1.000mv. temps around 75c. with cooler boost on.

hacktricks2006

 

uninstall Dragon center. its garbage. throttlestop is great for undervolting your cpu and creating a battery profile for using web browsing and streaming. You can use afterburner to create gpu profiles as well and make an Overclocked and a super-undervolt to go with your TS battery profile. that way you dont need mux switch you get like 2hrs on a good battery.

 

you have to manually mouse point square slider and move to desired position. if clicked on you can use arrow keys to get the exact value. it doesnt have to be 0.9 mv it can be lower if need be for extra cooling or for creating a battery profile so you dont need mux switch anymore. undervolt the gpu with afterburner and undervolt cpu with throttlestop. takes time and patience. but helps make you better aware of whats going on with you rig when you get it down.

 

As @1GreyGhost1 says Remove Dragon center it cause more hassle then it's worth, then use throttle stop and slient options it's what I use on my GT72-6QD with its GTX1060 aftermarket upgrade and I stay under 80c on GPU and 75c when benching with my insane overclocking on GPU without the cooler boost.



Sent from my SNE-LX1 using Tapatalk

 

One more step, remove Dragon Center, then go to BIOS, load defaults, reboot, go into BIOS, unlock the extra menus, set a negative slope on the IA curve so the CPU thinks its using less power than it really is so it stays on max clocks for ever, then use TS to undervolt/tweak your CPU.

For the fan curves, I talked to Svet and got his EC editor and modified my EC fan curves, so I always have a cool running system without more programs running in the background. The stock fan curve is something silly that goes to like 60% max RPM at 90ºC..

 

Hello! I had .3A firmware, I flashed my 1070 ( rev1.0) with 86.04.5B.00.8A vbios with unlocked tdp and my voltage does not exceed 0.8620V. I read earlier reports that this vbios has a voltage limit of 0.881V and was disappointed when I saw a lower voltage. With this voltage (0.8620V) I was able to achieve stable frequencies of 1873 MHz, although my chip is capable of more. At .3A, I could take the frequency 1898MHz (0.881) through the AB curve. When flashing the .8A, I counted on this performance, but the applied voltage failed. There are several versions of bios .8A or should it be so? I would be grateful for your advice. thanks
P.S I'm using google translator. I apologize in advance for possible spelling errors.

 

A8 -max 0.893mV "no boost clock" ,,,for custom TDP need 56.A3

edit,,and max stable clock for all games "A8" + 205 core only ,,OR range 1850-1889 think MAX
1850mhz tested some years --full stable ,,vram OC only get more power limits

 

Thanks for your feedback. But I can't reach - 0.893V. I have a limit of 0.862V.

Spoiler: GPU-Z screenshots at google disk

https://drive.google.com/file/d/1wFQrZiNSVHp2apzcwMJW4PQPK71dESWM/view?usp=sharing
https://drive.google.com/file/d/1SOgZRtWwiBzE5Dve0LI5-jxA7YQ0qJa4/view?usp=sharing
https://drive.google.com/file/d/12crgRctBc81_t8obHoqZF8fY957hZaR2/view?usp=sharing

 

for MOD gpu TDP need same mod for EC fw or possible get limits

edit ,,find FALKENTYNE post in tdptweaker topic .P or PM
edit 2 ,,test stock curve + L "voltage lock in curve" for 0.881 or 0.893 for test only voltage
eliminate instability

 

C:\Program Files (x86)\MSI Afterburner\Profiles
fing gpu ,,VEN_10DE&DEV_1BE1&SUBSYS open in notepad

[Settings]
VDDC_Generic_Detection=1
enable in afterburner - voltage monitoring
add line and see voltage in afterburner or any OSD