Mobile Pascal TDP Tweaker Update and Feedback Thread

please help
I have been planning to use this tool for a long time, but I'm still worried about EC for msi
i don't know if there is a problem ec or ac adapter but gpu under the cpu + gpu test falls down to 75w, that's normal ?? I didn't notice it before
can only be a problem ac adapter ??? it on the A8 bios no boost is 115w is ok, but not fall to 75w

I just did an update win 1909,, can this only be done by the gpu driver or something ?
trying default setup ,, without oc it's the same ,,temps no more 70C ,,card 1BE1

bios 86.04.5b.00.8a ,.230w psu ,,gt75 7re

 

Which OEM vbios do you need for them types?

Sent from my SNE-LX1 using Tapatalk

 

GPU down to 75W? Unmodded BIOS?
Looks like you're running on battery? Or the battery level is below 30% or you removed the battery internally?

EC draw limit can be increased to 330W by changing the value, using RW Everything, in EC RAM Register "E3" by increasing it by 1 (note: if the default value is 80, change it to 91, not 81. Some boards have the 1070 ID as 80 and 1080 ID as 91, while others have the 1070 ID as 90, and I don't know which boards do that), but the 230W power supply will shut off if it exceeds 250W.
But this clearly isn't your problem.

If you're going to mod the 1070 to 200W on a GT75, you should use the 330W adapter and make a script to change the EC RAM E3 register whenever you boot (batch file or whatever)

 

Thanks
so it's not normal
the battery is still around 90-100, more it starts when charging
i have suspected ac adapter for some time, before ,,i used 120A year -now problem ,.now is maximum 112-114A for avx ,,but tested without OC for gpu and cpu "still same"
I'll try another driver, in new windows I found nothing else
the adapter i i measured is normal for 19.5V to get up to 19.8V?
battery wear now 6. /. sometimes 10./. ,, is still ok?


some better measurements than hwifo ,, nvidia-smi ?? but all seem ok ,,only tdp is problem cmos same reset EC or need use ec button ?
,.I have updated ME fw and some things are different now. can it affect gpu? but I promise

I'll try a few more things



no must stress cpu+ gpu ,,only gpu + gpu mem ,,kombustor 6.5gb GL test ,,120fps games do the same to me

 

I don't need these vbios, I was just asking what was @RMSMajestic saying, I did not know GTX1060 have an hardware ID of 1C20 and 1C60

 

ok ,.this is really weird
tested several times ,,A3 bios ,A8 bios ,new driver ,,old driver ,,,all same
but after use nvflash recovery "InfoROM" on A8 or A3 bios ,,gpu work normal all time use 100w-115w really, what it is ??

inforom contains something about tdp ??? going to edit ??

I still try, but gpu is still ok, if I just update, again falls to 75w ,,,need more info about "inforom" ?????????

edit.............
is random problem temps ? is ok 60C ,,after flash work normal or some long test normal ,,cold work
after restart or random test ,,not work ,,all time drops to 75w ,,


I'll try to get a 330w adapter, I can't think of anything else power problem or temp on gpu "mem" ?

but I think it's power ,,some combo test cpu + gpu,,come tdp down more frequent

 

1C20 - vortex
1C60 - msi gt72 vr

 

Yep I own a GTX 1060, it's the 1C60 version as well.

Sent from my SNE-LX1 using Tapatalk

 

I tacked the resistors down with some liquid metal, as I left my soldering irons at a friends house...

The result was only ~4% increase in performance for the 13% increase in power. It's possible one or both of the resistors were not making the best contact.

Good enough for #3 on 3dmark for 9750h and 2070 mobile. https://www.3dmark.com/compare/spy/11387794/spy/11356856

 

Did the shunt mod on my GS75 a few weeks back as well.

Now it performs as well as a 150w 2080

 

Hey mate can i ask you how you oc youre mem at 3200mhz??
thank you a lot

 

GamerNexus recently made a video about it

As you can see using LM isn't reliable, sometime it could be nearly as good as a good soldered shunt and sometime it's like stock shunt, it depend on the type of load

 

It's been consistent. Not sure it's worth the heat, though.
Stock. the VRM's get to 90°C. I'm sure they will be > 100 C with a 130W mod.
I'll only leave the LM temporary until I pickup my soldering iron.

When I remounted, now I'm having CPU temperatures problems :-(
I'll need to clean the LM again and reapply once more arrives.
Anyone have any tips on mounting the HSF to get good contact? I used alot of LM. Hopfully, it's a piece of fiber or hair in there..

My Core 0, Core 2 and Core 4 are getting too hot. Anyone know which side of the die they are on?

 

Top side.

This is only 4 cores since it's kaby, but the numbering is the exact same sequence. It even applies to 8 cores (0,2,4,6 at the top).

*edit* first picture was horrible.
Maybe this helps better.

1) Kaby Lake with die showing.



2) here is what an 8 core would look like, assuming the same orientation.
You can see a 6 core is thus the exact same sequence.

 

I overclocked it in the bios. Msi has a hidden bios where you can overclock ram, undervolt cpu, adjust loadline etc.

One you're in the bios, hold down Left Alt, right control and right shift then press F2 to access the hidden bios. Then go to Advanced, overclocking and then memory to set memory clocks and timings.

 

@Falkentyne Have you experimented with using more powerful AC adapters? I want to get a dell 330w adapter and a barrel connecter adapter to use with my GS75 but im not sure whether my GS75 will accept it.

 

amps no problem ,,voltage is same ,,not get noting ,,still need mod ec ,maybe stability is better ,,now tested 330w
,.ram oc ,,is easy but exist sa voltage monitor ? my not exist ,,but without offset ,,still get 2933mhz 18.20.20.xx oem ram ,.max 16.20.20. 1570b in xtu
this bad thread for this .P
gt75 7re

 

Can you post in your native language so we can try translating it ourselves?

 

sorry this not my language ,,i use same translator ,.you probably don't know Czech ?
loadline use 1 --0.01 or 0.05
ia vr setup
AC -1
DC -1
max vr ? my is 112.5amp 450vr "vr- amp x4"
first use offset ,,test ,,find voltage .,use hidden "additional" vcore ,,oc tab not work .,no need offset
,.i run speedstep or max all time and all is ok ,.only better or test custom ,..c-states "random app close"
admin bios .,f2 + left alt - r ctrl + r shift .,in bios combo key ,.not must work for all notebook

@Falkentyne will tell you better

 

F2 must be pressed last (Press the other 3 keys first, then F2)

 

hi
when we're here, the 330w adapter is still the same ,.although it may have improved a bit
Have you ever tried pcie limmit 75w? or similarly
I can't get a programmer all the time
thanks

edit ,. or any better way for 1BE1 1070gtx ,,A8 Bios or exist better?
think "rweverything" ,, but I don't know at all, I can not read "hex"? in it to change something
where you can learn to read it

 

My CPU gets really hot as well, but inconsistently. The temps peak really high without reason and for not even a second. I've repasted many times but it's still here. Any idea what could cause that ? I never noticed any fps drop in game or something tho, but I'd still appreciate it if it was fixed.

 

I'll ask while conversation seems to be around that..
Anyone knows which exactly are shunt resistors on Clevo 2080/2070 MXM board?

 

Reeks of bad heatsink to die contact. When I had good contact, I could hold 99W TDP indefinetly.

Quick way to check your cpu core to HSF thermal contact:
datalog the following
Core temps 1...6
CPU Package power

Set the logging rate really fast, like 200ms

Open Prime95 and prepare Small FFT and 6-threads

Note temperature just prior to starting. Start test. Note temperature ONE SECOND after starting test. Stop test.
Calculate DeltaT / DeltaW

For me, I get (58°C - 44°C) / (98W - 13W) = 0.16°C/W. That's my best core.

A good number is probably 0.16 - 0.20 if you're using liquid metal.

My bad core is getting:
(92°C - 44°C) / (85W) = 0.56°C/W

By doing the test so quickly (less then 1 second) eliminates the HSF and gives the die to HSF contact thermal interface only.
If you want to evaluate your die to AMBIENT to die thermal contact in °C/W do the same test, except you will use infinite time (~20minutes+) instead of a really fast time.
If you want to understand why this works, you will have to understand The Heat Equation, which is usually taught in Partial Differential Equations and Heat Transfer.

 

Accidentally shorted the shunt with solder... sucks because I got a great mount, good temps.

I decided to see what it would do with some benchmarks...
I limited it to about 0.925V otherwise the power consumption will spike like crazy.
This setting was peaking around ~200w for the GPU

Got #1 score on Timespy for 9750h+2070, even #6 score for all 2070 mobile, competing against 9900K CPU's

https://www.3dmark.com/spy/11488801

https://www.3dmark.com/newsearch#advanced?test=spy P&cpuId=2453&gpuId=1243&gpuCount=0&deviceType=ALL&memoryChannels=0&country=&scoreType=overallScore&hofMode=false&showInvalidResults=false&freeParams=

 

Niiice

You dethroned @Prema as well, good job.

 

How check mosfet temperature?

To upload vbios to RTX cards I can do it using nvflash or they are blocked like GTX 1070/1080 and I need a programmer?

 

all need programmer ,,rtx or gtx ,,no way bypass nvidia cert in nvflash ,,not use mod nvflash only brick card

mosefet temp ,,exist diode ? i no ,,only core temps is positible check

 

Told you these mosfets are reliable lol, nice bench btw, I couldn’t find the resistor in the end so I’m gonna have to buy it... or I’m probably gonna sell this laptop

I can’t wait to get my hands on an rtx 2080 super laptop, imagine shunt modding one of those lol. You’d get near 2080 ti performance!

 

could I ask for compatibility with the Clevo 2080 vbios? Vbios in attachment.

 

You'll reach desktop 2080 super OC Max. No way it's getting near 2080ti performance unless you consider 25% slower "close"

 

Need some help guys..

I dropped a ball of solder on an IC located near the bottom nvme slot, to the left of the battery connector. I could not get the solder off the pins (impatient...) And ended up destroying the chip.
Anyone with one of these systems can ryt to identify this chip for me? You may need a microscope or good macro camera and try different lighting to read the text.

BTW, if you use liquid metal on something, don't expect to be able to solder to it in the future . The liquid metal imparts some metalurgical changes to the copper / solder on the existing component, it was impossible to get leaded solder to stick, no matter much I cleaned / fluxed it

 

For those who want to calculate needed shunt resistance, there is the formula


I have a similar formula to calculate thermal pads (and maybe thermal paste) quality needed to cool down a chip with a wanted delta Temp

 

I just follow buildzoid's formulas lol

And can you please share the thermal conductivity to delta T formula as well? Its easier to read a post than scroll through a video.

 

It should be the same formula as buildzoid, it's a basic formula

Very good thermal could be fujipoly at 17W/mk, there is also some alphacool at 17W/mk, otherwise there is lower grade but still very good at 14W/mk, 12W/mk, 11W/mk, grizzly thermal pad are only 8W/mk, which could be quite limiting if you are like me and you want to cool stock mofset of a 580 at 1.3v and more
Some result here :
https://www.overclock.net/forum/67-...insanely-low-temps-must-have-recommended.html

 

Higher W/mk numbers doesn't necessarily means better. I would take Gelid pads any days over the overpriced stiff FujiPoly 17w/mk pads.


And alphacool is in short same brand as FujiPoly.

 

I don't really know what really happened on this review, my only theory is that is wasn't making good contact on the die and on the heatsink, I think they are really 17W/mk but poor heat transfer between the chip and the thermal pads it not ideal at all, maybe if we add thermal paste it would help ?
However this hardwareluxx review show that there is improvement by going for these 10+W/mk instead of the common 3~7W/mk, that's why we should use these "high performance" thermal pads on laptop, I already see too many dead/shorted mosfet because of heat :/

Do you have any review on the gelid thermal pads ? I only find review of this pad on nvme SSD which are not comparing these with regular thermal pads.

 

Pads can't be compared to pastes, it's not that simple.

Pads are meant for components that have gap to bridge whereas paste is used in between direct contact surfaces.

Thickness plays a huge part in heat conductivity which is why the review shows that the pads performed so so much worse than paste.

So what if the pad is 17w/mk but is a 100x thicker than paste.

The same goes for pad thickness, they come in 0.5mm, 1mm, 1.5mm and 2.0mm thicknesses and the 2.0mm one will perform the worst thermally because the heat has to transfer through a lot more pad.

 

I know that pad can't be compared to paste, but can we increase heat transfer between the die and the pad or the pad and the heatsink by adding thermal paste in between to fill the gap ?

 

That's extremely wasteful and not effective at the same time.

 

paste work fine ,,but week or first full stress
paste on pads ? why not ,,no need ,,need good pads ,,the right pressure
.,but gelid is ok ? I need better, and the gelid pads goes to get it ,,only 12w/mk

 

Gelid has also 15W/mK pads but 12W/mK are very good and last long. Used them many times for different laps.

Both 15 and 12 have best performance per price.

https://gelidsolutions.com/thermal-solutions/gp-ultimate-thermal-pad/

 

nice .,only 12w.mk extreme is available in this country ,.using minus pad 8 thermal grizzly,,is good ,,but vr limit on paste or some better is different ,,problem is time
all works day or week
,.only grizly work now half year ,.still same result ,,stable

 

Grizzly Carbonaut has 62,5 w/mk. It's almoste performes like Grizzly Karbonaut. Bur nothing beats thermal paste

 

I use carbonaut on my 2080 die and it performs just like liquid metal.

HOWEVER, this is because I have lapped my heatsink to a mirror finish and so this presents the perfect surface for carbonaut to perform well.

In desktops carbonaut doesn't perform nearly as well due to the concave/convex shape of amd/Intel heat spreaders which will cause the carbonaut to be pressed too thin in some parts and nor have proper contact in others. The more uneven the surface, the worse it performs.

 

Grizzly Carbonaut ??,,tested ,,for me not ideal ,,only Conductonaut is best
,,carbonaut pads is ultra thin ,,need mod "heat sink" on my gt75 this not work
and still conductivity ,,only good for fast rebuild ,,not more

 

I was thinking about using 2 pads, on top of each other. What do you think about that?

 

Bad idea, it'll be too thick in some places.

 

Not sure how are you all doing the thermal pad installation, but for me this takes at least 10-30 mount/unmount of the heatsink and verifications, especially that most of heatsinks are unified with the GPU/CPU making it lot harder. In the old days, clevo did it nice on p17x/p15x laptops, where GPU heatsink had separate die and vrm/memory cooling, which was very easy to work with.

 

You are right, I learned this the hardway, shorting few mosfet on my 970M before realising that Kepler K4000M thermal pads placement and thickness weren't ideal for Maxwell card.
Unlike on desktop card, it's necessary to cool a mosfet properly on laptop because they are just enough to power the card, while on desktop card it's "oversized" on most card.