Mobile Pascal TDP Tweaker Update and Feedback Thread

A guy shared it on a reply to a thread I posted a long time ago recently - I guess I'll give it a go when I pull the heatstink off to re-apply LM. The worst that can happen is I need to flash back the default VBIOS.

 

2070? 2080?

 

I meaned RTX2060 Mobile (from 2020) which is using different chip ID than 2019 variant.
Flashing RTX2070/80 should be a lot easier because there was plenty of those bioses from max-q and max-p series with same chip ID as I remember.
Just use nvflash with checks disabled.

Flashing different BIOS could give you two issues :
1st - will not work at all (error 43) - so it basically mean that you need to reflash your stock/working bios (this is mainly related to RTX2060 Mobile with 115W because it might not work until properly tested),
2nd - some of your GPU Video output might now work, because different bioses are configured for exact laptop hardware, sometimes you will lose DP/HDMI or NV panel will show you extra output that doesn't even exist.
Also newer bios could provide newer GOP, that sometimes might be usefull with some monitors/TVs - this is basically an FW dedicated for video output.

Few years ago nvidia provided an update for pascal that brings compatibility for newer display port revision trough GOP update.

 

Even CH341A will be enough - for that task I recommend you to use AsProgrammer - which is free, providing more flexible and simpler UI and also larger database for 24/25 chips.
There will be no issue for chips like 25x from Macronix (which is popular in MSI laptops).
Default CH341A programmer app is quite clunky and not always working properly.
You also need to be aware that for programming NV BIOS chip might require to use 1.8V adapter or source 1.8V from additional voltage regulator etc.
Before you do something just take a look into datasheet - anyway 1.8V adapter it's a good thing to have.

 

I prefer flashrom in Linux over anything windows. I have the 1.8v adaptor as well!

 

I'm unaware about linux tool for doing such thing.
But on Windows AsProgrammer will do the trick for ya - anything from dumping system/EC/vbios is possible, also flashing, checking and setting protection bits.
After you flash new bios be sure that you set protection bits on - this might be non necessary required but still there might be some checks for those - so it's better to set them locked again - it will take a second or two.

 

Hello, i tried to flash my Acer Predator G9-793-79HC with a Mobile GTX 1070 bga card today. The vbios is on a MX25U4033E Chip.
I used the CH341A with a 1.8V adapter.
I used the stock vbios firmware and edited with mobile pascal tdp tweaker.
I unprotected the chip, erased everything, and wrote my edited vbios to the chip. The veriffication was also okay with AsProgrammer. But i forgot to set the protection bits on again.
The pc is booting (Acer Bios sound works), but no video at all on the laptop display even on a external monitor.
Its a big thing to set the protection bits on again? Or do i have a another Problem?
I have a Backup. Should i reflash the original vbios again?

Edit vbios dump from nvflash:
https://mega.nz/file/n49yyCIL#na1RsQjh21C-VpSwc6LuNpAVqck3izLnnBMefT9JoGw
Original vbios dump made via nvflash:
https://mega.nz/file/D89yFIIA#UHq2eWsoBXmTnxR2Dn9YWKvPi4LyhnXqMloptje9LC0
Original vbios dump from asProgrammer:
https://mega.nz/file/Wx9iVKwL#ISQ0sa5UtDFVN9_X8D_bMlBncz9tmOwIoYlhTgUyElA

 

Is setting protecton back on required when you flash an edited chip?

 

I cannot tell you about BIOS (is it right or wrong), however there is a very important and simple thing that I'm doing everytime I flash or dump something.
Because many BIOSes could be dumped via software on linux/windows I always recommend you to check MD5/SHA from that file and compare it to dump from programmer.
Bad dumps are nothing new especially when you're using poor quality clip or program is not detecting properly flash chip/or set wrong device to work with.
You can even do this on Windows context menu (there is an option to add powershell script for hash etc.) or using HEX editor like HXD or Flexhex with ability to compare two files side by side.

Edit: from what I see software dump is 256kB and programmer dump is 512kB.
Programmer will always dump every bit of information for chip (even filled with 0, to match it's real world capacity).
I do really recommend you to mod software dump from nvflash and also make a second one based on what AsProgrammer did - it might work.
Unused space on flash chip will be filled by 0 anyway if you erase that chip and program it back - modding HW dump might be useful because it might contain something that hasn't beed dumped by nvflash.

Hard to tell, but from my point of view it should be done just for safety reason.
Also software flasher are definitely aware of checking protection bits even without informing you visually about it via GUI/CMD window.
Setting protection bits will cost you no more that second or two - I do really recommend to that (even if I'm unsure about it necessity) - when something goes wrong it will be one less thing to worry about.
Especially when you need to disassembly your whole laptop to simply get to that chip for a second or two to make it done.

 

My main problem is, i get a blackscreen on my laptop monitor and also on my external monitor via displayport.
No Videooutput at all. But im know my pc is booting, because i can hear a post boot sound from the speaker.
Maybe i did something wrong in the flash process?

 

I'm using AsProgrammer_2.0.0a14 - if you launch any version there is a link to project site, you can download newer version anytime.
Very important thing is to be 100% sure that there is no clip issue and it's sticking to the chip all the time - I had many bad dumps or broken flashed mostly due to clip quality (most of them are crap), however desoldering IC every time will take to much time to play with.

If your clip, programmer and also adapter is perfectly fine - this is what you need to do :

1) press hardware and check CH341A (if you're using that programmer, if not pick other option),
2) select SPI (because we're working on 25 series chips in that case),
3) press read ID - program will try to detect your chip, sometimes it will show few similar (different companies etc.) pick one that correspond to your IC or it's closest too (comparing datasheets might be sometimes required),
4) load your desired BIOS,
5) press on that small arror visible with Program IC (unprotect -> erase -> program -> verify) - if everything is ok, then set the protection bits.

For dumping you simply press Read IC and after you done three first steps and voila you can save that file.

Because **** and issues happen sometimes it's not a bad idea to dump bios twice or more and check MD5 of each one to be sure that there is everything ok.

Hardware issues in my case are mostly related to that poor quality clip, software side will mostly be related to wrongly detected/picked flash chip.

Everything you need (apps and also updated drivers) is listed here :

https://4pda.ru/forum/index.php?showtopic=884713

 

Im already tried to flash my chip and dont have 5 dumps. So i cant compare the MD5 Hash anymore.
I only have 1 dump from AsProgrammer and 1 dump from Nvflash and for my bad these two are not the same.
and btw. a big thankyou for your tutorial!

http://prntscr.com/tersit
Attempt Black Screen Fix?
Should i choose Preserve or Remove Header Section?

 

You need to try each option, because not all systems are the same.
Given that you already have the programmer, testing that will take you 10 minutes.

 

Maybe techpowerup database will help you find proper dump (just in case) ?
If both dump are different please compare only 256kB block, because rest will be probably FF.

 

http://prntscr.com/tes9cg
Everything from the adress from 00028790 below is different.
The FF FF FF starts @ 0003A4D0

169EEA3E73E189962172C17AF713FDCBBFD8790FEDD5769C69B62A8988231B36
D24D776D43259798A3E67D26EBBA0F59DCCF2A862285A272086D8E2281DA8D86

 

So many differences mean that something went wrong while dumping - now we need to exclude which method fail.

I looked into TPU database and I found bios with same build version - which correspond to Acer notebook (Triton or Helios, dunno ).

Link to download :

https://www.techpowerup.com/vgabios/189329/189329

Edit : I compared it to your nvflash dump - 100% same bios (via MD5 and SHA checksum).
So probably AsProgrammer dump is broken.

I do really recommend you do this :
Flash TPU bios via AsProgrammer - after you do dump it somewhere and compare everything up to 3F1FF register (everything above that point should be FF - not used).
If your new dump will be correct then you will be absolutely safe for future modification.
I'm always storing software and hardware dumps just in case

 

I cant get a connection with my soic clip anymore, i will end up soldering.

 

Hello, did you restore your PC to working state ?

 

Not yet, i cant get a connection to the chip. Maybe i need another sop8 clip. The clip delivered from the ch341a sucks.

 

If you're familiar with soldering then you can try to use PCB adapter for CH341A like I did several times.
This might be good idea when you're trying to use clip designed for SOIC packages on MSOP chips which are little wider, but enough to lost connection.

 

I ordered now the POMONA 5250 clip with some female/male wires and i hope i can get it work

 

Most of the times it's not the SOIC clip issue but rather that the PCH interferes with the BIOS chip.

 

wow... it's really amazing. When I see you, I feel like I'm looking at myself. I'm in the same trouble as you. CH341a, GTX 1070, MX25U4033E, BlackScreen... all elements are the same as me. The only difference between you and me is the laptop model. My Aorus X5-V7 had the same problem as you and now my laptop is in the Gigabyte headquarters in Taiwan.

 

I see this place is still as interesting as ever.

Also to note from reading the last few pages. 1 does not need to do a shunt mod as you can't do anything with it even it it actually worked. The 20 series are hard locked at a max core clock. 2080 is 2100 Mhz core.
The real problem is these cards are using entirely to much voltage at the clock speeds they are running at. Try running lower voltage and see if that helps.

@Dantean
Also looking at the 3 different bios files.....1 of them has a different checksum. The original one dumped with the programmer. I would flash that one back and see if your machine will boot correctly.

If so, then take that one and edit that file. Make sure the checksum matches the original. And it should theoretically work. (As I just tested it and it came back with same checksum - I don't have a card to test it though.)

As to the clip.... Just get a nice good hold on it and hold it with your fingers. This should be enough for the contacts to touch and hold. I have had to do that on quite a few occasions. (Just read chip first, once a good read is had, then flash)

 

With a MAX-q card it helps a lot. On certain games it goes from 1300-1400mhz to 1900mhz.

 

Which part helps a lot?

 

Shunt modding

 

Do you have some benchmarks with this shunt mod? It has struck my curiosity.

edit:
Okay, was thinking you had a better score than the one in your sig.
Are all these guys using shunt's as well?
https://www.3dmark.com/compare/fs/22564653/fs/22853374/fs/22642443/fs/22953676/fs/22933172

 

I'm being limited by msi's 230w total device power limit.

And no, they are using eluktronics laptops which are eluktroboosted (tdp boosted).

DrewN is using a 2080ti egpu if you didn't notice.

 

One of the reasons why I didn't go with MSI. Super locked down!

Ah yes. I knew he was using a 2080 Ti. Got a lot of folks pulling that move and jacking up the scores on 3dmark.
Meant to pull his score.

 

Can you limit Cpu power to increase the graphics score? Shunt mod doesn’t help on the Cpu. And higher graphics score should easily offset the higher physics score to maintain total score. 3DMark benchmark software doesn’t give much credit to Cpu as for the graphics cards.

 

I am curious how these eluktroboosted rtx cards would behave with shunt mod??
Because I read somwhere that this eluktroboost is some low level software injections.

And I am thinking of buying Polish brand hyperbook GK7NPFR with R7 4800h and RTX2060 which is the same as RP 17

 

Yes, I'm aware of that. I've tried lowering the cpu clock to 3.9/3.8ghz but the gpu scores will also be lower.

Firestrike runs at 150+ fps so it's hitting a cpu bottleneck.

My timespy scores are better as it's more gpu bound. Its in first place for i7+2080 Max-q

 

Try running it with just one core to see if you can get a higher gpu score only.

 

He can also run Cpu at max clocks but cap the power limits because the Cpu won’t max out package power in the graphics parts of 3DM Fire Strike.

 

True.

 

I could limit the cpu to 4 cores in the bios.

Tried it, at its peak boost the cpu and gpu combined power draw will trip the total device power limit.

Really need to find a way to bypass it.

 

Then try put Cpu PL power cap limits close to trip point for higher graphics score. 3Dmark software don't max out the Cpu power. Neither in physics or rest of the sub tests.

 

Unfortunately i don't have that fine a control over the power limits. They can't be changed in throttlestop and my only option to get the cpu to boost to 4ghz all core during the physics test would be to use the imon slope trick and completely bypass the power limit.

Edit: I COULD fine tune and find the optimal slope value to give a specific value but that'll be too much of a hassle.

I haven't gotten to benching while connected to an external display so there's still some headroom. Both by increasing the avaliable power now that the internal display and igpu is off and from connecting directly to the dgpu.

 

And you can't bypass total power limit on the laptop ? I can already see there is a shunt resistor on the battery, it should be the same around the power connector. Last hope would be to solder a wire from power connector to the shunt resistor or the RTX2080, making the integrated total power limit partly bypassed.
To get the best performance/power ratio on the cpu you need to fine tuning using constant voltage (and not offset)

 

If only everything was measured using shunt resistors. I don't think it worth the risk. @Falkentyne has run into this problem before.

As for the cpu I've already tried that, using a constant voltage doesn't have any tangible benefit over just using an offset. If anything it crashes when a single core tries to turbo above 4ghz.

 

I tried limiting cores in bios and that kinda bricked my laptop til I finally got my spi programmer and was able to re-flash a backup dump.

Mines an insydebios. Yours was ami if I remember from the other thread - behaviour may be different but if u don’t know that it’ll definitely work, wait til u get your spi programmer if it hasn’t arrived already.

See this post by mr fox for more info on why this may happen - http://forum.notebookreview.com/thr...-power-button-led.833565/page-2#post-11032819

 

Limiting core is not beneficial on 3Dmark, physics and combined score depends a lot on CPU multithreaded performance, in addition you can limit core with windows (msconfig then startup then advanced settings)

 

You should be using a Pomona clip, not those cheap worthless chinese bulk made clips.
Use this.

https://www.amazon.com/CPT-063-Test-Clip-SOIC8-Pomona/dp/B00HHH65T4/
Then male to female jumper cables (might as well get all three types like this:
https://www.amazon.com/cables-multicolor-alambre-tableros-arruinar/dp/B01EV70C78/

To hook up the Pomona clip to the 1.8v adapter or to the programmer (only if it supports native 1.8v).

 

You need the power ID of the laptop that has a 330W limit, if it even exists anymore. But I don't know which one of the new ones have a 330+ or higher limit because of MSI using 2x230W PSU's and stuff from what I've been seeing instead of 330W PSU's. EC RAM register "E3" (In RW Everything --->click the EC tab) controls the master system power limit. On GT73VR, changing this value higher by "1" on GTX 1070 systems changes the master limit from 230W to 330W+ (10 to 11, on skylake GT73VR, or 90 to 91 on GT73 kabylake. Funny that "11" also works on kabylake (the old system EC values are still present). What's funny is "21" works too--a GT72 (or GT62?) setting, however the GT72 did not come with an unlocked CPU SKU, so the CPU will throttle to 45W after 28 seconds, even if you have a "HK" CPU! This is because there is a 330W sku that was apparently never released.. Or there MAY have been one GT72VR with an unlocked HK CPU but not sold in USA.....anyway,

For GT75VR, 80 was GTX 1070 230W and 91 (again) was GTX 1080.

I have absolutely no idea about any newer laptops from MSI. Never seen or used one, and I don't know if the old EC values work anymore. Someone tried 91 on a GTX 2060 system but it still throttled.

Does this laptop use 2x230W bricks?
Do any use 330W anymore?
Try to find a MSI RTX laptop with a single 330W power supply. Have that person post their "EC RAM" in RW Everything-->EC tab.
Then try using that value in register "E3".

Someone would have to literally test ONE VALUE at a time, starting from the original value shown and increasing it by 1, and seeing if the CPU throttles hard. That will take all day. But there's absolutely no other way. Zero. (I know nothing about shunt mods).

IMON SLOPE/Offset mods to make the CPU report less watts doesn't work for this. The power the EC itself reads seems to come directly from the VRM itself.

I remember some similar issue with the GT73VR and 230W power limit on GTX 1070 cards. Models which had SLI 1070's still had the same EC value (90 for kaby, 10 for skylake), but if SLI was installed the power limit was now 460W. So something must be detecting either the 2x230W AC capability or the EC sees that two 1070 cards are installed and ignores the 230W limit. But you could just get a 660W limit by changing to "11" or "91" (2x330W) for the SLI 1080 system

Can't help or suggest anything else. Find someone with a 330W PSU system with RTX MSI SKU similar to your model and ask them to post their EC RAM map for you in RW Everything. Maybe it will work or maybe it wont. Or you can try entering 1 value at a time and test throttling yourself. You can start at your base value and increase it by 1. It's hexadecimal so like if you start at 80, you go 80, 81,82.......89, 8A, 8B,C,D,E,F, etc, then after 8F, then 90, 91, 92, etc etc. Then 9A,9B, etc, then after 9F, then A0, A1, A2, A3, .......AA, AB, AC, AD, AE, AF, B0, B1.............and so on.

Have fun.

 

I can check on my GT75 with a 2070 what value is on that register, a bit funny that the laptop ships with a 330W charger, but I have a locked CPU and a 115W GPU, so I think the 330W is just because the high end model with the 2080 and HK CPU uses two 330W and its just cheaper to just buy a lot of 330W than to send a 230W on the lower models.
I still even haven't changed the IMON slope but for some reason the CPU holds 63-68W during games, so yay for not ultra crippled firmware at least.

 

any background info on how register E3 was determined to be the power limit? Curious as to how to apply this to my laptop (non-msi)

 

Comparing various models and trial and error.
@Falkentyne spent a lot of time to figure it all out, I admire is patience to find it all.

 

MSI laptops have NOTHING to do with non MSI laptops. EC code is NOT similar between them unless it's a MSI whitebook (e.g. MSI 16L13, WT75, etc).

 

Here it goes, a print with and without the charger connected.

On battery:


On charger: