Mobile Pascal TDP Tweaker Update and Feedback Thread

Nah. Should have given it the shoe treatmant. Or the cats sitting on your videocard treatment. If I owned a cat. Maybe if I can tame a garden lizard like I once did 4 years ago. Those things look at everything you do and sit there.

 

@CheeZuShRicE I see the version of nvflash you used isn't compatible with the GTX 10 series and that might be the problem. Joe Dirt hasn't been able to bypass the certification and has given up. The vBIOS version you have is correct but try the nvflash that comes with the vBIOS in the link.

https://drive.google.com/open?id=0B28ZFBJtCZWubm5pRTREdWRSZmM

This is what I used to flash my 1070 and modded to run with a higher TDP.

Try this first with the programmer as a last resort.

 

Hello mate, a couple of things.
The moment you are on intel graphics mean you cannot flash your dGPU. Since MSI notebooks to not have optimus but instead a muxswitch.
To help even further, the reason why youre having issues is not because you killed the card, but the vBios is corrupted. On MSI cards you need to solder the vBios chip off, program it, then solder it on again. A clip is not going to be enough since other components seem to interfere with the low voltage chip.

You can download the newest nvflash and flash the default fixed BIOS Prema send you via that tool. Use teamviewer to see whats going on on your screen. If you don't have a 2nd PC or don't feel like you can manage, hit me up I can make it work again.

 

is the 1070 BGA or MXM?

 

Do *NOT* listen to Danishblunt under any circumstances!! PERIOD!!!!

YOU DO NOT NEED TO DESOLDER ANYTHING. The only time you need to desolder is if the legs are damaged and you need to do a reflow.
100% of modders here have used Pomona Clips (or non chinese high quality basic clips that at least worked a few times), 1.8v adapters and wire harnesses. Trust us.

@CheeZuShRicE
BTW if that's a GT73VR, you do know that when you do get your TDP mod working, if you want to go past 150W, you will NEED:

1) Delta 330W PSU. The 230W will power trip past total system draw of 250W.
2) Need to modify the EC RAM with RW Everything to trick the EC into thinking you have a GTX 1080, or you will THROTTLE SUPER HARD.: Note: this also works for GT75VR users. The same EC RAM register has the same value for GTX 1080 identification that is changeable live in windows. No idea if this will work for the new 6 core GT75 Titans however.

You can push 200W safely through that poor 1070. JUST REMOVE THE ORIGINAL THERMAL PADS AND REPLACE WITH ARCTIC 1MM THERMAL PADS! Do not risk 200W+ TDP with crappy stock MSI pads!.

https://www.amazon.com/gp/product/B00UYTTXSM/ref=oh_aui_detailpage_o08_s00?ie=UTF8&psc=1

Proof:

 

its a MXM card.

 

Then the vbios I linked should work because it worked for me.

Sent from my Honor View 10 using Tapatalk

 

The 'board' on top of the 1.8v adapter is just an adapter board, because the prebuilt cheap clips have those plastic blocks that plug into pins. Since they are fixed sizes, it has to attach to something.
With the pomona clip, you are building your own clip connection. Now wiring the pomona clip to an soic8 adapter board will still work, sure, but is unnecessary to do. You would only do that if somehow you didnt have male to female cables (e.g. if yours were male to malefemale to female only).

 

Those people never care. They just want money and more money. They don't care about anyone but themselves. They would watch the world burn while they got rich ....yet if something "tragic" happened to them, they would be the loudest whiners, first.

 

What commands are you using to flash using nvflash?

 

Are you freaking serious????

 

How did the police get involved so fast?
Don't they need a warrant?

 

Here's what I did.
Open stock default MSI bios (the one thegh0sts gave you).

Press Presets button. This will change your default ranges, as well as increase extreme power limits setting to allow more than 115W to work.

Change the throttling temp (affects boost clocks) from 91C to 80C, TBH i don't even know if that setting actually works, long story on that, but I accidentally disabled CPU/GPU temp reporting to the EC, not realizing automatic fan curves depend on the temp, and my GPU was at 88C and rising, with sparkles on the screen....UH.....so much for my 80C throttling temp....whatever. ok.

Keep base TDP at 151W (151000).
Change target (maximum) TDP to whatever. 170000 is desktop range. If you have a 330W PSU you can go higher if you hack the EC (Embedded Controller) RAM (extremely easy to hack stop getting scared). I've gone up to 220W without blowing up my 1070 but I don't recommend it unless you're as stupidbrave as I am.

NOTE: If you are above 150W TDP, even if your max target is *BELOW* 200W TDP, please go to EXTREME power limits and change the 16200 to 19200. You will be glad you did this.

PUBG will throttle at 155W and flag power limit even if your target is set to 200W TDP (132% in afterburner, if you are using 151W-200W range) unless you do this. Why? I do NOT know.
I suspect people with GTX 1080's will also have the same throttling because the "Default" stock 1080 values are the same as the "preset button" values for the 1070. I only saw this throttling in PUBG with 16200 for the first extreme limit. Nothing else (but you can bet something else will throttle). I did not see this in Valley or Heaven benchmark at all. Only in PUBG. When I changed 16200 to 19200, with a 200W target (132% in afterburner), PUBG now acted identically to Heaven. PWR flag would appear starting at 124% TDP, but clocks/voltage would not drop except at 132% or higher..

Then use "Auto" in Skypro and off you go....

 

the 16200 value is what is required for the GTX 1070 to reach 150W TDP. But there is more to this value than what I think @Coolane , @bloodhawk and others realize.
The preset button changes the extreme power limit presets to match the "stock" values for the GTX 1080 MXM vbios. Can't speak for desktop cards (barely any testing from anyone with those).
But for SOME reason, PUBG will start Power Limit throttling at 155W TDP if this value is at 16,200. Changing it to 19,200 fixes that.

I noticed no 3dmark nor final fantasy 15 benchmark decrease when using 16,200 and 19,200 at same TDP. Only PUBG was acting weird like this.

The 19200 value comes from opening a GTX 1080 vbios and clicking preset on THAT, which raises the TDP range to 215W-258W, and changes the 16200 to 19200.
When I saw PUBG throttling despite anything I did, I got the idea of using this at 19200, reflashed, and PUBG was fixed. Then I was like...screw it. Since GT73VR supplies the MXM slot directly from the PSU, and since it HAS to allow at least 230W of total MXM power (there are two MXM slots, because there is a 1070 SLI configuration possible), I knew 230W TDP would not fry the motherboard. The video card on the other hand, well...keep it cool. And don't run furmark at 230W TDP.

Bat file is easy. Just make your own, and make your own. RW file, or you can use these (should also work on GT75VR as well since 1080 has has same EC RAM ID in it):

@echo off
cd\
cd /d C:\Program Files\RW-Everything
Start /b rwe.exe /Command="WEC 0xE3 0x91"
timeout /t 2
taskkill /f /im rwe.exe

This bat file MUST be run as ADMINISTRATOR in the file properties, or the "timeout" field won't automatically close the command prompt.

 

Thats rather obvious lol.

And a lot of the other values are hard Falcon protected. And a few other than are not, are rather risky to fiddle with, without having monitoring hardware hooked up.

 

Yes, but why does PUBG throttle starting at 155W in a bad way (signaling PWR limit and dropping voltages and clocks) if 16,200 is also the default value for basic GTX 1080 ?
Someone messed up and I'm blaming Nvidia or MSI for it.

keep in mind MSI makes the Clevo 1070 card also (maybe 1080 i dont know).

I know you don't want to give information that will make people blow up their cards, but still, that's important to know here.
16,200 is fine for 150W TDP.
But if you even want the max "by MXM 3 design" TDP of 195W, you need 19,200.

My point is, that a *STOCK* GTX 1080 MIGHT ALSO be affected by this!
I'm again blaming Nvidia or MSI. Probably MSI. Who knows. But someone didn't do their job right. As you know, one version of the MSI GTX 1080 had a 150W TDP limit. Another had a 200W TDP limit. If they used the same "Extreme" value for both, it might not cause problems in Heaven, Valley or 3dmark /firestrike, but PUBG definitely picks up the problem!

Not trying to insult you, man.

 

I sit high 1800s at stock around in plebg, I don't get any weird throttle like that.

 

You aren't going to get clock throttling at 1800 mhz, @D2 Ultima, but again I can't speak for 1080's. I'm specifically referring to MSI cards here. I'm guessing the theory is that since there *IS* a 150W MSI 1080 card (with no power connector), I think used in MSI 16L13, the 16,200 value was intended for that.

Then MSI made the 200W 1080, bumped up the power limits, but didn't change whatever 16,200 is actually affecting (probably some sort of rail limit).

i'm pushing 2000-2025 mhz. If you dont know what to look for you won't notice anything.

let me try to explain.
Push 2012 mhz clocks constant at 60C in PUBG, voltage 1.050v. TDP set to 195W Or 200W.
PUBG signals power limit at 155W-170W+ off and on constantly. At around 165W (estimated), actual throttling kicks in (the PWR flag can turn on before any throttling ever occurs btw).

GPU voltage drops to 1.012v to 0.993v. clocks drop to 1987 to 1974 mhz, then goes back up to 2012 mhz, voltage goes back up to what it was before.
Then it happens again.
This was happening repeatedly and constantly (MSI Afterburner was showing PWR flags at 105%).

Then run heaven: NO power limit flags at all until 124%. No actual real throttling until 132%.

So change 16,200 to 19,200.

Now PUBG acts like it should.

Note: I am NOT the first person with a modded 1070 who noticed this in PUBG. @Timbabs123 also noticed this and "thought" it may have had something to do with that value. I was the only person to scientifically TEST it though.

Too bad NVflash can't flash modded Vbioses or taking a video of it happening would be trivial.

 

*shrug*

MSI?

I only mentioned my experience because you mentioned 1080Ns likely having it, to which I personally don't. Cannot account for basic MSI 1080Ns though.

 

As you well know now, I cracked MSI's power limit BS on their BGAbooks.
And even their LGAbook has the same cancer there. The 16L13 even has BATTERY BOOST. (Prema removed it on the Evoc version).
I looked at the EC Strings and the EC RAM.
MSI recycles everything. Just updates code, recycles base code. Plug in some values for new hardware, add some stuff here, remove stuff there, you know what I'm getting at.

Most likely, MSI's default "extreme power limits" for the 1080 without a power connector were based on the 16L13, which has 150W TDP.
The GT73 and GT75 versions all have external power connectors and have 200W TDP.
Most likely, MSI simply extended the range of the Target and Limit TDP's, from 150000 to 200000, said "ok you're good to go" and recompiles. But didn't think, or didnt properly test that the 16,200 (default for 200W 1080 MSI card) that may have been used on the 150W card (I do not have nor did not look at the 150W vbios), would have had a negative effect. I mean, think about it. If you used Heaven, valley and 3dmark, saw your scores increase properly, and no power limits under 190W, you would say 'mission accomplished' also, wouldn't you? I doubt these people test tons of games. . Just a few benchs and bam: done. Then something comes along (PUBG) and then side effects happen....

 

No, because I don't play many games right now. Unless you want to count RPG's that use 30% GPU power. Only other thing I tested was Witcher 3, and only the tutorial.
A game -itself- isn't a trigger. Even if the Nvidia drivers are using a feature which is flagged and throttled when the game is running, still, changing a value caused a huge difference.
When you disable gsync/anisotropic filtering, power off the laptop, power it back on, test and get the 105W+ throttle, it's clearly the game using a feature on the card. Then you run Heaven and Valley, ZERO issues. I even tried installing another Nvidia video driver just in case. Same problem.

Then you take apart the laptop, change to 19200, repaste LM (after sanding heatsink with 3000 grit again), and so on, and now PUBG acts like heaven does.
That's not a fluke.

 

Whoa what?
Laptop will not boot without GPU?
Did you wait 60 seconds?
If a full CMOS clear is done, you will have a black screen for 2 minutes, with a 1 minute power button "cycle".

Do the following:
Even with the dGPU completely removed:

Hold down the power button for 60 seconds. Do not release it.
Then release it then press it once to power on
Wait 3 minutes.
You should see boot loops with the MSI logo appearing and disappearing.
Then it should boot to windows.

Are you in California?
I'll fix your laptop if you're anywhere remotely close to here. But I won't be at this location for much longer. BTW your steam ID does NOT WORK. it says no users found.

 

Ah I see. Sorry

 

errmm.. Can you link me to the Original BIOS dump , the BIOS you are trying to flash.

And list the exact steps you are doing to flash.

Because im not sure how the first bit is messed up for you every single time.

Also what happens when you click the detect button with the clip connected to the chip?

And a picture of all the parts you are using as well.

Yours is the first case i have seen of someone bricking their GPU trying to flash a modded vBIOS, the original backup shouldn't have any issues as well, provided it flashes properly.

(Please quote my post when replying or tag me)

 

You're not supposed to use any version of nvflash to flash a modded vbios. It will always fail with a broken digital signature.

Sent from my Honor View 10 using Tapatalk

 

I agree.
I had a verify error occasionally when doing the "Auto" sequence, but the software said it was a verify error
Then you just hit auto again until it says successful.

In fact @CheeZuShRicE you didnt tell us exactly what you did in the skypro software (follow @bloodhawk 's instructions).

Here is what I do:
1) Press "Detect". if no "hex error popups appear", and it says 1.8v adapter is required, good, load modded bios
2) press Auto.
3) If verify error, press auto again.
4) otherwise done.

It's that simple.
@CheeZuShRicE are you using the latest skypro software?
There was a buggy package awhile back but I thought that was bugged on Skypro II (the LCD version).

 

Pretty much.

Just to be triple sure i always add these steps -

Here is what I do:
1) Press "Detect". if no "hex error popups appear", and it says 1.8v adapter is required, good, load modded bios
* Check the chip model you are selecting IS the one on the board. Similar GPU'sometimes have different chips. But they function the same. However the correct model needs to be selected no matter that.
* Do multiple reads of the stock Chip data. And save it.
2) Erase the chip, and then read.
* press Auto.
3) If verify error, press auto again.
4) otherwise done.
* do a few more reads just to be sure.

I flashed my vBIOS' about 20 times over the last 2 weeks. Maybe had the first bit messed up 5 times at the most.
The single chip 1.8v adapters sometimes have issues with the initialization voltage, causing the first bit to get messed up.

 

Where is the GT73VR?

I could unbrick it (not even sure how it's bricked to begin with), but I don't know how you would get it to me and I may be moving from my current location rather soon.
Jeanlegi is much closer to you and I think he has a SPI programmer. Would be extremely elementary to force flash the system bios (I already did a read on it, without 1.8v adapter; its a normal bios chip).

 

Just send it it the mail LOL

Sent from my Honor View 10 using Tapatalk

 

Does the power button light up at all?

I had this problem when I had mismatched DDR RAM causing system incompatibility and failed POSTS (Permanent black screen). Removing one RAM dimm (Just 1) fixed it. Had to get a matched set later.

Just send the laptop to one of the trusted forumers here.
There are many people who can be trusted with hardware here.
Bloodhawk, Prema, Mr Fox, Papusan, Jeanlegi, many others.
But Papusan wouldn't touch a BGAbook even if you paid him....

 

Sorry to hear that, I take it there is not a decent local shop nearby that can do the basics on testing it?

 

Does the POWER LED Light up?

 

There is no secondary bios chip, the EC is in the keyboard controller.

 

@Meaker@Sager

SVET disagrees with you.
He said the EC is stored on a 1024 mbit flash chip. NO NOT THE ENE CHIP.
So SOMEONE NEEDS to do a READ ON THAT CHIP and find out what the hell is in there. Don't expect a disabled man with a horribly swollen back to take apart his laptop. Do I have to show you a nasty picture of my swollen back? Kittys has seen it. Do you people need to start vomiting all over and see it too? I don't think so. Seems like the only way anyone believes anything these days is if you get grossed out in real life first.


No secondary chip?
So what is this red dotted chip behind the mainboard?

SVET said this is where the EC flash is stored.


I'll say it one more time.
Someone needs to do a READ ON THAT CHIP and find out what is in there.

 

Oh your talking about an MSI machine, not kept up with those *shrug*

 

@Meaker@Sager Yeah. SVET made a risky EC mod for me today, changing one byte in hopes that the EC doesn't brick, and allowing an unsigned EC to be flashed by his flasher. But I'm not going to flash it until my best friend can help me disassemble the laptop to access that chip, because if it's the EC and I brick it, my skypro is going to have a say in the matter. I can't bend over to service anything right now. But first a verified read has to be done on that chip to make sure that's the EC flash and not a backup bios copy. If that's the EC, then recovery is possible. If it's not ggwp Where's my Clevo 8 core 9700K at?


Mod is brainless and a prayer.
one byte changed, complete untested conjecture based on the power of the Gods.

CPU_CrtT = 1.CPU_ThtlT = 1.CPU_ThtlT = 0.SYS_CrtT = 1.SYS_ThtlT = 1.SYS_ThtlT = 0

to

CPU_CrtT = 0.CPU_ThtlT = 1.CPU_ThtlT = 0.SYS_CrtT = 1.SYS_ThtlT = 1.SYS_ThtlT = 0

Then hope:
1) the EC doesn't see "oh look you messed with my checksum: CYA LAPTOP, BGA LOVER" = bricked.
2) that string controls CPU Current Trip being enabled and setting it to 0 will allow me to finally run 5 ghz @ 1.45v WITHOUT THE EC SHUTTING OFF THE LAPTOP, and completing a cinebench r15 run BEFORE I reach 99C.

 

@CheeZuShRicE
Sad to hear, you should have not listenend to people who don't know what they're talking about and try something you have no experience with. Normally you should have done a proper backup with nvflash before doing anything weird. SkyPro and even the cheap CH34A1 offer features such as "verify" which tells you if your chip is being read properly. If your reading of the skypro was correct then the 1.8V is also an issue, since the GIGA chips (which are also on GTX 780M / GTX 880M cards) actually need 2.7 - 3.5voltage to be read and written properly to. I have a modded clip which makes perfect contact, but some chips will have some issues when reading and writing on it, despite having the right voltage on the adapters and clip, which will get solved by desoldering the vBios chip, same happens on some mainboards as well. I gathered this knowledge thanks to some russian programmers and my own experiences, doing lots of different BIOS mods, they even offer mods where your Z170 mainboard would support coffee lake, so if you plan on getting a coffee lake CPU, you can hit me up for a modded BIOS for that.


Also you killing your notebook is not the notebooks fault but your own. If you don't have the money to pay for repair for the hardware which you broke, then don't attempt a mod, especially when you don't know what you're doing.

 

@CheeZuShRicE


Is there any data at all in that chip?

Svet said the EC is a 1024 mbit chip
However the EC flash itself is 128k size

Try scrolling down and look for something that starts off like this:

(text): €... ...ENE Haru..$àT.p...€ä“`.´ÿ.£ä“´ÿ....£ä“p.u.)..9......àD.ð

Hex: 80 0E 13 00 20 14 06 12 45 4E 45 20 48 61 72 75 90 10 24 E0 54 01 70 1E 90 0F 80 E4 93 60 0C B4 FF 14 A3 E4 93 B4 FF 0E 02 10 00 A3 E4 93 70 06

might not be identical.

You should see this in there somewhere though:

CPU_CrtT = 1.CPU_ThtlT = 1.CPU_ThtlT = 0.SYS_CrtT = 1.SYS_ThtlT = 1.SYS_ThtlT = 0

 

BTW, you need to remove the 1.8v adapter when reading or writing main bios and EC.
The bios itself is a 3.3v or 3.5v chip.
Pretty sure EC is the same.
If you press detect and it does NOT say "1.8v adapter required" then you should remove the 1.8v.

Let me know if it says ENE HARU or any of that CPU_CrtT fields I mentioned

 

*edit* got the file.
Um.....can you try reading it without the 1.8v adapter?
This data doesn't make sense....

 

Thank you. Going to send a dump to SVET. I need to find out that this is.
If this is the decrypted image, then perhaps ECflashwin is decrypting the data and storing it here. I honestly don't know.
Brother @Prema would probably know. But that would mean force flashing the 128k original EC binary is useless.

 

Yeah I fogured you have the same chip as I do on my GTX 1060 MXM. Only way to get a proper read is to desolder the chip, tried with skypro and CH34A1 and it would keep distorting on reads.

Also the winbond is a 1.8v chip.

 

@CheeZuShRicE There is a problem.
The data you read with the first read (with 1.8v adapter) is completely different than (without) it.

This data should not be different.

 

@CheeZuShRicE

Ok now we need to do several consecutive system bios reads and then file compares.
Because the system bios 8MB is the control.
This data must NEVER, EVER change.

If the system bios reads different data each time, then there is a problem with the skypro or the cables or something.
This will be the smoking gun test.

If the system bios returns random data each time, that could be why the video card didn't flash. Doesn't explain why there is no power LED when powering on though.

 

Yeah. You did everything fine. Thank you.
Now time to see what SVET says about this.
But it's clear the raw EC file (17A1EMS1.108) is NOT safe to flash if every MSIbook will return random stuff from that chip. However if your EC is actually CORRUPT somehow (why would the data change?) that explains why your power button doesn't work. Because the EC controls the power button actually turning on in the first place. But at this point this is pure random BS by me. I'm in over my head.

That chip should not even be powered by the battery itself. AKA NVRAM
or CMOS.

CMOS has system time/date etc but that's all. Can't be that.
I hope someone else can find out why that chip is reading random data.

 

I am on my mobile right now, but the hex in that picture shows the TBT3 firmware not the EC.
Never dump that while the chip is still on the board or it can get corrupted beyond repair!

 

And there we go......brother Prema has spoken.
TB3 firmware.....which now means svet's mod is impossible to do.
@Prema thank you.
btw he did this because the laptop was already dead. Wouldn't even power on (no power LED, nothing). And no idea how it could die.

 

Since SVet said the EC is a 1024 mbit flash chip...there's another chip back there that we all missed. Unmarked.
Are there any other 8 pin SOIC chips?

 

This could be the 128kb SOIC of that boards EC if it's not the BIOS/ME!
If it is the BIOS then check the backside of that board region.
Otherwise your EC is probably hosted in the ENE controller itself.

 

Brother Prema has been busy lately

That chip is the bios chip. Already did a read and sent my backup of it to Cheez.
Emailed Svet. He doesn't know which chip on the backside is EC because he doesn't have this mainboard. But on older MSI, one of them (1024 mbit size) stores the EC. I'll checked the unmarked chip myself if I ever run into a brick. I circled it for reference. no idea if there are any others.