Mobile Pascal TDP Tweaker Update and Feedback Thread

Then there is no flash chip close enough to the ENE controller to be the firmware.
Looks like it's hosted in the controller.

 

CMOS battery one seems to be off in la-la land. Could be anything.
the one I circled in red is much closer to the actual ENE and still not physically far away from the system bios chip on the front side.

@Prema ugh ill just do a read on it later :/

 

Most SOIC8 chips are not for hosting a firmware!
Be very careful and research every chip properly or you can permanently damage the board by connecting a programmer and sending voltage through them.

 

Understood. Thank you for your help, @Prema. Have a good weekend.

 

Have you checked none of the main rails are shorter? Usually death is something more electrically simple than corrupt firmware unless you were playing with it.

 

Sent a PM real fast, while I'm feeling despondent right now.

 

The EC is in that SOIC8 with the pink-ish mark on it, but as I have said in lots of places, forcing power into a Flash chip while its soldered in a mobo is never a good idea, if you want to read it, dessolder, put in zif socket, read it, and solder it back. And it runs at 3.3v, not at 1.8v, so you shouldn't be using the 1.8v adapter.

 

Wait, how do you know for certain that is the EC Chip?

@Prema said that is the TB3 (thunderbolt?) chip....

He tried 2 reads on it. One seemed...bizarre but at least part of it made sense (but the raw EC file was not there) and his 2nd read was pure garbage data.

@Prema we sorta need you if you aren't too busy

Why does it not have the RAW EC data if it's the EC? it doesnt match anything even remotely close to 17A1EMS1.108 (which i also attached here).

Really need proof because I'm sending SVET all this information.

Yes I know I'm being push and maybe I'm being rude and i VERY MUCH apologize for it.
But i'd rather be rude and pushy than lose $2500 dollars. You know?

So what is in there? The DECRYPTED image of the EC? Prema how do you know that's the thunderbolt and not the EC?

Does ECflashwin_041.exe read the file 17A1EMS1.108 and decrypt it and then store it there? So the decrypted data would not match the "Raw" file? And what about this brilliant text in the original EC that is not showing up in that read?

EmSD=1.EmSD=0.AC_Out.AC_In.SusOn=H.Batt_Out.Batt_In.BATT_OFF.BattThrottleST = 1.BattThrottleST = 0.BattCurrThlFlag = 1.BattCurrThlFlag = 0.BattCurrentCrtFlag =

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

And since i know EVERYONE here wants to know WTF i'm doing, well, SVET made a ECflashwin mod which allowed flashing of UNSIGNED EC files, to force flash a 1 byte EC change: changing CPU_CrtT = 1 to CPU_CrtT = 0.

To see if it work or brick. I have NOT run the flasher mod yet!

SVET said that an EC that fails a checksum will 'revert' to a backup.
And I don't want to 'test' that theory, hahaha.

----------------------

Is that what we're reading here?

attached both files. But that read is most likely corrupt. I don't know.

 

Already gave you all the information:
The EC firmware is probably hosted in the ENE controller.
The chip you dumped is the TBT3 firmware.
(Looking at the picture of the hex is enough to see that).
The other chip is the BIOS/ME.

 

Thank you @Prema . I trust you.
Why is @senso lying to us then? @senso : proof please of 'that chip being the EC?
You're going DIRECTLY against what brother Prema is saying here. And it's causing friction here. That's not good.


All this is demoralizing me. Feel like just bricking my damn laptop and throwing it in the trash. The 9700K isn't out yet. If I had a clevo right now I'd just burn the MSI up for science.

i wanted a world record 5 ghz CB run on a 7820HK. NO one else has done this.

Sigh..its like telling a person who climbs mountains "Don't climb that mountain, you may die"..

 

Relax Dude, nothing going on. He's not lying, he probably simply just guessed without looking into it...

 

Let me fix that real quick:

Spoiler

hugs

 

I'm not at the repair center anymore, but the BIOS Flash was usually painted with a yellow dot, and the EC with a red one, refub mobos also had green dots on the Flash chips.

I didn't call anyone selfish, nor did I receive any PM..

I only gave a quick look to the pick, I'm used to only seeing the kapton tape on top of the EC, now that I have a bit more time, yeah, its right near the USB-C port/TB controller, so it makes sense that its firmware for the TB3 thingy.

I was a guess/going by what I'm used to see, but I will just refrain from commenting on any topic, because anything that can be said in this forums will be used against me...

 

Hi everyone,

I don't know if this project is still going on or not but I am trying to undervolt my GTX 1050M in my dell xps 9560 (laptop), it took me long to figure out how to extract the vbios from the bios but I think I got it, I can send it on request.

Here the screenshot of what I see in Coolane's application :

My goal is to be able to go under 800mv which is the actual limit and also to underclock the VRAM more.

Thanks!

 

Bloody hell, a lot of different people getting defensive and upset in this thread - nothing bad has really happened, seems out of the blue, sleep more & drink less coffee I'd say.

 

You know, I already blocked the guy for his behaviour for obvious reasons, but it doesn't help that you act as childish as he does. Simply block him and you won't have to deal with his crap, he's probably mad because he saw that some could get a 7820HK on 5.1ghz (I also managed to do this) and he doesn't understand how people do it, hence hes frustrated.

As I already pointed out, you probably did something wrong and killed something off. You clearly have no experience in this and fiddling even more on it will only result in making it worse. As Prema already pointed out if you keep on making reads and apply to high voltage on chips that require much less, you likely will break something even more and make the repair more difficult for the person who attempts to repair your board. I say, leave it as it is, let it be a lesson and get it repaired, if you still have warranty, RMA it and don't say anything, you might get it fixed for free within your warranty, if not, pay for the repair, sell the notebook and you gained some money which you can use on your desktop and a less powerful notebook on the side for portability things.

Being angry and tell someone to delete your account is somthing a 7 year old rage boy would do because he didn't get his lollipop. Take a chillpill and appreciate the fact that people like Prema spend time into trying to solve your problem. It's not our fault that you did something wrong and being angry at people on this forum for mistakes you made is the wrong way to go on about it, instead be angry at yourself for doing something stupid and learn from it.

I was in the same situation as you were, I broke a notebook which I spend a lot of money in and was financially rather stuck, and instead of being angry I found a solution which basicially let me not only get a much better notebook for free, but also made me some money as well, because I realized I done goofed and had to find a solution. Anger blinds your reason and blocks paths which you could have taken to make something great out of a dire situation.

 

Just so ALL of you know ,he's lying again.
This guy is a creep who makes up complete BS when he doesn't even know how the embedded controller works on this laptop, nor does he understand any of the thermal characteristics. He doesn't even know which code is responsible for limiting the EC. HINT: it's not CPU speed at all. The MSI 16L13 is affected by the exact same thing.

This guy is a pretender and a fraud.

He doesn't know half of what I know about MSIbooks
I put in my work and study in using the scientific method to hack MSIbooks, even though I am not a coder.

There is a reason he does not give 'proof' of his EC hack.
You can't even reprogram the EC in a GT73VR by force. it's stored in the controller chip
You can't force flash it. You need specialized hardware. And if you mess up one byte. YOUR ENTIRE LAPTOP IS BRICKED FOREVER.

i'll say it again

5.1 ghz on a 7820HK is impossible outside of a CPU-Z suicide shot.
The EC can NOT allow you to reach that speed on a GT73VR And put any sort of load on the processor.

And even if a EC hack (which danishblunt doesnt know about, this kid doesn't even know what a checksum is) were available, you would reach 100C without Liquid metal paste. He doesn't use liquid metal. read his old posts. Even in his broken tests his Lm tests always had HIGHER temps than traditional paste.

Do you believe he knows how to hack 32 bit SHA-1 checksums and has the MSI decryption key to their EC, when he was sitting here telling people here to desolder their vbios pascal chips to flash the modded vbioses? he doesn't even know nor understand why nor when you need to desolder chips.

Vbios and system bios chips are *bus isolated*. Meaning they don't need to be desoldered. Just cut the power to the laptop and remove AC plug and battery..

TB3 (thunderbolt 3) chip is NOT (the chip in question earlier).

I'll say it 1 more time

5.1 ghz at 1.270v STABLE <--this is his claim.
on regular thermal paste :/

Even a 7700K can't do this.

He is a liar and a fraud.

@Papusan @D2 Ultima @Prema

You people can NOT allow people like danishblunt to spread lies like this. ever.

 

Come on guys can we put this thread back to topic, vBios and EC chips are two different things (Does different tasks) which only one can be worked up on by the tool supplied by the thread. My suggestion for your EC Hunt is to make a new topic about it.

There shouldn't be any fighting going on at all, after all we are a community and not a school play field. Each person has a different way weather or not if its the correct way to go about things doesn't warrant a public outburst. I take no one's side but i hate to see the community i have grown to love get involved in a argument. Now please back to the topic of this thread.

 

http://forum.notebookreview.com/thr...rvolting-overclocking-and-overvolting.817033/

 

PUBG is hopelessly bugged with unlocked GPU's.. It's like it's tripping some rail or something which triggers PWR limit even when you're nowhere close to the power limit, and the clocks and voltage drop as if you were leapfrogging the power limit. Or maybe its making the Nvidia drivers trip a rail. Only seems to happen on Erangel too. Setting extreme power limits from 16200 to 19200 seems to prevent this until the GPU reaches around 70C, then all gloves are off and it randomly does it again. It's almost like some PARTICULAR map structure or object is causing it. If you don't set the first extreme power limit to 19200, you can get this even in the main menu, starting at 160W....

Besides this Nvidia BS, no problem with PUBG otherwise rocking 230W through a GTX 1070. VRM's be damned. And don't play with Shadows on Ultra unless you want 50 FPS in Erangel houses, especially around doorways.

Nvidia...the way it's meant to be fried.

Peons wanted.

 

It could be some coding goes into a loop and hammers a particular section of the core like a power virus.

 

Yeah that's what I was thinking.
But that's why I want to know what the first value in "Extreme power limits" actually does.
All Coolane said is that this value must be increased from the default of (14,000 something) to 16,200 to allow a modded 1070 to draw 170W of power.
Yet the GTX 1080 (MSI version) uses the EXACT same defaults ! I'm assuming that's because the original MSI 1080 card (the one used in the 16L13) has a default TDP of 150W! Only the GT73VR (and I think GT83VR) versions have 200W TDP. So MSI increased the TDP but not the rails. Nice...

I KNOW its some sort of rail, but i don't know what precisely.

I remember in the Kepler editor, there was 'faint' discussion about these values, ONE Of them was a "PCIE voltage or amps" rail.
Another was something else. There were like three or four for "PCIE".

I'm assuming that everyone with a MSI 1080 (who has a default of 16200 on the first extreme power limits rail) will have the same problem in PUBG.

 

Ok where are all the techheads here?
@Meaker@Sager @Prema (the God), @bloodhawk

The Texas Instruments IC's are rated for 40A operation (under recommended operating conditions) and 90% efficiency at 25A.

http://www.ti.com/lit/ds/symlink/csd87350q5d.pdf

So that means if a MSI GTX 1070 (MSI makes the Clevo cards also) has FIVE PHASES, that would seem to imply the card is 100% stable and guaranteed to run at the full MXM specification of 200W (195W) without having Chernobyl 2.0 in your living room. But can one of you please explain this VRM layout?

there are 5 phases, if each choke counts as a phase.
But why are there two mosfets connected to one phase? What exactly is going on there?
Is each mosfet a pair (e.g. Sync Fet + Control fet?), or is there some sort of doubling going on?

How much is each pair exactly? Are we looking at 200W of mosfets or 400W?
( @Khenglish mentioned once 400W, but this has to be a complete mistake).

It also "seems" like they can burst operate up to 60 amps (in which case 60 * 5 =300 watts), but not sustained. Only 40 amps (*5= 200W) is guaranteed sustained.

Anyone want to explain?

Look at the CPU MOSFETs here for this BGA: 5 phases=5 chokes:
You see clearly 1 mosfet linked to a phase.



Now look at the 1070 MXM card:
Using TI mosfet: csd87350Q5D





someone want to help a brother out here? @Papusan i want to swim please.
*Edit* please don't get me started on how X watts become Y amps...and pascal supply GPU voltage (1.0625v) i'm confused.

 

It's just 2 power FETs per phase to double the max current.

1 power FET = 40A
one phase = 2 power FET = 2 * 40A = 80A
5 phases = 5 * 80A = 400A

 

You can also have 2 inductors per phase to trick people into thinking there are more phases than there really are.

Doubling the VRMs per phase increases capacity, you are still mostly as efficient (small extra losses) but you can now handle double the current at the same efficiency (20% of 100W is still bigger than 20% of 50W of course for instance).

It does not improve ripple or other electrical characteristics.

 

Why would MSI even bother to do this?
Why would they design a card that can handle 230W+ (or up to 330W) easily without frying, and limit it to 150W (original design specification) TDP ,and 115W-125W TDP (MSi and Clevo boards--MSI makes the Clevo models) on release? What's the point of overspeccing a card like this? Doesn't it just add cost? Especially since no one would ever be able to run a card this high? (without both bios mods, shunt mods and hardware voltage mods?)

Or did MSI just copy part of the design for its desktop cards and chop them down with a chainsaw?

 

Stability, reliability, ripple, dI/dT, keep MOSFET's under SOA, EMI/RFI, performance/behaviour when its running with low loads -> low current, usually done with discontinuous mode, two FET's in parallel give you a lower RDSon, so it also plays into what inductor you can use, to maximize Q.

Tons of reasons.

Just like everything, the first page of a datasheet is marketing wank..

At 40A that MOSFET is dissipating nearly 10Watts when it switching at 500Khz, thats a LOT to dissipate with that small footprint.

And then there is the SOA, page 5, look at the SOA curves..

RDSon goes up with temperature, the more current you drawn, the higher the junction gets, and the higher it gets, the more it will heat, its a positive feedback loop, so using 400 marketing Amps for a real world use of 100-150A seems a decent design.

If they went by the first page "specs" you would have a card that would blow up if you even typed furmark on the search bar..

 

Because the efficiency is a curve generally. By "over" speccing the VRM full load is much closer to optimal efficiency and there will also be typically less component failures and less heat dumped into the heatsink and PCB (spreading the VRMs out by having more also makes cooling easier)

You also have to consider trace sizes and the abilities of the VRM heatsink/supporting components and consider it as a full system.

 

Tried it on my extracted vBios (using UEFI Tool) from Acer VN7-593G (1060) System Bios-> Unsupported Model GP107 with Bios Version shown.
Is there any chance for a mod? Is Prema offering vBios mod for Pascal?

 

Not that I am aware of at the moment.

 

And does Prema offer vBios mod services of Pascal GPUs?
One more thing that's odd is that my Notebook has a GP106 GTX1060 and I found only 1 vBios in System Bios with UEFI tool that reads GP107 in header.
Size is 161kb.

 

This probably concerns about 5% of users who used mods here, but that 5% should still be aware of something.

Increasing power limits by the change of the advanced rail settings (presets) with this tool while it works great on AC power, actually completely screws up the way the system throttles on battery power.
While the Nvidia vbios still throttles the GPU *hard* on battery power, if you reduce the TDP down to 115W (default) via MSI AB slider, and then run on battery power with the rail changes, the throttling point on battery is actually *STILL* raised. Before my laptop repeatedly shut off by itself, I was greeted to my GTX 1070 pulling 65W on battery power (!!!) even with the TDP forcibly set to 115W default. The first advanced rail was set to 19,200, but the problem had actually originally happened with the basic "preset" way back for 16,200 and 151-170W slider. I remember trying to play Tekken 7 on Battery power last year (basic preset, 1st rail at 16,200 rather than the default 14,500, if that was the default) and the laptop simply shut off after a little bit of gaming, and I thought nothing of it at the time. But now I was able to find out what's going on.

On a 75W battery, the GPU should NEVER be allowed to pull 65W by itself! The EC (Embedded Controller) was shutting off the battery due to overdraw (rather quickly).
I had to set the MSI AB TDP Slider to 35% (even 50%, which is 75W TDP, wasn't enough) to keep the GPU from drawing more than 50W.

Something silly, but occasionally people may be stuck without an AC charger and may want to game on a TDP modded video card. Just something interesting to remember!
Again everyone's specifications are different and no two vbioses or competing systems are created identically. But keep that in mind. Game on!

 

@Coolane

Any chance you could release the source code for this?

 

Did you manage to flash your vBios with the skypro?

 

Hi everyone, first thank you so much for the work on this utility!

I made an account to ask possibly a silly question though, regarding this with a desktop card.

If I've already done a power input shunt resistor mod (by adding another 5MO resistor on top of each of the 6 and 8 pin connector resistors) is there any reason to bother modifying the vbios to increase tdp?

Specifically this is for a pair of Titan Xp cards I'm installing to one of my desktop rigs.

I have an SPI flasher that i was trying to use on Titan V cards but I must have been doing something wrong because even reading the vbios chip the data was slightly off vs extracting with nvflash so my flashing attempts did not function.

Anyway thanks in advance and keep up the good work!

 

Great utility! I'll have to try it out.
Low temps keep me a happy boy.

 

You have to set up the read/write parameters of the programmer to match the eeprom you're attempting to program. These settings differ between chips and using a mis-match results in exactly the kind of symptoms you're experiencing.

Usually the programmer has a library with dozens of eeproms to choose from (manufacturer and model are printed on top of the chip). Some allow you to set each parameter individually. In this case you can google the model nr. and use its specification sheet to find out the required settings.

 

It would let you more easily track your real power consumption if you were to remove the extra shunt resistors.

 

@Falkentyne just wanted to ask does the eurocom tornado f5 have the same limitations of hard capping a gtx 1080 to 200w and throttling the cpu?
I know the mxm slot can handle at least 200w on load with no issues since it has many vrms and mofsets for power delivery and also the gtx 1080 was made for the machine, question is will modding the tdp actually make the gtx 1080 run significantly better considering I would have to buy all the stuff to do the mod in the first place.

At the moment I'm messing around with the bios trying to inject the coffee lake microcodes into the system so it runs a 8600k. Having the full potential of both gpu and cpu will mean that this can become my ACTUAL desktop replacement laptop.

 

Considering I'm running my GTX 1070 (!) at 230W TDP, I don't see any reason why you're going to have problems with the 1080. The 1080 VRM's can handle it.

Are you talking about the CPU being TDP power throttled if you exceed 330W of total system power (e.g. if you modded the GTX 1080 to something like 250W?).

There is some sort of internal power trip that happens if the GTX 1080 is running at 250W with the CPU running overclocked high, which causes the system to just shut off. But then you just set the TDP slider lower. Like set it with a manual rage of 100% (150W to whatever 250W is, with the "preset" button also pressed in the Pascal editor).

No one has tested (on a GT73VR or GT75VR) what happens if you exceed 330W of total system power. I suspect the 16L13 has the same type of limitation, except you're in much better shape since you would be capped to 91W TDP rather than 45W (unless the EC tries to limit you to BELOW 91W). I don't think there is a single user who has TDP modded a 1080 model MSI outside of China. I've only seen Clevos modded with 1080's past 200W. Pretty much everyone (including most of the chinese users I saw on baidu) were modding 1070's, since obviously they were much cheaper to buy. However regardless, you shouldn't have any problem. Your 16L13 already has a 91W TDP CPU power cap which ignores any bios power limit 1/power limit 2 override settings anyway, so you're already screwed on that front.

But you should be able to bypass both problems by simply using the IMON SLOPE and IMON OFFSET tweaks as mentioned already in the 16L13 thread.
Set IMON SLOPE to 50. And set IMON OFFSET to -31999 (negative).

This will cause the CPU to report *less* than half the TDP it's actually using in watts, and then you have absolutely no problems whatsoever, no matter what you do with the 1080.

This same trick works on a GT73VR/GT75VR BGAbook too.

However disabling "battery boost" (battery drain at full load) requires EC RAM register tweaks, the internal battery has to be unplugged first, then EC RAM register 31 must be set manually to 09 (with RW Everything) and EC RAM register 42 must be set to 64. That will trick the EC into thinking the disconnected battery is connected and will "re-enable" NOS, which will stop premature CPU throttling (from bypassing the 180W AC system power limit barrier when the battey is removed).

BTW one warning about the "preset" button when using the 19,200 value preset for GTX 1080 in extreme power limits.

This increased value causes the GPU to also use more power on battery power, which can exceed the battery limit, then the EC will just shut the laptop off (seems to suspend windows rather than a direct power off). I saw my 1070 pulling 62W on battery power with the 19,200 value (far more than it's supposed to be pulling on battery).

 

OK I will look into that thread, thanks for the info!

About the gtx 1070 pulling more power, did you adjust the power states such as p8 (standby), p0, p1, p2 and p3?

Also my alienware 18 with a gtx 1070 shuts down instantly when playing a game and accidentally unplugged. Funny thing is the card keeps taking the full 120w load from the BATTERY and shuts of the alienware in 5 seconds with 25% battery drain! It would even blue screen telling me its a GPU state error. For some reason this happens only on battery which makes me think that the card starts to malfunction on lower power.

 

That's normal behavior, the SLI systems of MSI and some higher end gigabyte notebooks behave the same way.

 

Mine seems to handle slowing down on battery in SLI just fine. We are talking minimum clocks though.

 

How have I not found this thread for this long?

I'm planning to get myself a Dell Precision 7530 with a Quadro P3200 ( supposedly not Max-Q, but performs as such), and it has extremely low power limits of 78 W, and the memory clock is ridiculously low, too. The form factor is apparently removable and upgradeable, but it's some strange proprietary stuff (see here).

I understand that this tool doesn't work with CFL notebooks—I can't exactly find the reason for that here. Is there some firmware-related issue that has to be ironed out? I'd really like to increase the power limits on that P3200 to ~110W, and see what comes up, as well as overclock the memory and core...

 

78W is the normal P3200 like in the Precision 7730. The Lenovo P52 P3200 Max-Q is 70-75W. The Precision P3200 Max-Q is ~65W.

Anyway, even if you could mod the TDP to let the card boost higher, an equally big problem is the Quadro vBIOS/driver block that locks out all overclocking and undervolting ability.

 

Ah, yeah. 65 W, my bad.

I suppose it's possible, (if and) once this tool is updated for CFL notebooks, to unlock the voltage and clocks as well, if TDP can be unlocked?

 

This tool only changes the power and temp limits, nothing else.

 

Of course. However, is it not possible to otherwise modify the vBIOS to enable core and memory overclocking?

I get ISV certification and all, but it seems a waste of good Pascal cores and GDDR5 memory only to clock them so low.

 

I do not know. Pascal is extremely locked down.

 

Pascal is more locked down than the Great Earth Serpent around Thor's Hammer !