Successful MXM GPU Upgraded Laptops

The eDP output is directly connected to the mxm port, there is just passive component between so it should work with the right connector installed and a cable compatible with the connector on board (in this case the last issue would be incorrect wiring of the cable on the display connector).
I don't know about the M15x, it could be only 2 lanes eDP like on some actual 15.6" laptop, I don't know if 2 lanes eDP display existed at this period.

 

Here an SLI run with 240W Adapter:
https://www.3dmark.com/3dm11/13788722

Not much better then your one.

 

Check throttle reason in gpuz. If reason is PWR it probably needs vBIOS mod from @Prema to solve.

 

Power play table regedit? I pushed a 7970m as far as the weak vrms would go when testing my first watercooling mod, 110W iirc up from 84W, no vbios flash required

The throttle never showed itself in clocks or anything else, only bench numbers.

I know it's not 8970m, there were a few tweaks around the edges, 4gb memory, edp support, fixed function hw etc, but basically the same core

 

I have no trouble with 8970M they run fine, I was comparing to his M5000M scores to show he has significant throttle in SLI mode probably due to vBIOS problem.

680M did not work properly in my M17x R1 without svl7s vBIOS

 

So install a Dell XPS panel to our m17xr2 and pop in the maxwell m5000M?

That is it?

Sweet mother of pearl...

 

Have you seen the wiring behind the screen? It's far faaar away from 'that's it' but I am extremely curious on how to do the wiring and what exact components are necessary

 

All we need is a 8bit to 6 bit adapter.
And then we have maxwell to play with.

 

Hello @anyone may be some1 have VBIOS QUADRO P4200 from MSI? i want to solder to HP card.

 

Does anyone know if gp104 GPU can be forced to work in 192 bit mode (6gb vRAM)?

 

Did you swaped the core of a 1060 and it doesn't work ? did you tried 192bit vbios as well as vbios that work with the specific GP104 you put, even if it is 256bit ?

 

I noticed that with 1060 vbios it still asks for last 2 memory chips, so I think maybe the board config resistors can do that, but did not found any clue of this anywhere.

Although I saw gp104 chips used in some desktop cards as gtx1060, but there the chip physically could be different. Also there is no knowledge what those 6 config resistors are for.

Want to try this chip swap in the alienware 13r3, if possible.

 

What do you mean by "1060 vbios it still asks for last 2 memory chips", what resistor are you talking about ?
For your information, 13r3 use GP106 core, not GP104, I think they share the same size, like GM204 and GM206, but nobody know if it would work, furthermore you would probably need a new vbios, and thats where you could have compatibility issue with the motherboard bios.

 

I upgraded a 1060 mxm card to a 1070 with adding extra vRAM chips and missing SMD components. With original vBIOS from 1060 it still reads all 8 memory chips.

That's one problem with dell what their BIOS is totally locked down.

 

Can you share picture of this please ? What is the current hardware id of the card and what was the previous one ? Does it works in the end ? How the card is recognized by GPU-Z/Drivers (like a 1060 or a 1070) ? What about frequency, does they are the same as the 1060 or they are different ? That's great to know GP104 and GP106 are pin compatible.
Where are located config resistor ? and how did you find out they are config pin ?
If I remember correctly for maxwell there was like 2 resistor for hardware id and 1 for memory brand (and timings)
@Khenglish might be instered at reading this

 

Several pictures are here
https://imgur.com/a/1pArQDc

Flashed a modded vBIOS of 90W on it, which is without g-sync, but it worked with some other vBIOSes. Card working as a normal 1070 except that 90W is a bit low power for it, scores drop at this rate a lot.
Config resistors - I do not know where are located, I did not change the board cofig at all, just the new die and extra memory plus memory SMDs.
I bet @Khenglish already knows of such pascal compatibility, since boards are the same and the p106 and p104 have same pinout except that p106 has physically only 192 bit memory support.

I also have a maxwell MXM board but it has many ripped off pins and that killed the GPU since only after the GPU pull I found this out. Maybe I will get a 980 core on it sometime.

 

Did you tried the 1060 with all the ram chip soldered ? did it work ? about hardware id it change from 1060 to 1070, right ?
I don't think he ever tried to swap a 1060 to 1070, I'm only aware about his "multiple" 1070 to 1080/P5200 GDDR5 with no sucess, he said it was because of 1080 chip not compatible with GDDR5X while we can clearly see that 1060 and 1070 share the same core number for GDDR5 and GDDR5X card
Now I wonder if we can flash a vbios from 1060 to any 1070, so as P4000 to P5000 or P4200, in my case I have a P4000 which is locked at 1250MHz max while the P5000 only have double the ram (double the chip) and the P4200 have the same amounth of vram but they can boost up to 1700+MHz.
Did you tried to flash a 1070 vbios on your card ?
I think that config resitor are located near the eeprom, I have a Zotac 1080 which is more or less the same card as your, as far as I can tell from you picture, my bottom and left side is the same as your card, do you think I need to unsolder them to mesure the resistance properly or it should +/- the same ?

 

This is physical difference, extra memory for 1060 is useless.

Yes.

Yes, now has custom msi vBIOS with 90W TDP.

1060/1070 vs 1080 boards are different, so there is no clue that config resistors are pointing to same GPU pins, for this we need schematics or desolder the GPU and check it manually.

1070 with 1060 vBIOS rarely posts, but for testing I did not care about vBIOS only about the GPU die and memory.
P5200 might ask for 16gb of vRAM, though I cannot tell how the board memory configuration is done, could be that there are pairs of vRAM chips, one on top and another on bottom, so memory mapping would not be linear on the 1060/1070 MXM board with 8 memory banks resulting that it will artifact.

Try a p5000 vbios on the p4000, maybe it will work. TDP and clocks for pascal GTX are controllable from vBIOS, maybe Quadro as well.

 

So to resume : you started by swaping the core from a 1060 to a 1070 without adding memory chip, still with the 1060 vBIOS, it doesn't work because "it still asks for last 2 memory chips", so you soldered the last 2 memory chips + SMD and it worked, but with the original 1060 vBIOS or a "random" 1070 vBIOS or a custom 1060 vBIOS (customized to work with the 1070 ?) ?

We don't really know if memory bus width is tied to the core or to the vBIOS on Pascal card, no one ever tried to solder "missing" memory chip on a 1080ti or a 1060 GP104 (GP104 is 256bit memory bus on higher model but this 1060 is 192bit)

As for clocks for Pascal controllable from vBIOS, where did you get your information ? did you tried it yourself ? I even asked coolane (author of Pascal TDP tweaker) about it and he said that it ended up to error 43 with a 1070 in device manager, I also tried too but I get same error 43.

 

Not really:
Initial status of 1060 - core die chipped, no activity on vBIOS soic.
Next step was adding extra vRAMs.
Next step was the chip swap. Initial boot - artifacts, there was the original cards 1060 vBIOS.
Next step - debug the problem. Tested lots of vBIOSes, most posted with same result, artifacts. Artifact nature was related to memory.
Next step - chip pull, reball and resolder. Result - no changes.
Next step - nVidia mods tool. Errors on 3 memory banks, 1 of those was already existing on the board the other from the new ones, another existing one was completely in errors on all addreses. Identified the broken vRAM chip, repaired broken traces on board, soldered new chip. Result - positive, no more artifacts.
Next step - flashed msi custom vBIOS 90W. Result - EC panic. Flashed clevo g-sync, same result, EC panic.
In between - soldered missing SMDs for the 2 new added memory chips.
Final step - added missing SMDs for thermal SMbus communication.

p106 has no connection to the last 2 memory banks, there is no connection there to the chip memory controller. Same thing will be for 1080ti, the missing one probably will stay in the air.
I have no clue about the nVidia internal configuration of the memory controller and the memory itself.

What I find very interesting is that a gtx780 will boot with no memory on the board but with artifatcs, but the pascal will also boot with no memory but will show nothing on the screen at all, only backlit or monitor initiation.

The TDP is in the vBIOS, maybe I wrote it wrong, not an vBIOS expert, if the clocks can go high on a less power consumtion these will get as high as possible.
In short, my 1070 went up to 1690MHz and even above 1700MHz, consumed less than 90W, in some heavier 3D, stays on 90W but cloks vary, being about 1500MHz. As far as other tried to shunt mod the current measurement resistors, they got some higher cloks but not too much.

 

Thanks for all the details you gave, could you tell us more about "Nvidia mods tool" ? I never heard of such a tool that can detect if a vram chip/bank is broken, I'm very very interested in such a tool
The 1080ti is based on the GP102 core which in fully unlocked chip is a 384bit memory bus width chip, in the 1080 ti it's a 352bit, but we don't really know if it's "sleeping" or if it's laser cut from factory, but since you are able to use 1070 core with a 1060 vbios it's more likely a laser cut.
That's kinda funny a 780 still work with no memory, I have some gpu which gave me artifacts, I suppose it's memory related and you reinforce my supposition.

Indeed increase TDP help to increase clocks if we are power limited, however I'm far from the 100W with my P4000
a 1070 with 1700MHz below 90W seems very impresive, is this with undervolting ?
Thanks you

 

I am not sure if this is what nVidia uses itself for testing cards, but the program was not for free at least several months ago, so I had it from some specialized service which used it for desktop cards, not for free.

Measure the empty memory bank, there should be voltage drop on address/data lines.

That is some easy load like GPU-Z render test, and sometimes it rated 1700MHz but just for a moment.
Heavy load is about 1500, I will check it more in detail, right now the 7400 i5 can't hold the load, looking for some 8th gen i7.

 

adding this to the list

p170em + nvidia 1070 mxm

http://forum.notebookreview.com/thr...nstall-mods-nvidia-inf-mod-with-photo.832679/

 

Well done. OP upgraded

 

Now the Quadro is working like i want but the low power mxm Slot is still a problem.
Next step is, to get sli working correct.
https://www.3dmark.com/fs/22463578

 

You da man. Fingers crossed on how it works out

 

Hi.
Has anyone tested graphics Quadro P series like P3000 P4000 or P5000 with HP Zbook 17 G3. Is that works?
Thanks for answer.

 

Is it possible to get it also working in an M17X R1?

 

I don´t know. Not tried it yet. But if the maxwell card starts with inverted color, it should work also.

 

I have also a screen from an M17X R3 and AW17 R1 here. So I need a Display Cable from the old M15X too?

 

AMD 6970M upgrade for 2009 iMac 27" [Video Card]

 

Yes, you need the M15x cable. And you have to mod the lvds on Mainboard. Power for r2 Display is 3.3V, the 17,3 Displays Need 5,0v.

 

Where to get the right schematics?

 

Maybe someone can add that zotac MXM cards are not compatible with laptops which need GPU temp via SM BUS?
These need hardware mod to work fine.

 

Help.
I have seen GT72VR (I have 7RE) can be updated to P5000, but it does not say if I would need a special BIOS or something so I am scared to buy it and then won't work. Would I need special BIOS or VBIOS?

Also if I have 1070, and I update to P5000, can I sell the laptop at a good price or nobody is going to buy a i7 7700 + p5000 for 2500eur at this moment? Because I see the update cost 2500eur on eurocom, https://www.eurocom.com/ec/upgrade(2,444,0) (Taking in mind that I will buy P5000 from different seller at cheaper price)
but M5000M also cost 2000eur and it is very bad perfomance card, and that confused me.

 

Quadro P5000 are like downgraded 1070, no overclocking is possible with them, they are a lot more expensive than a 1070 for only 8GB.. For 2500eur there is laptop with 9750h and rtx2080.

 

Yeah, not for gaming. But I saw a person who bought 9 of these laptops to resell them with a 20% benefit, (it cost 5000, it is sell at 3000, so probably he is selling at 3500) and he said already sold them. So people is really buying this for some reason?
https://www.amazon.es/gp/product/B07G7C5B6B?psc=1

P5200 is not much better than p5000

So wondering if I should buy a 1000eur p5000 , change my 1070, and sell my laptop for 2500eur with p5000 instead 950eur with 1070 (and sell 1070 separatelly)

 

try if you want, good luck selling such joke.

 

What makes you think they are jokes. Maybe people will buy them for professional use huh

 

Because almost no professional will buy a laptop that requires modded drivers to work.
Why would I buy a laptop with a Quadro that uses certified drivers and then I need to disable driver signature enforcement to install a driver, if you make a part, the part fails, you say that the simulation was OK, but you are using what is in fact tampered drivers, you get the full blame.

 

I don't think professional would buy a "gaming laptop" when they can buy for the same price a more "corporate looking" laptop like Dell or HP workstation which already include a quadro like that and a better cpu with no issue with driver update.. some example :
https://www.ebay.com/itm/DELL-PRECI...838099?hash=item3427ab5093:g:9z0AAOSwLXBcXzH7
https://www.ebay.com/itm/DELL-PRECI...649400?hash=item34270096b8:g:9z0AAOSwLXBcXzH7
200 $ more with 4k screen etc : https://www.ebay.com/itm/HP-Zbook-1...579884?hash=item1f0f3c0cec:g:DrEAAOSw0vperA2O

edit: and the quadro P5200 is a lot better than the P5000, the P5000 is equivalent to a 1070 while the P5200 is equivalent to 1080 max-q (which is superior to an overclocked 1070)

 

Max-q they aren't nearly half as good as the mxm formats

 

P4000 for gaming from my experience is about 10-15% lower in effectiveness to a 1070M. But the upside is that the temps are extremely low at full load. So there's that. If you can get it cheaply then definitely do (I got mine for about 300$, recently got an M4000M for ~130), some markets sell these cards cheaper than their gaming equivalents, ain my case 1070 costs 500+, and Quadro cards can serve other purposes. I rent my main machines to a friend for video editing courses. It can pay you back

 

I have an m5000m and m2000m I dont use wink wink ;-)

 

... and I've got an m4000m I don't use anymore, more wink winketty wink

 

A Quadro P4000 is more similar in performance to a 1060, it's far from a 1070, the 1070 have 256 more cuda core (15% more), frequency is higher by 40% (1700~1800MHz vs 1250MHz) and memory is faster by 33%...
As for the P5200 vs P5000, it still have 25% more cuda core, a slightly lower frequency and faster memory by 15%, so it's faster than a P5000 anyway, that's why it's very difficult to sell a P5000 laptop at the same price of a P5200 one with better other specs.

 

Someone please help!
I would need an Nvidia Quadro P5200 Max-Q vbios.
MSI, but others may come.

 

Help!
I'm looking for this vbios file.
86.04.84.00.39 (max-q)
I can't download it, it's nowhere to be found.

https://szchangyou.en.alibaba.com/p...18E_Q5_A1_16GB_GDDR5_Graphics_Video_Card.html

 

Did anyone tested Quadro RTX 3/4/5000 in Clevo, Alienware or some other Laptops?

 

https://www.3dmark.com/fs/22229508
https://www.3dmark.com/fs/21130593