GT 550m MXM-II Project

[triturbo]

I've used GT 240m as a base, and GT 550m core. I kinda overlooked that GT 550m's core has 4 more balls in its BGA. When the guy that was about to swap the cores told me, I decided to give it a go anyway. He also has flashed Lenovo Y470 vBIOS (thanks to a guy of MDL). The result: it outputs mixed/scrambled color matrix (and stays like this) only through VGA, no digital output so far - internal display/HDMI. I can hear the Windows loading sound. It can't be blind flashed, and I'm sure that the USB thumb drive is OK and properly set. I'm going to try different vBIOSes later this week with his programmer. Possible Optimus interuption or something? Is it possible to make OptimusLESS (or disabled) vBIOS? I read somwhere that upgrade to 7970 had similar issue, and once flashed with Clevo vBIOS (Enduro disabled) it booted just fine. I'm probably going tomorrow, the guy lives 100+ miles away in one direction, so I want to be prepared as much as I can. I already have a handful of vBIOSes thanks to Prema (N43SN), capitankasar (l502x), Sharadeos (l702x) and a Quadro 1000m. Waiting niffcreature for MXM GT 540m vBIOS. Your input is greatly apreceated. I'll make a guide if it turns out successful. Fingers crossed

[svl7]

You put a 550m GPU on a 240m board? And expected it to work? Those are two completely different architectures unless I'm very mistaken. Why even try this? You can get a more powerful MXM card for a reasonable price... I guess all that experimenting wasn't free. I love crazy projects, but something like this is bound fail, sorry.

According to the title, the 240m was MXM type 2.1? The 550m is an on-board GPU, it most likely lacks the necessary MXM callbacks as most vbios of soldered-on GPUs, and even if they were there, they would use the 3.0 protocol and not 2.1, there are differences in the software interface, not only the boards and pinouts.

You can't compare that to the AMD 7970m issue, in fact I tend to say that said issue is due to the Clevo BIOS / system design and not due to the vbios.

It's pretty incredible that your card seems to POST... in this case the system can access the vbios, you have to be able to blind flash the card. What's the notebook you're using? Are you sure there's no vbios in the system bios of your notebook?

[triturbo]

Yes, GT 550m core to GT 240m MXM-II board. The architectures are different, but the BGA package is the same, well almost, GT 550m has 4 more balls. Why I tried - last attempt to bring some live in my 5+ year old 5920G, and it's 16:10, that matters. No, it's not free

I've also read long ago that some GTX 280m MXM-HE was able to run with GTX 280m MXM-B vBIOS and that it was the only way to run it in some laptop. This gave me some hopes as well Otherwise, yeah, I knew that it's pretty much impossible and that MXM 2.1 and MXM 3.0 are different standards and probably different signaling. BUT the fact that it POSTs and even boots got me more than surprised So now I want to check every posibility to properly output to internal.

My initial plans were for HD6730 MXM-II - HD6770 core (6770/6730 has the same core, different memory) and HD4650 MXM-II board - 1:1 match in the BGA. I ditched it because of the more parts I had to buy and the lack of HD6730 vBIOS, also as I said, I overlooked the additional 4 balls on GT 550m's core

[svl7]

I've also read long ago that some GTX 280m MXM-HE was able to run with GTX 280m MXM-B vBIOS and that it was the only way to run it in some laptop.

That's possible... many manufacturers don't adhere to the MXM specs and simply omit certain callbacks and methods... can be and advantage when it comes to upgrading, but it can also be a big headache since certain vbios (especially nvidia) require at least parts of the MXM software interface to work properly, and if it doesn't the card will only run in a low P-state or refuse to post and display an error message.

I'd suggest to try the blind flashing again... make sure you really boot from the USB drive, use an autoexec.bat with the commands so you don't have to type anything. Use all the force commands, and the 'no user interaction' flag. It might still not let you to override the board ID error which might come up though.

[triturbo]

This is the autoexec that I was using, and I'm sure that the priority is set to USB first:

beep

C:\nvflash -4 -5 -6 -A dell1.rom


beep

When I put it in my ATi machine it writes "No nVIDIA adapter was found" or something along the lines and beeps. It doesn't beep at all in the other machine. I'll give it another go, but I don't think that it's going to work.

[svl7]

You might need to disable the write protection, at least that was necessary when I first flashed my 260m.

[triturbo]

So it should look like this:

beep

C:\nvflash -r -4 -5 -6 -A dell1.rom


beep

[svl7]

More like this:

nvflash -r
nvflash -4 -5 -6 -A dell1.rom

[triturbo]

Sadly almost 3 months and still nothing, the guy is busy and etc. I guess there aren't many people that are willing to deal with my stupid things. He was the only one who was willing to swap the cores, and now even he's out, so I guess, I'm on my own.

The vBIOS is still not flashable. It recognises the thumb drive, as before, but that's it. Should I try with a different version of nvflash?

Using the "flipped"* version of this image, and this schematic, I have found out what the additional 4 balls do:

G16 - FBC_DEBUG1 R154 10K/J_4_NC /// Mem I/F C FBB_DEBUG1 
G20 - VMC_CMD31   TP12           /// Mem I/F C FBB_CMD31 
T29 - FBA_DEBUG1 R196 10K/J_4_NC /// Mem I/F A FBA_DEBUG1 
Y29 - VMA_CMD31   TP22           /// Mem I/F A FBA_CMD31

G20 and Y29 are leading to TP (Test Point?), so I guess they can stay disconnected. G16 and T29 are obviously for debugging, and there are resistors before they hit the ground. What difference could it make if there are no resistors, as it is now?

*I got it wrong since there's nothing related to the graphics chip at AJ16, AJ20, Y7, and T7. Also, comparing few boards, I have found out that there is no "I", or "AI" in the BGA row/column designation. Am I wrong?

Finally, what kind of programmer I'll need in order to flash the various vBIOSes? The chips is

Atmel 924
25F512AN
SU27

Thanks to whoever gives an answer to all of this

[svl7]

I think one of the cheapest options would be an EZP 2010 (check ebay) the chip probably isn't officially supported, but I'm pretty sure that it'll work if you choose a chip in the flash programm which is identical pin-wise. Or I can sell and ship you a couple of chips already pre-programmed if you want (it seems you already have some possible vbios files).

[James Michael Acierto]

i think you have to disable the write protect on the flash chip to get the card to be recognized by nvflash

[svl7]

I think he already tried this.

[James Michael Acierto]

maybe there are some relocated pins that causes the card not to be recognized ?

[triturbo]

Thanks, but I have tried it, it's not working. I dig some more, found other notebooks with this GPU, and their schematics as well, and on some there's no connection at this points, so I REALLY hope that it's BIOS related

[pchatzop]

Sorry to resurrect yet another thread(on a different forum though), but triturbo's crazy projects always pique my interest . I am sorry to hear that you did not have much luck with the 550m gt project (though its impressive you got the card to POST). I recently got a romless toshiba hd 4650 to boot on a 5920g via sbios mod and you inspired me to go for the best mxm II card that could be created(that is the romless toshiba 4650 card with the 6770m core and eeprom=6730m). I searched high and low to find a vbios for the 6730m and i found a obsene laptop bios in a russian forum, and after disassemblying it with phoenix tool i came with this: Download 6730(2gb).ROM from Sendspace.com - send big files the easy way . This is the 6730m vbios, but for the 2 gb version. If some talented guys could compare this bios and the hd4650ddr3.rom bios and incorporate the needed mxm 2.1 stuff, we could have the best mxm 2.1 card ever WITH DIRECTX 11! I hope you have the courage to keep going for these crazy endeavors, you have been an inspiration for the modding community!

[pchatzop]

It seems like i did a mistake identifying the bios and it is in reality the hd 6650m vbios with 2 gb ddr3 vram. So back to square one. I'm trying to extract it from the Asus A43SA bios but i am having no luck(does asus use any obfuscation techniques to hide the vbios?) ,as i can not seem to find the bios of the only other laptop that contains the 6730m, the Nexoc E643. If anyone has the bios file(or could dump it) for this laptop, i'd appreciate a link.

Edit: I got it. I extracted it from the Asus A43SA bios with aptio MMTool. Hope someone finds it useful: http://www.sendspace.com/file/7ekefw

[triturbo]

OK we have some ground for 6730m now! Awesome work!

This project is currently on hold, but this doesn't mean canceled at all. Just that I've concentrated on other supposed to be successful mods I hope that I'll call it dead/successful by the year's end.

[triturbo]

Sadly it's pretty much dead so far I've used this guide to integrate the vBIOS in the sBIOS. I've tried two so far, and the only difference is that the scrambled colors vary from vBIOS to vBIOS. It still boots and loads Windows. I have a few more as mentioned in the first post, but I don't have much hopes Hopefully the last GPU upgrade that I'm planing (HD8850m) would be successful

[pchatzop]

Have you tried the original 240m vbios? I could extract this 550m vbios from the asus n43sn bios: https://mega.co.nz/#!R9wiCACJ!SeeNvzcZLish0Y8nbfmAmz6ncb4PPgFJQG6sNQJ652c . I still dont think this will work because of the mxm parts missing but i would love to be proven wrong . Been searching for that 8850m vbios too, but the samsung models seem to be the only ones with ddr3 vram. And because of the way samsung supplies bios updates, have not been able to get a proper dump of the sBIOS to extract the vbios from. Hope you have more luck on that project.

[triturbo]

Yes, I've tried it, and no, it's not working as well. I have better hopes for the ATi

[bauto601]

Maybe the 550M bios has different memory timings than the memory that is on the mxm board? Because i think that one of the causes of the fail of the card is that the memory isn't controlled properly.

[triturbo]

As I dig deeper there are like 10~14 resistors that has to be changed. The problem is that I have all the schematics needed, the GPU one including, but it has no designations over it (so I don't know which component is which), and I can't find boardview of it. So it might be doable if one gets his hands on the boardview or component placement sheet or whatever. The GPU project code is M90GN, that's how I found the schematics.