1NOTE: If you have a 9600M GT or similar card, that works with the Microsoft standard VGA driver, but produces a lockup/BSOD/black screen as soon as you install recent Nvidia drivers, read this. This is a known bug in Nvidia drivers, and a workaround is available in the form of a VBIOS patch. This is not a problem with the BIOSes in this thread.
2
3Greetings!
4
5I think I found the reason behind GeForce 9600M GT's reluctance to work with Acer Aspire 5920G/6920G/8920G laptops. Apparently it all boils down to a corrupt MXM structure in the BIOS.
6
7The 6920G and 8920G <s>share the same BIOS</s> EDIT: They actually use different BIOSes, despite containing the same motherboard (though the MXM structure is the same in both), and indeed, their symptoms are also the same: no display on the internal LCD during POST. However, connecting an external display to the HDMI port allows POST messages to show up on that screen.
8
9So, I looked into the MXM structure embedded in the 6920G's BIOS (armed with the recently leaked specification for the MXM structure), and found the following:
10The output device structure for the HDMI port of the 6/8920G is 0x00F9F78A2120, which parses to:
11Device type: TMDS (DVI or HDMI)
12DDC port: DDC-B (MXM pins 230 & 232)
13Connector: HDMI Type A socket
14Connector location: External
15Link type: Single-link DVI/HDMI connected to TMDS link A
16HDMI audio source: No audio
17Digital signal drive strength: Normal
18Digital reserved field 1 (2 bits, should always be 11): 01 (incorrect, violates the MXM spec!)
19Output select GPIO: None
20DDC select GPIO: None
21Device detection GPIO: #0, active low (logic low on GPIO pin 0 indicates a display plugged into the HDMI port)
22System hot plug notification: Disabled
23
24The culprit is apparently Device detection GPIO (marked in red), which is presumably always low/0 on this machine, indicating a device plugged in! Because the boot display order is VGA first, then HDMI, finally internal display; connecting a display to HDMI (or fooling the MXM VGA into thinking one is connected, as it is happening now) suppresses POST display on the internal LCD, and directs it to the HDMI port. When Windows finally boots up, it recognizes the internal display and makes it primary (Windows always considers the internal display primary by default, regardless of the MXM structure), so display in Windows is normal.
25
26So, the MXM VGA basically thinks that an HDMI display is plugged in all the time, and proceeds to make that imaginary display primary.
27
28The 5920G (and its little brother, the 4720G) fares even worse: it flat out refuses to POST with a 9600M GT plugged in! Again, a quick look at the MXM structure (hidden in BIOSCOD02.ROM inside the BIOS image) reveals that something is amiss, even more so than with the 6/8920G:
29-The checksum of the MXM structure is wrong.
30-On this notebook, the boot display preference order is different: internal first, then analog VGA, then HDMI, and finally TV-Out. So, the problem is not the same as with the 6920G. However:
31-The internal display's output device struct claims that the internal display can also produce sound through a SPDIF connection, and that the link between the internal panel and the computer is long-distance, requiring extra signal power. A field identified as "reserved, set to all ones" is set to all zeros.
32-According to their respective output device structs in the MXM structure, the VGA, HDMI and TV outputs are all internal connectors.
33-The VGA output has a "TV format" field set to NTSC-M, despite not being a TV output. The Digital Connection field is set to "Reserved for future uses, do not use" instead of "Not applicable".
34-The HDMI link again has the reserved field zeroed out.
35-The TV out is identified as an HDTV YPbPr component output, but the TV format is "legacy 525-line NTSC-M" (should be "480i" for an HD connector). And again, Digital connection is "Reserved for future uses". But here comes the real bummer:
36-All outputs (including the internal display!) mark GPIO#0 for device detection, interpreting logic 0 as Connected, and 1 as Disconnected. This is wrong in many ways, e.g. one GPIO line can only be used by one display; the internal display has no device detection (it is always plugged in); the board has no GPIO expander, so GPIO#0 doesn't even exist, and is always read as 0; and finally the HDMI output also marks this GPIO as its Output selection GPIO - which is impossible, as a GPIO can't be in more than one mode at the same time.
37-The MXM connector claims to be MXM-HE, capable of delivering 3.2W to regular and 12W to HE cards. I the reality, the connector is MXM-II, so it doesn't support HE cards at all; and I'm pretty sure it can handle more than 3.2W when dealing with regular Type II cards. (I've modded this one to 32W.)
38-The final two fields are a vendor-specific field for PCI\VEN_0401 (a nonexistent vendor), and a weird 4-byte field that appears to be completely random junk. Disassembling BIOSCOD02.ROM in IDA revealed that this field is not really junk - it is the beginning of the x86 machine code that responds to MXM interrupts. Apparently when the BIOS was built, the end of the already-wrong MXM structure got overwritten with BIOS code!
39
40I've made an attempt to fix these errors, but I have no Acer 5/6/8920G to test them, <s>so the following modded BIOSes are untested</s> Update: 6920G+9600M GT confirmed working by lizzarddos, 5920G confirmed by multiple users to help 9500M GS, 9600M GS and Quadro cards, 5920G + 9600M GT confirmed by u6b36ef, 9920G personally tested by me, so there is no reason why the 4720G would not work.
41
42(BIOSes last updated 2011-05-29.)
43Aspire 4720(Z)G (Phoenix): request download ticket | ifile.it
44Aspire 5520G (Phoenix, with unlocked menus): http://ifile.it/bi2qtc6/IC50133_MXM.WPH
45Aspire 5720(Z)G, 7720(Z)G (InsydeH2O, with unlocked menus and integrated EFI shell, probably fixes external display issues, updated 2011-11-24): http://ifile.it/3wlxm6j/icl50_1.45_black_edition_40.fd
46Aspire 5920G (Phoenix, with unlocked menus, updated 2011-11-06): http://ifile.it/noxplab/5920G_MXM_V3.WPH
47Aspire 6920G (Phoenix, with unlocked menus): <!--request download ticket | ifile.it previous: -->request download ticket | ifile.it
48Aspire 8920G (Phoenix, with unlocked menus): <!--request download ticket | ifile.it previous: -->request download ticket | ifile.it
49NEW! Aspire 9920G (Phoenix, with unlocked menus, updated 2011-11-06, use with driver v275.33 and these VBIOSes): http://ifile.it/ej0c3oi/CV120_MXM_V2.WPH
50Extensa 5620G, Travelmate 5720G (Phoenix, with unlocked menus, experimental): request download ticket | ifile.it Link fixed 2012-03-28
51<s>iFile.it folder containing all current BIOSes: BIOS Mods | ifile.it</s> This folder contains newer BIOSes than the ones above, and one of them (5920G) has produced a bad flash. Consider these to be beta BIOSes until further notice!
52Please test them and report any results here!
53
54Note: <s>The 5520G, 7520G and 9920G are affected, too - however, the BIOSes on these machines don't contain an MXM structure at all, so unless someone figures out how to add an interrupt handler for INT 15h, EAX=0x5F80 to BIOSCOD02.ROM, they cannot be patched. (Note: Even if someone finds the interrupt handler for INT 15h, EAX=0xE820 - memory map, that would be of great help.)</s> 5520G and 9920G are available now. The 5520G BIOS may also work on the 7520G, but this is untested.
55
56The Aspire 5720G (ICL50) BIOS is NOT required for 9600M GT upgrades - it primarily fixes issues with ATI cards refusing to drive external displays. It is also usable on Aspire 5315, 5320, 5715(Z)(G), 5720(Z)(G), 7320, 7720(Z)(G) and eMachines E510 machines, including IGP ones, to unlock extra tweaking menus. It does not, however, enable upgrading IGP ICL50s to MXM cards - that is still physically impossible.
57
58Thanks,
59.NetRolller 3D