Testers for AM2CPU Phenom BIOS wanted

[killerbobjr]

All the AM2 and AM2+ motherboards made by ASRock use the exact same microcode which consists of microcode IDs 0414, 041B, and 1022. Other manufacturers (MSI and ASUS for instance) use these, plus additional microcodes. If I had to guess, I'd say the 0414 microcode is for the F2, F3 cores, 041B handles the G1, G2 cores, and 1022 is the B2, B3 cores. I did see one more microcode ID in MSI/ASUS BIOSes of 1020 that had the same date as 1022. It makes me wonder if that is the TLB patch. Since ASRock designs their motherboards for the economy market and does not intend for them to be used in server machines, I'd hazard a guess that 1022 is the non-TLB patch microcode. The only way to know for certain is to try both microcodes and see which one slows down the built-in test in Winrar.

As far as my method, it's extremely simple and very obvious in hindsight: I remove all the description strings that show up in the rightmost panel for each of the BIOS settings that have it. For things like "Legacy USB Support", which very obviously enables or disables support for old USB devices, the description string reads: "Enables support for legacy USB. AUTO option disables legacy support if no USB devices are connected." This is verbose, unnecessary, and entirely redundant. There's about 4K of strings like this that when eliminated, free up about 2K of BIOS space -- enough to add in one additional microcode. If I need more room for additional microcodes, the next step would be to abbreviate as many of the BIOS settings as possible. For example, all "Configuration" strings can be shortened to "Config", "Primary" to "Pri", "Secondary" to "Sec", etc. I found I could free up an additional 2-3K in strings doing this. The final method to get additional space (which I haven't tried yet since I didn't need it) would be to reorder the abbreviated strings so that all the similar ones are grouped together to enable greater efficiency in the LZW compression dictionaries. I'd expect this to only yield a few hundred bytes in additional space at most, but it's there if needed.

Since I took the microcodes from ASRock's AM2 motherboard BIOSes, there is full support for all AM2 dual cores. What surprised me was that this microcode also supported my socket 939 Opteron, but since it's a F2/F3 stepping, I suppose that would make sense. So the answer is yes, newer AM2 X2 cores are supported. The full list of cores supported in this microcode:

Single core - Orleans (F2, F3), Lima (G1)
Dual core - Windsor (F2, F3), Brisbane (G1, G2)
Triple core - Toliman (B2)
Quad core - Agena (B2, B3)

Finally, to repeat: The only thing I've changed in this BIOS version is the description strings and the microcode. All other blocks in the BIOS have been left untouched. This means NONE of the executable code in the BIOS, aside from the microcode, has been changed.

If you are unsure about flashing, then don't. To be absolutely safe, get another flash chip. You can do what I did which was to get my Hackintosh computer's chip (a 4Mbit chip -- only 2Mbit are flashed in the ASRock), pop it out then gently press it part way back in, start up the Hackintosh with a DOS/Windows boot disk that has a flash utility on it, back up the Hackintosh BIOS, then pop out the flash chip with a non-metallic tool (a toothpick will do if done carefully) and leave the Hackintosh on. I then pop out my 939SLI32-eSATA2 flash chip, gently push it partly back in the socket, start up the machine, load the flash utility, load this BIOS, pop out the flash chip, push in the Hackintosh chip, and flash it, flashing all blocks. To restore the Hackintosh, I pop its chip out of my 939SLI32-eSATA2, put it back in the Hackintosh, then reflash it with the saved BIOS, telling it to ignore the checksum/ID and to flash all blocks.

If you're going to experiment on your original BIOS chip and this BIOS doesn't work (mis-flashed, microcode doesn't work, etc.), then either a BIOS chip replacement or reflashing would be necessary. If you have a friend with an AMI BIOS based motherboard computer, you could persuade them to let you reflash on their machine. Reflashing an AMI BIOS on a Phoenix BIOS motherboard may be possible, but a blind write flashing utility would be needed. Or if you have a backup computer somewhere, you could buy a cheap Willem burner off of eBay (about $45 including shipping for one that has a PLCC socket) and reflash with that.

. . .

Okay, I just went over the SATA2 BIOS again. It looks like there's enough room for the entire AM2 microcode without having to remove any strings. In this new version, the ONLY thing I did was replace the old microcode with the new microcode. If you are currently using the AM2CPU board on your 939Dual-SATA2, then this modified 2.31c BIOS should work transparently. Since I know for sure this microcode works with my 939 Opteron, I have 100% certainty that it will work with any F2, F3 core cpu, and an almost 100% certainty it will work with G2, G3 AM2 cores. The only thing then to test is Phenom (B2, B3) capability.

[killerbobjr]

Originally Posted by Clworld

I tested your BIOS on my 939SLI32-eSATA2 board with Phenom 9850 Black Edition but It isn't works.
It does not POST and Nothing displayed.
It's same behavior as I tested before with old 1.40 BIOS.

With Athlon X2 6000+ I used before, This BIOS works correctly.

Okay, let me try using the other microcode (1020) and make up a second BIOS to test with. If that one doesn't work, there may be other initialization code in the main BIOS block that has to be added in for a Phenom to work. If it's just a matter of setting the default multiplier, that should be easy to add. If it requires specific code, I'll have to disassemble the ALiveDual-eSATA2 BIOS to find out what it is.

[killerbobjr]

I've been digging into Phenom documentation and reading some BIOS mod threads. It looks like there's additional code that is needed to initialize Phenom processors which may not fit into a 2Mbit flash, at least on the 939SLI32-eSATA2. The Dual might have enough room if I edit the strings into a very small size and only use one microcode. The disassembly of the BIOS code is going to be long and tedious. I might have more success if I had both 939Dual-SATA2 and ALiveDual-eSATA2 motherboards on hand to cross test with. Unfortunately, there is no AM2 equivalent for my beloved 939SLI32-eSATA2, so I'm operating pretty much in the dark with that.

I'll try to post progress reports here as I find out more.

[Armatoste]

Thanks killerbobjr. You are the man for doing what Asrock cannot or does not choose to do.

[bumblebee]

Originally Posted by Armatoste

Thanks killerbobjr. You are the man for doing what Asrock cannot or does not choose to do.

I second that motion. Would test but for lack of a Phenom. Keep up the good work!!

[Maggi]

Greetings everybody !

After hunting the web, mailing with ASRock back and forth, swapping my former Dual939-SATA2 with a Dual939-VSTA (mainly because ASRock told me the 2mb BIOS chip form the SATA2 would be insufficient for supporting Phenoms, but the 4mb chip from the VSTA could do), I finally found this thread and have hopes that it it indeed possible to get a Phenom up and running ... :)

Admittedly, due to lack of official support from ASRock, I didn't buy a Phenom just yet, but seeing this excellent community seems to be hard working to get the ball rolling, I'll start saving money right now.

I'll let you know, once I get my hands on a Phenom and hopefully by that time, there's a BIOS from you to tinker around with ... :D

So thanx a lot for taking care and keep up the good work !

Best regards,
Maggi

PS: right now, I'm running an X2 3800+ EE @ 2500 and applied a physical, adjustable vmod to my AM2CPU board to get it stable up there

PPS: my vmod is based on this instruction, but I used an adjustable resistor

[killerbobjr]

Well, we have a bit of a sunshine breaking through. Going over some of the BIOS threads at rebels haven, I found out that a good sized chunk of BIOS code is used for boot from LAN. In removing that and replacing it with a stub routine, I now have about 25K of free space to work with. That should be enough for adding all sorts of microcode and initialization code. I also found that uniflash will flash AMI BIOSes on Award mb's, so I can now flash the 939sli32 chip directly and not worry about recovery. That should speed up the development cycle.

[master381]

thanks killerbobjr! this is awesome; i was so disappointed when we found out that there wouldn't be official support. I don't have a phenom, but am thinking about buying one now because of this! I have a 939dual-sata2.

and for everyone, just fyi, newegg is selling 9550,9750,9850(243$) phenoms now.

[bluetooth]

Any interesting discoveries so far?

[killerbobjr]

Yeah, I just discovered that Mwave has the 9850 for $435 -- eight dollars less than the one I just ordered from Newegg on Saturday (when they weren't offering free shipping). Show ya what happens when one is impatient.