PDA

View Full Version : Pinmod for Phenom X4 9950BE? (Undervolt/underclock)



ziddey
05-24-2012, 11:04 PM
I'm trying to fix a machine for a friend. I get a feeling it never actually worked right. It's a Shuttle XPC SA76G2, which from what I've gathered, supports CPUs up to 105W TDP and has a 250W power supply.

The Phenom X4 9950BE (Agena) is 125W TDP with 1.30V VID.

The BIOS does not have provisions to decrease vcore or multiplier, unfortunately.

I managed to get Windows installed, but any sort of stress testing will immediately cause the machine to reboot. Hooked up to my Killawatt, I see that it's drawing 300W, well over the 250W PSU rating.

As a proof of concept, I'm using PhenomMsrTweaker to reduce multiplier and VID, and am getting rock solid stability, with power draw not exceeding 200W at the wall. I will try to find the limits, since it's "only" at 1.15V VID and 10x200=2000 right now.


I have not built (or really even touched) an AMD machine since the 462 days and know nothing about these K10s. Since there is no bios adjustment, is there a way to pinmod these CPUs for lower vcore/multiplier?

Again, it's not my machine, and I'm imagining the owner is not wanting to spend any money (although how a 9950BE ended up in here, I have no idea)

Thanks.


If there's no easy solution, I'll have to have it run PhenomMsrTweaker on boot and advise to get a lower end (or Phenom II if supported) and sell this 9950BE

CrazyNutz
05-25-2012, 05:46 AM
PhenomMsrTweaker on boot sounds to me like a very practical solution :shrug:

The Stilt
05-25-2012, 07:14 AM
You said there was no control for core FID & VID in the bios.
The XPC must have Award bios then -> i.e it is either SA76G2 or SA76G2L model.

Either way you may want to try this: http://www.multiupload.nl/WLRLT8TDGD

It is a modded bios I made based on the most recent S10J bios.

CPU FID & VID controls added under "Advanced Bios Features" -> "CPU Feature".

This specific bios is PITA to modify as the standard tools cannot be used (will completely corrupt the bootblock).
Therefore even I checked there SHOULD NOT be any of such issues I cannot be completely sure since I cannot 'field test' it.

So try it on your own responsibility.

ziddey
05-25-2012, 09:43 AM
Wow, thank you! I am extremely appreciative!

I will have to see how the owner wants to proceed, since I won't be able to do a hotflash currently.

I'm imagining it's not as simple as not flashing the bootblock?

Hopefully, I get the green light and it all works out.

Thanks!

edit: multiupload.nl seems to be down right now


Any chance you could provide a mirror? Looking through all my old parts, I think I have an extra chip to hotflash in case things go south

The Stilt
05-25-2012, 07:32 PM
http://depositfiles.com/files/xsqhkh3wg

http://www.filefactory.com/file/5tmcrwa0xih1/n/SA76G_S10JMOD.zip

http://oron.com/oga1gfh0p8pl

Flashing the bootblock should be fine since it has not been touched.

If this one works I can enable some other features too, plenty actually.

ziddey
05-29-2012, 01:21 PM
Thank you! I might have to order a spare chip to hotflash onto.

I'm imagining there's no way to use the bootblock in case of a bad flash since it looks like there's not even a fdd header?

ziddey
06-02-2012, 04:12 PM
Have any recommendations for a person/company that sells pre-flashed chips?

I was debating flashing in windows (with msrtweak to drop fid and vid), or use a usb key and awdflash. Chose the usb key route, and it rebooted itself midway through the flash. Come to think of it, I probably should have hooked up a more powerful psu, but too late now.

The chip is an ESMT 50
F25L008A
C70H2
0744

Never completed the bios flash to be able to tell if it'd have worked or not. Figure with an extra chip, I'd be able to hotflash. Don't have to give the computer back for a little bit longer, but need to get it done soon.

I'm assuming a usbjtagnt won't work.



Edit: checked all the spare boards I have laying around and found one with the same package chip. However, it's got a 4Mbit 3V chip vs the shuttle's 8Mbit 3V. Also, it runs amibios vs awd.

Winflash refuses to run because it's not an award bios. Awdflash also detects it's not the same part, but with /qi flag, it gets stuck at "please wait!"

Is there any way this would work? Or because of the chip disparity, it's not possible? The potential hotflashing board uses an amic a25l040-f chip

The Stilt
06-03-2012, 05:17 AM
Try this: http://ra.openbios.org/~idwer/flashrom/dos/0.9.5.2-r1517/flashrom.exe

Make a bootable DOS USB-stick and throw it in there.

It should be able to flash it even the hotflashing platform is using a smaller bios chip.

You might need to use "-p internal:laptop=this_is_not_a_laptop" switch incase the chipset is not recognized.

ziddey
06-04-2012, 11:20 AM
AWESOME! Program worked great. Unfortunately, it doesn't seem like the bios mod works. Hotflashed back to stock twice.

The Stilt
06-04-2012, 01:49 PM
Ok, the bios structure is then even more delicate than I first thought.

I rebuilt the whole bios manually this time.
If this does not work, nothing will.

https://rapidshare.com/files/1548167553/FINAL.zip

Flash in DOS mode, with flashrom.

ziddey
06-04-2012, 02:13 PM
Flashed without a hitch, but don't see any new options in BIOS.

Thanks for your help by the way. I really appreciate it. Perhaps, msrtweak will be the only way?

The Stilt
06-04-2012, 02:29 PM
You can see the "CPU Features" menu under "Advanced Bios Features"?

The Stilt
06-04-2012, 02:40 PM
Try this too: https://rapidshare.com/files/1763571557/FINAL2.zip

ziddey
06-04-2012, 02:56 PM
You can see the "CPU Features" menu under "Advanced Bios Features"?
Yeah, still only have cnq and secure vm mode options.

Try this too: https://rapidshare.com/files/1763571557/FINAL2.zip
Giving it a try now. thanks

ziddey
06-04-2012, 03:03 PM
Awesome. Disabling CnQ allows fid/vid adjustment in hex now. Thanks so much!

I'm seeing that each fid adjustment is 0.5x, so assuming that fid=0 would mean 13-5=8x (stock was 0a).
Time to play with vid.

You're a hero

Dropping vid by 1 increases vcore by 0.02v. Going the other direction, it doesn't seem to scale linearly. I'm imagining due to droop maybe? Either way, got it at 1.20vcore in the bios now. Time to play!

Assumedly, vid should be in 0.0125v increments. 1.175vid is translating to around 1.200vcore idle and 1.216vcore load

This CPU is showing itself to be a real runt. I'm told it was prebuilt with the 9950be, a 125w tdp part. The psu is only 250w, and the system calls for no higher than 105w tdp IIRC. Supposedly, it worked without issue in the past. Either way, it looks like there's really no undervolting headroom at higher fids.

Ok, going to have to settle with 2.1ghz 1.125vid. That's "stock" for a 9450e, except with 2ghz ht. It's a pig with vcore afterwards and becomes hard to tell whether I'm hitting instability from insufficient voltage or psu overload. Hopefully, they'll be happy the pc works again, even at the cost of 500mhz.

Thanks for taking the time to mod this bios!

The Stilt
06-04-2012, 03:07 PM
Sorry, just a minute.

The Stilt
06-04-2012, 03:22 PM
Desired FID = Multiplier (e.g 13.0x) * 2, dec to hex -10h = 0Ah (26 in decimal to hexadecimal = 1A, 1A - 10 = 0A).
Desired VID = 00h = 1.550V, 01h = 1.525V, etc (IIRC).