GA-MA78LMT-US2H, Phenom II 555, core unlocking, disabling core 0

[creisti86]

Hello,

I have the mb&cpu in the title, good aftermarket cooler and PSU. MB revision is 1.3.

When first trying to unlock the cores, I was using Windows 7 64-bit. When I set the ECC Firmware to Hybrid, and ACC to auto or per core or all cores, the computer would POST and show cpu as a x4 B55 but wouldn't boot (would reset or BSOD and reset somewhere during Starting Windows screen). I tried disabling some of the cores, numbered from 1 to 3, (or even all of them, using only core0) having the same result. This seemed a bit weird as core 0 and 1 would work perfectly when it is a dual-core, but wouldn't work when tried to unlock, and disabled all others. Tried all sorts of BIOS versions, form f1 to f9, and some betas. All the difference I got between these versions is that somewhere along the line the spelling for Core Control option corrected from 'Munual' to Manual and, later at that same option, Core 1 became ungrayed and could be disabled (on older BIOSes you could only disable Core 2&3), but Core 0 remained grayed on 'Enabled'. Also tried a lot of settings with overvolting, underclocking, ACC(-12% to +12%) etc., no result.

For some time I let it rest there until i read somewhere that unlocking works worse on 64-bit OSes. So I installed a Windows XP 32-bit, and I could boot as an x4  (the joy). What I found out when trying to test the cpu with OCCT and prime95 was that the unstable core was Core 0 that cannot be disabled form the Bios. The computer behaved unstable, some programs would crash, others didn't, games crashed always after opening screen, but prime95 was able to run on the other 3 cores perfectly for about 1h, when i stopped it. Prime95 was failing on the first core as soon as the test got started saying rounding error 3... should be less then 0.4, hardware error. Again tried overvolting and/or underclocking but the core would fail regardless. Also tried Easy Tune 6, it has the options of disabling only the last 2 cores. As the processor is perfectly stable as an x2 on both core0 and core1 it is obvious that core0 in x4 mode is another core than core0 in x2 mode.

Obviously, I would like to run my cpu as a x3 with the defective core disabled, but can't do that as it is shown as core 0 in x4 mode. I read on this forum about someone saying that for him the weak core seemed to float around all numbers (http://forum.giga-byte.co.uk/index.php/topic,1276.0.html), someone else said that MSI boards might be able to disable core 0, another that core 0 cannot be disabled because it would have some logic for the whole CPU. Somewhere else I've seen people having the same problem http://hardforum.com/showthread.php?p=1035474882

So, is there any way someone could disable core 0 or do something so that another (working) core would be seen as core 0 so you could disable the defective core from BIOS? Is there a chance of a BIOS update that could do any of these ideas?

Sorry for the lengthy post, but i wanted to be precise and say everything I tried so i wouldn't get replies suggesting something I already tried doing. Any help would be appreciated.

[Dark Mantis]

Hi and welcome to the Gigabyte Forum.

I deal more with the Intel platform and so don't know so much about the AMD side of things but however found it a very interesting read. I am sure that Absic or someone else from the AMD side will be able to help you. Good luck.

[bytheway_r]

Hi,

Unfortunately I don't think you'll have much luck. A quick search revealed other issues just like yours:

http://www.overclock.net/amd-motherboards/615510-gigabyte-cpu-core-control-option-awesomeness.html

http://forum.giga-byte.co.uk/index.php/topic,900.0.html  ( same ppl )

Though it is interesting. It'd seem that either core number can float around or AMD sometimes disables the original main core ( core 0 ). Then, when you unlock it, you're stuck with a base core that's actually faulty. Anyway, from these topics I mentioned above, it'd seem that only Asus and MSI can disable core 0. Maybe ask runn3R about that bios idea he was supposed to forward to the HQ.

[creisti86]

Hi,
thanks for the links, I also read those before, just couldn't find them when I initially posted.
I'll try PM-ing them to see if there were any updates on the issue.

[creisti86]

no answer by now...

also found this thread:
http://forum.giga-byte.co.uk/index.php/topic,596.30.html

and on a romanian forum:
http://www.mygarage.ro/overclocking/46740-deblocare-amd-phenom-ii-x2-550be-5.html#post681722
"I asked gigabyte for a bios with an option to disable the core 0, not just 2 and 3, and after I gave them all the information they requested, they sent my request for technical support. These days I received an answer from tech support and they said my Core 0 can not be disabled:

<<Dear Sir,

Core 0 is the original core of the processor. So you could no disable it. Sorry for the inconvenience.

Best Regards,
GIGABYTE TECHNOLOGY >>"

I don't really understand what they mean by "the original core of the processor"... (and it's not a translation)

So people have been having this problem for some time now. If other MB manufactures have this option available, Gigabyte should also be able to introduce it. They might just not find this important, since only a few people are interested, and unlocking cores might be for them a low priority issue, because it technically is running your pc out of spec., they might be more interested in solving other issues.

Edit:
meanwhile I might use this:
http://www.tomshardware.com/reviews/bang-dual-processing-buck,815-3.html

If I could set-up all processes to run on the 3 good cores, I might be able to get it stable. Will only have the time for that at the end of the week probably though.

[Dark Mantis]

Please ignore this if I am wrong but I would have thought that if Core0 was the unstable one then you should be able ot RMA it for a CPU that had a good initial core.

[creisti86]

Well, as i said in my previous post, as a dual core both cores are perfectly stable. Only as a quad does core0 become unstable, probably core numbering isn't consistent between the two modes (which might also mean that it could be changed, so I could disable the unstable core). Being sold as a dual core, that means there is no reason for any RMA, like it is said to the unsuccessful amateur AMD unlockers: I got what I paid for.
sorry that I couldn't ignore it...

[Dark Mantis]

No that's fair enough it was just that I was under the impression that the primary core was unstable.

[bytheway_r]

Who knows, maybe it actually IS unstable. Maybe the dual-core version is using cores 3&4 renamed as 0&1 in his case? So, when he unlocked the other cores the processor reverted back to original naming? We probably would have to ask AMD to figure this one out .

[creisti86]

All right, I tested some time with THG. I managed to run applications that crashed before (chrome, prime, etc), and some games, but some apps set their own affinity manually (to 4 cores) so those should be kept in check, otherwise would crash. Also it lacked a bit the functionality I needed. I needed a program that would set the affinity to all processes, so I made a little app myself. For anybody interested or having the same problems I included it here. The app just cycles through all opened processes at a given interval and sets their affinity to the one selected (if it was different). Some processes can't be set the affinity to because access is denied (although I have used something I found on-line to enable some rights: "SeDebugPrivilege" http://rongchaua.net/blog/c-how-to-enable-sedebugprivilege/). The app will slow down the PC if set on a low interval and with the 'try again' option.
A lower level control of cpu cores might have been wanted, but I am no kernel programmer. No harm should come to the machine from using this, but still, I don't want to take any blame, so use it at your own risk. I also included the source code for anyone to see what it actually does, and perhaps modify it as they please. If anyone is interested I could make a version that would start setting affinity right away when it is run and that anyone could put in Startup so it is started with windows (edit: a windows service would be better), that version would also have the possibility to save the settings and probably some filtering on processes by name (a process that is known not to be able to be set affinity to, would be ignored).

http://www.4shared.com/file/lB-8pC-9/MyAffinitySimpleExecutable.html
http://www.4shared.com/file/wzBFFh3X/MyAffinitySimpleSource_2.html
it requires .net framework

Didn't have time to test it on the actual Phenom II computer as I am now at work... I expect it to work well mostly, but because not all processes can have the affinity set, and some might try to change it while running (and would run like that until I change it back), the PC would probably remain a little unstable.

[creisti86]

update:
Tested the application on the machine in question here, did some stability testing and some benchmarking, even some overclocking (3766Mhz on default voltages, max temp 56C), stable so far. As far as games go, only tested Starcraft 2, which seemed a bit smoother then on dual-core mode, and GTAIV, because I knew it was very cpu intensive, and could take advantage of even a 4 core cpu, but it was stuttering a lot every 1 or 2 seconds, don't know why, because it didn't try to change affinity after it was run, so i could stop the affinity setter application, it could not be related, haven't tried gtaIV in XP on dual-core mode. Except that, I had no crash or weird behaviour at all, also, on that computer, all processes could be set affinity to. Even OCCT which would always try to set its own affinity couldn't do it this time. I could still see in task manager core0 being used some 1-2% from time to time though, but this didn't seem to affect stability. I might need more time to be sure of all this, as everything was done kind of in a hurry.

[Dark Mantis]

That's the trouble, thorough testing takes time

[creisti86]

the thing is that normal stress testing doesn't work in this case, because the problem is not whether the working 3 cores will withstand 9999 hours of 100% stress testing, but if certain applications, events, services, might use the defective core at a certain point, and that could result in crashes or corrupt data. The defective core isn't that defective in the sense that it doesn't cause windows to crash or BSOD or anything like that, it makes processes crash, that is why I got the idea of trying to run all processes on the other 3 cores. So the testing should have a lot of application launches, multi-tasking, data handling and stuff like that, and, the more difficult thing, a good way to verify for errors.

[j2o]

hi core 0/1 are X2 mode 2/3 are x4  mode i & friend have simmilar rigs ,we have the core value set at -2% all cores . why down grade to3 cores .o/s vista 64 is ok with any configuration we are still running DDR2 800mhz ram 4x1gig 555 18

[creisti86]

why down grade to3 cores

because my core0 (in x4 mode) is unstable, that's what I've been saying this whole thread, I haven't been as lucky as you have.

also what do you mean by

555 18

?