Fun Stuff > CLIKC
The PC-building/hardware knowledge thread
bhtooefr:
A lot of the BIOS POST code is very old code.
That first 64k RAM failure code likely dates back to the original 1986 AMIBIOS code, which was copying a feature that goes back to, I believe, 1982 (the latest BIOS update for the original IBM PC) or 1983 (the PC XT). The first 64k of RAM failing is legitimately a different failure from any other failure, because the system can't do anything, whereas a failure later in RAM would still allow the machine to boot, at least to ROM BASIC or PC DOS 1.x. And, back then, the first 64k were nine chips on the motherboard.
I *believe* that failure happens after cache tests - modern PCs boot in multiple stages, and I think the BIOS doesn't get control until after caches are initialized anyway. I believe that would indicate a failure somewhere in the memory path - could be the memory controller on the CPU, could be the RAM slots, could be anything in between.
Case:
--- Quote from: bhtooefr on 22 Feb 2019, 05:22 ---A lot of the BIOS POST code is very old code.
--- End quote ---
Yeah, I suspected as much - there are actually some ancient Intel helpsites that speak of 'three long beeps', which allegedly indicates some damage to the keyboard controller?
I don't think my heap of junk even has a dedicated keyboard controller? (I don't think it even has a Southbridge anymore, for that matter...)
--- Quote from: bhtooefr on 22 Feb 2019, 05:22 ---That first 64k RAM failure code likely dates back to the original 1986 AMIBIOS code, which was copying a feature that goes back to, I believe, 1982 (the latest BIOS update for the original IBM PC) or 1983 (the PC XT). The first 64k of RAM failing is legitimately a different failure from any other failure, because the system can't do anything, whereas a failure later in RAM would still allow the machine to boot, at least to ROM BASIC or PC DOS 1.x. And, back then, the first 64k were nine chips on the motherboard.
--- End quote ---
There was a time when I had memorized the entire layout of the 64k of RAM of my first machine... (the original Commodore Breadbox, of course), what address that little Motorola 6502 would mirror its registers to, etc.
Part of me still marvels at CPUs having an onboard L1-cache of 64k....
--- Quote from: bhtooefr on 22 Feb 2019, 05:22 ---I *believe* that failure happens after cache tests - modern PCs boot in multiple stages, and I think the BIOS doesn't get control until after caches are initialized anyway. I believe that would indicate a failure somewhere in the memory path - could be the memory controller on the CPU, could be the RAM slots, could be anything in between.
--- End quote ---
Are you sure about the memory controller being on the CPU rather than the board? I don't think it is the slots themselves - all for of them failing simultaneously would be plenty odd, so I figure the damage is a bit higher up the memory path ... of course, if the damage is to any part of the CPU, be it the cache or an onboard memory controller, I'm back to square one ... :-\
Ok, thanks bhtooefr. Now what wanted to hear, mind you, but I can't say I hadn’t suspected something like this already. Always better to have a second opinion, even if it's not what you'd hoped for.
Case:
I just hope it's either the CPU or the board, and not both. Either part would go for 100+€ on ebay, which would make an upgrade to a modern system affordable for me, but if it's both of them...
bhtooefr:
The memory controller moved to the CPU die with the Nehalem CPUs (although Westmere temporarily moved it back out to the GPU die, which was a separate die on the same CPU package, but Sandy Bridge moved it (along with the GPU) onto the CPU die for good).
Basically, the list of components between the memory controller and the RAM consists of:
* CPU socket
* The motherboard PCB itself (a cut trace can cause all sorts of problems)
* The RAM slots (I doubt this is the case considering that you said all slots are affected)
Any solder joints in between can also cause problems, as can power delivery hardware on the motherboard not supplying power correctly to either the CPU, the RAM, or both.
IIRC, keyboard controller is embedded in a Super I/O chip attached to the LPC bus. Some parts of it are IIRC emulated, though, for higher performance (specifically, A20 gate switching).
And, as far as a southbridge... you have one of those, it's the Z87 "chipset" itself.
pwhodges:
My bet is clearly on the motherboard; I've never experienced a CPU failure in my entire career, even when a fan failure caused overheating which stopped normal functioning (that was a Pentium Pro, a notable high spot in the progress of CPU design). Although I've had memory board failures, they've always been clearly diagnosable using one of the Memtest utilities, and generally from new rather than appearing later in time (though sometimes subtle, and so not apparent till later).
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version