Beep codes during POST - Are they made by CPU running, or m/b hardware? Q9650, P45mb

[BlindFreddie]

aka "How do I jump-start a comatose Q9650?"
or "Name 101 uses for a dead Q9650"
or "CPUs are more delicate than you might think".

Question 1: If you get beep-codes from the PC speaker during the POST, does this mean that the CPU is running, or are the codes generated by motherboard hardware?
If I remove all RAM, I get beep codes dah-dit-dit - 1 long, 2 short.
If I install a RAM stick that is known to work with another board & CPU, I get nothing.
Same thing happens in 3 motherboards: GA-EP45-UD3R, P5Q dlx, P5K-VM.

Question 2: What is max safe volts for a P45 Northbridge? Is 1.48V OK? (My suspicion is that P45s are lemons for oc'ing quads - 55nm technology hasn't got the grunt.)

Brief history: This chip ran fine for many months 24/7 in my P5K3 dlx at 467x9, Vcore 1.350 (BIOS), VNB 1.55V, VFSB 1.30. I swapped it into my GA-EP45-UD3R, and after a struggle got it up to 469x9, Vcore 1.3625 (BIOS), VNB 1.48V, VFSB 1.32V. RAM is 2x1GB Corsair PC-8500C5 at 2.14V. Corsair HX-520 P/S. I tweaked everything, but it responded only to more VNB. It ran fine for a couple of weeks crunching WCG 24/7, but then I got an invalid result or 2.
So, I decided to drop to 468x9. After hitting F10 to save the new settings, I got a blank screen, and sets of 22 shortish beeps, with a 2 sec gap. Clearing CMOS incl removing battery did not help. I finally discovered that it would boot at default 333x9 with RAM in Channel A only.
I then tried exploring its new overclocking limits, and bumped Vcore to 1.30V. Nothing! No sets of 22 beeps! And nothing since, except for RAM-error beeps as above.
I decided to risk my QX9650 in the board, so installed it with my spare stock Intel HSF ... It runs, but not with RAM in Ch B.
I was going away next morning, so then set LLC ON, bumped the multi up a notch, restarted BOINC, and it's been crunching away with 1GB RAM at 333x10, 24/7 for the 3 months or so until now.

The Q9650 has never had Vcore above 1.40, and that was only long enough to test that it didn't help. VFSB was never above 1.34, VNB never above 1.48V (UD3R) or 1.55V (P5K3 dlx). VPLL never above 1.57. I'm pretty sure the Realtemp has never hit more than about 60.
-----
[Edit] - Further info: I've been experimenting ...
Now only 1 of my 2 Corsair 1GB will work in the UD3R, and only in slot DDR2_1 (closest to CPU). I'm almost certain that either stick would work earlier.
The QX9650 would not boot in the P5Q dlx, but I was probably using the RAM stick that now seems bad. This motherboard had not been tested by me since it was returned from RMA. (The RMA was the reason for buying the UD3R).
The QX is back in the UD3R, crunching at 333x11. Successfully booting the UD3R seems a bit hit-and-miss.
Did installing the probably-dead Q9650 kill the P5Q dlx? I'll have to try the QX again I guess.
I'd go buy a new Q9550 to experiment with, but now is hardly the time to be buying more LGA775 gear.

It seems like I have at least 2 component failures ... CPU, RAM, maybe 1+ m/bs.
Has my hardware got leprosy or something?
----
Answers/opinions/theories of the failure welcome.

[Machinus]

Beep codes come from the motherboard.

[BlindFreddie]

Beep codes come from the motherboard.

Not the reply I was hoping for, Machinus, but thanks anyway.

Nobody has ventured a reply re. the max safe NB volts for P45.
I've cross-posted to the GA-EP45_UD3* thread: http://www.xtremesystems.org/forums/...postcount=4255

[Leeghoofd]

Lo Freddie, well on the P5Q Dlx and E boards I never needed over 1.3 to stabilise my quads with 4GB... the magic with P45 is the cross reference of the volts and the GTL tuning that needs to be correct. Some CPU's however need a total different setup than others...

P45 is very sensitive to volts and more isn't always better, it's the mix of NB, FSB Term and Ram divider that gets these boards flying....

If I was you I would try the two sticks seperately with some Memtest torture in all the dimm slots... it could be that one of the sticks produces errors... or that one dimm slot is borked...

I've hardly seen a CPU die... but the volts you pumped are a bit too much for my own 24/7 safe zone... and in fact they are a bit steep for the Q9650 (could be it was a dud ) When a cpu is dead the board usually powers on by itself once you flip the PSU switch....could be the P5Q was dead from the start. To test, buy a cheapo cpu mate like an e6300 or such, no need to waste cash on another quad cpu...

To me the giga board is a bit harder to setup but handles high FSB better. Beep codes are generated by the bios and put out via the speaker...

[BlindFreddie]

Thanks for your reply/advice, Leeghoofd.
I'll be happier if I can get the P5Qdlx to boot with just anything for the moment

Beep codes are generated by the bios

For the BIOS to generate beep codes, surely it (the BIOS flash-RAM) has to be read by a CPU, ie the CPU has to be running.
If I take out all RAM sticks, I get memory-error beeps. In this situation, what RAM is the CPU using, or is it just reading the ROM and executing instructions that only store working data in the CPU registers? Or is there some scratchpad RAM somewhere on the m/b, eg in RT clock?
These are details that are not discussed these days - Oh for my Rockwell AIM-65, KIM-2 or Z-80 system!

[Leeghoofd]

The bios checks all components and one that is faulty generates a certain beep code... just try the working dimm in all the slots and see if it posts or even OC's the slightest... Straight after the first post set the correct voltage for the rams plz, otherwise it might stop posting... also check if all power cables and co are properly seated... it can be many things...

I hope both ya mobo's are okay and it's just a dimm that's gone bad

[BlindFreddie]

@Leeghoofd - thanks for the advice, but you didn't answer my question as to whether the CPU has to run in order to get error beep codes (dah-dit-dit). (The UD3R mb manual says this means "monitor or graphics card error" )
AFAIK, the "BIOS" is just a dead program in ROM/flash that needs a CPU, or very tricky hard-wired logic, to bring it to "life". On the above-mentioned vintage 8-bit machines, generating beep-codes would have been achieved by having the CPU toggle on-and-off 1 bit of a parallel port, to which the speaker wire was attached. And you could have read how it was done because you got a copy of the BIOS source code, in glorious 8-bit assembler.
Those machines used the one-and-only main onboard CPU to run the BIOS code, but with today's gear it would be feasible to have a small auxiliary CPU (probably non-x86) run a boot-section of the BIOS ROM code to perform a motherboard check before starting the main CPU. That may sound strange, but it might be easier than building hard-wired logic to generate beep codes. Maybe we should ask Gigabyte or Asus.

Progress report:
QX9650 is in UD3R, working with RAM #2, 333x11, RAM at default settings - 5:8 ratio (533MHz), 5-5-5-15, 2.2V. Currently crunching WCG.
I tested the RAM sticks in it at 333x9:
* RAM #2 will run alone in any slot of UD3R. This is not what I remember from earlier. RAM #2 will also run at 5:6 ratio (400MHz).
* RAM #1 will NOT run in any slot at either 533MHz or 400MHz. Funny, but I marked it before with a sticker that says "OK". Sometimes I get the 22 short beeps with 2-sec gaps signal. Manual says continuous short beeps mean "power error". Bovine excrement!

Q9650 is available for testing in P5Qdlx and P5K-VM, with Intel HSF. Only working PS/2 keyboard connected and working PCI video card installed.
In P5Qdlx, it will not boot with either RAM #2 or RAM #1, with or without keyboard and/or video card. Screen says "No Signal".
With Q9650 installed and no RAM, I get dah-dit-dit beep codes.
** Big experiment *** With no RAM, and no CPU installed, I get silence. This suggests that the CPU does run to produce the beep-codes.

Next moves (?):
* Check P5Qdlx and P5K-VM with known good CPU. Risks: The suspect Q9650 has been in both, and could have damaged them in a way that will damage any good CPU tried in them now. Likelihood?
* Try suspect Q9650 in UD3R. Risk: UD3R could be damaged or further damaged by Q9650. But the Q9650 was in there when the initial failure occurred. Can it do harm or more harm?
* Or test with a cheap, used, expendable CPU. The Asus/GB websites show that these mb support older LGA775 CPUs like P4 Prescotts, Celeries, etc. Hmmm ... eBay ... Celeron D 331 - Buy it now, A$15 ...
Update: Just bought an E2160 for A$45 inc postage. These look like much more fun to play with than a doggy old D 331 And if it survives current testing, it might just crunch a few numbers for WCG until I've had a look at how the i7-860's stack up.
* Obtain 1-2 more RAM sticks and test 2-channel mode in UD3R.