From: nospam@needed.invalid
On Fri, 7/25/2025 7:55 AM, pinnerite wrote:
> On Fri, 25 Jul 2025 12:51:41 +0100
> pinnerite wrote:
>
>> On Tue, 22 Jul 2025 11:36:46 -0700
>> Mike Easter wrote:
>>
>>> pinnerite wrote:
>>>> A few weeks ago a young neighbour of mine asked i he could borrow a torx
sreedriver
>>>> if I had one (some). At the time I had never heard of them.
>>>
>>> Your GMT tz looks like .eu or .en; but around here we have huge
>>> warehouse stores called Home Depot which has a wealth of so many things,
>>> incl tools. The section which has screwdrivers and related tools is
>>> VERY extensive and a lot of fun to shop. I was there not too long ago
>>> because I wanted to improve my collection of 'little bitty' screwdrivers
>>> between what I had for eyeglass repair and an ancient sewing machine
>>> screwdriver passed down from my mother.
>>>
>>> It is a lot of fun to shop in that section; I discover all kinds of
>>> things I never 'imagined'.
>>>
>>> --
>>> Mike Easter
>>
>> None of the stores within a five mile radius stock an "8".
>> One is coming from China, due 10 August!
>>
>> Meanwhile a replacement Corsair 850watt has been delivered and tested.
>>
>> On the assumption that the failures were due to short circuits, I
>> adhered fibre washers to a new Gigabyte B650 Aorus Elite AX motherboard
>> (where do they dream up these names), mounted it with one of the Ryzen
>> 7 CPUs installed and powered up.
>>
>> The fans started up but a warning LED (the CPU one) illuminated.
>> I tried a second Ryzen 7, same result.
>>
>> How can I find an engineer with the equipment to test both CPUs in NW
>> London, UK?
>>
> I forgot to add the DDR RAM was also installed.
>
Ask a computer store "if they do builds".
Any computer store which sells computer-case, PSU, mobo, CPU, DRAM,
will have a technologist who assembles these for customers. For an
extra 100, they install an OS and drivers.
At Best Buy, this is the Geek Squad, and you can bring
a PC in for test at the desk there. Always back up the hard
drive, when presenting valuable materials to these sorts of people.
There have been many many cases of the C: drive "getting erased".
My computer store has room for two guys to build PCs, and the staff count
in the build area varies with time of year and state of economy.
When the technologist tests, they test the same way I do, "swapping" and
"fault isolate to nearest three components". They don't have bed-of-nails
structural test capability or JTAG scan chains or a 2.5D Xray machine for
checking BGA balls. They might have a multimeter :-)
To be a good tester, you have to know the bringup sequence of the equipment.
What parts matter. Sure, all sorts of unlikely faults can develop on a board,
and a bulk tester like the technologist can only do so much.
So lets see if we can kinda guess at AM5 bringup.
1) Board starts in RESET. A power bug watches the rails, noticed the rails
out of spec at startup, and asserts RESET. The pulse width of the raw
(starting) RESET signal might be half a second. (Anything more than 100mSec
is usually enough for PLLs to lock in on synths.)
2) User must take finger off RESET button. Defective RESET button keeps
the board reset determination in RESET. The computer can be operated,
without
the RESET twisted pair, and the RESEt switch, connected to the FP header.
3) The PSU must assert its Power_Good signal. The onboard switching
converters must assert their private Power_Good. These are all gated
together, to make a "Global_Power_Good". When Global power is good,
that is a reason to leave RESET.
4) Once RESET is deasserted, the clock gen is running. There is 33MHz for the
BIOS flash chip on the LPC bus.
5) We don't start executing right away. First, the PCH (Southbridge) copies
the AGESA code from the BIOS chip and shoots it via a serial link, to the
CPU. The signature of the AGESA code must be good, for the CPU to accept it.
Now, the CPU is ready.
6) Since RESET is deasserted, the CPU sends out its first instruction fetch
at a well-known address. This address happens to be the boot block of the
32MB BIOS flash chip. The boot block contains code which allows
"checksumming" the main BIOS code block. The main BIOS code block is
not executed, unless the checksum passes. On some special Tyan/Intel
or Tyan/AMD server boards, the CPU won't start unless the main BIOS
code signature also passes (says it's Tyan or whatever).
7) The boot block is executing. There is no graphics output capability in
the boot block.
8) Code can execute without DRAM being present. For example, once a few
instructions
of boot block code have run, the boot block code can clear the CPU LED.
(the CPU LED having been set via RESET).
9) Boot block hands off to the larger code block. DRAM commissioning begins.
Special CAS cycle is sent to the DRAM, telling it of its contractual
obligation to the CPU memory controller. Both devices must know and agree
with this parameter, to talk to one another (CAS is in the SPD table).
...
So what do we have so far ?
1) ATX12V connected ? MAKE SURE its in the correct hole.
There is a defacto correct connector, but the manual may not draw attention
to this.
2) Is BIOS content sane ?
3) Is AGESA version modern enough for Ryzen 7 ?
4) Is board Power_Good there ?
5) Is RESET deasserting (and AGESA transferring, and first instruction
fetching) ?
6) Soon after this, we should be programming the DRAM and fiddling DRAM LED.
Assumes CPU is properly seated, fans are running.
+5VSB did not overload early and cause fans to stop spinning.
Since your power is staying up, your fans are running now,
we're no longer asking you to "check for fan twitch indicating overload".
Paul
--- SoupGate-Win32 v1.05
* Origin: you cannot sedate... all the things you hate (1:229/2)
|
