Bloax
Slightly less n00b
- Joined
- Jul 17, 2019
- Messages
- 113
- Points
- 28
So here's a fun story for you:
Back in the start of December, I moved someplace, then I moved back.
Seeing that Noctua heatsinks have a nightmarish mounting mechanism, I didn't undo my heatsink - and my 3600 CPU got wrecked the second time. wHoopS
So, I replace my 3600 with a 3700x - after a two hour session of torturous hollow-screw-screwing with my favorite mounting mechanism, I finally get a mount after undoing the central mounting-mechanism screw and just manually forcing the hollow-screws to screw into the screwy screws on the mounting plates at the same time with both of my hands.
Nice, nice. But what's this; my system isn't POSTing?
First thing I check is whether the memory didn't get loose during transport (it can happen!), taking it out, and plopping it back in.
Nothing.
OK, so I remove the second stick to see if maybe one of them got wrecked during transport somehow.
It POSTs!
Alright, nice. It boots.
Now, how about the second memory stick; is it still functional, or is it busted?
I plop it into the same slot, and..
It POSTs.
So then, both sticks work. Time to test slots.
A1 POSTs, B1 POSTs, A2 doesn't, B2 doesn't
Well, well - looks like it wasn't just my CPU that got wrecked during transport! Half the **** slots don't work!
NUH; not so fast. After a bunch of fiddling, I got somewhere in regards to getting my old memory speeds and timings back:
but see - I found one thing peculiar, but didn't think of it too much:
On-Die Termination was set to High Impedance, and RTT/CADBUS settings were kinda weird.
What's more, is that I thought initially these were the board's Auto settings "knowing" how to get 2-DIMM Single-Channel (yes, populating A1B1 slots results in that) working properly on itself.
But as it turned out when I got another board to replace my "half-working" b450 Tomahawk (it's an AssCock x570 Phantom ITX, which has its own story), those were not actually the board's auto settings.
Indeed, what I also discovered is that populating both slots in this 2-DIMM board doesn't POST - -
or rather, it posts about 50% of the time. Bone-stock 2133 CL15 JADEC. Anything better bootloops, the other half of the time it doesn't recognize the second stick as existing.
At first I thought that mayhaps this was an artifact of thermal paste on the CPU pins; the previous nightmare mount was hardly a painless, clean affair - with some thermal paste gooping out onto the socket.
But having cleared off both the socket and CPU thoroughly with a namptha-based electronic degreaser/cleaner and even cleaning off any remnants with cotton-tips, the problems persisted.
So, to recap:
most importantly Scythe coolers are a pleasure to work with, can recommend despite not mentioning that during all of this
Apparently, you can have faulty IMCs on your CPU - and besides the obvious "dual channel no work wtf" symptom, you might be able to tell that it is definitely the CPU if it
only respects its own, strange ProcODT, RTT[..] and CADBus settings. It also tends to force 1.1v VDDP
Cheers, and have a good one.
bonus screencap of 1.58v single-stick settings
fun fact: 1867 FCLK has a serious performance penalty on a b450 tomahawk while running very fast memory settings, but does not on this board.
hehe ecks d?
Back in the start of December, I moved someplace, then I moved back.
Seeing that Noctua heatsinks have a nightmarish mounting mechanism, I didn't undo my heatsink - and my 3600 CPU got wrecked the second time. wHoopS
So, I replace my 3600 with a 3700x - after a two hour session of torturous hollow-screw-screwing with my favorite mounting mechanism, I finally get a mount after undoing the central mounting-mechanism screw and just manually forcing the hollow-screws to screw into the screwy screws on the mounting plates at the same time with both of my hands.
Nice, nice. But what's this; my system isn't POSTing?
First thing I check is whether the memory didn't get loose during transport (it can happen!), taking it out, and plopping it back in.
Nothing.
OK, so I remove the second stick to see if maybe one of them got wrecked during transport somehow.
It POSTs!
Alright, nice. It boots.
Now, how about the second memory stick; is it still functional, or is it busted?
I plop it into the same slot, and..
It POSTs.
So then, both sticks work. Time to test slots.
A1 POSTs, B1 POSTs, A2 doesn't, B2 doesn't
Well, well - looks like it wasn't just my CPU that got wrecked during transport! Half the **** slots don't work!
NUH; not so fast. After a bunch of fiddling, I got somewhere in regards to getting my old memory speeds and timings back:
but see - I found one thing peculiar, but didn't think of it too much:
On-Die Termination was set to High Impedance, and RTT/CADBUS settings were kinda weird.
What's more, is that I thought initially these were the board's Auto settings "knowing" how to get 2-DIMM Single-Channel (yes, populating A1B1 slots results in that) working properly on itself.
But as it turned out when I got another board to replace my "half-working" b450 Tomahawk (it's an AssCock x570 Phantom ITX, which has its own story), those were not actually the board's auto settings.
Indeed, what I also discovered is that populating both slots in this 2-DIMM board doesn't POST - -
or rather, it posts about 50% of the time. Bone-stock 2133 CL15 JADEC. Anything better bootloops, the other half of the time it doesn't recognize the second stick as existing.
At first I thought that mayhaps this was an artifact of thermal paste on the CPU pins; the previous nightmare mount was hardly a painless, clean affair - with some thermal paste gooping out onto the socket.
But having cleared off both the socket and CPU thoroughly with a namptha-based electronic degreaser/cleaner and even cleaning off any remnants with cotton-tips, the problems persisted.
So, to recap:
most importantly Scythe coolers are a pleasure to work with, can recommend despite not mentioning that during all of this
Apparently, you can have faulty IMCs on your CPU - and besides the obvious "dual channel no work wtf" symptom, you might be able to tell that it is definitely the CPU if it
only respects its own, strange ProcODT, RTT[..] and CADBus settings. It also tends to force 1.1v VDDP
Cheers, and have a good one.
bonus screencap of 1.58v single-stick settings
fun fact: 1867 FCLK has a serious performance penalty on a b450 tomahawk while running very fast memory settings, but does not on this board.
hehe ecks d?
Last edited: