A looked in DS again, IOs are not 3,3V tolerant except CLK input where it is explicitly mentioned. For other inputs the maximum is VIO+0,3V so if someone use separate VIO regulator at 2,6V then maximum on input would be +0.3 = 3.0V, still over limit of 3.3V. Reasonable compromise would be set VIO at 3.0V then input can take 3.3V and it would be under abs. maximum. That nobody sees his overvolted CPU failing is good but doesn't mean there's no long term degradation of internal structures. Of course no problem for occasional hobbyst.

So now I managed to solder a copper IHS (I used 2mm thick Cu plate cut from a sheet) on top side and hope it also melted the paste under CPU core, without pin foil popping off the PCB (fused by epoxy). Then only pin work remains... I added some macros. I inspected the PPGA370 pin and I can see that the part fitted inside PPGA370 interposer has groovy surface, not smooth. So it seems that pins was pressed with a high force inside interposer vias and then soldered just for secure/better conductivity. As I have larger holes in my interposer it fits easily.

Oh wow, you are making a very nice product there! I like the adhesive strips

What did that side look like before the ihs?

On the top side I left unmasked copper area witch cca 200 tiny vias for thermal conduction. I apply solder paste, put IHS on and solder by hot air...

Next time, get some low temp metal collected from under the IHS of a large, cheap xeon CPU for the die solder bond.

Cheers,

Ah, I also have some low-temp solder paste (indium based?) but I don't belive it would be stable as old good Sn37Pb63... And I even didn't think about it as I don't use it for normal soldering...

Can you liquid metal or solder the core to the pcb?

Probably pointless but, would be cool.

Hey, question, what if any of the tillamook mods did you incorporate into this, and where did you map bf2 to?

The best way to route the two bf2 pins would be to just link them together (intel bf2 and amd bf2) and give them a 1000 ohm pull up resistor

As for the other pins, may want to see if the same tillamook mods apply to your unit.

Re: Tillamook 266MHz and working L2 cache?

Damn, I'm wrecked. After soldering the PGA pins (it went quite well using precise socket as pins holder) I did the final test with multimeter and found shorts on both VCC2 and VCC3 to GND and also A13, A14 to GND. I cannot see anything wrong on the surface. Maybe it happen that during soldering of IHS some excesive metal leaked through vias to bottom side in the gap around the core and touched the golden traces on foil inner side (I should kapton tape it but didn't). I would need do x-ray image to prove it but as it's glued so I cannot fix it anyway...

Sphere478 wrote on 2023-10-25, 02:19:

Can you liquid metal or solder the core to the pcb?

pentiumspeed wrote on 2023-10-24, 17:56:

Next time, get some low temp metal collected from under the IHS of a large, cheap xeon CPU for the die solder bond.

to immediately corrode copper? https://www.gamersnexus.net/guides/3362-how-l … -corrosion-test

bummer about shorts 🙁

Maybe try burning them out by applying power across, in case they are just thin whiskers or very thin points of contact.

Not that liquid metal. The low temp solder from cheap large xeon does not corrode! The gap in the silicone glue is air vent.

Cheers,

I pleased a friend who took a few x-ray images, not much hi-res but we found that thick IHS makes core and foil wires invisible.
So I desoldered IHS and let him x-ray it again. I also tried to suck off as much melted solder as possible by litz wire - it completly cleaned the vias but all shorts remains. I guess solder blob somewhere on golden foil wires....

Damn. Do you have an idea of how to fix? Or start all over?

Good idea about kapton for next time!

The only chance is carefully grinding off the PCB area around CPU core by Dremel from top side, then suck off the excessive solder from gold foil wires. I cannot take the CPU off, the epoxy glue proved very good thermal stability 😜 I hope that x-ray image could help me localize the possible short so only small hole would be needed instead of grinding the entire center area. Yes, better insulation preparation with kapton tape (I taped only PCB surface with traces that could get in touch with foil) or epoxy varnish on CPU bottom side could save this pain, now I-idiot know for future, but not going to reproduce it anyway. 4 PBCs left for someone brave...

So you won’t try again? 🙁

I have only one such CPU and I cannot remove it from current PCB without destoying glued pins.

I'd be doing something like belt sanding the PCB off 🤣

I had a thought, when soldering the ihs, perhaps it should be done upside down with a clamp so all solder blobs flow down, if trying to solder the core I mean, if using paste, just do it before placing the core

As for your current cpu, I think perhaps you will have to make it a display piece or destructively remove the current cpu in order to place another one

Have you ordered more?

More of what? The interposer PCBs was made in count of 5pcs because it's a minimum Qty that JLCPCB do. The CPU I have was only one and I desolder it from a wreck of old Compaq Armada laptop (my 1st post). I was quite surprised by this as I never seen such package before and expected a socketed cpu but then I said it would be a challenge to make it working, even when PGA tillamooks are now hard to buy... Yes, soldering IHS upside down would.probably prevent this issue, maybe someone else take a lesson from this...

I'm waiting for next x-ray images so I hope it will tell me more if/where/how to fix it.

BTW it's a pitty nobody pointed me to old L2 cache issues and solution that I could add to this interposer. At least there are 2 things to be added: VCC2DET for proper dual.VRM detection and ADSC copy signals for cache...

Another cpu

They are all over EBay.