There are only 3 NC pins on SMD package so no chance for hidden SMP. Intel was not so generous to give hidden features for free like AMD...

Okay. Thanks 😀

Are you going to work on the bga interposer?

I may take it on if you aren’t

No, I don't have a BGA chip.

Right on 😀

I was noticed that JLCPCB now do 6-layers board up to 5x5cm for fancy price of 2L boards. I'm thinking about adding 2 solid VCC and VCC2 planes into stackup for better power impedance. It will be optional, still possible to manufacture as 4L. Does anybody know how long the JLC promo will be offered? Is it just xmass gift? I still need to finish some other PCBs to send the whole batch...
Also I would appreciate some discount for PCBs with one size over 10cm and other under 5cm (72p SIMM).

EDIT: the price increase of PCB over 10cm is nothing dramatic, $4,0->5,4, it's OK...

I think it’s going until January So not much longer but I do vote that you do a 6 L design

I just sent a batch of PCBs yesteray, manufacturing in progress. I hope there's no bugs. I checked pinout tables again.
The 2$ special prize is valid only for the 1st uploaded PCB of the whole batch. I looked at JLC web today and the discount is still present there, I feared about end of january deadline but it seems it will last longer, who knows...

Now I need to find some source of socket pins, individual or usable header that fits side by side. Also I have one precision socket S7...

What is these? First two and what the reason for recreating 72pin DIMM?

I'm interested to know.

Cheers,

First 2 are some of my experiments not related to retro computing.

The 72pin SIMM is parity (36-bit) version that's much more rare than common non-parity SIMMs (at least in our country). I need some more for my Alpha64. Mentioned in other thread. I plan to canibalize FPD DRAM chips from non parity SIMM to this new one (3 modules with 3*8 chips -> 2 modules with 2*12 chips).

RayeR wrote on 2023-02-03, 17:42:

I need some more for my Alpha64.

What kind of AXP do you own?

I just finished the major part of soldering job. It took about 4 hours of soldering and inspection under binocular magnifier. Maybe comparable to some neurochirurgical operation :p I had some short on vcc3 but after some cleaning it's gone. I hope no more shorts. Now I need to do next important step. Before start of soldering I applied a solder paste between CPU core and central pad. The paste must be melted but not the soldered pins. It's tricky because the CPU foil base act like a spring with force that lifts up the pins from interposer PCB. So I plan to fix it by epoxy glue that I hope it will withstand the temperature for a short time. But it will disable furure soldering rework if I would find another short.
Then I will need some pins. I didn't find better source than use some old S370 CPU and desolder pins from it. Std 2.54 precise headers don't fit in zig-zag matrix like it would fit on 486 to regular matrix.

https://www.digikey.ca/en/products/filter/ter … -connectors/323 this sort of thing? Copy/paste part numbers or descriptions to get something more local maybe.

Yes, Dgk has something usable (1mm body dia.) but insane prices, for such pin count I could buy whole cpu instead. Need some cheap aliexpress for $1 a 100pcs lot and free shipping 😜

Amazing. The fine soldering made me really doubt if it would get this far, but you are determined!!

I would like to procure a tested 300 mhz unit from you once you get the bugs worked out 😀

For soldering the pins, place them in a socket to hold them straight while you hot air them or whatever your method will be.

Epoxy glue applied, no step back...
I did some thermal design calc. One via should have Rth = 222K/W, then 200 vias in parallel gets 1,1K/W total. CPU TDP @266MHz is 7,6W so temp. gradient from CPU core side to heatsink side should be 8K, not bad...

I have 4 lame Celeron 400 PPGA and some lame Durons 650-900 that could serve as pin donors. I have some desoldered precise socket 5 that I can use as a holder for pins alignment. I'm not sure how to solder them on PCB, I can use paste or just 0.5mm thin tin solder wire put through via and soldered from opposite side than pin pad. I just need properly align the pins and nivelate them at same height...

EDIT the pins from PPGA are really hard to pull out even at high temp (hotair 500C), they are through-hole, tightly fitted in vias. I have greater vias on my interposer so I can insert themeasily. Opposite to FC-PGA that use SMD pins that are quite easy to solder off.

RayeR wrote on 2023-10-24, 03:51:

I'm not sure how to solder them on PCB, I can use paste or just 0.5mm thin tin solder wire put through via and soldered from opposite side than pin pad. I just need properly align the pins and nivelate them at same height...

Take apart a socket7 socket. You then have something you can space them properly with with the top part of the disassembled socket. I have a scrap motherboard I can try to rip the socket off and send to you.

I also have socket 5/7 desoldered from some dead MB but the plastic could melt during pins soldering so I'd rather use that precise socket.

BTW there's a bit problem with voltage of this mobile MMX CPU.
DS says abs. max, Core/VIO : 2,8/3,2V and operating values Core/VIO : 1,8/2,5V. Some newer motherboards has flexible Vcore regulator that can go down to 1,8V but I didn't see any MB that can lower VIO to 2,5V, probably because all other logic, chipset and memory runs at least from 3,3V. So that golden tigers must run heavily overvolted on common MBs...

I don’t believe the tillamook has any issue running vio at v3.3v every time I have run one it has been at 3.3vio on ppga

Obviously 2.5v would be nice but then you are undervolting your ram, cards and chipset

You may be able to incorporate a regulator into your interposer, but it doesn’t seem to be necessary

Exceeding the absolute max. rating is not good idea generally, esp. for long run. But for some retroplaying it probaby would not degrade too fast. A tiny SMPS would be possible, modern components like used in smartphones can easily fit on top side under heatsing not exceesing 0.8-1mm height. Also for perfect match there should be some voltage level translators on each IO pin (probably it's not 3.3V tolerant) but when VIO and logic high difference would be below 0.5V it would be ok. I didn't pay enough attention on this in datasheet, now it's too late to add it in design. But it's more just for fun project than daily workhorse PC so I don't care too much. Just to notice others there exist such issue...

Then also the power dissipation would be higher than I read from DS, maybe close to non-mobile MMX that was about 17w...

RayeR wrote on 2023-10-24, 12:42:

Exceeding the absolute max. rating is not good idea generally, esp. for long run. But for some retroplaying it probaby would not degrade too fast. A tiny SMPS would be possible, modern components like used in smartphones can easily fit on top side under heatsing not exceesing 0.8-1mm height. Also for perfect match there should be some voltage level translators on each IO pin (probably it's not 3.3V tolerant) but when VIO and logic high difference would be below 0.5V it would be ok. I didn't pay enough attention on this in datasheet, now it's too late to add it in design. But it's more just for fun project than daily workhorse PC so I don't care too much. Just to notice others there exist such issue...

Then also the power dissipation would be higher than I read from DS, maybe close to non-mobile MMX that was about 17w...

Basically everyone playing with tillamooks is doing 3.3v i/o and we aren’t seeing any failures

The p55 had 3.3vio, (on a larger process granted) but it came with chips rated down to 2.45 core and 3.3v io

Two possibilities: the tillamook is 3.3vio tolerant and was rated to 2.5vio for motherboard/thermal envelope reasons

Or two it is being over volted but we aren’t seeing any failures

In either case, a problem doesn’t seem to exist.

I would like to know which one, but I don’t think we would ever know for sure unless we got a intel engineer in here.

Some motherboards can do 3.5 vio, someone could run a tillamook on that for a while to see how close we are to failure mode. 🤷‍♂️ (for science)

Btw, my s5/7/ss7 voltage interceptor interposer is designed to allow vio to be divorced from motherboard vio if you set it up for that. If you wanted to do experiments