Best way is design the 3 phase VRM. Much more efficient and can be oscillating at higher frequency and smoother DC for the processor, most especially overclocking success.

The motherboard does not have enough capacitors for CPU, you need to add at least 3 polymer capacitors to the VRM's output. The ceramic capacitors can only do much and mostly for high frequency filtering, bulk capacitors especially polymer capacitors gives both low ripple filtering and absorb current spikes best.

Cheers,

pentiumspeed wrote on 2022-08-22, 23:10:

Best way is design the 3 phase VRM. Much more efficient and can be oscillating at higher frequency and smoother DC for the processor, most especially overclocking success.

The motherboard does not have enough capacitors for CPU, you need to add at least 3 polymer capacitors to the VRM's output. The ceramic capacitors can only do much and mostly for high frequency filtering, bulk capacitors especially polymer capacitors gives both low ripple filtering and absorb current spikes best.

Cheers,

The capacitors used are deceptively large for their size, ceramics have come a long way and the ESR is extremely low compared to polymers. Having the very low ESR means the buck can be run at a higher frequency which lowers the required capacitance to meet a given ripple target. The amount used exceeds the amount recommended by Vishay in order to meet the ripple requirements given by Intel for the VRM itself (including derating, tolerance, etc).

Cool

RussD, What's your background?

Cheers,

Edit: Nice work 😀

I kinda replied to this before fully reading it and realizing you had already made it 🤣. That’s what I get for multi tasking and posting in a rush.

If it helps here is a footprint and project that I worked on a little while ago if it helps you may use anything you like from it

Re: IBM Personal Computer 365 / M/T: 6589-17U / Voltage Regulator Module (VRM)

As far as a multi-phase VRM, it's just overkill for this. Measuring the in-system ripple with the CPU at full load:

The budget by Intel for the full system is +/- 5% or 165mV. This is the full budget though, including regulator set point tolerance. There is a ripple requirement for the VRM and it's +/-1%, or at 3.3V, 66mV. I'll need to test that outside of the system as part of my further testing, but things are looking pretty good.

Question: where did you source your pin header connector?

You don’t happen to have part numbers for this and the pentium 1 variant do you?

I decided to go with a external psu for a few reasons, parts availability, ease of assembly, greater range of parts quality depending on what user was trying to do.

I like the all in one solutions though they are cleaner looking, you definitely made a nice looking card there. It looks oem stock! especially with that tan pcb. Very good job.

pentiumspeed wrote on 2022-08-22, 23:52:

Cool

RussD, What's your background?

Cheers,

My day job is mostly embedded software, but I end up doing a lot of hardware stuff. My undergrad degree is a combination of cs and ee. My hobby projects usually involve a PCB somewhere.

Sphere478 wrote on 2022-08-23, 00:07:

Question: where did you source your pin header connector? […]
Show full quote

Question: where did you source your pin header connector?

You don’t happen to have part numbers for this and the pentium 1 variant do you?

I decided to go with a external psu for a few reasons, parts availability, ease of assembly, greater range of parts quality depending on what user was trying to do.

I like the all in one solutions though they are cleaner looking, you definitely made a nice looking card there. It looks oem stock! especially with that tan pcb. Very good job.

The part numbers have all changed from the original versions as they are all the new lead free part numbers. I sourced it from Arrow. There were actually quite a few components I ended up sourcing from there as their pricing has become much more competitive as late:

https://www.arrow.com/en/products/5-532956-7/te-connectivity

Unfortunately, as of now, all the stock is gone from digikey, mouser, and arrow. I have no idea who is buying these up and for what purpose. Digikey still has the one with 15uin of contact finish rather than the thicker 30uin version:

https://www.digikey.com/en/products/detail/te … 32955-7/1126722

Here's the 30 pin version, also only available in the thinner contact finish, don't know how much that matters:

https://www.digikey.com/en/products/detail/te … 32955-5/1126719

I resisted the urge to just use a normal right angle header in this case, as I think the TE connector has much lower impedance. A straddle connector might do the trick here, but it won't have the AMPMODU Mod II "ears" and would need to be shaped a bit differently in order to still fit. Amphenol ICC also makes what looks to be a drop-in replacement under their Dubox series that might be worth looking at.

The tan PCB is the oshpark after dark service. It's the first time I tried it, and yes, I didn't want a big purple PCB in my retro PC. I really like the black color of the fiberglass and if I end up doing a second run I'll delete the copper from the bottom ears as it's a pretty neat look on the one ear without copper.