I don't have any specs on them, sorry.
However if you have a way of measuring capacitors you can try and unsolder one of the others on this CPU and use it as a reference. They're most likely all the same value.

The picture looks more like somebody unsoldered the two rather than them being ripped off.
If that's not from you cleaning up the left over bits of broken caps I would assume the CPU was broken and has already been used as a parts donor so there's a chance it won't work even after soldering on new capacitors.
It also seems one solder pad is missing.🙁

Yeah, the missing pad could be a thing, If you're lucky, all the caps are on the same connections and you can connect across, although the extra distance and inductance probably mean any capacitor there is effectively disconnected. And it's possible the CPU won't work even if the pad was still there.

Might be worth a go anyway. They look to be reverse geometry MLCC and looking on Mouser, they list them with (reversed) after the case code. Testing one of the others to get a value is a good idea, but if you just want to guess then 0.1uF might be a good place to start (voltage shouldn't be an issue). You'll also need to measure the dimensions to get the case size (e.g. 2mm long and 1.2mm wide will be 0805). Mind the height for any replacements.

I took the photo after I did some cleanup and desoldered what was left of the caps. The one with ripped off pad - I figured that all caps on that side of the IHS are connected the same, with the side closer to IHS being all connected to ground. So my plan is to bridge from one to the other....

Of course they are all interconnected (and sit between supply voltage island and ground). These capacitors don't form any specific circuit. They are decoupling the input voltage to minimise noise. It is important they are as close to hi-freq parts inside the CPU as possible to provide a low impedance power supply path.

Bridging one to another won't achieve the purpose these caps are there for.

But might still work (even without any caps) if lucky. Just the CPU might be less stable at times. Particularly when driving the bus.

mpe wrote on 2021-05-21, 13:23:

Of course they are all interconnected (and between supply voltage and ground). These capacitor don't form any specific circuit. They are decoupling the input voltage to minimise noise. It is important they are as close to the hi-freq part inside the CPU as possible to provide low impedance path.

Bridging one to another won't achieve the purpose these caps are there for.

But might still work (even without any caps) if lucky. Just the CPU might be less stable at times. Particularly when driving the bus.

I didn't know whether there was a split voltage supply within the chip. If it had a couple of supply voltages (say 2.8V and 3.3V) then the capacitors could be decoupling those supplies to ground, and possibly between them. So I thought it would be a good idea to check that before risking connecting two separate supplies together. I agree that jumping across will make the capacitor less effective. It might even cause problems if the designers wanted to inductively separate different areas at the same voltage and directly linking across the capacitors might break that. And yes, they should be as close as possible to whatever is switching, where in this case (what with the pad missing) 'as close as possible' might be via the pad of another capacitor. As I said, likely to be sufficiently far away that the inductance in the wire might effectively make it appear to be missing anyway.

OTOH, I don't think it'd do any harm to replace both the missing caps as best as possible. I've seen designs with small decoupling caps placed a few cm away from the relevant pins, and they still work (might still work without the cap there at all...).

Had a bit more of a skim through Mouser, something like this might do, if the case size is 0508:
https://www.mouser.co.uk/ProductDetail/Murata … 2EdiUX8AQ%3D%3D

Giving this some more thought, would it be possible to just measure any of those caps in circuit and devide the resulting value by 6 to get a close enough approximation of the individual value? I mean they're all in parallel, right?

Maybe not, they could be part of different circuits. You can trace them with a multimeter to find out.

Doornkaat wrote on 2021-05-21, 14:37:

Giving this some more thought, would it be possible to just measure any of those caps in circuit and devide the resulting value by 6 to get a close enough approximation of the individual value? I mean they're all in parallel, right?

Not sure. uP datasheets will often specify a few different values (e.g. both 1uF and 0.1uF). There's something about the combination of the trace+pad inductance and component capacitance that gives a particularly low impedance at a particular frequency. And for decoupling you want that low impedance frequency range to be as wide as possible, hence having a range of capacitance values. Pretty sure smaller value caps work better for decoupling higher frequencies. I think it's part of the difference between bulk power supply capacitance and HF noise decoupling.

At this point I'd probably go for just giving it a quick try as is to see if anything happens, hopefully at least get to a beep code. If it works, great. Measure voltages on all the caps to find out if they are all on the same power supply or not, and see if there's anything on the bits of copper exposed by the missing pad. If those fragments are still connected, then try soldering a bit of metal tape on to them (tricky) and using that as a pad, then attaching the two missing capacitors. Might be easier to solder the tap to the capacitor first. If the fragments can't be used then tape over them to insulate them (don't want partial, noisy contacts), then fit the two missing capacitors and then use a bit of metal tape (I'm sure I remember from somewhere that for a given CSA, flat conductors have a lower inductance than round) to connect from the capacitor with the missing pad to one nearby.

It didn't even occur to me that the CPU in this state could work at all, but after reading your ideas I tried it.....well it has been running Prime95 for a couple of hours now, seems to be perfectly fine...