Of course you can't get a $6k CPU for $200. While Engineering samples are typically much more affordable, I would still expect prime condition 7702 ES chips to sell for
at least $1000.
The chips you linked for sale are that cheap because they have missing and/or damaged capacitor networks (or 'compactors ' as they're incorrectly called in the description). Capacitor networks are just several capacitors in one package, this helps keep the ESL down (important for decoupling CPUs, amongst other things).
So before you even address points a and b, you have your SMD hot air rework station, tweezers, years of practice, solder paste, and one steady af hand good to go, right? You'll need to identify the capacitor as well. I'd bet my right nut that they're probably one of these, though it is hard to get a sense of scale from the photos so I am not sure of they're 0612 or 0508 sized, and you'd need to measure the value and get appropriate replacements. I would guess they're as high as was available, 2.2uF or maybe 3.3uF though they might be smaller, it depends on what impedance they needed. Long story short, you'd need to measure a working network and know how to properly do that. It's not hard, but its yet another thing and another bit of kit you'd need to already own.
AVX IDC Series Capacitor Arrays & Networks | Mouser
But if you're comfortable doing all that, then it
might be workable. It's definitely a huge risk. The biggest thing you'd be doing is helping that poor seller out, since those CPUs are as good as worthless unless some one who doesn't mind taking a chance. What will probably happen is you'll end up with a $200 paper weight for your desk plus some lost hours trying to get it to work. But it could also work. And you're right, it is quite the payoff if it works.
I'm an electrical engineer, and would have no problem replacing capacitor networks like that myself. I do much smaller and closer lead pitches by hand regularly. I have all the equipment and tools on my desk right now. It would be trivial for me to identify a suitable replacement capacitor once I had a chip in hand. I actually might even have suitable replacements on hand right now, I wouldn't even have to order new capacitors and wait for them.
And know all of that, knowing what dealing with that chip looks like from my perspective, even then, I personally wouldn't do it. I don't want to buy that CPU. I think its just a unpleasant way to lose $200 and a few hours of frustrated failure after failure. I mean, take a close look at the photos, there is even a nasty gash near the capacitors on one of them, what if you get that one? Nothing to be done in that case. The green is a micro pcb interposer with a ton of layers and traces, a gash spells certain doom for the CPU ever being repairable except by the packaging part of the fab that originally made it - and by repackaging the dies onto a new interposer.
That said, there is also nothing saying that you couldn't get it working. They could work. Ripped off capacitor networks could be the only thing physically wrong with them.
Moving beyond the physical damage that would need to be repaired, to answer your point A, typically getting a working CPU to post is merely a matter of microcode. Either the BIOS has microcode that matches the stepping of the engineering sample, or it doesn't. Supermicro is usually very good about this, and you can just download one of their H12 bios images and extract the microcode (google it, there are tons of guides and tools on how to do this) and then you can insert it into other aptio V based bios images and bob's your uncle. Assuming the motherboard's chipset etc. supports a second gen EPYC of course.
I get the impression that microcode for that stepping do exist in at least one Supermicro bios, but I am not certain. You'd have to do some research.
As for overclocking, it would again probably depend at least partially on the motherboard and bios features, but worse comes to worse, you should be able to overclock it by writing directly to MSR registers. AFAIK Gen2 EPYCs are unlocked.
I wouldn't be to confident about reaching stock clock speeds, nonetheless higher-than stock clocks. If they could clock engineering samples as fast as retail CPUs, why wouldn't they? One of the main purposes of engineering samples, and why their clocks are typically lower, is because the process and/or silicon is still being tweaked to improve stability and reach the retail clocks at decent yield. If you clocked that thing up to stock speeds, it would probably work, probably post, seem ok. And randomly hard freeze (no BSOD, the computer simply...stops) completely at random due to a stochastic and infrequent instability somewhere. And why would it be any other way? If the chip was good to go at stock clocks as is, why would they make another stepping of the chip at all?
But, as long as you have suitable microcode for a given stepping, usually it will work. No one can say how much they change from stepping to stepping because microcode is encrypted, but the microcode abstracts any such changes out and makes those differences invisible to the hardware and OS.
Personally my recommendation on these is: don't.
But if you do anyway, there is a chance it could work. If you are setup to repair them anyway.
Good luck!
