Does anyone have a way to map a delidded k6-2+ down to the thousandth of an inch and diagram it. I was thinking a pcb that had clearances for all the parts and slid on like a puzzle piece

Mechanical specs should be on the datasheet.

BitWrangler wrote on 2024-04-04, 00:01:

Mechanical specs should be on the datasheet.

https://pdf.datasheetcatalog.com/datasheet/Ad … ices/mXvutz.pdf

Page 353 doesn’t have required measurements or diagram

Maybe you can just scan it and use a caliper to get the heights ?

This is form AMD K6-2+ official datasheet pdf:

As you can see there's no size of what you're looking for, but after some calculations it can be found with some approximate accuracy:

thickness from the ceramic to top of the die in milimeters* ≈ (Emin + Emax)/2 - (Fmin + Fmax)/2 - (thickness of the IHS cap's plate) ≈
≈ (2.91 + 3.63)/2 - (1.3 + 1.52)/2 - (thickness of the IHS cap's plate) ≈ 3.27 - 1.41 - (thickness of the IHS cap's plate) ≈
≈ 1.86 - (thickness of the IHS cap's plate) ≈ ??? [ mm* ]

You have to measure the thickness of the IHS cap's plate by yourself 😀

* - I'm using SI units only, but as you can see in the pdf there are given dimensions in inches

I knew it wouldn't have everything going on under the cap, but thought it needed to be seen that you're not going to be able to do it to the thousandth of an inch, when AMD thought 20 thou was good enough.

Wonder if I could scan the cpu’s image (I don’t have a scanner) or take a pic, then scale it to the proper size, convert it into a silk screen image and use that as a template to lay out the edge cuts. 🤔

Maybe a scanner would be best to keep aspects correct

Edit: I see Crazytiti had the scan idea also.

Caliper should work for height

I measured it and from the ceramic to the top of the die is 0.80mm 😀

I have a scanner, but only an old 350 or 380 already delid k6.

If someone have delided K6-2+ at hand, better off to use caliper to take direct measurements of CPU die.

analog_programmer wrote on 2024-04-05, 05:55:

If someone have delided K6-2+ at hand, better off to use caliper to take direct measurements of CPU die.

I did that, it is 0.80mm

Skalabala wrote on 2024-04-06, 15:40:

I did that, it is 0.80mm

Ok, this is height over the ceramic base, but what about length and width of the die?

analog_programmer wrote on 2024-04-06, 15:53:

Skalabala wrote on 2024-04-06, 15:40:

I did that, it is 0.80mm

Ok, this is height over the ceramic base, but what about length and width of the die?

Ok, will get that today and post tonight 😀

The die is 9.20mm by 8.30mm by 0.80mm with error up to 0.02mm

That’s the easy part, mapping the exact position of the smd components is the tricky bit. My idea was a pcb similar to a punch card with a bunch of holes in it. And it would just slide on like a glove. It could even have solder pads next to the smd parts as a way to attach it and a place in the middle for a heat spreader to solder to.

Probably more work than it is worth but it would be very cool.

We could add extra holes for mounting of extra parts. Just need to make sure none of them lie above any of the gold test points.

Next mod 512kb...

2048kb 1/3 clock L3 on the empty ceramic made out of chips stolen off crappy PIIs ... and tryna get Jan to write BIOSes for 256L2 plus 2048L3 plus 2048L4 ... ... ...

Lol

Cache interposers might be possible.

Still waiting for someone to check what’s going on with the jumpers on the non plus might be 512k hiding there… (doubt it)

So, I decided to try on my K6-2+ 500ACZ and it seems to have worked! I ran the custom range Memtest for a few hundred passes and got zero errors. Also did a 3DMark 99 run without issue.

I haven't tried overclocking or anything yet. And now that I think about it, I never tried it before the mod either. I suppose I could have been a little more thorough with some before and after testing along the way. I was just really excited to give it a try. 😀

Sphere478 wrote on 2024-04-10, 01:57:

Still waiting for someone to check what’s going on with the jumpers on the non plus might be 512k hiding there… (doubt it)

Agreed, it would be very unlikely that any K6 had 512k L2 cache, when AMD never released on that had it. Unless of course it was used as a test vessel for production methods. But since even the Athlon didn't have that much on board cache, not likely.