Supported PowerMac 6100 L2 sizes?

[Melkhior]

I would buy a cloned 1mb L2 cache stick

In theory, there's an alternate solution to the many chips implementation mentioned previously, but this would require a much larger budget...

Motorola did produce an integrated 256 KiB L2 cache for the 60x bus, the MPC2605. You can use 1, 2 or 4 of them to produce a 256 KiB, 512 KiB, or 1 MiB cache, and those would require minimal external support. And the chips are available from RocElec. However, those clicking on the link will have noticed the "this is for professional" pricing, at over $50 per unit...

 

[dr.zeissler]

puh! 50 per chip and 4 are required for 1MB ... I am not going to see that happen.

 

[Melkhior]

puh! 50 per chip and 4 are required for 1MB ... I am not going to see that happen.

After I posted that, I found some AliExpress vendor claiming they were selling the 66 MHz version for about $17 each. Between the unreliability of such claims, and the fact it would require being shipped to e.g. JLCPCB somehow for assembly (or be handled by someone able to solder BGA at 'home'), it probably doesn't help much :-/

 

[dust bunny]

So, how many $17 chips would be needed for the 1, 2 and 4MB sticks?

 

[Phipli]

So, how many $17 chips would be needed for the 1, 2 and 4MB sticks?

4, 8 and 16 respectively if all it is is quantity.

 

[dust bunny]

Which would make it $68 worth of chips for 1mb, $136 for 2mb and $272 for 4mb.
Which is a lot better than $200, $400 and $800!

 

[Melkhior]

So, how many $17 chips would be needed for the 1, 2 and 4MB sticks?

I don't think the chips can do more than 1 MiB out-of-the-box as they don't have the necessary inputs. It might be possible to add external logic to support larger sizes, but then timings would become a major issue. 1MiB might be feasible, but I don't know of a reference design, and assembling BGA packages is not easy...

 

[dust bunny]

I have a vague recollection, reading somewhere about a theoretical max L2 cache size of 2mb, but I don't remember where I read it, and it might have been the relate Power2/Power2+ chip.
Equally, my memory might be playing tricks.


 

[Unknown_K]

The old benchmarking software would probably all fit in 2MB of cache making the machine seem much faster than it should be.

I think I have a 512KB or 1MB stick in my 81/110 but it's been ages since I checked. My 6100/7100's are probably stock 256KB.