VOGONS


First post, by AlessandroB

User metadata
Rank Oldbie
Rank
Oldbie

Mine have 256KB of cache non expandable in any way. The rules is different from 486 system, and for socket 7 and K6III ecc ecc... what is the "Rule"? 128 for.... 256 for... 512 for...
Tnks

Reply 3 of 39, by rasz_pl

User metadata
Rank l33t
Rank
l33t

Does it really matter? we are talking Pentium 60-66MHz here. Best case scenario 430LX. 16MB ram is right on the money. Nothing requiring more would run smoothly anyway.

Open Source AT&T Globalyst/NCR/FIC 486-GAC-2 proprietary Cache Module reproduction

Reply 4 of 39, by AlessandroB

User metadata
Rank Oldbie
Rank
Oldbie
Doornkaat wrote on 2022-11-21, 22:13:

You'd help us by letting us know your motherboard model or at least what chipset it uses.

This one:

Attachments

Reply 5 of 39, by Sphere478

User metadata
Rank l33t++
Rank
l33t++

http://dosdays.co.uk/topics/cache.php

256k is good for 32mb in write back
256k is good for 64mb in write through
Motherboard and chipset may over ride this though

You need a late model 486 or a pod/or above for write back if I recall. (Being socket 4, I believe all cpus support it)

Write back is faster

I think I got that right… little rusty.

*Technically* socket 4 is adaptable to socket 5 and socket 5 is adaptable to socket 7 and socket 7 is adaptable to k6-3+ And that can support lots of memory. More than the motherboard can for sure.

But good luck finding those adapters

So, *technically* your max memory is whatever the max the mobo supports. To know that, you might have to start swapping sticks into it.

But to get best performance and being realistic, stick to max cachable. And forget about exotic impossible to find adapters 🤣

If you want k6-3+, get at least a socket 5 as a starting point. Minimum.

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)

Reply 6 of 39, by Horun

User metadata
Rank l33t++
Rank
l33t++
AlessandroB wrote on 2022-11-21, 22:57:

This one:

256k cache is fine for your board. With a 66Mhz CPU it should perform better than a good DX4-133 and board with 256k cache.
Only down side is the VLSI chipset. Saw some reviews and the Intel and SIS based soc 4 boards did fairly well.
Do you have the case and riser for it ?

Hate posting a reply and then have to edit it because it made no sense 😁 First computer was an IBM 3270 workstation with CGA monitor. Stuff: https://archive.org/details/@horun

Reply 7 of 39, by AlessandroB

User metadata
Rank Oldbie
Rank
Oldbie
Horun wrote on 2022-11-22, 05:04:
256k cache is fine for your board. With a 66Mhz CPU it should perform better than a good DX4-133 and board with 256k cache. Only […]
Show full quote
AlessandroB wrote on 2022-11-21, 22:57:

This one:

256k cache is fine for your board. With a 66Mhz CPU it should perform better than a good DX4-133 and board with 256k cache.
Only down side is the VLSI chipset. Saw some reviews and the Intel and SIS based soc 4 boards did fairly well.
Do you have the case and riser for it ?

Yes i have the genuine IBM PC330 P60, a really rare computer, i wanted it for years! So, having 256k on a P60 (there is no jumper for 66mhz bus) give the possibilities to install ONLY 32MB????? On mainboard say 128, too much difference

Reply 9 of 39, by AlessandroB

User metadata
Rank Oldbie
Rank
Oldbie
Doornkaat wrote on 2022-11-22, 09:29:
AlessandroB wrote on 2022-11-21, 22:57:

This one:

What's that PCB on the CPU?

Who know????

Attachments

Reply 10 of 39, by PD2JK

User metadata
Rank Oldbie
Rank
Oldbie

This looks rather rare. You might want to be more careful with the pins. One broke off, how did you manage that with a (so it seems) ZIF socket?

Last edited by PD2JK on 2022-11-22, 10:40. Edited 1 time in total.

i386 16 ⇒ i486 DX4 100 ⇒ Pentium MMX 200 ⇒ Athlon Orion 700 | TB 1000 ⇒ AthlonXP 1700+ ⇒ Opteron 165 ⇒ Dual Opteron 856

Reply 11 of 39, by AlessandroB

User metadata
Rank Oldbie
Rank
Oldbie

So, following this:

And knowing that from the IBM manual:

My P60 accept only 64MB cached? No issue with chipset? no possibilities to going to 128? and from 60ns to 70ns for the 72pinsimm, whats the real difference in this system?

Attachments

Reply 12 of 39, by AlessandroB

User metadata
Rank Oldbie
Rank
Oldbie
PD2JK wrote on 2022-11-22, 10:38:

This looks rather rare. You might want to be more careful with the pins. One broke off.

Yes, the photo comes from another unfortunate computer I had, the pentium60 subject of the post is perfect, the CPU has never been removed since it was installed by IBM.

Reply 13 of 39, by PD2JK

User metadata
Rank Oldbie
Rank
Oldbie
AlessandroB wrote on 2022-11-22, 10:40:

Yes, the photo comes from another unfortunate computer I had, the pentium60 subject of the post is perfect, the CPU has never been removed since it was installed by IBM.

Then it makes sense. Sorry for pointing fingers.

i386 16 ⇒ i486 DX4 100 ⇒ Pentium MMX 200 ⇒ Athlon Orion 700 | TB 1000 ⇒ AthlonXP 1700+ ⇒ Opteron 165 ⇒ Dual Opteron 856

Reply 14 of 39, by AlessandroB

User metadata
Rank Oldbie
Rank
Oldbie
PD2JK wrote on 2022-11-22, 10:41:
AlessandroB wrote on 2022-11-22, 10:40:

Yes, the photo comes from another unfortunate computer I had, the pentium60 subject of the post is perfect, the CPU has never been removed since it was installed by IBM.

Then it makes sense. Sorry for pointing fingers.

No problem, you did well instead! knowing the super prepared and competent people who frequent this forum I was counting the seconds to see how long it took you to point it out to me

Reply 15 of 39, by dionb

User metadata
Rank l33t++
Rank
l33t++
AlessandroB wrote on 2022-11-22, 10:38:

So, following this:

And knowing that from the IBM manual:

My P60 accept only 64MB cached? No issue with chipset? no possibilities to going to 128? and from 60ns to 70ns for the 72pinsimm, whats the real difference in this system?

128MB cached was a challenge in 1997 (here's looking at you, i430TX), this chipset is four years older, from a time where 64MB would have cost USD 4k... that said, i430LX could cache all its 192MB RAM.

This board has a very obscure VLSI 82C591 chipset. No one has been able to find datasheets for it, this is the best info we have:
https://theretroweb.com/chipsets/578

Now, based on that, you might even be able to install 512MB of RAM on here, if you can find 128MB 72p FP SIMMs.

As for caching - the table you posted above listed ideal cache vs RAM sizes, not max cacheable limits. For the limits, the interesting thing here is that this cache controller supports 32B and 64B cache lines. 64B cache lines are unusual, not seen in common use until the Athlon. That could potentially double the cacheable limit given a certain Tag RAM and cache size. Question is a) how big is your tag RAM (get a good pic of the IC to determine that) and b) what did IBM implement here (anyone's guess given lack of documentation. Tbh, I'd just try and feed it a lot of RAM (>=128MB) and test how much of that is cacheable.

Reply 16 of 39, by AlessandroB

User metadata
Rank Oldbie
Rank
Oldbie
dionb wrote on 2022-11-22, 11:50:
128MB cached was a challenge in 1997 (here's looking at you, i430TX), this chipset is four years older, from a time where 64MB w […]
Show full quote
AlessandroB wrote on 2022-11-22, 10:38:

So, following this:

And knowing that from the IBM manual:

My P60 accept only 64MB cached? No issue with chipset? no possibilities to going to 128? and from 60ns to 70ns for the 72pinsimm, whats the real difference in this system?

128MB cached was a challenge in 1997 (here's looking at you, i430TX), this chipset is four years older, from a time where 64MB would have cost USD 4k... that said, i430LX could cache all its 192MB RAM.

This board has a very obscure VLSI 82C591 chipset. No one has been able to find datasheets for it, this is the best info we have:
https://theretroweb.com/chipsets/578

Now, based on that, you might even be able to install 512MB of RAM on here, if you can find 128MB 72p FP SIMMs.

As for caching - the table you posted above listed ideal cache vs RAM sizes, not max cacheable limits. For the limits, the interesting thing here is that this cache controller supports 32B and 64B cache lines. 64B cache lines are unusual, not seen in common use until the Athlon. That could potentially double the cacheable limit given a certain Tag RAM and cache size. Question is a) how big is your tag RAM (get a good pic of the IC to determine that) and b) what did IBM implement here (anyone's guess given lack of documentation. Tbh, I'd just try and feed it a lot of RAM (>=128MB) and test how much of that is cacheable.

Here is the pic.
Basically i not want to exceed the maximum cacheable area, i want to maximize at the level of cache. Is better to know the amount of ram cacheable because i not have all that ram, i must to buy and is not really cheeper.BUT I have discover something interesting.. if you look the pics you see some hidden/unpopulated jumper: one for the bus 60/66 (the Socket4 produced by IBM in only one and is only at 60Mhz, never produce a 66Mhz speed) and another one for the cache (256/1M) but all chip are soldered, i not want to modify this rare Computer.

Attachments

Reply 17 of 39, by Sphere478

User metadata
Rank l33t++
Rank
l33t++

What does that pcb do? Looks like a signal terminator.. but why?

Sphere's PCB projects.
-
Sphere’s socket 5/7 cpu collection.
-
SUCCESSFUL K6-2+ to K6-3+ Full Cache Enable Mod
-
Tyan S1564S to S1564D single to dual processor conversion (also s1563 and s1562)