VOGONS


First post, by muon

User metadata
Rank Member
Rank
Member

Hi I've got a some curious motherboard. The motherboard is a PCCHIPS M321 Rev. 3.1. It has an empty socket for plug a 16K x 4 SRAM chip to enable write back cache (see pictures).

IMG-20190309-110952.jpg

IMG-20190309-111026.jpg

I have a IDT 7188 S15D chip (16K x 4 SRAM chip ) and I tried to plug thant chip in that socket, but it didn't mismmatch: the chip is 22 dip socket and the socket is 24 dip. Is it possible?

Reply 1 of 17, by mpe

User metadata
Rank Oldbie
Rank
Oldbie

It should be fine. TAG chips have often smaller capacity than data chips (they have less address bits). Just make sure it sits correctly in the socket (usually on the right side). It is best to identify +5V and GND pins using multimeter.

On the other hand 16kx4 should have 24 pins (14 address bits) if my math is correct.

EDIT: There might be actually different pin configurations of DIP28/DIP24/DIP22 16kx4 and multiplexed data bus pins. So you will most likely need to find out which chip exactly is required by your motherboard.

Last edited by mpe on 2020-01-03, 20:34. Edited 1 time in total.

Blog|NexGen 586|S4

Reply 3 of 17, by CoffeeOne

User metadata
Rank Oldbie
Rank
Oldbie
muon wrote on 2020-01-03, 19:40:
Hi I've got a some curious motherboard. The motherboard is a PCCHIPS M321 Rev. 3.1. It has an empty socket for plug a 16K x 4 SR […]
Show full quote

Hi I've got a some curious motherboard. The motherboard is a PCCHIPS M321 Rev. 3.1. It has an empty socket for plug a 16K x 4 SRAM chip to enable write back cache (see pictures).

IMG-20190309-110952.jpg

IMG-20190309-111026.jpg

I have a IDT 7188 S15D chip (16K x 4 SRAM chip ) and I tried to plug thant chip in that socket, but it didn't mismmatch: the chip is 22 dip socket and the socket is 24 dip. Is it possible?

yes, 24 pin types are more common, so keep yours it's rare 😁

Last edited by Stiletto on 2020-03-02, 06:28. Edited 1 time in total.

Reply 4 of 17, by GigAHerZ

User metadata
Rank Oldbie
Rank
Oldbie

I have the M321 board and when i researched the writeback capability, i came to understanding, that both 22pin and 24pin SRAM are supported in that socket.

NB! I also have both variants of the SRAM. The 22pin, which is currently not in the machine, is CY7C164-25PC. I didn't have a hard time to get that from ebay. 😀

"640K ought to be enough for anybody." - And i intend to get every last bit out of it even after loading every damn driver!

Reply 6 of 17, by muon

User metadata
Rank Member
Rank
Member
GigAHerZ wrote on 2020-01-03, 22:04:

I have the M321 board and when i researched the writeback capability, i came to understanding, that both 22pin and 24pin SRAM are supported in that socket.

Are you sure I can plug the 22 chip in the 24 pin socket?

Reply 7 of 17, by GigAHerZ

User metadata
Rank Oldbie
Rank
Oldbie
muon wrote on 2020-01-04, 10:31:
GigAHerZ wrote on 2020-01-03, 22:04:

I have the M321 board and when i researched the writeback capability, i came to understanding, that both 22pin and 24pin SRAM are supported in that socket.

Are you sure I can plug the 22 chip in the 24 pin socket?

About 99%, but please confirm it for yourself before doing it as i will not be responsible for any damages. 😁
But yes, kind of similar case like with cache memory chips on some boards, where you can fit 28 pin SRAMs into 32pin sockets.

"640K ought to be enough for anybody." - And i intend to get every last bit out of it even after loading every damn driver!

Reply 8 of 17, by derSammler

User metadata
Rank l33t
Rank
l33t
GigAHerZ wrote on 2020-01-03, 22:04:

I have the M321 board and when i researched the writeback capability, i came to understanding, that both 22pin and 24pin SRAM are supported in that socket.

Did you test for performance differences as well? I own an M321 board as well, which is populated with a DX33 and half of the cache possible. I did not add that extra SRAM chip, however, since I wasn't aware of its function.

Btw, the one I have is rev. 2.5 with the CPU and crystal in a socket. Also, the FPU socket still has the additional row of pins for a Waitek. Seems like they cut costs quite a bit in later revisions.

Last edited by derSammler on 2020-01-04, 10:45. Edited 2 times in total.

Reply 9 of 17, by GigAHerZ

User metadata
Rank Oldbie
Rank
Oldbie

@derSammler, yes, by adding writeback cache chip, the memory write speed jumps crazy high. Have to do it again, to make sure i'm correct, but speedsys general memory bandwidth jumped from twenties to fifties of MB per second.

All other benchmarks (from example phil's compiled benchmark set) also confirmed a great jump in performance. 😀

But i don't think there's performance difference between 22pin and 24pin writeback cache chips, but this is something, i haven't tested.

My board had 2.6 as revision on board - it had soldered cache and soldered wire instead of jumpers for cache config and other such things. I've desoldered and socketed/jumpered everything on it and it's now in perfect condition.

Last edited by GigAHerZ on 2020-01-04, 10:46. Edited 1 time in total.

"640K ought to be enough for anybody." - And i intend to get every last bit out of it even after loading every damn driver!

Reply 10 of 17, by CoffeeOne

User metadata
Rank Oldbie
Rank
Oldbie

Nope. No way. They are incompatible:
your chip: 22 pin: https://pdf1.alldatasheet.com/datasheet-pdf/v … T/IDT7188L.html
mentioned 22 pin from Cypress https://pdf1.alldatasheet.com/datasheet-pdf/v … 7C166-15PC.html
24 pin: https://datasheet4u.com/datasheet-parts/MT5C6 … .php?id=1351917

EDIT Buying a MT5C2564-15 should be possible.

EDIT I am sorry, my link for the 24 pin was a 64kx4, my apologies

Edited again, there are different pinouts for the 22 pin.

I am 100% sure, that the chip that you have is not working 😁

Last edited by CoffeeOne on 2020-01-04, 11:06. Edited 3 times in total.

Reply 11 of 17, by GigAHerZ

User metadata
Rank Oldbie
Rank
Oldbie

Found the discussion, i think, that i got my information of dip22 vs dip24: http://www.vcfed.org/forum/showthread.php?390 … write-back-chip

"640K ought to be enough for anybody." - And i intend to get every last bit out of it even after loading every damn driver!

Reply 12 of 17, by kixs

User metadata
Rank l33t
Rank
l33t
GigAHerZ wrote on 2020-01-04, 10:42:

@derSammler, yes, by adding writeback cache chip, the memory write speed jumps crazy high. Have to do it again, to make sure i'm correct, but speedsys general memory bandwidth jumped from twenties to fifties of MB per second.

All other benchmarks (from example phil's compiled benchmark set) also confirmed a great jump in performance. 😀

This would be great... But even on a 486 there is only 3% difference between WT and WB cache. Redo the benchmarks as mind can play tricks on you... 😉

Requests are also possible... /msg kixs

Reply 13 of 17, by derSammler

User metadata
Rank l33t
Rank
l33t

A 486 has a much higher memory bandwidth to start with and also two cache stages, so it's obvious that the effect there isn't as great. But it still would be nice to see "jumps crazy high" in numbers. 😁

Reply 14 of 17, by GigAHerZ

User metadata
Rank Oldbie
Rank
Oldbie

Yep, one must remember, 386 did not have any on-chip cache. The motherboard cache was the first level cache. (Yeah, those 486SLC/DLC did exist, but those are not exactly 386 cpus)

That is what makes this writeback feature of M321 boards so great!

"640K ought to be enough for anybody." - And i intend to get every last bit out of it even after loading every damn driver!

Reply 16 of 17, by muon

User metadata
Rank Member
Rank
Member

Hi again!!
Finally I received the memory chip with 24 pins and I plugged and it works properly.

kixs wrote on 2020-01-04, 15:29:

Benchmark results please (before and after) 😁

Here go:

"With chip"
1-Cachechk-con-chip.jpg
"Without chip"
1-Cachechk-sin-chip.jpg

Firs picture with chip, second without
2-Con-chip.jpg
2-Sin-chip.jpg

3-Con-chip.jpg
3-sin-chip.jpg

4-Con-chip.jpg
4-Sin-chip.jpg

5-Con-chip-DOOM.jpg
5-sin-chip-DOOM.jpg

Memory read access is better!!
Some applications improve (doom..), others not