VOGONS


First post, by Synaps3

User metadata
Rank Newbie
Rank
Newbie

So I am going to begin building an 8088 clone system - Sergey Kiselev's design.
http://www.malinov.com/Home/sergeys-projects/xi-8088

My goal is to get the CPU running at its specified speed of 16 MHz - so technically not an overclock for the CPU itself. The supporting clock generator chip (8284A) is rated to 10 MHz. I find it odd that this NEC is meant to run up to 16 MHz, but I can't find a clock generator that is rated high enough. How did they ever run it at that speed? There's got to be an NEC equivalent clock generator chip that is rated for that I'd assume. Anyone know what it is?

I've been experimenting with emulators and I've found that at 7-8 MHz, I can't really do much with the 8088; it's just too slow, but if I run the emu at 16 MHz, then I can run a ton more games and software just fine. I was even able to run King's Quest 6 at that speed. I don't even think it was designed to work on that CPU, says 386 or better, but it works well.

So I have a major urge to make this thing clocked to 16 MHz. I think the PC-sprint board would be the way to go, but I think I need a chip that is better than the 8284.

Systems:
BOARD | RAM | CPU | GPU
ASUS CUV4X-D | 2GB | 2 x PIII Tualatin ~1.5 GHz | Radeon HD 4650
DELL DIMENSION XPS 466V | 64MB | AMD 5x86 133MHz | Number Nine Ticket to Ride

Reply 1 of 16, by Anonymous Coward

User metadata
Rank l33t
Rank
l33t

82C84A is good for up to 25MHz.
https://www.renesas.com/us/en/www/doc/datasheet/82c84a.pdf

It's the CMOS version of 8284.

I've been interested in a faster 808x upgrade for a while as well. That PC Sprint design could probably work at 16MHz, but I wonder if it's really that much faster with memory still operating at planar speed.

It's also a real bummer there are no 8087 options over 10MHz. Nobody ever bothered to make a CMOS version of the 8087. There are rumours that Cyrix designed one, but it sounds like it was never brought to market.

"Will the highways on the internets become more few?" -Gee Dubya
V'Ger XT|Upgraded AT|Ultimate 386|Super VL/EISA 486|SMP VL/EISA Pentium

Reply 2 of 16, by Jo22

User metadata
Rank l33t
Rank
l33t

Interesting project!
What about using a fully 16-Bit CPU (V30**) and fast SRAM or pseudo-SRAM (DRAM with internal refresher)? 🙂

Edit: Some interesting 8086 board designs are shown here (S-100 bus, SRAM) :
http://www.s100computers.com/My%20System%20Pa … CPU%20Board.htm

"Time, it seems, doesn't flow. For some it's fast, for some it's slow.
In what to one race is no time at all, another race can rise and fall..." - The Minstrel

//My video channel//

Reply 4 of 16, by rmay635703

User metadata
Rank Member
Rank
Member
Anonymous Coward wrote on 2020-09-08, 15:04:

It's also a real bummer there are no 8087 options over 10MHz. Nobody ever bothered to make a CMOS version of the 8087. There are rumours that Cyrix designed one, but it sounds like it was never brought to market.

I long since wondered why the 80287 wasn’t a drop in replacement to the usually faster clock for clock 8087?

The 8087 had an MMU and I think the 80287 was a design error removing this to save what? A couple thousand transistors?

I wonder how one would interface an 80287xl to an XT given the missing MMU ?

Maybe one of the XT floating point accelerator cards would do it? I think those were usually application specific and not x87 based.

Ah well

Reply 5 of 16, by rmay635703

User metadata
Rank Member
Rank
Member
AlessandroB wrote on 2020-09-08, 16:40:

does anyone make this kind of cards?

Intel made a card to drop in the 8086 in place of the 8088

Considering the bus similarities I’ve often wondered if a dumb upgrade would allow half the bus to sit on the ISA card and the other half go to the main board 256k on the board 256k on the card for 16 bits.

Everything else would be stuck at 8 bits with a latch

Reply 6 of 16, by Jo22

User metadata
Rank l33t
Rank
l33t

The S100 bus was a spiritual predecessor of PC/XT bus, so to say. It was used in the 1970s. Many pioneer systems used it, before Apple II and the 5150 used "ISA". It was used in professional systems.
https://en.wikipedia.org/wiki/S-100_bus

Considering the bus similarities I’ve often wondered if a dumb upgrade would allow half the bus to sit on the ISA card and the other half go to the main board 256k on the board 256k on the card for 16 bits.

Or why not doing it the right way ?
Some soviet PCs had a 16-Bit PC/XT bus that was not ISA (AT-Bus). Unlike ISA, which was designed for the 80286, it was based on the 8086 frontside bus and had a small 16-Bit extension. This would be period-correct even.
Imagine CGA and Hercules cards running via 16-Bit i/o! 😁

Edit:
XT clone with 16-Bit slots ?

Here's a link to the Romanian (not Soviet) "Junior XT".
http://retroit.ro/product/junior-xt/

Here's also an older, but interesting thread.
It's about 8086 oddball chips, so to say.
The postings by the other users were very interesting, I think. Some of the NEC chips (V60?) even had EMS-style featured built-in.
8086 clones with more than 1MiB of addressable RAM ?

Anyway, that's all for now. I don't mean to disturb this thread any longer. 😅
I realized that the interest lies mostly in creating a very fast IBM 5160 style computer. 🙂

Last edited by Jo22 on 2020-09-08, 18:04. Edited 1 time in total.

"Time, it seems, doesn't flow. For some it's fast, for some it's slow.
In what to one race is no time at all, another race can rise and fall..." - The Minstrel

//My video channel//

Reply 7 of 16, by Synaps3

User metadata
Rank Newbie
Rank
Newbie
Anonymous Coward wrote on 2020-09-08, 15:04:

82C84A is good for up to 25MHz.
https://www.renesas.com/us/en/www/doc/datasheet/82c84a.pdf

It's the CMOS version of 8284.

I saw that, but I thought that was the crystal frequency, NOT the output clock frequency which would be / 3.
So am I wrong? is that the actual clock frequency for the CPU?

Systems:
BOARD | RAM | CPU | GPU
ASUS CUV4X-D | 2GB | 2 x PIII Tualatin ~1.5 GHz | Radeon HD 4650
DELL DIMENSION XPS 466V | 64MB | AMD 5x86 133MHz | Number Nine Ticket to Ride

Reply 8 of 16, by rmay635703

User metadata
Rank Member
Rank
Member

I was thinking more of if you were to upgrade a real 5160

The standard 16 bit ISA slot could work with an 8086, you would just need to either latch or NC the upper 4 address bits.

Trouble is 8086 only has a single 16 bit IRQ, rest officially are “8bit”

Reply 9 of 16, by Synaps3

User metadata
Rank Newbie
Rank
Newbie
AlessandroB wrote on 2020-09-08, 16:40:

does anyone make this kind of cards?

I assume you mean the pc sprint, not the SBC.
https://ctrl-alt-rees.com/downloads/ibmpc/PC-SPRINT.zip
https://ctrl-alt-rees.com/2020-03-02-overcloc … -pc-sprint.html

If not, that sergey board can be purchased on ebay much cheaper than from OSH park.

Systems:
BOARD | RAM | CPU | GPU
ASUS CUV4X-D | 2GB | 2 x PIII Tualatin ~1.5 GHz | Radeon HD 4650
DELL DIMENSION XPS 466V | 64MB | AMD 5x86 133MHz | Number Nine Ticket to Ride

Reply 10 of 16, by matze79

User metadata
Rank l33t
Rank
l33t

Why not just go 286 ?

On my XT ISA Clock = CPU Clock.
So no way i can go 16Mhz

https://dosreloaded.de - The German Retro DOS PC Community
https://www.retroianer.de - under constructing since ever

Co2 - for a endless Summer

Reply 11 of 16, by rmay635703

User metadata
Rank Member
Rank
Member
matze79 wrote on 2020-09-08, 18:06:

Why not just go 286 ?

On my XT ISA Clock = CPU Clock.
So no way i can go 16Mhz

I’ve ran 16mhz ISA Bus on a 486, trouble is the ide controller and sound cards.
An XT is much easier on ISA cards than a 286+

So I expect you Would need to cherry pick cards

Reply 12 of 16, by matze79

User metadata
Rank l33t
Rank
l33t

Try run some XT Class Cards.. you will find ones that would not even work beyond 4.77Mhz

https://dosreloaded.de - The German Retro DOS PC Community
https://www.retroianer.de - under constructing since ever

Co2 - for a endless Summer

Reply 13 of 16, by rmay635703

User metadata
Rank Member
Rank
Member
matze79 wrote on 2020-09-08, 18:22:

Try run some XT Class Cards.. you will find ones that would not even work beyond 4.77Mhz

I had an ibm CGa on a turbo XT, just got more screen sparkles in text mode

Reply 14 of 16, by Anonymous Coward

User metadata
Rank l33t
Rank
l33t
Synaps3 wrote on 2020-09-08, 17:54:
Anonymous Coward wrote on 2020-09-08, 15:04:

82C84A is good for up to 25MHz.
https://www.renesas.com/us/en/www/doc/datasheet/82c84a.pdf

It's the CMOS version of 8284.

I saw that, but I thought that was the crystal frequency, NOT the output clock frequency which would be / 3.
So am I wrong? is that the actual clock frequency for the CPU?

No, it appears you are right. Maybe get the military version and try your luck at overclocking?

"Will the highways on the internets become more few?" -Gee Dubya
V'Ger XT|Upgraded AT|Ultimate 386|Super VL/EISA 486|SMP VL/EISA Pentium

Reply 15 of 16, by bakemono

User metadata
Rank Member
Rank
Member

In the 1987 NEC microcomputer products catalog they list a uPD71011 clock generator for use with V20/V30 CPUs. However it outputs a clock with 50% duty cycle. Looking at the timing specs for the V20 itself, the allowed range for the clock duty cycle is 41% to 51% which is different than an 8088! Doesn't the ISA bus use 33%?

This document is too old and doesn't list the 16MHz parts, only 10MHz ones. It would probably be good to find a datasheet for the 16MHz version and see what kind of clock signal it needs, maybe it even suggests a part for the purpose

Reply 16 of 16, by Synaps3

User metadata
Rank Newbie
Rank
Newbie
bakemono wrote on 2020-09-09, 05:22:

In the 1987 NEC microcomputer products catalog they list a uPD71011 clock generator for use with V20/V30 CPUs. However it outputs a clock with 50% duty cycle. Looking at the timing specs for the V20 itself, the allowed range for the clock duty cycle is 41% to 51% which is different than an 8088! Doesn't the ISA bus use 33%?

This document is too old and doesn't list the 16MHz parts, only 10MHz ones. It would probably be good to find a datasheet for the 16MHz version and see what kind of clock signal it needs, maybe it even suggests a part for the purpose

Hmm. So maybe the way to go is to use two different chips on the PC sprint. One uPD71084 (or 8284) for the bus and then the uPD71011 for the CPU. That way the crystal frequency is only double rather than triple and not so far out of spec. The chips appear otherwise identical. The PC sprint uses two of these chips, so it'd be easy to try the 71011 in there instead I would think.

I checked the datasheet of the NEC V20HL and I can't find anything about the clock generator in there (and search doesn't work cause it's images). It seems very strange to rate a CPU at a certain frequency that would be driving their supporting chips out of spec. Maybe they used something different in their test system. Either way, it's not very professional.

I suspect it might just work fine anyway. I'll just have to try it when I can. It'll be a while cause the boards and components are shipping from Russia, but I will report back when I try it. I wonder if anyone has ever run one of these at 16 MHz anywhere?

Systems:
BOARD | RAM | CPU | GPU
ASUS CUV4X-D | 2GB | 2 x PIII Tualatin ~1.5 GHz | Radeon HD 4650
DELL DIMENSION XPS 466V | 64MB | AMD 5x86 133MHz | Number Nine Ticket to Ride