IBM AT 5170 Crystal and Processor Upgrade

[ButINeededThatName]

So earlier today a co-worker of mine gave me an IBM 5170! Aside from a few minor parts missing (I posted a wanted ad regarding them), the system is in excellent working and cosmetic shape! It's an earlier revision system with the type I motherboard with piggybacked RAM and it came with a Hercules MDA card, a fully populated Persyst Time Spectrum 384L and a TDK EGA card, as well as the IBM AT keyboard and a 5151 display (according to him his EGA display bit the dust a decade ago).

Anyways, my system being a 5170-099 and, for what ever reason having a 12.5Mhz C80287XL, I figured I'd swap the crystal and drop in a faster processor. So my question to you is, what is the fastest 80286 I can run in this system and what crystal would I need to run said faster processor at it's true speed?

 

[Ruud]

It is not only the processor that counts, can the other parts cope as well with a higher clock speed?

If you really want to give it a try, the crystal needs to be twice as fast as the CPU. So for a 10 MHz 80286 you'll need a 20 MHz crystal.

 

[Chuck(G)]

Crystal-swapping was quite a thing back in the early days of the 5170. There were even continuously-variable PLL setups to allow adjustment to the hairy edge of reliability.

 

[modem7]

It is not only the processor that counts, can the other parts cope as well with a higher clock speed?

When the 8 MHz version of the 5170 was released, various motherboard chips had been upgraded from a speed rating perspective. Listed in note 2 at [here].

If you really want to give it a try, the crystal needs to be twice as fast as the CPU.

Per the diagram at [here].

So my question to you is, what is the fastest 80286 I can run in this system and what crystal would I need to run said faster processor at it's true speed?

IBM did not support overclocking on the 5170, so you face all of the associated risks of overclocking.

FYI. The power-on self test in the type 2 and type 3 motherboards contains some code that looks for overclocking (and underclocking), and if found, stops the machine. See test/checkpoint 11 at [here].

 

[2icebitn]

I discern the need for water cooling before too long

 

[ATSystems]

FYI. The power-on self test in the type 2 and type 3 motherboards contains some code that looks for overclocking (and underclocking), and if found, stops the machine. See test/checkpoint 11 at [here].

How is that even possible? How does a clock determine it is running too fast if there is no other clock to test against? Does it test against the RTC?

 

[modem7]

How is that even possible? How does a clock determine it is running too fast if there is no other clock to test against? Does it test against the RTC?

A comparison is made to something derived from the 8254 timer chip.
Per [here], the clocks for the CPU and 8254 are derived from separate crystals.

 

[ATSystems]

Makes sense, thank you.

 

[Ruud]

FYI. The power-on self test in the type 2 and type 3 motherboards contains some code that looks for overclocking (and underclocking), and if found, stops the machine. See test/checkpoint 11 at [here].

I didn't know that. Never to old to learn. Thanks!

 

[modem7]

I didn't know that ...

I learn something every day.

... Never to old to learn.

Dementia runs in my family, and so when I get old, I'll be unlearning something every day.

 

[Malc]

I learn something every day.

Me Too, I never knew that either

Dementia runs in my family, and so when I get old, I'll be unlearning something every day.

Sorry to hear that, very sad, Back in march i lost my mum to Dementia and other health issues, Let's hope it misses you.

 

[ButINeededThatName]

When the 8 MHz version of the 5170 was released, various motherboard chips had been upgraded from a speed rating perspective. Listed in note 2 at [here].


Per the diagram at [here].


IBM did not support overclocking on the 5170, so you face all of the associated risks of overclocking.

FYI. The power-on self test in the type 2 and type 3 motherboards contains some code that looks for overclocking (and underclocking), and if found, stops the machine. See test/checkpoint 11 at [here].

Thanks for this information! Luckily my board is a type 1 so there's no speed-limiting code to worry about. Incidentally I came across a different post on here yesterday via a google search about someone who pushed their 5170 to 10Mhz by swapping the crystal and upgrading the BIOS to the newest AMI variant, and someone else had talked about replacing the 82288 chip with the 12Mhz variant, so now my plans are as follows;

-Replace the oscillator crystal with a 25Mhz one as to use the 12.5Mhz 286 I have
-Source a 12Mhz 82288 and chip carrier and put my soldering skills to use
-Upgrade the BIOS on my ROM chips to the AMI one (or maybe pay someone to do it for me as I don't have a way to do this currently).

Either way this should be a fun project, and if the absolute worst happens I do have a second type 1 and a type 3 board shelved.

I discern the need for water cooling before too long

We all know watercooling increases performance by a factor of three

 

[modem7]

Incidentally I came across a different post on here yesterday via a google search about someone who pushed their 5170 to 10Mhz by swapping the crystal and upgrading the BIOS to the newest AMI variant, ...

Yeah, I think I remember that one. They started off at 15 MHz, then gradually reduced that until the smoke stopped.

 

[Anonymous Coward]

Thanks for this information! Luckily my board is a type 1 so there's no speed-limiting code to worry about. Incidentally I came across a different post on here yesterday via a google search about someone who pushed their 5170 to 10Mhz by swapping the crystal and upgrading the BIOS to the newest AMI variant, and someone else had talked about replacing the 82288 chip with the 12Mhz variant, so now my plans are as follows

It sounds like it could have been one of my old posts. I definitely overclocked my Type III AT motherboard to 10MHz and swapped in an AMI BIOS. I also bought a 12MHz bus controller, and a 10MHz version of another support chip, however I didn't get around to swapping them in, and I certainly didn't smoke anything.

I lost interest in clocking beyond 10MHz, because in the 5170 the AT bus is infact NOT decoupled, and running faster than 8-10MHz might upset some ISA cards. At least, I've never had much luck with running ISA faster than 10MHz in any of my other systems.

From what I remember reading, most of the 6MHz ATs will overclock to between 7-9MHz, and the limiting factor is usually the RAM. Unfortunately for you, the RAM on a type 1 board is piggy backed, and getting replacements is much more of a pain in the butt. I think the stock ones are something like 200-250ns. The type II and III planars on the other hand use standard DRAM chips, and I was easily able to swap mine out for some 100ns parts. From what I understand, 8MHz is usually acheivable on a stock 6MHz system without modification.

My current plan is to swap an XT-286 5162 planar into my AT, and then overclock that to 10MHz. These boards are 0WS, and supposedly the performance boost is quite nice. I would guess about 35% faster than an AT of the same speed rating (6MHz XT 286 is said to be about equal to an 8MHz 5170). The only problem is that the 5162 board I bought has problems. It throws up some IRQ related errors on my POST card. I guess an IC went bad.