Wait for Atheatos.
He likes building stuff alike.

Nexxen wrote on 2021-07-08, 14:36:

Wait for Atheatos.
He likes building stuff alike.

You mean i can get more speed out of my system with that? 🙂

GuillermoXT wrote on 2021-07-08, 15:15:

Nexxen wrote on 2021-07-08, 14:36:

Wait for Atheatos.
He likes building stuff alike.

You mean i can get more speed out of my system with that? 🙂

If it catches his interest he might build an adapter to pimp all 286s.
Engineer with retro love.
If he could build an adapter with cache it would be nice.

Many things back then were just too expensive or costs were lower with new hardware.
IMO

Nexxen wrote on 2021-07-08, 15:56:

If it catches his interest he might build an adapter to pimp all 286s. Engineer with retro love. If he could build an adapter wi […]
Show full quote

GuillermoXT wrote on 2021-07-08, 15:15:

Nexxen wrote on 2021-07-08, 14:36:

Wait for Atheatos.
He likes building stuff alike.

You mean i can get more speed out of my system with that? 🙂

If it catches his interest he might build an adapter to pimp all 286s.
Engineer with retro love.
If he could build an adapter with cache it would be nice.

Many things back then were just too expensive or costs were lower with new hardware.
IMO

Can you tell him i am very interested? 😁🙏
It would be a dream if he may build an upgrade for my 8MHz plcc68 286 wich does not have an oscillator. I can only switch to 6mhz down on bios that's all.
In a german forum (Dosreloaded.de) there is someone I know who is able building a mod chip. With that you may get the FPU run at his maximum speed so mine in that case would run at 10MHz instead of 5mhz 🙂

The "mod chips" is actually pretty easy to assemble. You could probably do it yourself with some basic soldering skills. They sold these in kits in the 80s, and they're basically just an interposer with a few resistors, capacitors and a clock crystal.

How does your 80286 set the CPU speed if it doesn't use a clockgen, oscillator or crystal?

Anonymous Coward wrote on 2021-07-09, 10:31:

The "mod chips" is actually pretty easy to assemble. You could probably do it yourself with some basic soldering skills. They sold these in kits in the 80s, and they're basically just an interposer with a few resistors, capacitors and a clock crystal.

How does your 80286 set the CPU speed if it doesn't use a clockgen, oscillator or crystal?

I really have no idea, probably the CPU is clocked with the ISA bus?
In addition to the 14.318 and 1,832 MHz, there is an oscillator with 48MHz but whether that is intended for the CPU? As already mentioned, I can only switch between 6 and 8MHz in the BIOS.

Anonymous Coward wrote on 2021-07-09, 10:31:

The "mod chips" is actually pretty easy to assemble. You could probably do it yourself with some basic soldering skills. They sold these in kits in the 80s, and they're basically just an interposer with a few resistors, capacitors and a clock crystal.

How does your 80286 set the CPU speed if it doesn't use a clockgen, oscillator or crystal?

This is my Board

Caluser2000 wrote on 2021-07-08, 00:57:

I wouldn't personally touch 386s at all. For 16-bit operations they need to do twice the work. A 386SX20 is about the same as a 286/10 in that regard.

I'd like to know which datasheet this information came from.

I'm pretty sure the 48MHz oscillator is for the CPU clock. It probably divides by 6 for 8MHz operation, and divides by 8 for 6MHz operation. My guess is that you could probably overclock to 10MHz if you replaced the oscillator with a 60MHz one. That would probably give you 7.5MHz is de-turbo mode.

Anonymous Coward wrote on 2021-07-09, 10:31:

The "mod chips" is actually pretty easy to assemble. You could probably do it yourself with some basic soldering skills. They sold these in kits in the 80s, and they're basically just an interposer with a few resistors, capacitors and a clock crystal.

Do you have a blueprint? I'd like to try assembling one.

I'm pretty sure all you have to do is feed the 80287 its own clock signal.

This interposer only uses three components. I am pretty sure the resistor and capacitor simply turn that clock crystal into an oscillator.

Anonymous Coward wrote on 2021-07-09, 13:08:

I'm pretty sure the 48MHz oscillator is for the CPU clock. It probably divides by 6 for 8MHz operation, and divides by 8 for 6MHz operation. My guess is that you could probably overclock to 10MHz if you replaced the oscillator with a 60MHz one. That would probably give you 7.5MHz is de-turbo mode.

Could be there someone else with the same board who already tried this? I recently changed the 8mhz cpu to a 10MHz version just to check if the settings would change to 10/6 mhz for example.

Deunan wrote on 2021-07-09, 10:54:

Caluser2000 wrote on 2021-07-08, 00:57:

I wouldn't personally touch 386s at all. For 16-bit operations they need to do twice the work. A 386SX20 is about the same as a 286/10 in that regard.

I'd like to know which datasheet this information came from.

No data sheet requited at all. Just personnel experience. And it is well documented.

If you don't believe go over and have a chat with then greybeards over at the vcfed forums...😉

If your experience was a 20MHz 386SX merely matching a 10MHz 286 then your 386 system was Badly, and that's with capital B, misconfigured somehow. I couldn't even guess how.

And no offence to vcfed and their "greybeards" but perhaps they all got a bit senile if that is what they claim. Is it possible to write a short piece of code (or a synthetic benchmark) that would be almost 2x faster on 286? Yes. Even REP STOS is getting close to that. Would a whole program/game be 2x faster? Nowhere near that.

What's more, this claim might have some merit (but again, not at 2x factor) when comparing a very late 20-25 MHz 286 to early 386SX systems, which often even reused the same chipset as 286 - but it is not universally true. Early 12MHz 286 systems are just slow even under pure DOS and rarely the mobo sports RAM chips fast enough to lower WS (if that is even an option on said mobo). SX came after DX so 16MHz CPUs are actually somewhat rare, most are 25-33 MHz (Intel) and 33-40 (AMD), on improved mobos with SIMM slots and decent BIOSes.

And that's not even touching the subject of availability and prices of fast 286 systems, whereas 33MHz 386SX are plentiful. Again I have no idea how you've made a 10MHz 286 match a 20MHz 386SX but it will be hard for other people to match those results - so it's a terrible advice.

I'd say a 20MHz 386SX would better match a 12-16MHz 286 when running generic 8088/8086 code, everything else (chipset, wait states, etc.) being equal. 10MHz is a little too conservative. Now, if said 386SX was doing nothing but executing 386 code that requires a lot of 32-bit transfers, then yes, I can see a case being made for it effectively wasting 1/2 of its machine cycles doing things.

That being said, I've definitely seen my fair share of slow 16MHz 386SX systems, so I'd be remiss to say they were 'rare' in any capacity. My first 386 ever was a 386SX-16, and it was, frankly, a complete slug. (Not helped by the dying hard drive and slow video card it had, I'm sure.) But, ironically enough, a slug that I've been trying to reclaim. It was an Everex/AGI 3000G, forgot what model the mobo was.

On the other hand, maybe I'm being a little spoiled by my IBM 486SLC2 here. 😜

Deunan wrote on 2021-07-11, 19:03:

If your experience was a 20MHz 386SX merely matching a 10MHz 286 then your 386 system was Badly, and that's with capital B, misc […]
Show full quote

If your experience was a 20MHz 386SX merely matching a 10MHz 286 then your 386 system was Badly, and that's with capital B, misconfigured somehow. I couldn't even guess how.

And no offence to vcfed and their "greybeards" but perhaps they all got a bit senile if that is what they claim. Is it possible to write a short piece of code (or a synthetic benchmark) that would be almost 2x faster on 286? Yes. Even REP STOS is getting close to that. Would a whole program/game be 2x faster? Nowhere near that.

What's more, this claim might have some merit (but again, not at 2x factor) when comparing a very late 20-25 MHz 286 to early 386SX systems, which often even reused the same chipset as 286 - but it is not universally true. Early 12MHz 286 systems are just slow even under pure DOS and rarely the mobo sports RAM chips fast enough to lower WS (if that is even an option on said mobo). SX came after DX so 16MHz CPUs are actually somewhat rare, most are 25-33 MHz (Intel) and 33-40 (AMD), on improved mobos with SIMM slots and decent BIOSes.

And that's not even touching the subject of availability and prices of fast 286 systems, whereas 33MHz 386SX are plentiful. Again I have no idea how you've made a 10MHz 286 match a 20MHz 386SX but it will be hard for other people to match those results - so it's a terrible advice.

Ah no it is not! See the vogons post directly after your post.

People agree or disagree on variuous things. W hen Im say in" MY" experience that is "EXACTLY" what I meant. You have exactly no clue what I or anyone elsewhere on vogons has been though wrt using old computer system so please lighten up a bit. And please note the word "opinion" in the OPs thread heading thank you. You'll fell much better. Posts on vogons are not a points contest but a forum to express our experiences and help those folk who need it.

As well as point out other vogon members blatant inaccuracies, some of which are out wright false fabrications that could misdirect vogons members new to user old computer kit both hardware and software wise, without being rude I might add.

I'm gad we could have this very fruitful conversation...😉

I have tested 16-bit real mode code on 386sx and 286 on the exact same chipset and clock speed - both with zero waits. The 286 is a bit faster, but only to the tune of 3-5%.

A 386sx20 should beat every 286 except the Harris 20MHz and 25MHz systems, all things being equal.

null

I'd like to be the voice of reason here, and tell everyone; NONE OF THIS FUCKING MATTERS.

Bickering like children about whats "faster" as if it fucking matters. *Maybe* it did in 1990, but not now when we are all spoiled for choice on hardware.

Jesus tapdancing Christ.

Just build what makes you happy.

maxtherabbit wrote on 2021-07-11, 22:30:

I have tested 16-bit real mode code on 386sx and 286 on the exact same chipset and clock speed - both with zero waits. The 286 is a bit faster, but only to the tune of 3-5%.

A 386sx20 should beat every 286 except the Harris 20MHz and 25MHz systems, all things being equal.

That performances increase certainly wasn't worth the extra cost involved when I purchased my first 286/16 in the very early '90s...😉