Uncreative Labs

Joined: 20 Nov 2004 Posts: 589 Location: Shandong, China

I have dug out my IBM AT system once again, and I have gotten over the road block I had encountered with the Make-It-486 upgrade adapter I installed. This problem prevented me from doing soft resets. I did so by upgrading the standard IBM BIOS chips with an AWARD v3.03 that I ripped off a VLSI board. It seems to work well, and even contains its own CMOS setup program at boot without the need for a setup disk. The Make-It-486 is a drop-in replacement based around the Cyrix 486SLC chip. It contains some extra circuitry to make it behave like clock doubled CPU. In my AT system which runs at 8MHz, the make it 486 operates at 16MHz. Though honestly it doesn't perform much better than the 286 that it replaced. There is also extra circuitry included on the module to activate the 1kb internal cache without having to run usual software. The module normally seems to come with a 487SLC tacked on underneath, but mine doesn't seem to have this option. So, I decided to try using it with my Intel 80287XL instead. The manufacturers of the Make-It-486 claim that it should be able to co-exist with 287 coprocessors, however it doesn't seem to like my 80287XL. It might be because it has an 80387sx core. I'm not exactly sure, but I am tempted to buy a Cyrix 82S87 to find out if that would work better. It is also based on 387sx, but I am hoping it will work better, since the CPU on the Make-It-486 is also a Cyrix. I think a few of you guys have this upgrade chip, have you ever had math-co related problems like this before?

Joined: 28 Feb 2006 Posts: 127 Location: LI, NY

Sorry, I haven't tried it yet. My question is, if it's claiming "Make it 486" then why does it only run about double what you have? A better question would be, why does it have a 386 core? Razz

Joined: 20 Nov 2004 Posts: 589 Location: Shandong, China

In reality, it might provide about 30% improvement over the CPU it replaces. The only real advantage lies in its ability to execute 32-bit code. I guess some people could argue about whether or not the chip is really 386 or a 486, but the fact is that performance is closer to 386SX. Eitherway, it's not a bad upgrade for a 286 system.

Today I am going to order some new crystals and DRAMs for my AT, and see what I can pull out of it. My goal is to run clock doubled 16MHz out of a type 2 board.

Joined: 20 Nov 2004 Posts: 589 Location: Shandong, China

Anyone in here have a Make-it 486 with the built-in math co that they would like to sell to me?

I went ahead and ordered 20MHz, 24MHz and 32MHz crystals the other day. I wonder how hard my AT motherboard can be pushed.

Joined: 20 Nov 2004 Posts: 589 Location: Shandong, China

I succeeded getting my AT up to 12MHz using a 24MHz crystal and upgrading to 100ns memory. However, I was unable to reach 16MHz. For some reason when I install the 32MHz crystal my AT behaves like a 3MHz 286. Everything seems to work, just very slowly. I wonder if I just have a bad crystal.

Joined: 19 Apr 2006 Posts: 261 Location: WisConSin

There are utilities to check the clock speed of the processor, though my guess is it probably isn't clocking down. Usually when you are right on the edge so to say thing actually do run much more slowly.

Joined: 28 Feb 2006 Posts: 127 Location: LI, NY

So you're saying upgrading past the motherboards limit, the computer will still work but slower? (Your sentence was a little difficult to decipher.)

Joined: 20 Nov 2004 Posts: 589 Location: Shandong, China

The motherboard is designed to run at 8MHz. I've overclocked it to 16MHz, and the CPU is clock doubled so it runs at 32MHz. When the board it clocked at 16MHz it still boots, it just runs really slowly. Everything works just fine.

Joined: 20 Nov 2004 Posts: 589 Location: Shandong, China

The motherboard was designed for an 8MHz 80286 CPU, but I am currently trying to clock it to 16MHz. The system still works perfectly fine, just really slowly. Diagnostics seem to report it runs like a 5MHz system.

On other note, I did some pretty crazy things today. First I tried swapping out the 486SLC CPU and installed a 20MHz 286. This did not help my problem at all, so obviously it is not CPU related. However, it is interesting to know that a 286 is indeed slightly faster than a 486SLC at the same clockspeed...I benched at 12MHz in both configurations.

The next thing I did was replace the BIOS chips. I thought maybe the manufacturer of the 12MHz 286 I swiped my AWARD chips from might have implemented their own overclocking prevention. I only had two extra sets available. One turned out to be an earlier verision of the AWARD chips I was already running, and the other set was from DTK and they didn't work at all. I ran out of 286 boards to salvage so I though "what the hell, I'll just try an AMI 386SX BIOS....it's close enough anyway". Installing the 386SX chips did not work with the 286-20 chip installed, but after replacing the 486SLC it worked beautifully at 12MHz. Though, operating at 16MHz gave me the same results as the previous BIOS chips.

While I didn't solve my problem, I did learn something interesting about transplanting 386SX BIOS chips. I guess it doesn't do anything terribly useful, but it includes a few interesting diagnostic programs in ROM. I think there's a low level format program in there too. It gives me the standard AMI splashscreen found on 386/486 boards, and even reports my CPU accurately as an 80486. Finally it counts up my memory quite quickly.

If anyone is interested in doing this, I am using two (HI and LO) 1989/1990 AMI chips from an early model SX motherboard with a VIA chipset. I did the usual tests and wasn't able to get the computer to generate strange errors.

The behaviour of the system at 16MHz is still a mystery, and I intend to order some extra crystals to figure out exactly what's up.

Joined: 19 Apr 2006 Posts: 261 Location: WisConSin

To elaborate on my post above, true vintage systems can end up loading and operating much more slowly (not clock speed) in general when you are near their absolute maximum speed, I will give a few explanations but there can be many reasons.

Anyway for many years in the early through late 90's I overclocked and sold systems that were outdated cheaply and overclocked other folks machines (mind you safe overclocks). I've probably overclocked around 100 computers, not many but still a fair spattering of everything from XT's, 286, 386sx (never did a dx), 486, pentium original, I stopped at the k6-2 series, never did a cyrix or k5/k6 for anyone other than me.

Some behavior I noted on TRULY vintage machines AKA very old 386's and most 286 and lower systems is that the keyboard system and ISA bus did not have its own clock became very buggy when pushed beyond 14-15mhz. Most of my problems on old machines started right around 16mhz. On some machines I would find that the keyboard controller either made the machine unusable or very slow by resetting often, on some machines the type rate becomes glacially slow making your machine seem slow but graphics run fast, very very annoying, you can see stuff fly but can't control it. You can tell if the keyboard and ISA bus has its own clock just by looking to see if there is more than one crystal on the board, many 10mhz 286's would have a 40mhz & a 24mhz or 14.xxx Mhz chips.

Many times ISA cards (especially memory cards & video cards) will become buggy and slow your system down when pushed too far. You can determine which ones by removing everything and trying different card combinations. Some ISA video cards will run as fast as 25mhz like one relatively modern Trident ISA card I had in a overclocked 50mhz 486sx I had, too bad the sound card wouldn't work correctly and made the machine slow to a crawl whenever a sound would try to play, I did benchmarks and it outdid most VLB machines on graphics. Other 16bit ISA cards won't even go past 10mhz without issues, old 8bit cards many times would wig out beyond 6mhz.

The reason a buggie ISA card can slow your system down can be for a variety of reasons, one of which is a lack of a wait state which leads to a reset which leads to a glacial delay. (many ISA devices were overengineered to continue working even after an error but it wastes time) Many disk controllers become glacial when overclocked also or just don't want to work, a few would work but corrupt all the data on the hard drive.

Something I've often wanted to try was a POD 63mhz chip with the fan disconnected from the chip and connected to an external supply. Then drop on my m912 486 board with the unofficial 60mhz bus and see what type of performance a POD gives when it is run at the correct bus speed. Never had much luck OC'ing the POD 83mhz chip, everyone said it would run @ 100mhz but I could never get it to load windows in that state.

Good Luck
Ryan

Joined: 20 Nov 2004 Posts: 589 Location: Shandong, China

It's definetly not the ISA BUS. This is an original IBM AT motherboard. The ISA bus runs at the same speed no matter what kind of crystal is installed. However, other parts of the system board running off the CPU crystal could be a possibility.