PDA

View Full Version : AT 5170 problems



framer
September 16th, 2010, 08:33 AM
I have an old beater 5170 MB that I use to test old boards with. It started throwing off all sorts of error messages then would start runing very slowly....and on and on...

After checking everything that I could think of I noticed the cpu seemed hotter than I thought it should be. I put a good size heat sink and fan and it runs fine. So what could cause the 80286-6 to run this hot it's not over clocked. Too hot to hold your finger on it.

framer

Chuck(G)
September 16th, 2010, 08:54 AM
Checked your power supply voltages lately? If they're good, you might want to try replacing the 82284 clock generator.

per
September 16th, 2010, 09:04 AM
I have no idea what might be the problem, but I may make a guess:

I assume the MB has been inside a chassis and well grounded.

If not, external moving charged objects may have partally broken some isolating material between two lines inside the CPU, causing a short-cirquit which might not be signifficant enough to acutally break the operation of the chip. This is a risk with all finer electronic devices, and this is also why nobody should ever use foam-peanuts when shipping such devices.

I suggest you run some kind of burn-in check that tests all the instructions and registers of the CPU for an hour or two, in order to guarantee relaiable operation.

mikey99
September 16th, 2010, 09:31 AM
Did you try plugging in another 286 chip to see if it also overheats ?

I recall there was a hardware hack to replace the crystal on the 5170 MB to force it to run at 8Mhz.
I'm sure some of the 6Mhz chips would run okay at 8Mhz, could that be the case here ?
Does the crystal say 12 Mhz or 16 Mhz ?

framer
September 16th, 2010, 09:50 AM
I have no idea what might be the problem, but I may make a guess:

I assume the MB has been inside a chassis and well grounded.

If not, external moving charged objects may have partally broken some isolating material between two lines inside the CPU, causing a short-cirquit which might not be signifficant enough to acutally break the operation of the chip. This is a risk with all finer electronic devices, and this is also why nobody should ever use foam-peanuts when shipping such devices.

I suggest you run some kind of burn-in check that tests all the instructions and registers of the CPU for an hour or two, in order to guarantee relaiable operation.

Interesting answer. I said it is a beater. I bought it for a late third party BIOS that it held and was used elsewhere. It came from a junk bin and has been without a case for years. It now has an old 1st generation BIOS that came from the board that got the new one. I will get my hand on a new CPU and see what happens. I'd bet you're right.

framer

Chuck(G)
September 16th, 2010, 01:00 PM
It's doubtful that styrofoam peanuts could have nuked a CPU installed on a board--there's plenty of static protection afforded by the connected components. Still, the CPU could be suffering from other physical problems.

Reminds me of when MOSFETs made their first appearance. Some where shipped with a wire bridging the leads; it didn't take much to punch through the gate barrier. After you'd installed the FET, you clipped the wire between the leads. Once in a circuit, the danger of ESD was much lower.

james1095
September 16th, 2010, 01:05 PM
In addition to checking the power supply voltages, I would check the ripple. An oscilloscope is ideal, but a multimeter set on AC will work in a pinch. The 286 CPU runs directly from 5V from the power supply, unlike modern PCs where the CPU voltage is derived from dedicated onboard voltage regulators.

If that looks ok and swapping the CPU doesn't fix it, the only other thing I can think of is some weirdness with the clock generator.

per
September 16th, 2010, 01:38 PM
It's doubtful that styrofoam peanuts could have nuked a CPU installed on a board--there's plenty of static protection afforded by the connected components. Still, the CPU could be suffering from other physical problems.

I didn't say styrofoam peanuts was the cause of the problems, but what I said was that you should never ship finer electronic devices (like chips and ISA cards) using them.

Regarding the CPU, it could almost be anything that's causing the problem, and my guess was not limited to just styrofoam peanuts as I was reffering to anything that can store charges (even ionizised air, but that's very unlikely).

modem7
September 17th, 2010, 12:39 AM
Too hot to hold your finger on it.
Note that that is normal for an 80286-6 in a 5170. I'm not saying that your CPU isn't overheating; I'm simply saying that the normal running CPU temperature is such that it is, "Too hot to hold your finger on it". If you have a temperature probe, we can do some objective measurements.

framer
September 21st, 2010, 01:38 PM
To bring those interested up to date. I can run that system all day with a heat sink and fan then remove and it will throw an error within minutes.

Not having another 80286-6 to play with, can I install a faster -8 -10 or -12 processor or will it void my warranty and expect it to work at the slower clock setting.

framer

Chuck(G)
September 21st, 2010, 01:54 PM
A faster 80286 is fine. An Intersil 80C286 CMOS version would use less power.

per
September 21st, 2010, 02:11 PM
To bring those interested up to date. I can run that system all day with a heat sink and fan then remove and it will throw an error within minutes.

Not having another 80286-6 to play with, can I install a faster -8 -10 or -12 processor or will it void my warranty and expect it to work at the slower clock setting.

framer

By the way, what kind of socket is is? It may be as simple as a leg which don't make proper contact with the socket. When the chip is heated, the metal slightly increases in size and that may cause improperly connected pins to loose contact completely.

Raven
September 21st, 2010, 05:47 PM
In my experience, yes. I don't guarantee that there's something I don't know, though.

Chuck(G)
September 21st, 2010, 09:05 PM
Most 5170s used PLCC 80286s. Perhaps there are a few that used ceramic LCCs, I don't know.

modem7
September 22nd, 2010, 01:16 AM
Most 5170s used PLCC 80286s. Perhaps there are a few that used ceramic LCCs, I don't know.
All of the 5170 motherboards I have seen use ceramic PGA 80286s.

modem7
September 22nd, 2010, 01:19 AM
Here's a consolidated list of possibilities suggested so far:

1. Dud 80286-6 (suggested by several posters).

2. Overvoltage (suggested by Chuck), e.g. 6 volts instead of 5 volts.

3. Overclocked due to upgraded crystal (suggested by Mikey99), i.e. something greater than 16 MHz (since an upgrade from 12 to 16 is a well known 5170 upgrade for the first models).

4. Overclocked due to 82284 failing (suggested by Chuck & james1095) in a really unusual way.

5. Poor connection (suggested by per).

6. Operation of motherboard in ambient air temperature above what the motherboard was designed to handle. The 5170 Technical Reference specifies that temperature for around the 5170 chassis, not the motherboard. For the 5170: Maximum ambient air temperature of 32 degrees C (90 degrees F).

framer
September 22nd, 2010, 05:00 PM
Cross off #2 voltage was 4.85 V under load.

Crytal is 12 Mhz so #3 is gone.

framer

doramide7
September 30th, 2010, 09:01 PM
It's doubtful that styrofoam peanuts could have nuked a CPU installed on a board--there's plenty of static protection afforded by the connected components. Still, the CPU could be suffering from other physical problems.

Reminds me of when MOSFETs made their first appearance. Some where shipped with a wire bridging the leads; it didn't take much to punch through the gate barrier. After you'd installed the FET, you clipped the wire between the leads. Once in a circuit, the danger of ESD was much lower.

In addition to checking the power supply voltages, I would check the ripple. An oscilloscope is ideal, but a multimeter set on AC will work in a pinch. The 286 CPU runs directly from 5V from the power supply, unlike modern PCs where the CPU voltage is derived from dedicated onboard voltage regulators.

If that looks ok and swapping the CPU doesn't fix it, the only other thing I can think of is some weirdness with the clock generator.

__________________
burn wii games without a mod chip (http://www.youtube.com/watch?v=m6CDpTh1ds8)
copy wii games to dvd (http://www.youtube.com/watch?v=m6CDpTh1ds8)

Chuck(G)
September 30th, 2010, 09:11 PM
All of the 5170 motherboards I have seen use ceramic PGA 80286s.

Quite right. Most 286 motherboards used the PLCC or LCC versions. IBM was a bit of an aberration. Never did figure out why--we had early 80286 steppings (pre-production silicon. circa 1981-82) and all of those were ceramic LCC. When I saw the innards of a 5170, I was a bit surprised at the CPU package. Most people don't realize how far behind the state of the art the 5150 actually was.

framer
October 1st, 2010, 08:45 AM
I'm still following this thread. The PS I feel is not the problem. I've got 3 5170 PS's and they've all been tried and power levels checked. I've been looking for another cpu with out success so far. I really hate to pull a cpu out of another good machine and risk causing problems with it. It still runs fine with a heat sink and fan.

framer

framer
October 9th, 2010, 01:32 PM
I got another CPU, it's an 80286-6. It's in the MB and booted up. Running time is currently at 1 1/2 hour without a problem. Even played Stellar7 for 30 minutes. I declare this fixed.

Bad CPU, per was correct.

framer

framer
October 11th, 2010, 03:54 PM
Has not skipped a byte over the weeked. It's been loaded up w/cards and feed old games for two days. Could this be considered computer abuse?

framer