29 Feb
2008
29 Feb
'08
6:50 a.m.
The same CPU and FPU are sued on the KA630cpu (uVAXII) and in early
microvax3100s so any of them should compare.
Is it possible that someone has overclocked the that uV2000???
FYI: gate leakage would kill the chips not make it run hot. The
usual reasons for hot running are:
Over voltage!
Overclock
Undercool
Excessive bus loads (capacitive or resistive)
Since the machine is a closed system for the most part the first is
most likely and the others are least likely.
I may add that over 60C is way too hot at the heatsink and the die
due to thermal resistance will be hotter.
I'm trying to pull down one of my uV2ks and fire it up.
Allison
Subject: Re: VAXstation/MicroVAX 2000 CPU/FPU overheating?
From: "Dave Dunfield" <dave06a at dunfield.com>
Date: Fri, 29 Feb 2008 08:18:31 -0500
To: "General Discussion: On-Topic and Off-Topic Posts" <cctalk at
classiccmp.org>
I assume these CPU and FPU chips are MOS devices.
Is it possile that
excessive leakage across the gate oxide layer in some transsitors of that
chip would cause it to run hot, but still work? I'm pretty sure I've seen
chips that seem to work, but get hot and then stop working, althoguh
cooling them with freezer spray keeps them running. And these were not
chips driving high pwoer laods -- they were things like the clock/timer
microcontroller in a VCR.
Yes, I'm wondering if it's something like that, however how likely is it
that both devices would experience the same fairly-uncommon failure mode
at the same time?
Perhaps I've been engaging in the persuit of an undomesticated ornithoid...
Is it possible that the devices normally run this hot, and the failure is
occuring for another reason (possibly a side effect of the heat, as cooling
them does allow it to keep running).
Allison said "they do run hot" - does anyone know how hot?
I did some further tests last night - I dug out my thermocouple and made
some actual heat measurements with the system assembled, but with the color
frame buffer removed - there should be better airflow, and this appears to
be the case, since the machine did not die after 10 minites of operation.
Ambient temp was measureing about 25 degrees C at the start of the test.
From power on, the CPU and FPU rose rapidly
(within 3-4 minites) to
the 70C range, then more slowly rose to temperatures of 76C
for the CPU
and 82C for the FPU after 10 mins. At this temperature they seemed to be
getting stable - having done nearly 1-1/2 mins without an increase. It's
possible/likely that they would climb anothe few degrees in extended
operation.
Once this temperature was reached, I held the system in RESET, and
observed that the temps dropped back after a few minites to about 64C
for the CPU and 68C for the FPU.
To my mind, a processor running at 80C is damn hot - most of the embedded
devices I work with list absolute maximum running temperature as 70C or
75C - but perhaps the DEC devices are designed to run hotter - I note the
Intel 486 datasheet lists absolute maximum (under bias) as 110C ...
Anyone on the list with a VS 2000 or uVAX 2000 that can do some temperature
measurements?
Dave
--
dave06a (at) Dave Dunfield
dunfield (dot) Firmware development services & tools: www.dunfield.com
com Collector of vintage computing equipment:
http://www.classiccmp.org/dunfield/index.html 
1 Mar
1 Mar
6:11 p.m.
The same CPU and FPU are sued on the KA630cpu (uVAXII)
and in early
microvax3100s so any of them should compare.
I have a uVAXII which I'll be getting to eventually - right now it's
not running, but I hope to change that eventually :-)
Is it possible that someone has overclocked the that
uV2000???
I don't think so, as all the clocks (5Mhz, 40Mhz and 69Mhz) are
right where they should be in frequency.
FYI: gate leakage would kill the chips not make it run
hot. The
usual reasons for hot running are:
Over voltage!
Overclock
Undercool
Excessive bus loads (capacitive or resistive)
Since the machine is a closed system for the most part the first is
most likely and the others are least likely.
I've checked the power and it's well within spec.
Clocks appeare to be where they should be.
Cooling fan is not binding, appears to be running at the
right speed, there are no abnormal obstructions to the cooling
flow.
I may add that over 60C is way too hot at the heatsink
and the die
due to thermal resistance will be hotter.
I'm trying to pull down one of my uV2ks and fire it up.
I pulled out NETBOT:: a 12mb uV2000 with a RD53 and booted it with the
case off to get the temp and the hotter of the two chips stabilized
at 58C +- a tiny bit while running VMS 5.4-3.
Thanks! that at least tells me I'm not chasing an imaginary fault. At
82C mine is running significantly hotter than yours.
I've partially convinced that the FPU is bad, but I don't have hard
evidence/specs to confirm or deny this theory. Here's some of the
factors which lead me to this hunch:
There is some loading of the 40Mhz clock line going on (this is the
one going to the CPU and FPU). With W4 removed (disconnects the 40Mhz
clock at the osc, can) I measure a peak voltage of about 3.4V - with
the jumper in (connected to CPU and FPU) it drops to about 2.3v ...
Both chips have a CLKO line - on the CPU, this outputs a nice fat
20Mhz signal. On the FPU it outputs only about 130mv - but the technical
manuals says this pin is not used on the FPU and it is a no-connect
(so why label it CLKO - it's a different pinout than the CPU). This
could be normal - I won't know until I get a chance to compare with
another VS2000.
There are a couple of no-connect pins from the FPU which show what
appears to be a logic waveform at a very low level - possibly an
indicator of contention somewhere within the chip, possibly normal,
I don't know.
The CPU is obviously running, but I have no real proof that the FPU
us running other than that it passes the power-up '9' self-test. I
have no idea how comprehensive this test is, and how much functionality
is required to pass it. I do note that there are two series of accesses
which occur during the self-test, measured on the !EPS (External Processor
Select) line - this is a line running directly between the CPU and FPU
which is used to communicate with the FPU.
What I see are two accesses 200ns apart, then 800ns idle, followed by 13
acesses 200ns apart, followed by 1000ns. Then a long gap - I didn't measure
it, but it's several ms. The second series of accesses is identical except
that there are 14 accesses in the middle stretch instead of 13 (all 200ns
apart). This would suggest that during the test it has at least issued a
command ... But I really don't know.
The extended system tests 101, 102 and 80000106 will not test the FPU -
this may be normal, I don't know. The docs say they will test all hardware
which is installed - this may mean that the FPU is not being full detected...
I don't know.
The CPU is obvously running - the FPU may or may not be, however it is
directly connected to almost ALL signals in parallel with the CPU - if it
were failed in a heat-generating way, it may well be loading the CPU and
causing it to get hot as well - the CPU does run slightly cooler than the
FPU.
I think I've done about all I can do at this point - I see some suspicous
signals, but without real signal diagrams, or another system to compare it
to I can't confirm or deny a particular fault - I think the system is going
to have to stay on the shelf until I locate another one to make comparisons
with.
Thanks for everyone for the great ideas and assistance.
Regards,
Dave
--
dave06a (at) Dave Dunfield
dunfield (dot) Firmware development services & tools: www.dunfield.com
com Collector of vintage computing equipment:
http://www.classiccmp.org/dunfield/index.html
7:02 p.m.
On Saturday 01 March 2008 21:11, Dave Dunfield wrote:
The extended system tests 101, 102 and 80000106 will
not test the FPU
- this may be normal, I don't know. The docs say they will test all
hardware which is installed - this may mean that the FPU is not being
full detected... I don't know.
Is the FPU socketed? If so, can you remove it, and what the boot and
test output looks like?
I've not had my uV2000 opened up (or even turned on yet...), but that'd
be my next step.
If you can't remove the FPU (likely), the diagnostics would probably be
complaining about it not working...
Maybe if I have some time, I'll dig mine out and try playing with it.
Pat
--
Purdue University ITAP/RCAC --- http://www.rcac.purdue.edu/
The Computer Refuge --- http://computer-refuge.org
2 Mar
2 Mar
6:35 a.m.
The extended
system tests 101, 102 and 80000106 will not test the FPU
- this may be normal, I don't know. The docs say they will test all
hardware which is installed - this may mean that the FPU is not being
full detected... I don't know.
Is the FPU socketed? If so, can you remove it, and what the boot and
test output looks like? If you can't remove the FPU (likely), the
diagnostics would probably be
complaining about it not working...
Which is the main reason I'm not 100% convinced the FPU is faulty. But I
can't get the extended disgnostics to test it, and I don't know how
"good"
it has to be to pass the basic startup test.
I've not had my uV2000 opened up (or even turned
on yet...), but that'd
be my next step.
Maybe if I have some time, I'll dig mine out and try playing with it.
If you do open it up and have it to the point where you can scope signals,
please contact me ... I've got a few questions about the signals I see in
mine. I've also got a couple of questions about the diagnostics and the
FPU that could be answered with a running system (only the console ROM
required).
If anyone has a running VS2000 or UV2000 and is willing to open it up to
help me do some comparison testing, here are the things I would most like
to know:
[For best results, it would be good if the system were configured to match
mine - all expansion boards, drives, peripherals etc. disconnected, and
the uVAX/VS jumper located about 3" directly behind the BNC ethernet connector
set to the uVAX position, which is AWAY from the BNC connector.]
[To run the extended diagnostics 101, 102 and 80000106 you need a DB-25-F
connector installed (on the DB-25-M connector on the board) with pins 12
and 18 connected together]
a) Whats tests are run by the "T 101" command?
Mine runs: C, B, 7, 6
b) If you run "T 80000106" and then enter '9' at the "?
>>>" prompt,
do you get any test output other than the CPU designation title?
[For the following, I would like a description of the waveforms
observed, waveshape, amplitude and frequency where appropriate]
c) The 40Mhz clock at the W4 jumper (right beside the 40Mhz oscillator
"can") - with the jumper installed?
I see 40Mhz sine wave, about 2.3v p-p
d) The 40Mhz clock at the W4 jumper pin closest to the oscillator
"can" with the jumper removed?
I see 40Mhz sine wave about 3.4v p-p
e) Pins on FPU
44 - I see about 160mv irregular square wave
59 - Same as above
15 - I see 40Mhz at about 160mv
Dave
--
dave06a (at) Dave Dunfield
dunfield (dot) Firmware development services & tools: www.dunfield.com
com Collector of vintage computing equipment:
http://www.classiccmp.org/dunfield/index.html
6139
days inactive
6141
days old
3 comments
3 participants
participants (3)
-
ajp166＠bellatlantic.net -
dave06a＠dunfield.com -
pat＠computer-refuge.org