🚀 go-pugleaf

In article <901576753.4197.0.nnrp-05.c2de7290@news.demon.co.uk>, Padsu
<adam@littleangel.demon.co.uk> writes
>Any suggestions for thorough testing? I've played a bit of Quack2 which
>seems to push the processor a bit, but I'd like something that I can leave
>running that will really test things thoroughly.
Try Unreal, it beat my K6 233 at 83*3.(occasional hangs). I've never
seen  anything else test a system so intensely. Not a bad game either.
If you check out their website www.unreal.com it goes on about k6s
suffering if they are overclocked.
Best regards,   Petier     http://www.secondrow.demon.co.uk

In article <901576753.4197.0.nnrp-05.c2de7290@news.demon.co.uk>, Padsu
<adam@littleangel.demon.co.uk> writes

>But I want to test it fully, really hit the hardware to make sure everything
>is as hunky dory as it seems.

It's a perennial problem at work, where we do need to be able to stress
hardware quite severely to find faults.  In my experience, diagnostics
just don't cut the mustard as they test only one component at a time in
a one-step-at-a-time manner, which doesn't stress the hardware enough.

I tend to write self-looping DOS batch files to copy huge (300Mb) PKZIP
files around, unzip them with -t, delete them, copy them to the network
server, etc. and run those concurrently with a few .mov's or .avi's,
while loading, for example, Word 97 and Excel 97 simultaneously.  We
also have a hot-box into which I can chuck a troublesome PC for a few
days running burn-in diagnostics.

A real bug bear is where a customer puts in a PC reporting intermittent
GPFs or lockups but can't or won't tell us what they do to provoke the
fault, making it impossible to find.  It's akin to putting your car into
a garage for an intermittent pinking problem, but not telling the
mechanic when it was last serviced, whether the pinking occurs at speed
or when decelerating, when the engine's hot or cold, what sort of petrol
you're putting in, etc.  All one can do in those sort of situations
really, is adopt the shotgun approach to troubleshooting.

> I've played a bit of Quack2 which
                       ^^^^^
<grin>  Actually, Quake does seem to be quite a severe test of hardware.
I've read several reports of it crashing on marginal or overclocked
machines which were otherwise OK once reverted to rated speed or running
standard apps.

-- 
Mike Tomlinson
http://i.am/miket/

On Mon, 27 Jul 1998 22:58:38 +0100, "Padsu" <adam@littleangel.demon.co.uk>
wrote:

>Okay, so I've got a big new pootah, a k6-2 on a FIC 503+, running at
>112*3@2.4volts.
>
>
>But I want to test it fully, really hit the hardware to make sure everything
>is as hunky dory as it seems.
>
>
>Any suggestions for thorough testing? I've played a bit of Quack2 which
>seems to push the processor a bit, but I'd like something that I can leave
>running that will really test things thoroughly.

Well apart from running really big C++ builds there is one utility which
does run a real "Stress Test" - QuarterDeck's WinProbe 95.  It costs money
but if your system can run 30 mins with it it's probably golden.  As for
"free" stuff there's Cachchk V6 which will bash the memory (in DOS) to the
point you're sure there are no problems but it will not test above 64MB.


Rgds, George Macdonald

"Just because they're paranoid doesn't mean you're not psychotic" - Who, me??

I have found BCM diagnostic very helpful. You can try it for 30Days it's on
$29.
The 30 day trial version has a limit of 10 iterations of a selected group of
test to perform. You can run the selection concurrently or sequentially .

http://www.bcmcom.com

--
gregorys@one.net
Padsu wrote in message
<901576753.4197.0.nnrp-05.c2de7290@news.demon.co.uk>...
>Okay, so I've got a big new pootah, a k6-2 on a FIC 503+, running at
>112*3@2.4volts.
>
>
>But I want to test it fully, really hit the hardware to make sure
everything
>is as hunky dory as it seems.
>
>
>Any suggestions for thorough testing? I've played a bit of Quack2 which
>seems to push the processor a bit, but I'd like something that I can leave
>running that will really test things thoroughly.
>
>
>
>TIA
>
>Ad.
>
>

> It's a perennial problem at work, where we do need to be able to stress
> hardware quite severely to find faults.  In my experience, diagnostics
> just don't cut the mustard as they test only one component at a time in
> a one-step-at-a-time manner, which doesn't stress the hardware enough.

Are you sure? I've used a professional diagnostic tool with a system
load command, where it runs everything for a set time (e.g. 1 hour, 20
mins etc). I think it was called PC-CHECK.

On Wed, 29 Jul 98 02:48:36 GMT, peter@cdsl.demon.co.uk (Peter Hayward)
wrote:

>In article <35bdc2a6.1045941@news.garden.net>
>           fammacd=!SPAM^nothanks@garden.net "George Macdonald" writes:
>
>| On Mon, 27 Jul 1998 22:58:38 +0100, "Padsu" <adam@littleangel.demon.co.uk>
>| wrote:
>|
>| >
>| >Any suggestions for thorough testing? I've played a bit of Quack2 which
>| >seems to push the processor a bit, but I'd like something that I can leave
>| >running that will really test things thoroughly.
>|
>| Well apart from running really big C++ builds there is one utility which
>| does run a real "Stress Test" - QuarterDeck's WinProbe 95.  It costs money
>| but if your system can run 30 mins with it it's probably golden.  As for
>| "free" stuff there's Cachchk V6 which will bash the memory (in DOS) to the
>| point you're sure there are no problems but it will not test above 64MB.
>|
>   It's a nice though but wrong. CACHCHK is designed to check the speed
>   of memory - it's doesn't _validate_ the data is sends around the
>   machine.  Provided the code for CACHCHK itself doesn't become corrupt
>   it's will run to completion and print out an answer no matter how
>   much data corruption is going on.  I learnt this the hard way trying
>   to use CACHCHK to tune some of my BIOS settings.  I can assure you
>   that getting an answer out of cachkc does _not_ mean the m/c will be
>   stable when running a proper OS.

 I did not say that it will completely validate your "memory" but it *does*
hammer the memory and its interface, including the L2 cache interface.
There *is* a good chance that you will get some kind of failure if things
are not sound.  Perhaps I overstated things by saying that you can be sure
there are no problems but it *is* true that if it fails then you *do* have
a problem.  It's just another tool.  Hell I've run exhaustive diagnostics
for hours on many computer which it turned out had bad memory and never
gotten a memory failure from them, even with ECC and parity.

As for the Cachechk code, it obviously has to be read from main memory a
number of times depending on the L2 cache size, since it will thrash
against the data in L2 cache.


Rgds, George Macdonald

"Just because they're paranoid doesn't mean you're not psychotic" - Who, me??

What about benchmarks ? I'm running Windows 98 and I've just upgraded from a
K6 200MMX to a K6 300MMX via an Intel 233MMX. I'd be interested to find a
*good* benchmark to see how much difference there actually is. I've been
using WinCheckit which thinks I've got a AMD K6 601MMX and WinTune which is
quite good but very out of date.

Can anyone suggest anything ?

Andy

In article <901721090.14985.0.nnrp-09.c2dea8e2@news.demon.co.uk>, Andrew
Taylor <andy@hot-shot.com> writes
>What about benchmarks ? I'm running Windows 98 and I've just upgraded from a
>K6 200MMX to a K6 300MMX via an Intel 233MMX. I'd be interested to find a
>*good* benchmark to see how much difference there actually is. I've been
>using WinCheckit which thinks I've got a AMD K6 601MMX and WinTune which is
>quite good but very out of date.
>
>Can anyone suggest anything ?
>
>Andy
>
>
Microscope Diagnostics is a very, very good package. Used to use where I
used to work. very *intensive* set of utilities. Last I knew, it was
Microscope 6000 approx 1996. Not cheap, but damn good. Does mem and
throughly thrashes a system as often as you want. Even comes with its
own command.com to totally bypass Dos/Win.
-- 
Paul Hornshaw

In article <35BE2127.F4CF2B9F@markdnet.demon.co.uk>, Mark Dixon
<mark@markdnet.demon.co.uk> writes

(diagnostics)

>Are you sure? I've used a professional diagnostic tool with a system
>load command, where it runs everything for a set time (e.g. 1 hour, 20
>mins etc). I think it was called PC-CHECK.

Yes, I use that, and its successor, Troubleshooter.  I still stand by my
comment that while they are useful for finding fairly gross faults, they
are not so useful when trying to find intermittents or faults produced
by some weird combination of hardware and software interactions.

-- 
Mike Tomlinson
http://i.am/miket/

In article <901721090.14985.0.nnrp-09.c2dea8e2@news.demon.co.uk>, Andrew
Taylor <andy@hot-shot.com> writes

>What about benchmarks ? I'm running Windows 98 and I've just upgraded from a
>K6 200MMX to a K6 300MMX via an Intel 233MMX. I'd be interested to find a
>*good* benchmark to see how much difference there actually is. 

I use Wintune 97 from www.winmag.com.  This lets you save your results
and pick a selection of machines to compare  (It also has a database of
results built in.)

Wintune98 is available, but I don't like it.

-- 
Mike Tomlinson
http://i.am/miket/

In article <35bdc2a6.1045941@news.garden.net>
           fammacd=!SPAM^nothanks@garden.net "George Macdonald" writes:

| On Mon, 27 Jul 1998 22:58:38 +0100, "Padsu" <adam@littleangel.demon.co.uk>
| wrote:
|
| >
| >Any suggestions for thorough testing? I've played a bit of Quack2 which
| >seems to push the processor a bit, but I'd like something that I can leave
| >running that will really test things thoroughly.
|
| Well apart from running really big C++ builds there is one utility which
| does run a real "Stress Test" - QuarterDeck's WinProbe 95.  It costs money
| but if your system can run 30 mins with it it's probably golden.  As for
| "free" stuff there's Cachchk V6 which will bash the memory (in DOS) to the
| point you're sure there are no problems but it will not test above 64MB.
|
   It's a nice though but wrong. CACHCHK is designed to check the speed
   of memory - it's doesn't _validate_ the data is sends around the
   machine.  Provided the code for CACHCHK itself doesn't become corrupt
   it's will run to completion and print out an answer no matter how
   much data corruption is going on.  I learnt this the hard way trying
   to use CACHCHK to tune some of my BIOS settings.  I can assure you
   that getting an answer out of cachkc does _not_ mean the m/c will be
   stable when running a proper OS.

-- 
Peter Hayward, Cambridge, UK.      | In a democracy the public will only
peter@cdsl.demon.co.uk             | respect laws if they respect the
tel +44-1223-312-923               | politicians that make them.

On Wed, 29 Jul 1998 15:03:45 +0100, "Andrew Taylor" <andy@hot-shot.com>
wrote:

>What about benchmarks ? I'm running Windows 98 and I've just upgraded from a
>K6 200MMX to a K6 300MMX via an Intel 233MMX. I'd be interested to find a
>*good* benchmark to see how much difference there actually is. I've been
>using WinCheckit which thinks I've got a AMD K6 601MMX and WinTune which is
>quite good but very out of date.
>
>Can anyone suggest anything ?

Wintune is useless and its "results" are a joke - try posting them and see
the reponse.  The best seem to be Winstone and Winbench for general Windows
benchmarking and Threadmark for Disk I/O and they are essentially free.


Rgds, George Macdonald

"Just because they're paranoid doesn't mean you're not psychotic" - Who, me??

In article <901680516snz@cdsl.demon.co.uk>,
  peter@cdsl.demon.co.uk wrote:
> In article <35bdc2a6.1045941@news.garden.net>
>            fammacd=!SPAM^nothanks@garden.net "George Macdonald" writes:
>
> | On Mon, 27 Jul 1998 22:58:38 +0100, "Padsu" <adam@littleangel.demon.co.uk>
> | wrote:
> |
> | >
> | >Any suggestions for thorough testing?

I have always found that compiling the Linux kernel is a fantastic test.  If
the least little thing goes wrong, the gcc compiler will generate a trap.
Perhaps another good test is the RC5 cracker at distributed.net.

FWIW, I ran the Linux kernel compile continually for 24+ hours without a
hitch, on my new 503+ and K2/300 clocked at 3*112, with 60ns EDO.

Ray

-----== Posted via Deja News, The Leader in Internet Discussion ==-----
http://www.dejanews.com/rg_mkgrp.xp   Create Your Own Free Member Forum