Re: Odd crash problem

The message <oCMZU0Bn+0fGFwfU@xxxxxxxxxxxxx>
from Mike Tomlinson <mike@xxxxxxxxxxxxx> contains these words:

In article <3130303037373036467F0F8211@xxxxxxxxxxxxxxxx>, Johnny B Good
<jcs.computers***@xxxxxxxxxxxxxxxx> writes

In retrospect, despite the potential benefit of the larger slower
spinning fan, it was no surprise that it was less efficient than the
half speed rear panel mounted 80mm fan configuration it had temporarily
replaced.

That's interesting. It makes me wonder whether the power supply makers
actually model and/or test the airflow in these designs or whether they
just build 'em, call 'em good and ship 'em. Some Compaq SFF models have
a PSU with the fan on a corner. Those are proprietary models, of
course, and HPaq is a company with the kind of resources necessary to
design and test such a unit. They also built machines that had passive
heatsinks on the CPUs (at a time when everyone else was fitting active
heatsinks) by dint of designing cases with proper airflow from the
single fan in the PSU. These, of course, didn't conform to the ATX
layout. The CPU tends to be at the front of the machine rather than the
rear, making use of cool air drawn in via the front vents.

I think the psus with interior mounted fans work best when the fan is
mounted 'outboard' of the psu casing. This reduces the resultant
turbulence (and loss of efficiency) somewhat.

The small 150 watt unit in the Gateway 2000 case that houses my FreeNAS
4 disk drive server has its interior mounted 80mm fan mounted 'outboard'
of the psu casing (right next to the CPU socket, as it happens :-). The
only mods I did with this was to add a 3.3v zenner to the fan speed
control cct to raise the minimum fan speed a little and twist the
exterior vent slots to about a 45 deg angle to reduce turbulence and
increase the effective CSA.

This fileserver origionally had a super socket 7 MoBo with a K6/2/500
underclocked to 250MHz and undervolted to 1.35v to keep the whole power
consumption down to 50 watts. I had to use the spare KT600 MoBo and
XP2500+ cpu as an upgrade on account the Gigabit network adapter upgrade
was being throttled by that old MoBo.

I've configured the CPU to run at 665MHz instead of the normal 1836MHz
and this has allowed me to set the minimum core voltage setting of 1.1v
(even this amount of cpu 'horsepower' is gross 'overkill' for a
fileserver box). In spite of the huge drop in cpu power consumption, the
box now uses 20 watts more power, all for the priviledge of getting a
useful boost of network speed out of my Gigabit upgrade.

Although you might consider a small 150 W ATX PSU to be marginal for a
box with 4 hard disk drives, it seems to cope just fine in spite of the
fact that the spin up surge at startup does top the 150 watt mark.

I origionally picked the Gateway 2000 box to house my fileserver on
account of its more efficient and remarkably quiet PSU. This box is
still so quiet that you need to place your ear within inches of the vent
slots to hear anything at all whilst my reasonably quiet desktop machine
is running. In the dead of night when everything else is switched off,
it's just possible to hear it running if you stand close and 'listen'
very carefully'.

Mine doesn't employ such a quick release arrangement but, in any case,
I prefer the more solid directly mounted fastening arrangement over the
quick release designs which tend to be rather flimsy ime.

I get the impression that noise is a big issue for you. It's not for me
- I have a study in which my PCs live. The main one (one I'm typing on
now) is noisy, yes, but it doesn't bother me. Horses for courses, etc.

You might think that a constant background noise is of no significance
but, if you feel any sense of relief on shutting your PC down at the end
of a session, then it's significant enough to be a source of stress that
warrants some attention.

Even as quiet as my own PC is, I still feel a _mild_ sense of relief
when I shut it down at the end of the day (often, this is in the quiet
hours of the morning, around 3 am, so the contrast is magnified by the
almost utter silence of my nearly silent fileserver).

To give you some idea how quiet my PC is, I can tell you that even the
quieter examples of customers' PCs usually totally drown out the noise
it generates. I have on very rare occasions had the pleasure of working
on extremely quiet machines which don't drown out the noise from my own
PC, but this situation is exceptionally rare.

Many of the PCs I have in for repair have been so noisy, I wonder how
their owners can tolerate them for any lengh of time, especially if I've
had to run chkdsk /r from the recovery console for the, typically.
couple of hours this usually requires. When I've finally finished
working on _these_ machines, it's an absolute joy to shut them down and
return to tranquility once again.

The main source of the noise is two 80mm fans at the top rear of the
case, above the PSU. The 4 hard drives are mounted here and these two
fans draw air across the top of the machine from front to back to cool
the drives. That's the reason I chose this case; it's an AOpen HQ08, a
compact full tower.

True, but it was simply a source of irritation that the three dimm slot
full ATX design imposed such a reduction in performance with merely a
pair of dimms over the single dimm configuration.

Are you referring to the way PC3200 is downclocked to PC2700 if two
DIMMs are fitted? If so, I assume you're saying you tried various
brands of memory until you found two that would work reliably with
DDR400 forced in BIOS?

It wasn't the clock speed being reduced, it was one of the other timing
parameters that created a reduced "benchmark" reading in memtest86's
information panel.

Why not just fit one larger DIMM? IIRC, you said you're using an XP-
M2500 (nice chips), so you don't have dual channel.

I don't think a single 2GB dimm would necessarily get round the
problem, plus, AFAIR, 1GB is the limit per slot. Even if a 2GB dimm was
an option, I still wouldn't choose to invest in what, even by then, was
becoming an obsolescent technology.

Although water cooling can be very effective, when you consider the
growing list of chips in need of a water cooling block and the ensuing
complicated plumbing _and_ the fact that you still need some airflow to
keep the low grade heat sources and the PSU within safe temperature
limits _and_ some means of cooling the water, you have to question the
escallating costs involved

Indeed. Water cooling belongs to the modder community, I feel, and
isn't viable for mainstream desktop PCs.

True, but this is uk.comp.homebuilt so an appropriate comment.

and consider whether a better solution would
simply be an air conditioning unit for the room the PC lives in

But what about your Carbon Footprint (TM)? And the ickle bunnies and the
dolphins! And the childruuuun! You'll have the green contingent foaming
at the mouth with comments like that.

I'm nothing if not pragmatic. :-)

Then again, people tend to forget that a PC left running saves on the
cost of heating the room it's in during the colder months. If the
radiator in that room has a TRV (thermostatic valve), it may even not
come on except in the very coldest weather. With 1Kw+ PSUs now
available (rolls eyes)...

That might be the peak (or, being generous, the short term max loading)
rating but I doubt such PSUs could sustain half that loading
indefinitely or even that any PC system box so fitted would peak much
above 300 watts.

Any PC that demands more than 200 watt will be quite noisy even when a
'quiet' design of fan cooling is employed.

and a
case designed with proper airflow based cooling in mind.

I think the ATX board layout limits the ability to improve the design
much. BTX was an attempt to improve that, but seems to have vanished
without trace. Wonder why? I have a memory that Intel demanded licence
fees for BTX which the case makers didn't want to pay, but may be wrong.

The trick to good reliability, in a conventionally fan cooled design of
PC, is to have a large intake vent slot area and arrange that the hard
disks containing your priceless data have first sniff of the incoming
cooling air

I believe cooling for hard drives should be considered in its own right
and not as a by-product of cooling of the rest of the machine. This is
particularly important as drives increase in capacity and running
temperature. A reduction in temperature of only a few degrees can have
a massive effect on reliability and longevity of hard drives, and can be
achieved with modest airflow - it's simply not necessary to have a
howling gale blowing on the drive(s). The trick is not to blow air onto
the top or bottom of the drive, but _across_ it, so that it carries away
heat radiated from the larger surfaces.

Unfortunately, many cases encourage the practice of jamming drives
together in 3.5" bays with little or no space in between to allow heat
to dissipate, so they simply cook.

You're quite right about considering the cooling requirements of the
hard disks. A lot of the 'traditional case designs simply leave this to
chance and, as you mentioned, compound the issue by having hard disk
drive bays that force the two drives into very close proximity.

I agree that you don't need massive flows of cooling air to keep a
drive comfortably cool, just at least 2 cms of airspace both above and
below each drive for the air to flow, unobstructed past each drive.

The four drives in my file server rely purely on the air passing them
from the ventillation holes drilled into the front of the metal case
through to the CPU heatsink fins and into the PSU fan at the rear of the
case.

The original hard drive bay does present a problem in that it does
stack the two drives a little too close for comfort (even after modding
it to gain another couple of millimetres seperation) and I've had to
mask the vent holes near the other two base mounted drives (on MoBo
stand offs) with cellotape in order to balance the airflow for more even
cooling.

Even so, in spite of this, the drives manage to stay at less than 12
deg C above room ambient which is good enough to allow the fileserver to
remain safely operational at up to a 38 deg C room temperature (my
desktop is rated for a 40 deg room temperature limit at 100% cpu
loading).

Reducing temperatures and noise need not involve the expense of a new
case, if you're prepared to hack away at your exisitng case with a pair
of tinsnips and make ducting out of stiff card stock.

However, new cases are very cheap nowadays, even branded ones. You can
buy very good cases with decent cooling and features designed to make
life easier, not just bling aimed at the modding community.

The problem here is that it's not always obvious which cases are
reasonably well designed from a cooling and noise point of view until
you've actually got one set up on your workbench.

Spending a bit of money at the outset and buying a decent case that will
last you over several upgrades of the internals is the way to go IMHO.
I've lost count of the upgrades I've done on my main machine, but the
case has outlasted them all. And it still looks new and fresh; the
plastics haven't yellowed at all.

Coming back to the original subject of the failed memory, I have
experimented with heat spreaders on the memory fitted in this machine,
and found that it ran much hotter. Whether this was because the metal
spreaders tended to "collect" heat from the CPU HSF, I'm not sure, but
the memory certainly runs cooler without the spreaders fitted. I didn't
carry out exhaustive testing though; I don't like plugging/unplugging
memory repeatedly as it weakens the DIMM socket contacts and reduces
reliability. This box, despite running Windows 2000, can easily achieve
uptime of several months (forced to reboot by Windows updates), so
you'll gather that reliability is an important factor for me.

I think reliability is an important factor for _all_ users (even if
it's only FSVO in the case of gamers resorting to extreme overclocking
measures :-)

We obviously agree about the importance of employing effective cooling
measures in a PC. The difference in our attitude to this problem seems
to be that you're not prepared to compromise merely for the sake of
addressing noise issues, whilst I'm prepared to take the trouble to find
the quietest possible way to get the job done effectively enough.

--
Regards, John.

Please remove the "ohggcyht" before replying.
The address has been munged to reject Spam-bots.

.