IBM Model 95

The first set of photos shown below represent my first attempt to address this question.

The photo above shows the fan correctly installed on the side panel of a 9595-0KF.

Here's a closer view of the fan terminals and rear mounting bracket.

To install the fan, you must first orient it properly in the rear support (the white piece with the electrical contacts).  More on this later.  Insert the front mount into the front mounting bracket, and swing the rear clip with the electrical connections upward into the rear bracket until it snaps in place.  Take care that the wires are routed so they won't get pinched or crushed, as this could cause smoke!

The electrical contacts don't always make good connection, so it's essential to verify that the fan actually runs after the side panel is attached to the computer.  If the fan doesn't run, check for dirt or corrosion on the mating faces of the contacts.  Also, check the voltage supplied to the fan.  The best way that I have found to make sure the fan is running is to place your ear against the side panel and power the machine up.  You should be able to clearly hear the fan spin up.

While you've got the fan out, clean the dust from the blades.  It's truly amazing how much a little dust can cut the efficiency of a fan!

Update, Jan. 24, 2000; the rest of the story:

9595-xNx, -xPx, -xQx, -xYx, and all 9595A Array Models

After a period of time, and much debate, it began to dawn on me that there might actually be two correct orientations for the fan.  Several independant sources have indicated that the photos shown above reflect correct mounting of the fan.  However, one person has come to me with anecdotal evidence that an IBM service rep told him the alternate method (shown in the photo below) is actually the correct method.  He has also threatened to produce the IBM documentation to support that claim (more on this as it happens).  This mounting would seem to violate Charles Lasitter's well-reasoned description of airflow around the CPU complex.  After all, why would you blow hot air on an already hot CPU?  But on the other hand, why would you place another fan to pull against the very strong pull of the power supply fan?  Which is correct?

The most convincing empirical evidence for this alternate mounting is shown in the next set of photos: the white plastic rear support from a 9595-0PT Type 4 (a.k.a. "95a") has a set of notches in it to accomodate the flange on the side fan!  The fan from a "486xp" case lacks these notches, which makes it more difficult (but not impossible) to install the fan with the narrow slot facing the CPU complex.  The photo on the bottom shows a fan from a  9595-0PT.  The top photo is of the fan from a 9595-0KF Type 1 machine.

Conclusion:  Both methods are correct!

The first set of photos is correct for the "Model 95xp" case style used for all 8595, and 9595 non-Type 4 , machines.  These machines have only one serial and one parallel port.  (I should note that these machines can house a Type 4 complex as an upgrade; check the number of ports on the back to make sure of what you have).

The second set of photos is correct for the so-called "95a" (both array and non-array) Type 4 machines with two serial and two parallel ports.  There is also a white plastic air deflector attached to the top of the power supply in some of these machines, that would seem to angle the airflow from the side fan up towards the processor complex.  The presence of this deflector may have some bearing on which way the fan is oriented in the 95a box, according to Don Peter.

 But it's not over yet...

Also interesting (although I don't know how relevant) is the slight difference in shape of the slot immediatly above the fan.  I hadn't noticed this until I had the opportunity to compare "95xp" and "95a" side panels side-by-side.

From the godfather Don Peter himself:

    I think a lot confusion is caused by misunderstanding how "tangential fans" (technoslovakian for the squirrel-cage fans) work.

If you look at Jims pictures of the fan you will notice the fan blades and their angle.  The fan rotates "from up to down" corresponding to the shown orientation. This causes air to be shovelled from the inside of the "cage" to the outside. On *axial* fans, which have an open cage to one end it is obvious - the air enters the inside and got pressed out by the fanblades. But the tangential fan used on the 95 is closed on both ends (more or less) to hold the axle.

This fan-type takes air from the "lower pressure side" (towards the top) and presses it through the cage to the "high pressure" side: that side where the deflector is open - towards the cpu complex.

Terrance fans the fire again:

Peter's explanation just doesn't hold water -" this fan takes air from the "lower pressure side" (towards the top) and presses it out through the cage to the "high pressure " side: that side where the air deflector is open- towards the cpu complex.  That is not where the air is blowing out of my fan. In the position shown in the pictures the air exits the fan
closest to the side cover blowing down, not at the cpu complex. The air is entering the fan where you see the blades in the picture, not exiting from that spot.

"Empirically Louis" Ohland's observations:

OK, propped my side cover up, jumpered the terminals, then "Power On" (a la CPT Power, good show to watch when drinking rum and coke).

Fact. The fan is installed correctly. How it can be mounted otherwise makes me cringe contemplating the brute force.

Fact. The majority (if not ALL) fan exhaust DOES go pretty much straight DOWN. (That hurt, but the hi-tech thread showed the flow)

Fact. The air is being sucked in the top and side of the fan. The thread was blowing into the blades.... The exhaust is down the side of the side cover.

Preliminary conclusion- the fan sucks air from the upper part of the case. The bottom edge of the fan housing mates up (well, close) to the underside of the complex sucking air away from it.

Confusion. The PS has no vents on the outer side of the system where the air blows down. The air has to make it's way around the side of the ps to the front and top of the ps to be exhausted by the ps fan.

My own non-scientific investigation:

I used the cigarette-smoke method on my spare 9595 today.  Most of the airflow into the cabinet seems to be through the upper full-height drive bay.  From there, it is drawn through the power supply and exhausted out the back.

There doesn't seem to be a lot of airflow into the bottom bay, but there is  signifigant airflow through the half-height bay immediatly above the FH bays.  This airflow appears to be mainly pulled across the CPU complex and into the side fan, with some being pulled into the power supply as well.

This experiment was intended only as a quick and dirty approximation of airflow around the CPU complex, and is not menat to be a definitive answer or the final say as to what is actually happening.  The addition of drives or drive blanks to the drive bays will almost certianly result in more balanced airflow

Charles Lasitter thoughtfully adds:

While I find much of this analysis thoughtful and well considered, I think this matter of circulation within the case is troubling because of a paradox that beats scientists with some frequency.

That is, if you remove something from it's environment to study it and find out how it "really is", then by so doing you kill it and discover only things that are untrue or misleading.

It's the "fish out of water" syndrome all over again.  Want to study a fish by taking it out of the water and lay it on the
dock?  Soon it will be studying a dead fish, and little about it's nature IN THE WATER will be revealed to you.

The cooling system of the case is a SEALED SYSTEM, operating properly only when correctly sealed.  Beyond being sealed, it is compartmentalized as well.  The compartment seals don't engage until the side access cover lid is closed!

The compartments and the seals, and how they work in conjunction with the side louvers tell a great deal of the story.

Look at your side access covers folks!  Look at every piece of steel or foam that's perpendicular to that nice big almost square flat surface.

Take a close look at how they interlock with the drive support structure immediately below the FDD's & adapters to keep the hot air from that area right where it is.

You've got several air movement patterns in this case, and the IBM engineers spent a lot of time making sure that they didn't work at cross purposes to one another.

Up at the top of the case, air passes thru the inlet louvers and goes straight back over and thru the adapters.  Why?  It has no choice!  Turning downward it's blocked by the flanges from the drive support structure so that it can't escape around the sides of the "big plastic" and satisfy the pull from the PSU exhaust.

After making a trip down the length of the adapters, the air moves into a void between the foam and the side access cover.  It moved in here because this is mostly a low pressure area.  Why else would it come here?

In this area it is exposed to an enormous surface area of the side access cover fan.  It gets pulled thru the fan and exhausted above the PSU, where the hot air turns a corner to seek out an even larger low pressure area created by the PSU.  The evacuation of the hot air is very slick at this point, because it comes across one side of the PSU and around the corner without coming anywhere near (and thereby heating up) the SIMMs or the HDDs.

Now, for the other major air circulation area.

The drives are cooled in a fairly straightforward way, as air is drawn directly across the drive surfaces into the waiting vents of the PSU.

But to make sure that TOO MUCH air did not take this route, IBM supplied each unit with two blank HDD trays, which did a fantastic job of filling the 2nd HDD bay.  This forces inlet air to take a different route.

And what route would that be?  Passage thru the top HDD / half-height bay, which just happens to be -- Ta Daa!! --
immediately opposite that smoking hot CPU.

The airflow runs along the length of the processor board being exposed along the way and eventually taken up by the side cooling fan and the incredibly long expanse of louvers on the PSU itself.

If cooling the CPU was your greatest cooling concern, and you weren't even going to equip that cpu with a CPU fan, why on earth would you blow how air on it from above that had just come from all those toasty warm adapters?

Other proof of air movement near the CPU:  Which way do the fins on the CPU point after it's installed on the processor board, and the processor board is installed on the planar?

If the side cooling fan were blowing down on the CPU, and then that air were moving across the CPU and out the PSU in a straight path that way, the fins should orient in the opposite way that they do now.

Other observations:  The "slot" next to the side fan.  Which way does the air move thru this slot?

Take a look at the inlet louvers again, and where the slot is relative to the processor board.  Finally, look at the location
of the foam barrier relative to the slot.  My best guess is that air moves UP the back side of the processor board, thus cooling it, into the low pressure area on the other side of the slot, then being shot across to the PSU.

The plastic flange dividing upper and lower drive support structure occurs about the level of the processor board.  Some
louvers are immediately adjascent to this area but can't "communicate" well with the main PSU, exposing them to the pull from the side fan thru this slot.  To intensify this pull, it appears that the engineers have made this slot as long and wide as possible.

Joe Kovacs enters the fray with an interesting bit of trivia:

At Comdex in Las Vegas a few years ago, an IBM engineer who worked on PS/2 design told me that cooling and air flows are of major concern in every design.  They design everything for a _reliable life of ten years.

And it continues...

(Jan 16, 2000)
I hate to bring this up again.  I really do.  The side fan (in my newly acquired 9595-0PT) was mounted with the small slot aimed towards the complex.  This machine is actually in pristine condition, and was BIOS level 01, so I doubt it was fiddled with much.  Truth is, it looks like it could have been a spare.  Added to that, fellow auction hound 'Cigar Steve' tells me that he cornered an IBM service rep recently who is intimately familiar with model 95's, and he said the proper orientation is to have the small slot pointed at the complex.  Steve claims to have an IBM hardware manual that pictures this; I will browbeat him until he brings it over for me to scan.  I'm starting to wonder if the orientation was changed at some point by IBM, and each way is correct for a different type of complex...

From Peter Wendt, Feb. 5, 2000:

The "fan problem" is not yet fully solved. My own T4 P60 9595 for example has
the fan orientated in the "old fashioned way" as for the 8595. The fan
retainer however has the notch, which allowes the alternative fan orientation
as well. Recently I disassembled a T4-P66 machine at a customer, which has
never been taken apart since it was installed in '93: and it has the fan also
installed in the "old" way. So I really wonder what that notch is for. Guess
that requires more research before we finally can put down the quest to an
end.

More from Peter, Nov. 5, 2000

To be precise: all my Type-4 systems I have around here (and all I know of) have
the sidewall fan installed in the "95XP" orientation.

However: none of these have the plastic air-baffle installed on top of the power
supply. So my best guess is, that there might be a relation between "95XP
orientation / no air-baffle" vs. "Non-95XP orientation / air-baffle".

The air-baffle came with the 95A "Array" models with the 2 x 3 drive cages and
factory installed CD-ROM. So there might also be a dependency, which I cannot
figure out at that point.  (Ed. - non-array '95A' boxes have been sighted in the US
that have the air baffle; I have one, a 9595-0PT).

The orientation on the other hand is of lesser importance as long as there is a
sidewall fan in a Mod. 95 at all. The T4-platforms definitely need one - the cache
chips, the memory controller, the CPU no less and the "companion chips" on the
platform emit a lot heat. From the "smaller" platforms the Type-3 DX50 platform will
positively die without a sidewall fan. Had that once.

Al Brandt chimes in, Jan. 2, 2002

No one knows for sure which orientation is correct for the side fan. A
short while back I came across a late-model 9595-SOC. I opened the
case and low and behold there was a thin white sticker from IBM on the
fan stating that "this sticker must be visible if the fan is in the
correct orientation" (or something of the sort).

The fan is in the orientation pictured on Jim's page for the 95 XP.
The sticker is on the small strip of metal on the bottom facing
forward. It wouldn't be very visible any other way.

I'm glad I'm somewhat useful.

Very useful.  Thanks, Al!

More trivia, from Al Brandt via comp.sys.ibm.ps2.hardware:

The manufacturer of the lock cylinders for the Model 95 is:

    Rexnord Camloc
    601 Route 46 West
    Hasbrouck Heights, NJ 07604
    USA

And, for hardware hackers:

Convert your 512k cache SCSI controller to 2 MB with the "Alfred" Mod!

    Here's how a pair of SIMMs I converted look....