CT60 - A Heated Question?
There's no doubt that Rodolphe Czuba's CT60 Falcon accelerator is a well thought out product which works admirably well with the 1993-released Falcon030. There is also no doubt that not all CT60s, or Falcon030s for that matter, were created equal.
Rodolphe did a hell of a good job making a high-powered upgrade that is able to blend in with old hardware, and cope with a range of differing hardware tolerances dictated by Atari's notions of quality control.
Especially where they chose to source the cheapest job-lot components available at a given time.
For various reasons, there are a lot of variations on the basic CT60 theme. The default specification, CPU-wise, allows for a 66 MHz 68060, based on the Revision 1 68060. This is the maximum to which the Revision 1 can be safely pushed.
There are also a smaller number of higher revision numbered 68060s in some CT60 boards, culminating in a privileged handful of the rare 100 MHz capable Revision 6 CPUs.
On the other hand, a small number of the second batch of CT60 boards were throttled down to the default 50 MHz, as Rodolphe was ultimately unable to get hold of 68060s of sufficient quality to overclock even up to 66 MHz.
For my own situation, I've got a first batch CT60 with one of the average boards sporting a Revision 1 CPU. It is clocked to 66 MHz. This was tested by Rodolphe before coming to me, as all CT60s were, so it works correctly.
From my own experience, I'd say that at 66 MHz, the 68060 in mine is right on the edge for safe performance in overclocking terms. At the same time, it was not subsequently installed in a totally ideal environment.
It lives in a standard Falcon030 case, which is based on a standard ST case
with a few new holes in the back, as we all know. Now engineering an upgrade
of that level of power, which is able to be shoehorned into a miserly case
like that, is a huge achievement. Lyndon Amsdon (Stimpy to his friends) also managed to make it sit comfortably in that case. To do so, he recycled the original 40 mm Atari-supplied cooler fan, rather than fit the 50 mm fan supplied with the CT60.
So we have a situation where the CPU is slightly under-cooled from what Rodolphe recommended. I'd be interested in someone more technical who can tell me a bit more about the relative cooling efficiency of 40 mm versus 50 mm CPU fans? (I
think the bottom line is look for the air-flow
measurement for any given fan, it will vary
even between fans of the same diameter as a
function of specific blade geometry and RPM - Tech Ed).
Then we compound this lower cooling performance with a confined internal air space in the tight standard case. Thus any build up of heat is not able to disperse too easily.
I've become a CPU temperature obsessive, one of those sad cases who takes a morbid interest in the Pentium-like behaviour of their hardware. When left to its own devices in normal ambient room temperature, the CPU typically peaked at 41-43 degrees Celcius. Now for most situations, that wasn't a problem. However, one of my major software players, MagiC 6, proved to be fragile under those conditions, tending to collapse in a sweaty heap of exception errors after an hour or so from first switching on.
This became an increasing problem with the hotter days of last summer, as the time from first switching on to MagiC 6 exception error failure decreased greatly. It even got to the point that the CPU temperature hit 47 degrees Celcius in high ambient temperature heat wave conditions. At that point, the
errors and crashing were far more frequent. This even started to contaminate matters in plain TOS mode. It was clear that there was an overheating problem with this CT60.
So it was time to do something about it, and short of a complete recasing I looked around to see what was to hand. Now what you are going to see below is a classic bodge done in a hurry. It consists of a room heater, switched to a temperature neutral air blower mode, with an extremely crude air redirection! This is set to blow through the part of the casing where the CT60 is mounted. There should be an improvement in cooling, partially through a direct cooling effect on the whole board, but also partially from flushing out the build-up of hot air inside the case.
This was done from the materials to hand, which happened to be mostly cardboard and Sellotape!
In spite of the crude and gimcrack nature of this so-called "engineering",
it did make an immediate difference. In normal ambient conditions, the
internal CPU temperature went down a few degrees, typically pegged back to
around 35-38 degrees Celcius. This had an immediate effect, as MagiC
stopped falling over. It seems that the "safe" working temperature for it was
just a little bit below what it had been running at previously.
I stuck with this solution for a while, after all, why fix what isn't
broken, but pending any recasing I eventually decided on a slightly more
elegant version of what I had already done.
This took the form of a silent 120
mm fan, purchased new from the local
friendly PlasticWorld retail barn. This was intended to be mounted in a PC
casing originally, but in my hands it has ended up with a more exciting
fate. I sat it on top of the section of outer casing with the CT60 in it. It
was a simple enough matter from there to tap into one of the spare power
lines coming out from the ATX power supply, and off we went.
Still a bodge, but a neater and quieter bodge!
In operation, this fan has proven to be more efficient than the adapted air
blower, by managing to nudge the CPU temperature down to a range of
31-34 degrees. It might be interesting to see if it can keep things
reasonable when the really hot weather comes along? Confidence levels are high.
The success of this method of cooling opens up some possible suggestions
for people with CT60s in standard ST-style cases. If you can bear the idea
of mutilation, airflow efficiency could be even more enhanced by cutting a
hole in that part of the case. There could also be an extended mounting for
the fan a little way further up to form a space for future add-on cards like
EtherNat and the SuperVidel.
As a final point of interest, I've found that the CPU temperature reading is very
sensitive to changes in the ambient room temperature. I can even get a
drop with something as simple as opening a window to colder air. Maybe it
is a good job that Rodolphe is leaving off the temperature gauge on his latest proposed CT63 third batch? You can spend too much time playing with this temperature thing!
Now it's your turn! Fellow CT60 owners, why not relate your CT60 cooling
experiences to the MyAtari team? What unusual methods have you used to get
things running smoothly? And Rodolphe, Didier, and the other clever engineering people, what are your thoughts on this subject?