CTCM and Rev 6
060
CT60 Turbo-charged!
CTCM and revision 6 considered
or CiH and gwEm write in a Chaos Manor style!
Bless Didier and his restless mind. He's always trying something new. The
CTCM seemed like a good idea even before Rodolphe got his hands on a new
supplier for the revision 6 100mhz capable 68060. The original attraction of
the CTCM was that it allowed for detuning, especially if you felt you were
running right on the limit. Also it is good to have the option as a
safeguard in hot weather. You will recall from previous articles how I got
round the problem with extra cooling. The CTCM would add an extra layer of
insurance. Shortly after, Rodolphe came up with the new batch of revision 6
cpu's, so the possibilities then became very exciting.
For those people not following the CT60 story too closely, the CTCM replaces
the fixed frequency oscillator clock, allowing you to adjust the speed of
the '060 cpu on the fly in tiny increments. I doubt if you can do with too
many other systems, Atari and Non-Atari computers alike. Traditional peecees
were limited to the either/or crudity of the misnamed 'turbo' button, but I
doubt if there is any call for a 8mhz MS-DOS 8086 mode anymore!
The CTCM is controlled from the CT60 config cpx, with a slider bar adjusting
the processor megahertz. You can have a huge range of options from 33.3mhz
as an extreme 'safe mode' operation, all the way to 110mhz for the most
robust revision 6. In practice, those people still living with revision 1
processors, are rather unlikely to reach the top end figure. Those people
who invested in the upgraded rev 6 cpu have more scope to experiment.
The benefits are virtually self explanatory. Every Falcon and CT60 have
their own individual tolerance limits and quirks. The CTCM allows you to
find your own safe limit for processor speed, specific to your set-up. The
revision 6 benefit needs even less explanation, effectively providing a
hefty leg-up from a not unacceptable 66.7mhz anything up to the magic
100mhz. So I didn't find it too hard to say yes to Rodolphe's offer.
Presently these arrived, the rev 6 with a heatsink and low profile 40mm fan.
There was a minor glitch when the heatsink and fan didn't arrive in the same
package as the cpu, but Rodolphe soon sent these, with some heatsink glue to
attach it all together.
The fan bolts directly into the heatsink, and can be swapped out if needed.
The rev 6 did not look pristinely new. I guess these were extracted from
some obsolete formerly high-end boxes, but thoroughly tested to make sure
they were okay by Rodolphe of course :-) The CTCM was more of a revelation,
being a tiny half-inch circuit board, almost overwhelmed by the polystyrene
block in which it was embedded.
Yes this is it! The CTCM close up.
It wasn't quite a case of 'plug and play', as there were a couple of
additional wires to be soldered on the underside of the CT60, and of course,
the desocketing of the existing rev 1 cpu. Happily, this coincided with
Gareth 'gwEm' Morris getting a CT63, and needing a cpu to go with it. So we
arranged for a visit to gwEm ghq in deepest London to sort out both
requirements, him to grab my old cpu and me for fitting the new one, for one
fine and sunny Sunday morning.
Apart from when I drove through what looked like a CSI style crime scene in
a narrow alley-like street in central London, the journey was trouble-free,
and I got to gwEm's studio flat in Shoreditch. Only pausing for croissants
(thank you Manou!) and also to wonder at what looked like a "Dance dance
revolution" inflatable piano keyboard rolled out on the floor, we got down
to work.
The changeover was straightforward beyond belief. This was the easiest
upgrade that I have ever had. Firstly gwEm extracted the old cpu, and
replaced it with the rev 6. We managed to lever it out without the cpu
removal tool, and had to re-align the heatsink fan as I had stuck on the
heatsink with the wires slightly away from the connecter. Apart from that
there were no problems. The CT60 with its newly acquired revision 6 cpu was
then booted up. The boot-up info duly told me that there was now a rev 6 in
there, still clocked at 66.7mhz, but of course there was still some more
work to be done.
Then the heartstopping moment comes, as gwEm carries out the requisite
under-board soldering to allow the CT60 to recognise the CTCM. Not long
after that, the newly altered board is back in my machine. We were
exceedingly lucky, as the fitting instructions I obtained in a hurry, didn't
include a photo of how the CTCM should be mounted on the CT60. Apparently
doing this wrongly has nasty consequences for the CTCM! However, we did have
gwEm's CT63 with its CTCM already installed, as a useful reference tool.
Once it is all back in place, we switch on again, and we're off! Well, apart
from the small matter of flashing the most up to date CT60 TOS file, and
replacing the CT60 config CPX with the latest version. A further reboot
later revealed a new start-up sequence, with a choice of o/s (TOS, Magic or
Linux), then the normal boot-up sequence took over. The next logical step
was to explore a radically altered CT60 config cpx. A slider bar for the cpu
was found under the memory options menu, and with some excitement we started
to play with it. I opted for a 'safe' 90mhz, after saving this setting,
rebooted again, and immediately, I got 89.9mhz for the processor speed. Also
the ram access speed had increased dramatically to some 93-94 mb/sec in line
with the processor boost. We also got the data transfer speed of the hard
drive reported to us which is around 4.4mb/sec.
I left this sitting quietly, after the obligatory playing around with the
small catalogue of CT60 demos for a while, and it just minded its own
business, without a hint of falling over. this hardly constituted a major
test, but was an encouraging sign. gwEm was fitting his CT63 whilst this was
going on, and he was a bit less lucky there, as it booted, but proved to be
massively unstable afterwards. This was fixed a couple of days later, after
an exchange of emails with Rodolphe Czuba cured a clash with the pre-
existing Centurbo 1 card in there. Now they are both running happily
together, father and baby are doing well!
Whilst my machine was opened, we also sorted out another longstanding issue
with the 3.5mm headphone audio/out jack. Mr Pink will recall this one well
from the last Outline party! This had never easily given its optimum sound
quality since day one, so I decided to get it replaced. gwEm had sourced
replacements, but it seems that there was a badly done solder which was the
cause of the problem all along. We replaced it anyway, and that has been
fine since then.
I tweaked the speed a bit further at home to 96mhz, I'm waiting for that
special occasion to go all the way to 100mhz! UPDATE! - It does not seem to
like 100mhz, as it crashes as soon as it boots up. It was also not too easy
to get it back from there. After ten scary minutes, I found out that the
CT60 conf CPX works as well in '030 mode, and I was able to adjust the CTCM
slider bar back to a safer limit from there, save the new setting then
switch back to CT60 mode. There were sighs of relief all round when it
started working again! It should be possible to get closer than 96mhz, Maybe
97 or 98mhz is possible, but I'm waiting for the day when I'm brave enough
to try!
There is some debate on the CT60 bbs about this topic. It seems that not all
machines effortlessly get to 100mhz. Some have problems with different ram
cards, there is some debate but no consensus over whether there could be a
problem with cooling on some systems. Others suggest the CT60 ram controller
as a potential culprit, or some other feature of the CT60 design, or even a
combination of all of the above. There is even a school of thought that
argues a replacement power supply can help, by replacing a worn one with a
new one to improve stability.
On reflection, 96mhz works for on mine for a substained period, but with a
slightly higher risk of some more fragile software crashing than would be
the case at 90mhz? I've got to double check this, but the GemAce replayer
seems to be a good 'control' for this sort of test.
With what is effectively a new machine, by the time the firmware has been
updated, there are some new bugs and quirks to get used to. For example, I
need to remember to set up the boot screen if changing from a VGA to a TV
screen. When it first boots up, there is also a mysterious keyboard failure
message right at the start (which didn't affect anything subsequently.) Also
when transferring back to my LCD monitor using Centscreen, the extended
screen mode needed redoing, as this displayed slightly 'differently' from
before. The top part of the screen was creeping in at the bottom! After a
little fiddling on the margins of its vertical hz tolerances, I was able to
get the display back to how it was before.
The operating temperature for the new cpu ranges from 37 to 41 degrees
celcius, depending on ambient room conditions, and using the same casing and
cooling arrangements from before. Continual running tests seem to indicate
no problems with stability. An extended test at 90mhz and Magic 6 with
various apps tried, at a warm ambient room temp showed continual safe
operation at 41 degrees c. This is the point at which my Magic 6 desktop
used to fall over on my old cpu.
Also there is a boost to the MIPs figures, showing 180.15 mips at 90mhz, The
Zbench performance increase for the cpu is in line with the speed boost
provided, that is a 40 to 50% over what came before. There are greatly
improved figures for the graphics and VDI, over 50% in some cases, but using
a dissimilar display from before. In this case, an RGB screen set for
truecolour mode. Particularly impressive is what happens on the realtime SDL
raytracing, as this is now displaying completely smoothly in realtime.
What use is it apart from having the wow factor of a faster machine? There
is some benefit in the multimedia area. I managed to get better playback on
movies, with 320 x 240 mpeg and Quicktime just about there. Again a large
part of this is due to a favourable RGB and truecolour mode.
With that, the improvement is better than 50% from before. It looks like bus
speed and the Falcon videl is even more of a factor than realised. When I
switched back to an Svga screen, much of this improvement was lost. One
Quicktime movie running at 112% in 256 colours showed no improvement even
when going to 90mhz! Even on Truecolour mode, video screen size makes a
difference, with the same movie playing back anywhere between 65% and 87%
for 340 x 480 and 640 x 480 mode respectively. It is increasingly apparent
that a graphics booster is moving rapidly up the things to do list!
It's a minor detail, but there is some small improvement on streaming audio
with a 128 kb/sec mp3 file, down to around 60% cpu use with no dsp and with
Aniplayer set to good quality mode. The equivalent mp3 replay with the dsp
running, was typically around 5% of cpu time.
I tried out Didier Mequignon's port of Mplayer. That is the subject of its
own review in this issue. In brief, even a boosted CT60 isn't really fast
enough for this unoptimised straight Linux portover.
The snazzy image gallery browser zView, in it's 6th beta came along at the
right time. With an RGB display, it shows an impressive speed-up for
thumbnail generation, and loading times in general, over how it was before.
Again, with Svga and 256 colours, much of the speedup is lost! It is still
quicker than before. As with the media players, it will be interesting to
see how this type of application will benefit from a graphics card in the
future.
And we're not forgetting those SDL games which have been ported by Patrice
Mandin.
Doom was playable, not too bad for speed, but no faster than when I last
played it on the CT2. I think bus speed is an issue here as well, with the
CT2's 25/50mhz bus proving superior in that respect.
Hexen was playable, but not really a speed freak.
I finally got around to sussing out the Doug Little 'DougQuake' engine. At
90mhz, it runs at anything from 6 to 20-odd frames per second, depending on
how busy the screen is. A typical average would be in the order of
12-13 fps. DougQuake feels perfectly playable on here.
The 'Yeti 3D' Amiga ported Quake style engine from Gildor, is definitely
improved from before. The 240 x 180 mode is practically 50 frames a second,
whilst the 320 x 240 mode whilst not inadequate before, was quicker as well.
Whilst considering what Gildor did for the rest of us, I also played with
the SDL Wolfenstein. This was always fine before, now it was creamy smooth
and lightning fast. I can personally tell if the 3D is too quick, as early
symptoms of motion sickness, recalled from early PC Doom deathmatches, start
to appear!
The demos made to date all worked. I didn't notice too much difference for
the most part, but Deez's 96ktro 'Dazed' seemed to have faster and more
fluid generation and dispersal on the bubble stream screen. I also played
with a few non-CT60 demos. There were various outcomes, usually good. But
the T.O.Y.S 'Wait' demo tends to break with the extra speed! The parts of
'Revert' that chose to run were finally at their proper speed!
There was one application of potential interest. This is the multi-arcade
emulator Xmame. Probably the most ambitious thing that Patrice Mandin has
ported so far. Evil stepped in to do some beta-testing, and described this
as unplayable even on a CT60! He suggests that XMame was no use on sub-
500mhz machines, which would limit it to Aranym and higher end X86 running
JIT routines. At the same time, I seem to remember something like Mame
running higher end arcade games quite well on Wiztom's Pentium II Peecee,
back in 1998? Is this a bloatware syndrome, even getting into freeware and
open source? Or more prosaically, a question of native intel code simply
working better?
The news was better on the other emulations which work on the CT60. Both the
Gnuboy and Atari 800 emulators clearly benefited, with the GNGB running
decently. And I had the heartening experience of getting the awesome 'Numen'
demo from Taquart to run through without any problems. So that makes this 8-
bit demo the most asskicking thing about the CT60 at the moment! Now there's
a challenge to beat!
Conclusion - Most trouble-free upgrade I've had. Full benefits still await
further add-ons, such as the EtherNat, Supervidel, and CTPCI.
CTCM - rev1 68060
=================
gwEm's Bit!
A CTCM module is by default fitted to all CT63 boards, including mine. To
reduce costs I opted to order my CT63 without a CPU, and sourced my 68060
myself. CiH offered my a good deal on his old rev 1 060, I'm not sure if out
of charity or as a motivation for me todo something with it :) Anyway,cheers
mate, maxYMiser is now CT60 friendly and I did some music for a CT60 demo,
so hopefully this is some recompense for the nice price.
The rev1 68060 was mainly reliable at the default 66MHz but after a few
hours it would become unstable, and I generally got the impression, as did
CiH, that this particular 060 would be happier at a slightly lower clock
rate. This part of the review focuses on my experience with CTCM and a rev 1
060.
The current version of CT60 TOS does supports rev 1 060 very badly with
CTCM, but CT60 TOS maintainer Didier kindly sent me a preview of the next
version which is significantly better, and this review is carried out using
this up coming version.
The supplied CTCM module offers two clock ranges for rev 1 CPUs - 66 to
75MHz and 33 to 66MHz. Thoughtfully the CT60 configuration CPX detects a rev
1 060 and restricts the upper CTCM range to a safe 75MHz.
Sadly the programmable clock generator in the CTCM does not allow a more
useful 50 to 75MHz range, which would be ideal for a rev 1 processor.
Working in the upper range I managed to achieve a relatively stable
overclock to 68MHz, however this was simply a fun exploration and I knew
that something like a 60MHz clock would be more appropriate to the
processor.
Cutting a trace on the CTCM simply and easily switched to the lower clock
frequency range. However, my machine operates with a speeded F030 bus, and
refused to boot. A discussion with Didier revealed that for the lower CTCM
frequency range and a speeded Falcon bus there are some special
considerations.
At boot up the machine boots with the default CTCM frequency (either 33 or
66MHz depending on the range). The CT60 TOS boot then reprograms the CTCM
for the requested frequency in the CT60 configurator. This is done before
the CT60 super speeder graphics appear. The CT60 will not work if 68060 freq
< 2*F030 freq. So for the lower range, and a speeded Falcon bus the default
CTCM frequency must be changed to ever get to the CT60 super speeder screen.
Luckily, the CT60 configurator provides this option by 'hidden' double click
option on 'save' button. This is simple to setup, and is no hassle at all.
Using a default CTCM frequency of 55MHz, and a usual CTCM frequency of 66MHz
the machine booted as expected even with the speeded bus.
Bizarrely, reducing my 060 frequency to much below 64MHz with the F030 bus
speeded resulted in error ridden IDE transfers. I suppose I am the first
user to try this, and I obtained special permission from Rodolphe to replace
the CTCM with a fixed oscillator. There was no improvement, so this is not
a problem with the CTCM, and so I am now running my machine at 64MHz.
The CTCM seems not quite so useful or essential for rev 1 '060 as it is for
rev6, but it remains a very worthwhile addition to the CT60. Unlike the
upper limits of the rev 6 '060 overclock range, a few fixed crystal
oscillator modules are available suitable for rev 1 overclocking. The
frequency ranges of the CTCM are also not completely ideal for rev 1 060.
Together these things reduce the percentage slightly.
A special mention has to go to Didier for his fantastic support regarding
the CTCM software side.
Overall - '060 Rev 1 rating = 15% less than Rev 6!
A joint CiH/gwEm production, for Alive! Mag,Nov '05
|
