CT60 FIRST VIEW!
CT60 first impressions
Will you still feed me, will you still need me, when I'm 060?
###########################
## Blah blah yawn, etc..##
###########################
The epic tale of the CT60 68060 accelerator is wearily familiar to all and
sundry. The fact that it arose from the ashes of the tragically cancelled
Phenix computer, the minor consideration that it is a year overdue. There is
no need for this writer to recount the struggles, and tedium of waiting.
We're here to objectify that which has become reality, to classify, and
maybe even celebrate a little?
I might add it seems to have survived the going down of the Milan '060, and
we never lost hope. The CT60 was always there, a holy grail, shining at the
end of a long tunnel of tribulation, possibly more potent in the symbolism
of never giving up on Atari, than it ever could be in the flesh? Eventually
Rodolphe Czuba managed to gather his forces and resources, to put a
determined effort into finishing the thing last summer.
Some of us were on holiday in parts sunny and Greek, when the first wave
broke. I got mine in July/August too, but held off from fitting it, until
things became a bit clearer as to possible recasing options, general ease of
use, getting hold of a donor Falcon reliable enough to take it, and the
discovery of any nasty bugs lurking in the plasterboard by the other beta
testers out there. Eventually, I persuaded one Lyndon Amson, aka Stimpy, to
do the deed in November.
Getting the machine to him proved to be nailbiting, as the parcel service,
Parcelforce, or 'Parcelfarce' seemed to be on extra special go-slow
deliveries. Because of the tense nature of such a wait being unbearable for
the return journey, I opted to collect the machine when finished. It seemed
like a good idea at the time because I fancied a nice pleasant drive out
into the wilderness where Stimpy lives.
Oh the many varied and self-mutilating methods we use to deceive ourselves..
The drive to Stimpy's proved to be a) wet, and b) lame. It was nice to
remeet Stimpy on his home ground. The short time I had to examine the goods,
passed all too quickly, and I was under some major time pressure to get back
home again, fortunately which I did.
Later that evening, I took my first cursory look at the Rodolphe Czuba
mutant genetically enhanced Falcon.
################
## Let's see!##
################
The box itself, yes, it's still in a standard Falcon case, but Stimpy has
replaced the original keyboard with something that is still Atari issue, but
seems to have a nicer tactile feel, it doesn't feel as 'worn in' as before.
The power supply is an ugly brute of an ATX, which is umbilically joined to
the CT60, so I'll have to be very careful if and when I travel with this
machine! Not to mention the mess of leftover unused supply cables emitting
forth from the ATX ;-) I wonder what other peripherals I can run off them?
He has done a nice job with the reset/shutdown buttons on the back. these
respond in the same way as the push buttons on the front of most PeeCee
cases. There is an 030/060 mode select lever on the side which feels
reassuringly chunky and solid.
I would say that I will be recasing this machine eventually, more especially
when the Super Videl is to hand. There literally isn't any more space in the
standard case, even with the RF shielding removed. But at the moment, there
is something almost sexually perverted about having that much power, in what
is a slightly glorified ST case!
I like the idea of a machine that can take its own temperature, and this one
holds steady at around 41-43 degrees celcius, even when it was left on for a
while. Further lengthy examination suggest that the machine holds up at that
level, other people run happily with higher temperatures. The fact that
Stimpy managed to cunningly locate the original 40mm Falcon cooling fan over
the heatsink, with a little surgery to the keyboard, probably helps.
Hmmm, things are hotting up!
#############
## Engage!##
#############
Booting up is impressively quick. There are warnings about using X-Boot and
creating user-defined sets, so I went and pre-defined my user sets
beforehand, in '030 mode. In practice, I don't think that this is one of
those unfortunate set-ups made fragile by the presence of X-Boot. Time will
tell me more, but I think it will be okay.
When I first kicked off, there was an early issue with Magic randomly
crashing a short way in, usually when I was moving the mouse. I had a look
at the Auto folder running order, and the stuff that was in the CPX, then
got rid of any duplications etc. I don't think there were any really serious
issues with the Auto folder. I double-checked the patches, renamed one of
the 'old' files left behind by the Magic patching process to something less
likely to be picked up by mistake when Magic boots? And I think that might
have cured it?
A tendency to fall over at random reappeared after a few quiet days, and
this may well have been down to my installation of the Jinnee desktop
introducing a few extra duplications. Check through the install again,
remove any duplications, hope for the best, and at time of writing, it seems
to be okay again.
Standard TOS on the other hand, does seem to be pretty solid, unless you are
doing something very silly with a program file.
The only other issue on this install, seems to be some possible timing
problems between the CT60 and DSP when running audio streaming such as Mp3
from Aniplayer 2.20 and Falcamp. Aniplayer mostly works, and can replay
Mp3's *without* the DSP, just that it kicks most of the cpu time in the head
by doing so, unless you compromise with reducing to mono etc.
Running it with DSP 'on' reduces the overhead impressively, to a sub 10%
figure. At theat point though, there is a tendency for Aniplayer to 'stick'
in mid-tune. You can pause and resume, so the error handling is good in that
case.
Poor old Falcamp starts off well enough, then stops on a DSP timeout error
message (generated by the program.) The average Mp3 is compressed to play
back at 128kbit/sec, and makes these programs fall over. The Tempest 2000
soundtracks are MP2, and 112 Kbit/sec, managed to play okay with cache and
DSP, so only just off the pace then. Turning off the cache, but keeping the
DSP seemed to help some more, as Falcamp started to work again, and
Aniplayer was less prone to 'sticking', but not entirely free. As Didier
said that Aniplayer/Falcamp etc should work with the DSP and cache talking
to each other, then there might be a problem somewhere?
###############################################
## Then CiH slows things down just a little!##
###############################################
I then remembered that I had a 40mhz oscillator, which would knock the bus
speed down to 20mhz, a last resort really, but suggested in case this Falcy
didn't like life in the 25mhz fast lane. I opened up the machine on the
second Sunday after getting it back, and admired the class shoehorning act
that Stimpy had accomplished. It seems that he had 'carved' a little
indentation into the keyboard base to accomodate the cooling fan on top of
the heatsink! I soon found the oscillator in question, and it was socketed,
to make plug and play swapping easy. I swapped the oscillator, and put the
machine back together. It still worked even after going through my unskilful
hands...
Better still, it seemed that the minor, or perhaps not so minor hardware
related issues, had all magically cleared up. Firstly, Falcamp co-operated,
without sticking or stopping. Spurred on by this success, Aniplayer was
given the treatment, and has had lots of time to get used to the idea of
playing mpeg music files! There has not been one single stutter or halt
since the changedown. It plays MP3's quite happily, at a typical value of 8-
10% cpu, when running with DSP. Falcamp is even more optimised, I think,
which means that multi-tasking with background audio is now a real
possibility!
I noticed changes in other areas too, Deez's benchmark screen, with the
spinning 'CT60 Rules' globe always ran before, but flickered. I put this
down to the program being a two hour hack. Now, it ran completely smoothly,
without any flickering. Also, there had been pronounced pixel noise or
'dirt' on any non-Nvdi screen, now this had gone. It now looked like the
worst of the crashes when Magic was on, were finally done as well.
With a 20mhz bus, it runs a degree or so lower in temperature terms too.
After a complete reinstall of Magic et al, made necessary by me munging up a
later Jinnee install, I can say with some confidence that after two weeks, I
am roughly where I want to be with this machine in its early stages. It just
remains to build on that in due course!
In general, the 'burning-in' period for the CT60 has gone far more smoothly
than the early days of the CT2. When that was first fitted, it went through
three distinct stages of woe. Going from a screwed-up Fastram controller and
a cocked up 'whiting-out effect' desktop, it matured quickly through a
bugged DSP which didn't work at all. I then carried a working fast '030 GEM-
Box, with intermittent DSP operation, and no serial port working, for the
best part of 5-6 months. In the end, it took Rodolphe Czuba, in a rare visit
to the UK, when he was showing the Phenix, to fix the thing permanently!
It has been fine since, but in the CT60, it seems that Rodolphe's techniques
and debugging skills have refined from when he designed the CT2.
I might add, that the fortnight after getting the machine has seen it
operational, to at least 80-90% user satisfaction until it was set up and
run in properly. It has not been put out of action, or had to go away for
repair, or had major components of it not work at all for months at a time.
Maybe the extended waiting time had some benefits after all? To a large
degree, I would like to thank Stimpy for doing such a good job on the
install in the first place. At the same time, he accomplished the arguably
bigger miracle of bringing my Nemesis powered dodgy keyboard machine, my
original first Falcon, back to life to take the CT60. The Nemesis was
removed and bagged up, and is currently waiting for a passing bus to jump
under!
The CT60 seems to be missing some rough edges that came with the CT2. The
serial port is available 100 percent of the time, there are no occasions
where you have to reboot several times, particularly in Magic, to get Sting
to talk to the modem. Aniplayer seems to be a lot happier with the same
level of extended resolutions on CT60 as used previously on CT2. There has
been extensive testing of the image viewer, and lockout of keyboard input
when trying to come back to the desktop from the file selector has happily
gone as well. This made using certain other GEM apps in Truecolour extended
mode especially, a bit of a risky business on the earlier hardware.
Another thing I found a little bit 'non-standard' about Magic and its
interactions with my CT2, was the fact that you have to perform a soft reset
(as in 'Ctrl-alt-delete') on my machine to get it to continue loading, after
first booting. Most of the time, that wasn't a problem, but once in a while,
it took it as a personal insult and stopped working altogether! Combined
with the serial port recognition that was sometimes not there, just
occasionally, getting online with this combination could be frustrating
indeed. Still, I can laugh about it now!
#########################
## So, what's it like?##
#########################
The main place where I'm getting a good impression of the extra underlying
power, is with the speedier operation of the 'web. This is a good thing, as
it is my main daily use cpu-intensive operation. Even a slow old dog such as
Cab showed a great improvement over how it performed before, even compared
with the CT2. Highwire was even better. Web page HTML formatting was
virtually instantaneous. It wasn't too bad a performer before on CT2, so the
difference was less obvious to me, than it might have been to the owner of a
standard Falcon tortoising its way across the internet. The graphics loading
in both cases was greatly improved too. Now it seems the only thing slowing
down the web, is that 56k piece of string connection!
Saying that, it is still great to get an internet connection more responsive
than the crappy old Wintel box at work! (ancient 150/200mhz Pentium hooked
up to a server and running web from there.) It takes a-a-a-a-a-ges for it to
display any large pictures on a page onscreen.
########################
## Slick or sluggish?##
########################
Look at all those Mips! It compares nicely with 3.93 Mips on a base Falcon.
Or even 12.29 Mips on a Centurbo 2. The power, the power!!
Of course CT60 benefits the handling of the GEM environment and desktop
greatly, over that of a standard unexpanded machine. It is more difficult to
tell compared with a CT2, because that hardware, in combination with Nvdi,
is a pretty decent ride on a GEM based desktop too.
I carried out a rough rule of thumb test to compare the two, which was to
load in an identical picture file on Smurf, on each machine, and drag it
around the desktop, to see how smoothly it travelled with the mouse cursor.
The picture was a fairly big image, in the region of 800 x 600 pixels in
size.
There wasn't too much difference between the two actually. Both moved around
quite well, and the screen updated at roughly the same rate on each
configuration. In the case of the CT60, you might consider it disappointing
that screen updates in GEM weren't so smooth as to be invisible to the naked
eye? There are good reasons why.
There is the issue of the slower bus speed and ST-Ram. You might consider
the CT60 as having greater potential, so far not seen that effectively, as
it is pinned down by the limited nature of the original Falcon design in
that area. If or when the Supervidel Design by the Nature brothers gets into
production, expect to see that aspect of CT60 operation improve hugely! In
my case as well, the CT60 is running a lower rated bus at 20mhz, as opposed
to the CT2 turning in a healthier 25mhz on the bus speed.
Of course, if you have been used to a 16mhz Falcon for years, with its
underpowered approach to system compliant screen access, then even a CT60
bottlenecked by a slow bus will be vastly better than what came before.
#############################################
## Software stuffs, a hasty person's view!##
#############################################
I had a few problems getting Smurf 1.05 going at first. It loaded okay, but
crashed out when trying to display a picture in plain TOS. However, it works
fine under Magic, and I did have a little fiddle with several of the effects
modules. What a nice fast response they've got, even with a larger picture
to work on!
Evil pointed out that Apex, or at least the later versions, worked. This was
down to the fact that Doug Little did code for his Afterburner '040. And I
checked out my copy of 'Apex Intro', and yes, it did! As with any other
Falcon, you will need a 25mhz bus for it to display in higher resolutions. I
might also suggest that it is put under the category of 'working but buggy
in places'. The image magnifier can give some garbled results, when working
in the highest setting. Also, I had a memorable time, when Apex insisted on
loading a second picture on top of the existing image, when I had already
specified another empty screen to go to! With that in mind, I would proceed
carefully, before doing anything mission critical with it.
One which did seem to work with little trouble, was Crackart. This was
subject to setting to ST-low mode (Caches off might be a good idea too.)
Well you never know when you need to work on some 16 colour artwork, and
there isn't an STe to hand!
Aniplayer was one of the tested programs that gave me a good impression of
the underlying power. The DSP has no problem with Jpeg depacking, but
Aniplayer is quicker if you turn the DSP *off*! It is a good idea to select
some really large jpegs to notice the difference though! For one such
picture, I finger-counted to four when running with DSP (which seems to have
no problems in this instance.) Loading again without, I just got within a
finger count of two!
I managed to transfer the extended Centscreen modes from the CT2 Falcon
without a hitch when it was running at 25mhz. I had to redo these with
Centvidel when going to 20mhz. It seems that I got the Centvidel patch from
Didier at exactly the right time. It did not prove too difficult to get back
to the sort of 800 by 600ish modes I am used to, even with a 20/40mhz bus,
so we are working in extended resolutions again.
I also got online quickly, my Sting installation transferred without any
problems, Newsie, CAB and Highwire 1.5 were working first. Aftp and Atari
IRC needed a bit more work, but soon played along. Atari IRC settings
transferred over from the CT2, Aftp needed setting up from scratch. Still,
no problems there.
Music stuff, I've already partially discussed earlier with reference to
streaming audio, other DSP based sound players have worked. I gave Flaysid a
very extended road test, and it came up smiling, even on the 33mhz highest
quality playback. Jam Player seems to work happily with caches off, and I've
tried Oszi (non-DSP) modfile player, and the 'Dump' player as well. I'm
personally looking forward to CT60 friendly versions of Grauomf tracker, and
of course, the latest version of Ace Tracker!
I've taken a look at a tiny amount of demo stuff. I'm personally amazed that
anything not written prior to CT60, can work on this, but the Dead Hackers
4ktros work. I've got the early version of the Orneta 4ktro, which is not
framerate-synched and runs like the clappers, but I think it has decided to
run as a 66mhz '030 in that case!
I tried the two big demos from the Second EIL party, the T.O.Y.S 'Wait' demo
looks for ST-ram, and needs 14mb, unless you've got the new version which
streams the soundtrack from disk (Yes, I've got a 4mb STram board in) but I
it does work. It seems to be better in my case, with caches off, but the
overall speed is still good. The Dead Hackers 'Don't Break the Oath' gets
about a third of the way in, then stops on the plasma or credits screen.
This matches the report on the Satantronic web page that I've seen.
Oh yes, the Alive/UCM diskmag shell seems to be happy on '060 !! And, get
this, the two issues of the Maggie Falcon shell that I tried, also seem to
work too! Chosneck issue 3 was tried. I got it to boot up very smartly, It
loaded and depacked the first article very quickly which was excellent. It
fell apart from there, as it followed the CT2 practice of hanging up and
crashing the DSP music player when trying to exit back to the desktop. I
really hope that the Chosneck shell coder has got to grips with this whole
accelerated hardware topic, by the time issue 4 hits the streets?
I've still yet to try out a couple of the heavyweight applications, that is,
Quake, and the Joysoft Atari 800 emulator. Now that will be for slightly
later on!
##############################################
## We're still at the start of the journey!##
##############################################
There is no way that I could do a full review of the CT60 justice at this
early stage. That is partially down to the fairly small timescale since I
powered it up for the first time. There is also the issue of the vastness of
the subject matter itself. With the amount of extra power that a properly
unleashed '060 can deliver, my understanding of it is very incomplete, and
will need a decent timescale for other people to release productions that
show off the full potential too.
I also get the impression that a lot of the real power of the CT60 is still
hidden away under the slowness of the original Falcon design. It seems to be
more correct to refer to its performance in areas like desktop handling, as
similar to that of an enhanced '030. SuperVidel, in my view, will be as
important an upgrade as the CT60 itself, transforming screen access and
handling. Where you do get a kick from the '060 to tell you it is the real
thing, and not a fast '030, is where any sort of major maths is involved.
I'd point to picture depacking as a prime example, especially very large
Jpegs, which hardly show any time lag at all. Of course, this translates
into benefits when loading in a graphics heavy web page. So my major use for
this new machine has been online so far. CT60 leaves everything else in the
dirt, especially combined with a recent version of Highwire!
The performance of DSP applications is another shock. Previously the DSP was
a general purpose 'wonder-chip', used by the most skilled elite coders to
make impossible things happen on a poor 16mhz based computer. Now the DSP
used in conjunction with the CPU will slow down the CT60 for most things!
Roldolphe Czuba correctly suggested a realignment of coding priorities when
writing for the CT60, as the DSP should be relegated to its original
function of audio processing, the '060 being more than capable of the fancy
3-D stuff.
In conclusion, I would say that the CT60 hasn't disappointed, it has lived
up to its early promise, and I have an anticipation that the best is still
to come. The only regret I have is that it wasn't finished a year ago, but
part-time hardware developers can't pick and choose their timescale ;-)
I said earlier that this was only the start of the journey. There is the
possibility of some further hardware in the pipeline, software in
preparation will get my full attention as and when it arrives. I've still
yet to try out a lot of existing stuff, to see how well it performs with the
extra power in any case. So I'm sure to be revisiting the CT60 in future
editions of Alive.
CiH, for Alive Mag,Dec '03.
|
