Alive
News Team Current issue History Online Support Download Forum @Pouet

01 - 02 - SE - 03 - 04 - 05 - 06 - 07 - 08 - 09 - 10 - 11 - 12 - 13 - 14

Alive 4
                              T h e   T O S t e r
                         - ------------------------ -


                  Is it the dawn of a new era or just the dawn ?



  Hello and welcome to todays menu.
  Today we offer a bunch of new Atari clones, several new emulators
  and maybe you'd like some ultrafast 2 GHz accelerator for your
  trusty old Mega STE, too ?

  Now unfortunately, these sentences only occur, spoken by attractive
  young women with long, blonde hair, long shapely legs in far too
  short miniskirts and tight V-neck shirts with a very large V,
  in our wildest dreams and not in real-life.
  Reality looks slightly different and if at all, not-so-young but
  slightly hairy men tend to at least offer 80 MHz accelerators,
  a Coldfire-based clone (both under development) and several updates
  of emulators sometimes on the internet.
  So honestly, it could be worse. With the X-Tos project moving
  along, with the ct060 being on the edge of complete, with Tempest
  still not dead the future of Atari might be dim, but not all dark
  yet.
  But what do we see if we browse the web open-eyed and absent-minded ?
  An all new Atari-Clone, hold it, even better, an all new Falcon-clone,
  100% compatible, easy to design, cheap to produce yet fast, quick and
  realiable, being developed currently ? By the famous FlashLabs Crew ?

  By who ?

  Yes, by FlashLabs. Don't know them ? Me neither.
  Eyebrows raised and confusion increased the careful reader reads on.
  Now the amount of people capable of designing an Atari-compatible
  computer more or less from scratch is very limited and without
  wanting to exclude anyone unfairly, the amount of people willing
  to sacrifice time, money and devotion for such a project is even
  smaller. Memories of Wizztronics announcing a Falcon040 every now
  and then and finally skipping it arise.

  And after witnessing a vivid discussion between the "founder" or
  "lead person" of FlashLabs, Anders S. and Johan Klockars, author
  of fVDI, PCI-drivers for Eclipse, involved in the Aranym-project
  and several other gadgets on Atari, i can here finally and proudly
  explain to you about the background of this fantastic-sounding
  Falcon-clone project named TOSter:

  There's not going to be one.

  The specs on the page are hillarious and in the discussion with
  Johan, Anders S. fell back into insulting Johan and clearly
  displaying that he hardly knows what he is talking about.
  Even though the page just loses itself in wild promises not
  mentioning about HOW it's all going to work out - especially
  not about "when" it is going to work out and "how expensive"
  it is going to work out - after reading the content it is
  obvious that this project is not possible. Clearly to be seen
  even for people less involved in hardware-design - like me.

  So what's on this page ? Let's go through it.
  Beginning with a "jump-menu" that does not work - Why btw ? -
  okay, beginning with the "overall specs" in the upper right
  corner:
  - Motorola 68040 66.5 MHz 133 MHz SDRAM
  - Optional G3/G4 card up to 850 MHz
  - DSP 56321 - 360 MIPS 200 MHz
  - Double Videl 1600 x 1200 36 Bit
  - PCI, USB, Ethernet
  - 100% TOS compatibility

  ... yeah, and it's going to fit into the Falcon standard case, too.
  Sure thing.
  An 68040 CPU - nothing against that - with 66.5 MHz - nice - and
  a RAM at 133 MHz. Cool. Have the RAM wait for the CPU this time.
  And please, turn off all caches on the 68040 since the RAM is faster
  than the cache anyway. Not to mention that RAM has to be lowered
  to CPU speed anyway when the CPU accesses the RAM - otherwise there
  will be no signal-stability.
  Optional G3/G4 card. Since the RAM runs at 133 MHz anyway, this
  shouldn't be hard to add, should it ?
  DSP56321 - Now that's cool. And the specs are right from the
  Motorola page. Not to mention that this is one hell of a powerful
  beast, it's also one hell of an expensive beast. Makes you wonder
  why the Deesse card only incorporates a puny 56301 ...
  And the Double Videl, now that's a cool extension. 1600x1200 -
  pardon me, at what scanrate ? 1600x1200 is possible on the Videl
  right now, you just need a monitor capable of a scanrate low
  enough. And what is 36 Bit going to mean ? 24 Bit colour encoding,
  8 Bit Alpha Layer and extra 4 bits in case some other get lost ?
  Or 4 Bit Beta-Layer ? Or just a typo ?
  PCI, USB, Ethernet - Maybe they're cheaper when you buy them all
  at once ? And why does FireWire miss there ? Or is FireWire only
  available with a G3/G4 card ?

  And how come that i honestly doubt ANY of the data above ?
  Let's read on and find out HOW they're going to make this machine
  work ...

  CPU:
  Now FlashLabs claims that there is only the choice 68030 and
  68040. 68060 is obviously no option for them for reasons i do
  not understand. Their ongoing reasoning is that the Falcon uses
  its FPU - if fitted - only in 16-bit mode. Now i didn't know that
  you could switch an FPU into several modes, but i am afraid the
  FPU to CPU channel is only 16 bits wide since the Falcon was
  intended to carry an 68000. This does not mean "half of the FPU
  is wasted", but who cares. FlashLabs considered this reason
  enough to build a new computer (next sunday). Oh yes, and the
  68060 is too expensive and misses some instructions. And with
  such an extremely powerful DSP as the 56321, there's no need
  for a powerful CPU. Yes. I can hardly wait to run the FlashLabs
  Word processor on the DSP56321 or the first DSP56321-based MiNT
  kernel. But FlashLabs prefer MagiC (on the MAC) anyway ...

  So the FPU in the 68040 - Yes, they want a fast FPU with a slow
  CPU because you don't need a fast CPU with a strong DSP - is so
  terribly slow that they decided to give it an extra-68882.
  Come again ? The 68RC040 has its own FPU which is a lot more
  powerful than the 68882 ? Ah, yes, not compatible with the 68882
  because them instructions it lacks are badly needed and make
  CentBench look bad.
  That really sucks. IF FlashLabs build a new computer, they want
  at least CentBench to run terribly fast. So give this 68040, that
  cannot OPERATE an 68882 directly, an 68882 externally and just
  imagine how fast CentBench might run on it.
  If it ever works.

  DSP:
  Yes, let's go for the brand-new and totally inexpensive DSP56321.
  FlashLabs also announces to provide standard libraries for
  MP3, MPEG4, audio, video, JPEG, GIF and TGA decoding of the
  APEX series ... Hold on a second. APEX GIF and TGA decoders are
  purely CPU-based, the TGA viewer officially even runs on the STE
  and frankly, the GIF and TGA format bear really nothing that
  requires DSP code (JPEG does because it needs Fourier-Trafo which
  is well done by the DSP), but obviously, FlashLabs decided to
  convert them to DSP-code.
  Well, yeah, because they got a slow CPU, a fast external FPU and
  one powerful DSP, you need to decide which component does what.

  Then they want an 8 MB ROM to put a lot of PCM-sounds in it so
  the DSP56321 can operate as a MIDI-instrument right after boot-up.
  I am especially fond of 8 MB ROM. Ask Nintendo what they cost, they
  know because they avoid them like hell in their N64 cartridges.

  VIDEL:
  Now FlashLabs says the Videl is no graphics processor but a pixel
  pump, a bunch of registers in a programmable LSI chip.
  Sure. That's why Atari spent about 2 years on developing it.
  It's just a silly set of totally unsynchronized set of Digital-
  Analogue converters that just turn digital RGB values into
  the according output signal.
  FlashLabs says it only has 6 data lines to each DAC and that's why
  there's only 18 bit (262144) colours. So "18 bit is TC there".
  Actually, on the VIDEL 18 bit is NOT TC. And who the hell wants
  18-bit chunky pixels ?! Does that make any sense at all ?
  No, it does not which is why ANY resolution using 18-bit colour
  encoding is limited to 2,4,16 or 256 colours on the Videl so
  you can have 1,2,4 or 8 bits per pixel which fit nicely into
  memory. But Atari were suckers anyway, not capable of building a
  G4 card for the Falcon.
  So FlashLabs goes for 18 Bit True Colour - No, not yet, it gets
  even better, they want their VIDEL to access the memory with 72 Bit
  so it can do 36 Bit True Colour.
  36 Bit ! Think of the possibilities. Think of the colours !
  Think of the memory map. One longword + 4 bit for each pixel !
  Now let's skip a discussion about "72 Bit access ?!?!", "36 Bit
  per Pixel ?!?!", "Just a Pixel Pump ?!?!" and "Just 3 DACs ?!?!"
  because FlashLabs says, it might sound a bit raw but it's not
  hard to do.

  Now, i remember the Videl as a quite nice device. It is capable
  of synchronising itself totally, reserves bus access when it
  needs to (it cannot be put into "wait, bus busy"-mode though),
  it can do horizontal and vertical skips on screen, it offers
  both planar and chunky modes, frees the bus when it's done again
  and it does all of this by itself, without being timed, triggered
  or operated from the outside.
  But once again, Atari designers and of course XTOS-designers are
  suckers because they don't see how easy it all is. Especially that
  this new "VIDEL" would be 100% compatible and can do 1600x1200,
  which sounds impressive enough.

  SDRAM:
  Yeah, let's have some RAM. Does a Gigabyte sound good enough ?
  No ? Then FlashLabs even tops that. 2 8 MB 16-Bit FlashROMs
  are going to be fit as standard RAM-disk within this computer.
  Bah ! Lame ! Why only 16-bit ? And IIRC, FlashROMs were not
  really fast either.
  But luckily, these FlashROMs are going to act just like a harddisk,
  connected to IDE via DMA. Which is especially cool because IDE
  usually hogs the CPU bus and not the DMA, but FlashLabs are the
  experts, right ? And controlling RAM via IDE is cool too.
  Accessing FlashROM in cylinders, heads, tracks and sectors is
  a mighty cool idea. Probably FlashLabs has overseen that FlashROM
  is quite slow because they say it's going to be measured in
  NANOseconds access time, but FlashROM and FlashLabs both have
  a flash in their name and that's good enough.

  USB:
  Yeah, we want USB ! We want USB !
  So FlashLabs gives it to us. Cool. Not lacking any driver software ?
  Nah, no problem at all. FlashLabs writes us a nice software, that
  "converts" Windows USB-device driver to Atari !
  Now why don't people at X-TOS think about this ? It is all so damn
  easy, isn't it ? Just hack your Windows driver CD and you can use
  everything on your Falcon.

  PCI/G3-cards/Ethernet:
  Oh ... Dull ... FlashLabs doesn't say anything about this.
  Too bad, because i am sure it would have topped everything else.
  I am sure, the FlashLabs PCI-bridge operates at 850 MHz to give
  the G3/G4 card the speed necessary to really rock this new
  computer. And of course, the PCI-bridge has access to the 68882
  FPU for compatibility reasons. And it's 72-Bit wide so the Videl
  can access PCI-cards too, so you can hook TV-cards to your new
  computer and have the VIDEL access it.
  And of course, Ethernet would have to be at least 10MBit ... No,
  100 MBit, no ... we want more, we want 1 GBit !!! No problem at
  all with a 850 MHz PCI-Bus. And you can simply use your Windows
  driver that came with your PCI-ethernet card.
  Yeah, and the G3-card, that's of course only the low-cost
  solution. The G5 card is being planned currently, but so far,
  the suckers at Motorola haven't finished it yet. Until then, you
  can simply use your time-consuming software on the DSP56321.


  Now, honestly. I would have taken this whole page as hoax and wouldn't
  have mentioned a single word about it until i read the discussion
  between Johan Klockars and Anders S. of FlashLabs on the Atari.Org
  16/32-Bit forum. The tone was rather unfriendly and Anders S.
  persisted in this machine "coming along nicely" and "to be
  finished soon".
  I would have believed this page is just a wind-up for the Atari
  hardware-users that are still not using MagiCMac until i saw
  Anders S. vividly defending the concept of putting an external
  68882 to an 68040 because CentBench gives low FPU-performance on
  the AB040, ending with insulting Johan Klockars as a "PC lamer"
  who obviously has something against a "better TOS machine".

  But there not going to be a better TOS-machine from FlashLabs.
  This page is just summing up what people having neither any
  experience in coding nor any experience in hardware-design
  dream of during lonely nights. It's shouting "I have no idea
  what this all means but it all sounds so damn good that it
  MUST be true" all over.


  Well, i just thought i reported this to you.
  So in case you already saw the page and were wondering, please,
  don't waste any further thoughts on all this. It's nothing but
  pure lies. Please support the X-TOS machine, Tempest, the ct060
  or maybe Eclipse rather than any of this junk.
  These projects need support. The "TOSter" doesn't.


  Thank you for your attention.


  The Paranoid




Alive 4