home *** CD-ROM | disk | FTP | other *** search
-
- Doug here,
-
- The following message contains quite a bit of text related to Nemesis, but
- I hope it is MORE useful as a guide on Falcon SCSI problems and help prevent
- rumours about what this is and what it isn't.
-
- I hope I haven't made it look too much like an advert - I made an effort to
- be unbiased without omitting to explain Nemesis doesn't suffer in the same
- way as most accelerators apparently do. This is the kind of rumour I hope
- to prevent before the board gets (unfairly) filed along with some other
- less reliable boards.
-
- > It is the system with the Nemesis "stable"?. I read that in some cases the
- > FX board appears to do strange things with a Falcon (problems with SCSI,
- > video trouble, crashes) in certain speeds. I believe that to many ad-ons
- > (with the Afterburner) can do troubles to the Falcon.
-
- The more you add, the greater the risk - that is true. However, there are
- certain things which whould be made very clear about the Falcon, SCSI DMA
- and the Nemesis card...
-
- Fact 1:
-
- Speeding up the Falcon bus amplifies inherent problems already present in
- the Falcon's PCB. The faster it goes, the more amplified the problems get.
-
- Fact 2:
-
- These problems usually become significant around 20MHz.
-
- Fact 3:
-
- The problems themselves express themselves most obviously in SCSI transfer
- errors, and audio 'clicks' when replaying sound via the SDMA chip. When things
- get really bad, the video may glitch or 'scroll' during disk access, and the
- machine may even crash completely. Floppy drives may suffer too, but IDE's do
- not.
-
- Fact 4:
-
- Some Falcons suffer even at 16MHz - but normally only in video modes which
- are sufficiently 'expensive' to place a little strain on the video bus. 256
- colours 640x480 60Hz (and upwards) is usually where problems begin on these
- machines.
-
- Fact 5:
-
- The cause of these problems is not simply under-rated components - it's actually
- down to bad PCB design and bad circuit design. The components themselves actually
- have as much as 60% tolerance on almost all machines (fortunately for us!).
-
- Fact 6:
-
- Nemesis is rock-solid using a 50MHz system clock (25MHz bus & CPU). We have tested
- 10 separate installations, and one or two of these required some tweaking before
- performance matched the others - but this is because the PCB's are variable. I
- am not talking about motherboard revisions - this actually has relatively little
- impact on the stability of accelerators. Two supposedly identical PCB's will not
- act in exactly the same way no matter how similar they claim to be.
-
- Fact 7:
-
- Pushing Nemesis beyond 25MHz results in problems in most cases, but this does not
- involve SCSI or audio problems. What happens is the ST-ram gets cycled too fast
- and starts to fail due to lack of time for RAS (refresh) cycles. Since 80ns ram
- is common on standard Falcons, 25MHz is the standard expected limit. Machines with
- 70 or 60ns ram may tolerate up to 30MHz, but no promises are made.
-
- Fact 8:
-
- Problems caused by slow SIMMs results in video 'noise', followed by crashes if
- pushed too far. Most machines cannot cope with a 55 or 60MHz system clock for this
- reason.
-
- Fact 9:
-
- SIMMs are not solely responsible for RAM cycling problems. Memory upgrade boards
- are notorious for causing such interference. Simply swapping the memory board can
- make all the difference. Again, this is due to the PCB layout and quality. All
- brands of memory board are gulity of this! Be warned!
-
- Fact 10:
-
- All Falcons tested with a partially installed Nemesis at 24 or 25MHz produced SCSI
- and audio errors almost every time. Not one machine was problem free. This means
- the problem is consistent, but variable in 'intensity' between machines.
-
- Fact 11:
-
- All of these Falcons were succesfully modified to remove these problems at the
- final stages of the Nemesis installation.
-
- Fact 12:
-
- The tests used to determine the safety of Nemesis at 25MHz were performed
- under the following conditions:
-
- 640x480x65536 60Hz 16-bit VGA video, running Diamond Edge continuously on an
- EZ-135 removable media SCSI drive. Tests involved 'test disk structure', 'optimise
- disk', and 'map bad sectors' - repeated at least 5 times per drive partition.
-
- Additional tests for audio quality were performed using Cubase Audio 2.06, also
- running in an excessively outrageous and bus-hungry VGA video mode.
-
- These conditions are far beyond what is expected of any normal Falcon, and exceed
- what many users would even dare attempt in 800x600 256-colour mode on a stock Falcon.
-
- Fact 13:
-
- We do NOT claim that Nemesis works on all Falcons. What we DO claim is that no
- Falcon which HAS been tested has proved to be unsuitable. This includes motherboard
- revisions C, D/C and even '---', which appears to be the earliest model! It is also
- important to mention that all Falcons did initially produce SCSI errors and this
- was successfully rectified - in other words, not one machine was 'perfect' before
- we started.
-
- ----
-
- All of the above notes are unbiased results from our experiments and tests during
- the Nemesis development cycle. i.e. I have not made anything up!
-
- Make what you will of it...
-
- Doug.
-
-