Re: TWGS DMA and RamFast SCSI--HELP!!

[supertimer@aol.com (Supertimer)]

Liam Busey <liam1@ix.netcom.com> wrote:

>Sagochangi wrote:
>> 
>> I read a while ago (from Supertimer) that the TWGS is fully DMA
>> compatible with the RAMFast SCSI controller.  And that certainly isn't
>> the first time I've heard that. But my experience has been different.
>> I bought a TWGS a while ago, anxious to use all the new speed. But the
>> card absolutely refuses to boot with my RamFast in the system, even
>> with the RamFast set to TWGS: YES, and DMA off.
>> 
>> Nathan Mates' old accelerator FAQ lists the -2B GAL as being a 
>> necessary DMA fix, and the RamFast manual itself (p.40) says of the 
>> -2B GAL, "The RamFAST/SCSI will not function in your computer until 
>> you get this replacement part".  And the TWGS works fine without the 
>> SCSI card in the system, so I can only assume that I do need the chip.
>
>I wouldn't put much seock in Nathan's  Accelerator FAQ. It's way out of
>date. As for the 2B GAL requirement, I believe Tony Diaz is correct in
>his assessment that it's unneccessary. I've seen a a couple of posts
>from people Ramfast Transwarp difficulties even with the 2B GAL
>installed.

After a closer look at Tony's information and my old RamFAST
manual (I no longer have the card), I've come to the conclusion
that Tony seems to be using a complete set of "A" GALs except
for 2B, which he has.

This is the same configuration as several others who have
posted here.  Thus, the GALs seem to meter instruction flow
in order to allow the old, flawed 65C816 chips to work.  The
redesigned Sanyo type WDC 65C816 units don't need the
new GALs because all the instructions now work at higher
speeds.

Except GAL 2B.  This GAL seems to be a work-around not
for the CPU, but for the RamFAST.  Note: this 2B GAL is
definitely not required for the Apple High Speed SCSI card,
which also does DMA.  It seems to be a RamFAST specific
thing.

>> Here, also, is the information I've gathered on my TWGS: version 8H, 
>> rev. 1.4x. Performs at 6.3 mHz in Transwarp mode.  Card is installed 
>> in slot 3 on a ROM 03.
>
>Your TWGS is appears to be significantly different from his though.
>Tony's has the improved 65816, a different ROM, an expanded cache, and a
>faster oscillator.

Yup.  For the card to work above 7Mhz requires a ROM
v1.5 or greater.  He has v1.4, so he's one of the few TWGS
users who needs to upgrade to a newer ROM from SHH
Systeme.

>> Can anyone tell me if there is a work-around for this problem?  If 
>> not, please, *please* (he wrote with his own blood), can some kind 
>> soul tell me where I can get the -2B chip?  I can't give up my hard 
>> drives, but I can't keep going at 2.6 mHz, either. Replies by email 
>> are appreciated.
>
>The only place your likely to find a 2B GAL these days is on another
>TWGS. So unless you get lucky your stuck.
>
><hypothesis>
>If I recall correctly, Tony Diaz's TWGS Mod site said that the GAL
>upgrades on the TWGS accelerator where made to compensate for the bugs
>in the design of the original 65816 when running at higher speeds. I
>suspect that the 2B GAL (which supposedly allows compatibility with the
>RAMFAST controller) is not extra logic to interface with the RAMFAST but
>rather logic to clean up flakiness in original 65816 DMA. Perhaps the
>redesigned 65816 no longer needs the GALs? That could explain why Tony's
>TWGS was compatible and yours currently isn't.

That is a correct assessment.  The GALs (except 2B which was
specifically to deal with the quirks of the RamFAST) are
traps specifically designed to deal with the flaws of the original
65C816.  The original chip simply could not go above 7Mhz or
so and reliably execute all instructions.  The GALs trapped
certain instructions and slowed down the accelerator for those
instructions to complete, thus avoiding crashing the system.

After Sanyo redesigned the 65C816 for WDC, things changed.
The 65C816 units were now rated for 14Mhz, so ALL instructions
performed as expected at 14Mhz.  Thus, the "slow down" GALs
were no longer needed and indeed they should be avoided.
They'll just slow down processing speed a little.

Sadly, the ZipGS is locked into slow down operations
because these traps are designed into the single ASIC
chip.  That's why for a given Mhz, ZipGS is a little
slower.  This is a disadvantage today, but back during
the accelerator wars, it allowd Zip Technologies to
claim 9Mhz and 10Mhz operations when the TWGS
was doing 6.5Mhz and 7Mhz.