GVP G-Force 030 32-bit RAM not being seen with GVP HC+8 in place

[blakespot]

I have just acquired a G-Force 030 (Impact A2000-030 Combo Series II) 40MHz accelerator for the A2000, which has 4MB 32-bit RAM soldered on the board and 4MB more in one SIMM slot. The Amiga (mobo rev 6.4) is recognizing the CPU (it seems...) and FPU but is not seeing any of the on-board RAM. This was an eBay auction purchase.

I also have a GVP Impact A2000-HC+8 Series II card which has 8MB 16-bit FAST RAM on board and serves as the system's SCSI interface. (And, I also have a first rev X-Surf and an ASDG Dual Serial board installed in the slots.)

Both boards (from different sellers, years apart) have identically marked ROMs:

C 1992 GVP
V4.5
DF24

::: PART 1:

Before I installed the 030 board with its 8MB 32-bit FAST RAM, the system booted fine (3.1 ROM, AmigaDOS 3.1) and the 8MB FAST is recognized. I want to run the 030's 32-bit RAM in extended space, above Zorro space, and I intend to run a switch to the jumper that entirely disables the 030 board, for cycle-exact demos, so I want to keep booting off of the HC+8 and keep RAM on it. I have set the 030 board's jumpers accordingly:

- J12 to open for all on-board RAM in extended space
- J14 to closed to indicate no SCSI drive connected (to the 030 board)
- J9 to open to disable the autoboot ROM on the integrated SCSI controller

So, with everything in place, I boot the system and get to Workbench and I see only ~8MB FAST in the title bar (+ 1MB CHIP). (At this point I have not added GVPCpuCtrl to s-s.) Things do not feel faster. SysInfo v3.24 identifies the 68030 and 68020, but running the SPEED test shows the system to be only 1.10 times faster than the B2000 68000 7MHz... And after running the test, SysInfo shows the clockspeed to be 16.80MHz (not 40MHz, as the board should be).

SysInfo reports 1MB CHIP and 8.0MB "FAST RAM PUBLIC KICK 24BitDMA" at $00200000 - $009FFFFF, only.

I then downloaded the accompanying software from AHDB (guessing at which is the most recent) and got GVPCpuCtrl into the s-s as "gvpcpuctrl FASTROM" and I get an error something very close to "Not enough 32-bit FAST RAM to store ROM Image."

::: PART 2:

Trying to quickly determine if the 030 board is fried, to deal with eBay return, I removed the GVP HC+8 and booted WB off the GVP install floppy and I saw 1MB CHIP, 0 FAST RAM in Workbench.

I then closed jumper J12 to put the 030's RAM in Zorro space, rebooted, and I see 1MB CHIP and 4MB FAST (I removed the additional 4MB SIMM from the 030 board, to simplify things) -- all the RAM is there.

I downloaded a more recent ver of SysInfo on an ADF and booted it (the system feels very fast now) and it shows things as they should be, it seems: 7.42x B2000 68000 7MHz speed (1.12x A3000 68030 25MHz speed), and the clockspeed indicated is now 36.30.

(I had, earlier, set the jumpers on the HC+8 to 6MB rather than the full on-board 8MB, and booted up and saw 6MB FAST total. I then closed J12 to try and go to Zorro space on the 030 board, and in this config, the system just hung -- I know there's only 8MB addressable in Zorro space - but I wanted to see what it did.)

........So, is it unexpected that GVP's HC+8 and G-Force 030 would be incompatible in this way?

I definitely want to be able to drop back to 68000 mode (switch to the disable jumper), so I don't want to rely on the 030's SCSI controller and I want 16-bit FAST RAM in the system.

Any advice here? It seems the 030 board is not dead, but the problem lies with the HC+8 or some other situation in place here.

MANY thanks.


bp

[SpeedGeek]

::: PART 1:

Before I installed the 030 board with its 8MB 32-bit FAST RAM, the system booted fine (3.1 ROM, AmigaDOS 3.1) and the 8MB FAST is recognized. I want to run the 030's 32-bit RAM in extended space, above Zorro space, and I intend to run a switch to the jumper that entirely disables the 030 board, for cycle-exact demos, so I want to keep booting off of the HC+8 and keep RAM on it. I have set the 030 board's jumpers accordingly:

- J12 to open for all on-board RAM in extended space
- J14 to closed to indicate no SCSI drive connected (to the 030 board)
- J9 to open to disable the autoboot ROM on the integrated SCSI controller

J9 to open was your mistake. You need the ROM enabled to add the Extended 32-bit RAM. If you had a Guru ROM installed on the HC+8 then it would have worked that way but not with a GVP ROM. 

[blakespot]

** SOLVED **

That was it. All works fine now. And a piece I didn't see in the GVP manual or on Amiga Resource's page for the accelerator is this little note I found on BBoAH:

J9 - The ROM appears in the card's I/O space when shorted (high memory detected, autoboot possible).  Dissappears when open (no RAM/no SCSI boot).

In the manual I see mention of J9 for disabling autoboot on the SCSI controller and elsewhere for "disabling ROM," but no mention of it being needed for extended RAM visibility. Thanks.

[blakespot]

I will mention that when the Impact A2000-030 Combo Series II board is disabled with the jumper (or switch, in my case), on boot I get a non-deadly Guru / Software Failure

Error: 88C7 08C7 Task: 0020226A

Which looks like it's part of the system init routines.

Asking AI, I was told:

Disabling the CPU jumper often leaves the GVP onboard Boot ROM and SCSI controller active on the expansion bus unless explicitly isolated.

Kickstart 3.1 is strict when probing the Autoconfig expansion space. When it checks the GVP board, the GVP ROM attempts to report its onboard memory or CPU capabilities. Because the 030 CPU/RAM is disabled at the jumper level, the card fails to respond correctly to Kickstart's memory/configuration probe.

Kickstart catches this mismatch, throws the non-deadly 88C7 alert, bypasses the card's problematic resource, and continues the standard boot process using your motherboard's native 68000 CPU.

I've got the GVP HC+8 in there alongside the accelerator, and they have identical ROMs (as far as their markings - v4.5) -- dunno if that is part of it, but the HC+8 and the whole system works fine after that one startup Guru.