A530 and WHDLOAD

[CrashMidnick]

hello there,

I have a little issue (or not) with my GVP A530 (no mmu, no fpu, 8mb fast) when playing WHDLOAD games and demos. In most case I have to put as tooktype "cache" in the icon iformation otherwise games and demos are very slow. It does not really matter when we know that. But some of them remain slow. For example, 9fingers with tooltypes "cache" and "NoAutoVec" is slow and has graphic errors. My workbench (classic WB light 16 colours) is fast, working games and demos are playing normal....

I have the gvp map rom and burst cache enabled (tried to remove burst cache but same results).

Any Idea ?

My A2000 with GVP 030@33Mhz has no problem. Is the bus or Ram/chip ram access are different on A500 ?

Thanks for your help.

[Lockon]

Post some more specs on your A500 (MB revision, Kick, Workbench, gvpscsi.device, WHDLoad version); startup-sequence/user startup listing. That's only for SW side. If there's still enough room for guessing after that, I'm afraid we'd need to look inside A530 for jumper settings.

[CrashMidnick]

It is an A500+ so last rev MB
Kistart 3.1
GVPSCSI 4.15 (I have a SCSI/IDE bridge board)
WB 3.1 (WB Classic lite with only burst mode and maprom activated in the user sequence, I did not do anything in ss sequence as I only want to play games with this Amiga)
Jumper settings are set to default... I only disabled MMU as I do not have one. If activated it works too. Same for the CPU cache jumper.
8 mb 70ns gvp fast ram (tried 60ns ones, same result).
Last WHDLOAD version

90% of games are playing as they should (with cache tooltype). For demos I stopped my test because too many are unplayabled : If cache is activated, it is too fast, if disabled too slow...

I have tried with a full 030 from my Blizzard 1230 MKIV : same results


If someone around here has an A530 and could tell me if he has the same problems, it could be usefull

Thanks.

[Lockon]

Yes, I have A530, that's why I guessed I could be helpful...
Please, send a PM with list of affected WHDLOAD games/demos.
My setup is currently disabled due relocation, but I think I could run some tests this weekend.

[CrashMidnick]

Ok I will do it on friday evening as I am not at home this week. As far as I remember :

Games :

Lotus III (slowdown with 2 players, even in the presentation)
Pang : Slowdown a lot when many balls or playing 2 players (try to grab a dynamite when there are 2 or 3 big balls)
Rick dangerous 1 : Graphic errors
Agony : graphic errors when many enemies (it arrives 1 or 2 times per level), little slowdown


Demos :

9fingers : very slow even with cache activated, major graphic errors (if this one one works on your A530, it means that I have a problem somewhere).
Batman vuelve : graphic errors, sometimes freeze
Decaying Paradise : too slow or too fast if "cache" is enabled/disabled

Please note they are all last slave versions (except Lotus III which had a new one few days ago). Tool types used in most case :

Cache
Noautovec
Novbrmove (For the demo "242", it is necessary)

I will try to reinstall my os3.1 on another CF, who knows...

many thanks for your help Lockon !

[JHawk]

Hi,

I've also been having some slowdown problems with my A530 and WHDLoad and had been meaning to post about it for some time.

My A530 has full 030 with MMU, FPU and 8MB, used with A500+ with KS2.04, 2MB chip. I think I've tried with a rev 6 A500 with KS2.04 and 1MB also. WHDLoad 16.8.

I'm using the same tooltypes as CrashMidnick. Haven't tried that many games on it so far, but I've found problems with all of them:

Turrican - some slowdown, particularly when using the surround.

Turrican 2 - slow music on the intro, frequent slowdown in the game when there is more than a few things on screen (not the extreme cases like the fish in level 1-2 where it slows down normally). Music in the game normal though.

Rainbow Islands - slowdown when there is little more than just Bub on the screen.

I also have an A4000/030 with FPU, which seems to run games much better than the A530. I played Turrican 2 to the end on it recently and don't recall any slowdown at all. Most of my WHDLoad-ing has been done on an unaccelerated A1200 and that performs wonderfully. I prefer the A500 and big-box machines though.

Now that I come to think of it, I think even loading Turrican 2 from disk with the A530 on had the same problem.

I was disappointed by the A530 because of this (to me, any system that doesn't play Turrican 2 properly is unuseable) and got a HD8+, but I think it may be faulty - that's probably best left for another post though.

I'll put some more games on and see how that goes.

[CrashMidnick]

Thanks a lot for your post mate I will try those games on my side to see if I have the same results

It is like the ram access are slow or something like that. As far as I know, Whdload games are fully loaded in Fast and Chip ram before running so Hard drive access cannot be faulty.

Before, I had only 2mb fast RAM and whdload took the needed ram into Chip ram and slow chip ram (A502) : it was completely unplayable !!

The strange thing is that some "more powerfull' game are running as they should and a "little" game like rick dangerous has problems

It could be an instruction in the slave itself which the gvp530 does not like...

[Lockon]

In order to rule out HD access completely, you should edit WHDLoad tooltypes in order to include 'PRELOAD' option. That will end up with disk image/files being read right at the slave init, so you spare some disk access lag later, at the expense of FastRAM.

About Rick Dangerous, I clearly remember both WHDLoad versions worked great for me, that few months ago.

[Lockon]

I got some results...

Lotus III - slowdown confirmed
Pang - no slowdown
Rick 1 - no graphic issues
242 - fluid, no graphics issues
9 fingers - fluid, some graphics glitches, less than 10% of total demo time
Decaying Paradise - graphic glitch on early scene after loading, speed OK
Batman vuelve - graphics glitch on scene#1, stays stucked on scene#2

Test specs:

A500+ rev8.1, 2MB Chip, 8MB Fast, 68030@40, full-MMU, 68882@40, GVP 4.15
2GB CF + IDE CDRW through Acards, clockport, Subway USB, Dlink USB-Ethernet bridge
OS3.1, OS3.9 BB4, WHDLoad v16.9, no maprom active.
NoAutoVec is decalred globally in WHDLoad.prefs since quite a few games throw up an error requester box.

[CrashMidnick]

Thanks for the information.

I have just finished to set up a complete new system under winuae. I have noticed that skick was not installed. I have activated Burst, cache and maprom modes as well in the user startup. We will see this week end what hapens.

Did you try it on 3.1 or 3.9 ?

I will try another CF too, it is a kingstone and I know that some of them are not working well on some systems.

I will check all jumpers again and set the CPU to 40 mhz (I have overclocked it to 50MHZ).

Could you tell me what is your transfer rate in sys info for your HDD ? I have around 1.1 mb/s with FFS file system...

Thanks.

[JHawk]

I've tried the same games too:

Lotus III - graphics seem to run at about half framerate on the A530 even in one player mode.
Pang : I also get slowdown when a lot of balls are on screen
Rick dangerous 1 : was fine for me. Didn't get very far though - just past the room with the 3 guys and the mask on the bottom level.
Agony : the screen flickered a lot but didn't notice any slowdown.

Demos :
9fingers : Hadn't seen this before so tried it on the HD8+. Great demo! Definitely getting graphics errors on the A530.
Batman vuelve : Another one I hadn't seen before. Some glitches in the graphics, and froze on the screen with the cityscape and the electronic billboard. Sounds like the same as Lockon gets
Decaying Paradise : didn't try this

Others:
Formula One GP - seems ok, but wish it used the 030
Stunt Car Racer - while since I've played this. Thought it just has a low framerate anyway, but after seeing Lotus 3 I should try this again on a plain 68000.
Head Over Heels - fine, but hardly pushes the hardware.

Leander - some slowdown
James Pond 2 - slowdown
Desert Dream - some parts in the second half of part 1 were definitely slow and it lost sync with the music. I've seen it lose sync on other machines in the first half with the pyramids though - maybe that was a different slave?

I seem to have the HD8+ working now - it has a bad PSU which causes the disk to spin up and down occasionally which may have been causing reported disk errors. Works fine with the A530 PSU though. Had failed to get 8MB to work previously but got it to work tonight.

Can anyone confirm that the HD8+ should always be powered on with fan and disk spinning when it has power connected, unlike the A530 which only turns on when the Amiga is powered on?

[CrashMidnick]

So same problems as me for most games and demos... Humm did you try CACHE tooltype ?

I have emailed the whdload team if they did hear about it before... Will see if I have an answer.

For the HD+8, I do not know but I have seen one on ebay a day and it seems to power on even if it is not connected to the AMIGA (I remember the green light activity). If it can helps.

Thanks you so much guys for your precious help and taking time to test all those games

Cheers.

[Lockon]

Sorry, I misstyped, Kickstart 3.1, OS 3.9.
Transfer speed from various benchmark programs report a bit over 1MB/s.
That is valid for current Sandisk 2GB CF and was also same for previous Seagate HAWK SCSI drive. Filesystem is not a factor here, although I have 4 partitions of different filesystems (old FFS from KS1.3, FFS from OS3.9 BB4, latest SFS and latest PFS3).

Overclocking to 50Mhz might open some extra space for timing issues, notably for GAL chips, but maybe even for a RAM too.

Regarding power up sequence on HD8+ and A530, I expect that they both behave in same manner since, A530 is basically HD8+ with extra provisioning for CPU/FPU and 80286 bridging. Those expansions should be turned on before Amiga anyway. Exception from this assumption might be the fact that HDD SCSI interface in HD8+ is CPU independent - it doesn't touch CPU bus mastering role which happenin case of A530 (68030 hijacks 68000 role) so at the end you might end up with spinned up disk while A500 is still powered off.

In case of A530 I can confirm that powering up brings only +12V to main relay, which waits for A500 edge connector signalling (A500 power up) in order to pass +5V to the rest of A530. I've modified my A530 to drain power directly from A500 so the overall system power up with single switch (that was not possible while SCSI HDD was inside A530 since motor spinup takes too much current and thus cannot provide sufficient amps for the CPU/logic).

[CrashMidnick]

So I have tried everything but still the same problems.

What I have done :

- Take another CF card
- Tried on OS3.9, WB classic 3.1, my OS3.1
- Unabled all GVP software (FastRom, BURST...)
- Different jumper settings
- Install SFS
- ...

I have tested 209 games :

Here is my list :

S = Slow or Slowdown at some points
GFX = Flickering or graphic errors
F = Freeze sometimes
C = Crash at boot

- Agony : GFX (very few when a lot of enemies are on screen)
- Alien 3 : S
- Bubble Bobble : S (during the intro)
- Crazy Cars : C
- Crazy Cars III : C
- James Pond 2 : S
- Jaguar XJ 220 : S
- Jurassic Park : S
- Lotus 2 : F (Random freeze, it happens around after 5-10 minutes)
- Lotus 3 : S
- New Zealand Story : S
- North & South : S
- Pang : S (only when there are a lot of balls)
- Overdrive : S & GFX
- Risky woods : S
- Saint Dragon : S
- Shadow of the Beast III : GFX
- TOKI : S
- Zool : S
- Zool 2 : S

So we can say that around 10% of games are not working on the GVP530

Turican 1, 2 (some very little slowdown) and 3 are working fine on my system. Rick dangerous is working now (I took a new version), 242 demo is working if novbrmove is activated. Rainbow Island is working as well.

@ JHawk : do not forget to put "CACHE" as tooltype for those games

[Lockon]

Only 209 ?

LOL...seriously, that's really amazing.
10% isn't so bad if we rule out eventual specific HW issue bound only to your A530 unit.
This accelerator design, inspite high reputation (read: market demand) praised at all times, is somewhat left incomplete. It unique provision for PC bridgeboard might caused it being unable to support Fast RAM and SCSI interface in 68000 fallback mode; too bad there's no service manual explaning what those numerous 'service only' jumper really do.

[zipper]

The first thing with an 030 in case of incompabilities is to switch the data cache off - as 020 doesn't have one the programs made just a 000/020 in mind may behave badly with a processor having a data cache.

[CrashMidnick]

Thanks Zipper for your answer but just tried it and nothing better... I am pretty sure there is something in some slaves that the 530 does not like but what ?...

[Wepl]

I doubt that the memory speed is the reason because several of the games named are using only chip memory. To verify you could run 'src/memory-speed/speed.slave' from the whdload dev archive and compare the result.

More likely for me are some interrupt problems. Are all the games require NoAutoVec? What errors is displayed without?

[CrashMidnick]

Thanks for your help Wepl !

I would say around 40-50% games need the NOAUTOVEC tooltype (pang, dune, james pond 2...

The main error is :

"Exception "NMI Autovector" ($7C) at $DDE8 occured" (the $DDE8 numbers can change depending on the game of course)
"Exception "NMI Autovector" ($7C) at $8E03EE (expmem $293EE)"

I had other errors that have been solved by putting NOAUTOVEC but cannot remember their code. What I know is that there is no AUTOVECTOR word in the error description.

Concerning the speed.slave, I am not sure how to use it. Could you explain me ? I have tried to double click on it and I have an error, tried to rename it and replace a game slave by it and then lauch the game : I had many code lines with a lot of numbers for chip ram, Fast RAM...

Could you tell me why it is the only accelerator I have that need the "cache" tooltype on whdload games/demos ? Sometimes I put "dcache" and it is a little bit faster but often results in corrupt graphics...

Thanks.

[Wepl]

Originally Posted by CrashMidnick

"Exception "NMI Autovector" ($7C) at $DDE8 occured" (the $DDE8 numbers can change depending on the game of course)
"Exception "NMI Autovector" ($7C) at $8E03EE (expmem $293EE)"

if there are many NMIs this could slow down, the question would be how many of them are occuring.

Originally Posted by CrashMidnick

Concerning the speed.slave, I am not sure how to use it. Could you explain me ? I have tried to double click on it and I have an error, tried to rename it and replace a game slave by it and then lauch the game : I had many code lines with a lot of numbers for chip ram, Fast RAM...

just run 'whdload speed.slave' from the command line, then save the picture and post it here

Originally Posted by CrashMidnick

Could you tell me why it is the only accelerator I have that need the "cache" tooltype on whdload games/demos ? Sometimes I put "dcache" and it is a little bit faster but often results in corrupt graphics...

CACHE is not the solution for the problem with this acc. only a try to work around the performance problem