ACA630 Review (well, first impressions in my A600 portable)

[Photon]

Mailman came around today So I changed plans from 8 hours of coding to 8 hours of modding. No, it didn't take 8 hours to install the ACA630. But I just thought it was time to finish the surgery for V2 of my A600 Portable so I'd have it ready for action (as in demoparties), finally. It had lain sprawled, guts open, in my solder-room for too long.

Installation was very easy, I just removed my A6055 8MB fastram expansion and snapped on the ACA630. Still with the case open, I powered it on and it worked fine. I ran a few demos (some bugged, as expected since they were written way before 68030s were put in Amigas), and then fired up SysInfo.

Raw boot:


I then soldered on the fastmem on-off switch wires to the accelerator on-off pins of the ACA630 and tested the 68000 mode. All fastmem was lost, which was a bit of a letdown. Maybe it will be available with some future ACATune option, but I'm not holding my breath as this is a feature of most accelerators - memory is connected to the 32-bit CPU bus only.

Actually it guru'd on the first boot in 68000 mode, but that was just a problem in my existing startup-sequence on the Lexar CF-card. It ran AsmOne and a few demos perfectly (I think by this you can see what use my Amigas get all day eh? . But WB/WHDLoad test follows, don't worry...)

With the hardware test done I secured the accelerator with the two screws that came with it, making sure I didn't overtighten. I added dabs of hotglue to the black standoffs and another on the CPU clip-on socket after making sure it was flush with the motherboard. It didn't feel very necessary, it already felt like it sat more securely than the previous fastram expansion.

It took perhaps 30 minutes includes testing and soldering, and I then closed up the case, see the picture below.

A600 Portable II:


See my blog post on scoopex1988.org for what else I added.

With the raw boot specs I got a measly 1.1 MB/s from the CF card, but I must blame the Lexar card. While working flawlessly, its performance has always been noticably below Sandisk cards.

My normal "accelerator" startup-sequence on Sandisk Extreme III:
This gave a jump to 1.20 x A3000.

ACATune:
I copied that floppy (tsk tsk) to my CF card. Installation consisted of copying ACATune to C:. This CLI command did not list its options when supplied none, but the .guide provided the necessary info. The parameters I ended up with was

ACATune -maprom * -chipcache on -z2cache on -fastchip on

While there were more options, like testing out memory timings for fast/chip memory, these were what I felt was compatible maximum.

And I was happy to see a bump from "Cowabunga!" to "Smell the Rubber?" and almost 6 MIPS Actually, I was happy already when I saw the 1.06 x A3000 but it the tweaks took it from 1.20 x A3000 to 1.23 x A3000, at least. Not bad But with the above commandline, the biggest jump was the disk speed - from 1.88 MB/s to 2.4 MB/s!

See the picture for specs:


It has now been on "burn-in" for 5.5 hours and while the case (the underside, mostly) is warm to the touch, it's not warmer than the back of the LCD screen in the picture above.

I must say I was a bit skeptical when I saw the tiny distance (steel-plate?) between CPU and the nice Alu-cover of the ACA630, knowing that there should be as little (none) between CPU and cooler, even a too thick application of cooling-paste is a big no-no for overclockers like me - but it seems to work out.

Maybe I'm overreacting, but I would have liked to see a zero gap between CPU and cooler, and some way of making sure the cooler was snapped tight onto the CPU. But we are talking non-Intel/AMD CPUs here, so

The only other negative thing I can think of so far is that when changing an ACATune option, any option, from the command line, it all works fine but no programs can be started because all chipmem has been allocated and not released on exit. But I'm 100% sure a newer version will fix that.

(All the settings set by ACATune are preserved by a normal reset, and that's why I could test faster settings with any programs... )

Workbench + WHDLoad:

The ClassicWB install I use on all my 3 Amigas worked fine, poked around a little in folders, played games for about an hour and strangely the WHDLoad exit-key that had given me hassle on all 3 Amigas suddenly worked. At least in the 15 or so games I played. Played the usual State of the Art demo, drew a little in Dpaint IV...

Apart from the ACATune command thingy, I can't really think of anything negative. I don't regret spending the money, my main issue with the other accelerators was how secure they'd sit inside - since I bring my A600 Portable everywhere (which means from the TV sofa to the demoparties in Europe that I go to, basically ) I have to know it will never had bad contact or pop off. I don't think this one will. And even if I "only" see 31,208,504 bytes free in Workbench (!!!) I think the 25 MHz version is worth every cent.

30 MHz version owners should then see close to (but not quite) 7.16 MIPS in their SysInfo.

ClassicWB loads in 9 seconds from Ctrl-Amiga-Amiga, 3 seconds from first flash of the CF-card... I set the jumpers to check drives but map DF1 as DF0, since I have no DF0 in my A600. So maybe it could boot even faster...

Hardware: 10/10
Software: 8/10
Ease of use: 10/10
Value for money: 8/10

Now - I personally would give it 10/10 in value for money. It *is* expensive, but it offers features unavailable in old accelerators, such as all the rom and ram mapping, and it's 10/10 for me because I hold the strange belief that I have the best one money can buy for A600.

But maybe I should deduct 1 point because there was no "INDIVIDUAL INSIDE" sticker in the box!! Hehe. Please make some Jens

[Predseda]

Yup, I arranged that "This machine is powered by Individual computers" sticker for the general thread Do you have PCMCIA SRAM card as well? Could you test if it works together with ACA 630? My own ACA should arrive shortly, so I will test it then.

BTW: You should paint those screws bellow the LCD to black as well.

[Leffmann]

That is one serious A600

I have an A1200 with a B1230-50-50 myself, and while I would really like something like a maxed out A4000 I can't really think of what I would use it for. If I were to ever get another Amiga it would definitely be an A600 because of its small size and good looks.

I'd like to have this ACA accelerator in it too if so, but I dunno there seems to be a heap of problems with it reported in the bigger thread in the News forum. People are reporting it has crashed randomly from just sitting idle in the Workbench, and to me it's starting to look like an error in the ACA design.

I hope for the sake of all of you who have bought one that it's something that can be fixed with a firmware update.

Also, what's that small hinged panel next to the screen on your A600?

[chiark]

The panel next to the screen is, IIRC, a trackpad.

Nice work, looks fantastic!

[alexh]

Originally Posted by Leffmann

I'd like to have this ACA accelerator in it [A600] too if so, but I dunno there seems to be a heap of problems with it reported in the bigger thread in the News forum. People are reporting it has crashed randomly from just sitting idle in the Workbench, and to me it's starting to look like an error in the ACA design.

I just looked at the bigger thread and I think all the reports of problems are with the ACA 1230, not the ACA 630.

[Predseda]

So do I.

[DDNI]

looking at the big thread, saying ALL (plural) issues seems a little harsh.
From what I have read it appears that one person has issues and a few others are saying get out a hammer and smash the caps...!

[FOL]

Originally Posted by alexh

I just looked at the bigger thread and I think all the reports of problems are with the ACA 1230, not the ACA 630.

Correct, and its only one person reporting problems with his ACA 1230.
NOT ACA 630.

[fitzsteve]

This is a great project Photon

Thanks for sharing the pics and keep up the good work.

Like any other Amiga upgrade some people will have problems either because of some fault with the Amiga its self or because something is not compatible, software or hardware. Thats just the nature of the beast.

But I don't think its right to discuss that in this thread, this is Photon's project thread after all

Steve.

[Leffmann]

Originally Posted by alexh

I just looked at the bigger thread and I think all the reports of problems are with the ACA 1230, not the ACA 630.

Ok then things aren't as bad as I thought they were, and I hope no problems will surface later on either. It makes me want to get an A600 and an ACA630 even more.

[Photon]

Originally Posted by Predseda

Yup, I arranged that "This machine is powered by Individual computers" sticker for the general thread Do you have PCMCIA SRAM card as well? Could you test if it works together with ACA 630? My own ACA should arrive shortly, so I will test it then.

BTW: You should paint those screws bellow the LCD to black as well.

Nope and nope

An SRAM card for having "fast"mem in 68000 mode is actually not a bad idea, I'm sure there will be no problem with the accelerator turned off. And if it's on, you don't need the extra ram of course...

[Photon]

Originally Posted by Leffmann

That is one serious A600

I have an A1200 with a B1230-50-50 myself, and while I would really like something like a maxed out A4000 I can't really think of what I would use it for. If I were to ever get another Amiga it would definitely be an A600 because of its small size and good looks.

I'd like to have this ACA accelerator in it too if so, but I dunno there seems to be a heap of problems with it reported in the bigger thread in the News forum. People are reporting it has crashed randomly from just sitting idle in the Workbench, and to me it's starting to look like an error in the ACA design.

I hope for the sake of all of you who have bought one that it's something that can be fixed with a firmware update.

Also, what's that small hinged panel next to the screen on your A600?

They may look like hinges and panels but they are buttons and touch-surfaces It's a Cirque PS/2 touchpad. It works fine, at first I didn't want to bring a mouse because it didn't fit in the snug laptop-bag, but at one party I needed to draw some graphics and it was slow work...

Instead of switching 6-7 pins I put a PS/2 female connector inside the trapdoor, so by opening it you can plug in either the mouse or touchpad.

(I believe there's no firmware in the ACAs btw, at least there seems no way of updating it. Also, settings being preserved on reset but lost on poweroff suggests there's no flashable/permanent ram at all, but I didn't look under the hood.)

[Photon]

OK, burn-in lasted 7 hours yesterday and so far I've used it about 6 hours today. Still some setup things to solve but definitely a productive 6 hours.

Some more tests:

I played 15-20 more games (I know, it's a sacrifice but someone has to do it!), all worked fine.

The amount of speedup in 3D games was mixed - I guess it basically has to do with if there's a framerate cap in the games or not, and what the blitter/CPU ratio and syncing (amount of time-wasting) between them is. The happiest surprises were F-18 Interceptor, Frontier, Powerdrome, Vroom... Elite, Starglider II, Virus, and Stunt Car Racer not so much but played more fluidly and gave a nice experience.

It worked out well to put ACATune first in the startup-sequence with the desired options - but it seems if an option is changed, all chipmem is allocated but not freed. Or maybe the first time it's set. Or something like that. So a newer version of the ACATune where this is fixed is definitely needed!

In 68000 mode it had an issue with SetMap, but since there's no problem if SetMap isn't called I'll chalk it up to "old keymap used on new setup" or "wrong version of command". But it's strange that it works fine in 68030 - but that could also prove that the accel is even more compatible than 68000, I guess...

One definite difference is that my SD-adapter in PCMCIA works fine in 68000 mode, but always says "No Disk Present" in 68030 mode, with/without binddrivers, with/without ACATune. Tried -fastz2 off and z2cache off, no difference. But I'm sure Individual Computers didn't make an accelerator that forces all customers to throw away their network, CF card, and other PCMCIA expansions So I will definitely examine this even further... it's strange because the devs/libs etc shouldn't have any >68000-incompatibilities. I tried setpatch 43.6 and other setpatch versions, also two different binddrivers (not that that necessarily is relevant).

Advice to find where the error could be, or procedure to check off the stages needed to make it mount properly would be appreciated.

Startup order is

ACATune -maprom * -chipcache on -z2cache on -fastchip on
binddrivers
setpatch

And then mount SD0: from the commandline manually. But I've also started with empty s-sequence and just typed binddrivers, setpatch, mount sd0:.

Here's the mountlist entry:

  SD0:    /* Our desired DOS name, plus a colon */
    Device = compactflash.device
    Unit = 0 
    Flags = 0
    BufMemType = 1
    MaxTransfer = 0x1fe00
    Mask = 0xfffffffe
    Filesystem = l:fat95
    StackSize = 4096
    GlobVec = -1
    Buffers = 22
    LowCyl = 0    /* "Auto search ON" */

    BlockSize = 512    /* fat95 does not need this here, */
    HighCyl = 1    /* but the Mount command likes to see them ;^) */
    BlocksPerTrack = 1
    Surfaces = 1

    DosType = 0x46415401    /* first usable partition. Works almost always. */
   #

[Predseda]

Okay, try another PCMCIA cards. I am really curious about networking with A600. I also have a strange feeling that there is something wrong in your Workbench. You say "In 68000 mode..." which imho means, that the card is turned off by jumper, so it is like ACA is not present in your Amiga. And if there is any issue in this "mode" I would say that there is some conflict in your system.

For example - binddrivers must be AFTER all assigns and after Setpatch.

[fitzsteve]

Originally Posted by Predseda

Okay, try another PCMCIA cards. I am really curious about networking with A600.

I hooked my A600 up to our Wifi Network with a Netgear MA401, Prism2 Driver and MiamiDX, this was when I was using it with Apollo 630. Ok surfing the Web was not great but ok for text only or low graphic sites and oh Aminet

FTP and Chat/IRC clients should be good.

Steve.

[Photon]

Originally Posted by Predseda

Okay, try another PCMCIA cards. I am really curious about networking with A600. I also have a strange feeling that there is something wrong in your Workbench. You say "In 68000 mode..." which imho means, that the card is turned off by jumper, so it is like ACA is not present in your Amiga. And if there is any issue in this "mode" I would say that there is some conflict in your system.

For example - binddrivers must be AFTER all assigns and after Setpatch.

No, it is in 68030 mode that the mount results in Info saying it's there but contains no disk. In 68000 mode, all works well. And this is in the first few lines of s-seq, so Workbench is not loaded yet (nor is it needed to mount a simple disk).

Are you sure about binddrivers "last"? (Edit: or I should say, which specific assigns do you mean should be before it.)

[icbrkr]

I installed WB 3.9 on my 600 w/A630, and am using the Genesis stack with a cnet.device compatible card without any issues at all. Also, I have Amigakit's CF EasyADF kit mounting a 2GB CF card through the slot without any problems at all as well.

[Photon]

OK, updated the build/project post on scoopex1988.org with some "Finished!" pictures.

Attached a DF1 drive and boots fine as DF0 in 68030 mode (accelerator on), but many game loaders throw up on 68030, of course (at least with maximum settings and no compatibility settings in the pre-boot menu. Normal bootable DOS disks and IK+ original worked (only tried a few games).

With the accelerator off, DF1 is DF1 and still boots, but many old games don't detect which drive it's booting from of course and will stop after the bootblock. (Nothing to do with ACA630, of course.)

Played some more games, F1GP is zippy and very easy to play... Frontier intro is almost as fast as on my A500-040, the only place where you can tell the difference is when the two ships chasing are both on the screen and large in view. Of course the 68040 is 3x-4x faster but that doesn't always translate into 3x-4x faster framerate...

Haven't tried bug-hunting for the cause of the "no disk in SD card" error. I'm using it for coding. It's blazingly fast, even compared to my A500-040! It's perfect, it's just what I wanted... I expected "something faster" but got more than I expected.

[Zetr0]

@Photon

Have you tried this unit with ntsc mode (60Hz PAL) enabled?

I would be interested to see what Frontier would be like =)

[Photon]

Originally Posted by Zetr0

@Photon

Have you tried this unit with ntsc mode (60Hz PAL) enabled?

I would be interested to see what Frontier would be like =)

Flew to a neighboring planet, didn't find any NTSC-bugs, visual or otherwise. If you mean framerate, well, no noticable difference in the intro - in-game the framerate is usually lower than 60 anyway

Summary: 'The music played a little faster'

I guess if you have a VGA CRT the benefit could be a little less flicker, but 60Hz on VGA CRT is worse than 50Hz on PAL RGB CRT.


Additional info:

- NMI pushbutton connector could work both with accelerator on and off
- My existing NMI pushbutton had no conflict with the (unconnected) connector on the ACA630, worked both with 68000 and 68030.
- My existing df0: terminator worked fine with the DF1-as-DF0 jumper. No suprises, DF0 terminated both with accelerator on and off.
- How well any DF1-as-DF0 hack on the motherboard plays with the above jumper is unknown.

I have yet to make ClassicWB hang, guru, or go berzerk with the accelerator at max. So far it has worked with everything, it's perfect in WB.

When booting straight into CLI I've had 4 conk-outs and 1 'all chipmem allocated' in two days of use, I'll blame some command in the startup for those. Can't really relate it to anything specific since it happens so rarely, will post back if I find it.

Update:
OK, I found the culprit for the chipmem-allocation-bug, [Edit: the bpls command that sets the depth of the CLI screen.]

The bug doesn't take effect immediately, you can run some programs and suddenly all the chipmem is taken. Since I was testing ACATune options at the time, the chipmem was hogged after running the ACATune program. And since I've had no problems on other accels I suspected it had something to do with the ACA630 specifically. Not so.

[snipped out misdirected criticism of FastFonts]

Update 2:
Made a replacement command to set depth of CLI screen (and also sets the system font). Released here. Direct download here. Uploaded to Aminet.