Amiga 4000T - WHDLoad and demos

[highpuff]

A few months ago I purchased an Amiga 4000T (Escom) and moved all of the expansions from my Amiga 4000D to it. Everything works well, all but a few things that I try to run on through WHDLoad. Some old favorite demos that I like to run now and then is slow and no matter what I do, it will not be good.

When I run normal AGA demos without WHDLoad, they run without any problems. The demos with the most problems are:

• Nexus 7 - Slow as hell but works better with NoCache enabled.
• Complex A1200 demo - Slow and seems to point bitplane pointer to scrap data.

If I then test Enigma demo, it works.

The 4000T is newly recapped and has both a new Paula and CIA-A/B chips. I have also made some tests with LogicaRoms (3630 board sice it does not work with 060) and there are not issues found.

The following hardware was moved from my Amiga 4000D to the new A4000T

• CyberstormPPC - 060 72/50MHz + PPC 200MHz + 128MB memory
• CyberVisionPPC
• X-Surf 100
• ACard with 60GB SSD (via SCSI on PPC card)

Other configuration:

• 16MB Onboard FastMem

The following has also been tested:

• 3630 (demos works)
• Moved the CSPPC back to the 4000D (demo works)
• Other Fast Memory modules

Does anyone here have a A4000T with the same configuration that can confirm that it is a problem or not? That would really help me a lot

Thanks in advance

[Jope]

Most WHDLoad demo slaves are not very 060 compatible. This is not your computer's fault per se.

[highpuff]

Thanks for your answer Jope. But the main issue is that the same accelerator works with WHDLoad, both the old and on a new installation, on my A4000D.

That is what puzzles me.

[Jope]

Ah, reading comprehension at its best again.

Now that is indeed curious. Did you also transfer the hard disk so that you have the exact same software on the 4000T?

[highpuff]

I'm using the same drive as I did on the A4000D (with Workbench.library in LIBS and also made a new installation from scratch (AOS 3.9 BB2).

But! Everything else works good(!!!).

[roomeo]

I have the same problems with the demos you mention. Running a 4k with a3640. 50% of demos/games have small or big issues when launched from whdload.

Have found it best to ditch whdload. Nexus7 is an exe file, so no need to use whdload in that case.

When running an a1200 with a b1230iv back in the days everything just worked, but with a 4k/040 there are compatibility issues all the way. Some report minimal issues, so it might be some machines just beeing moody.

[highpuff]

roomeo; Tried the .exe version of Nexus7 and it was slow as when I did not use NoCache ToolType.

So moody machines? Wonder what it is that causes them to be moody. Nothing showing any memory errors either

[roomeo]

I might had to turn of caches on nexus7 myself. Don't have my miggy here, so I cannot test. I usually run demos from shell after booting with no startup sequence.

Dunno what makes a machine moody, but since some have no problems, that's my uneducated conclusion. Having a big collection of old-school demos as whdload is pretty much useless, since most have issues. Not fun when you are about to have a demo session with some friends.

Could it be that the whdload installs should target 040/060 better when they are made? Anyone who could comment on that?

[roomeo]

Originally Posted by highpuff

roomeo; Tried the .exe version of Nexus7 and it was slow as when I did not use NoCache ToolType

Btw, can you set nocache tooltype on a regular exe file? If so, I learned something new today. thought that was a whdload thingy.

[highpuff]

No I meant that the .exe version ran as slow as the WHDLoad without NoCache ToolType setting active

[jotd]

Demos are more sensitive than games because they use more timing hacks, and all, and they push the machine more than Tiertex.
NOVBRMOVE may help since whdload does not interfere with interrupts (and no quit key if no one has coded it specifically in the slave, like for 68000 machines)

[roomeo]

No I meant that the .exe version ran as slow as the WHDLoad without NoCache ToolType setting active

Ah.. I must learn to read properly.

[highpuff]

jotd: But why the difference between A4000D and A4000T?

[Psygore]

It seems to be the same problem than the blizzppc on a1200: slow access to chipram by the 68k cpu (on A4000T but not on A4000D).
Nexus 7 and Complex A1200 demo have code in chipram and was designed to run at least an 020/14 mhz with instruction cache enabled.
You can test your amiga with speed.slave (whdload package) and compare the result vs a plain a1200.
You should have the same issue with Super Stardust too.

[highpuff]

I will do that tomorrow morning and post the result here When you say it, I do have the exact same problem with my Amiga 1200 with BPPC.

Thanks for the reply Psygore!

Originally Posted by Psygore

It seems to be the same problem than the blizzppc on a1200: slow access to chipram by the 68k cpu (on A4000T but not on A4000D).
Nexus 7 and Complex A1200 demo have code in chipram and was designed to run at least an 020/14 mhz with instruction cache enabled.
You can test your amiga with speed.slave (whdload package) and compare the result vs a plain a1200.
You should have the same issue with Super Stardust too.

[Locutus]

Generally speaking, 030 accels have faster chipmem access then 060 boards, try if those demos are faster with the 030 board, i wouldn't be surprised if they are :-)

[highpuff]

Yes, that I can confirm right now. With my A3630 they run really good. But with 060 the run as TBL Rift would do on a 020.

Originally Posted by Locutus

Generally speaking, 030 accels have faster chipmem access then 060 boards, try if those demos are faster with the 030 board, i wouldn't be surprised if they are :-)

[highpuff]

Psygore:

See the result from running speed.slave with "whdload speed.slave" from CLI on my Amiga 4000T below.