A530 and WHDLOAD

[CrashMidnick]

Thanks damion !

So I Have an issue with my A500+. I have another MB here that need to be fixed. I will try on that motherboard at the end of the week. Hope I will find same results !

Damion could you tell us if you have full 030 and/or FPU ?

Cheers.

[Damion]

Originally Posted by CrashMidnick

Thanks damion !

So I Have an issue with my A500+. I have another MB here that need to be fixed. I will try on that motherboard at the end of the week. Hope I will find same results !

Damion could you tell us if you have full 030 and/or FPU ?

Cheers.

Full '030, but same behavior with the original EC part. No FPU yet (going to install one in the next few days and test again). Motherboard is a Rev 6, nothing too fancy, just an ECS Denise and MegaChip. Old SupraRAM 1/2 MB in the trapdoor. Software is basic 3.1 install with MuTools (mufastrom and mufastzero).

[Lockon]

@Damion
It's bit off topic, but now when we have at least 2 working A530 setups, can you also measure SCSI transfer speed ? I get 1Mbyte (1080KByte precisely) reported by SysInfo and few other benchmarks. I'm using GVPSCSI v4.15 through ACARD AEC-7720 bridged 2GB SANDISK compact flash card.
Thnx

[CrashMidnick]

I have made some more tests on my system(s) and discovered 2 things.

1) I launched bustest with no startup sequence option from boot. Chipram tests are quite the same as yours now. So why is it so different from the WB test ? Colours and background in fact. I set my system to 4 colours (16 before) and set no background. Now, results are better than yours under WB but still the same problem of chipram access. See pictures attached. Could you tell me the number of colours you set up on yours systems ? Could you make a test at 16 colours ?

2) I have tried to boot from a standard A500 (512ko fast, 512ko slow, rom 3.1, same HDD). The result is... suprising. In fact the slow chip ram is not recognized ! Bustest does not start the tests of chipram (the main window is here, but nothing starts). If I remove the accelerator, 1mb of chipram is well recognized. I do not understand why...

[Lockon]

Nice findings for 1)
My OS3.9 uses 8 colors, just tested 16 colors, results for chipram are 0,6-0,7 MByte/s.
Test with 4 and 2 colors showed no increase in chipram performance when compared to 8 color mode, so I guess 8-color could rest as sweet spot.
With respect to 2), where do you keep 512K fast and 512K slow ?
I thought you have 512K onboard CHIP and trapdoor 512K wired as CHIP via motherboard hack ?

[CrashMidnick]

No hack on this one. When I said "fast" I meant "standard" 512ko chip... + 512 ko slow chip with A501 expansion. I will try to hack this motherboard to have 1mb real chip

The other A500+ with mother board hack... is not working (one or more additional chips is/are dead). So they have been unsoldered to find original configuration again and now everything is working but have only 1mb chip...

Thanks for your tests of colours

[Damion]

Originally Posted by Lockon

@Damion
It's bit off topic, but now when we have at least 2 working A530 setups, can you also measure SCSI transfer speed ? I get 1Mbyte (1080KByte precisely) reported by SysInfo and few other benchmarks. I'm using GVPSCSI v4.15 through ACARD AEC-7720 bridged 2GB SANDISK compact flash card.
Thnx

I was getting around 2.2 MB/s with an old 1GB Quantum drive, synchornous transfers enabled, and a GURU-ROM, measured with RSCP.

CPU use was quite high. Some 15% free, vs about 98% free with the TekMagic. I'll post some more detailed benchmarks when I get a chance to mess with it again (hopefully) next week, with and without the GURU-ROM, and possibly a CF drive... I'd like to try the same setup as you for noise and heat reduction. Going to be a challenge to fit everything in there.

[Lockon]

Oh, yeah..

A530 is pretty cool outside, but also needs some magic to cram everthing inside.

[CrashMidnick]

Great job Lockon ! I need to do the same thing for mine as I only put an insulation sheet between the bridge and the gvp motherboard to avoid any contact...

[Wepl]

if still interested in:
here is a new speed.slave which should work on 68000 too

it counts the amount of memory accesses per 1/11 second

each row for different memory/size/read or write:
cia - cia register
cust - custom register
chip - chip memory
exp - expansion memory (maybe chip or fast depending on what is present)
slv - whdload slave memory (different to exp only regarding mmu/caches)
in the upper half the check code is in chip memory
in the lower half the check code is in expansion memory
in the head lines you see which address is chip/exp/slv

the columns are for different mmu/cache settings:
nc - no cache
ic - only instruction cache(20-60)
ss+bc - superscalar(60) + brache cache(60) + ic
dc-wt - data cache + mmu write through + ss+bc
dc-cb - mmu copyback + dc-wt
sb/nwa - store buffer(60) + no write allocation(30) + dc-cb

[springycz]

I recently bought a GVP A530 and now I encountered same problems with WHDload games. So let me revive this thread.
Did you have a breakthrough in the last six years?
Do you have any new insights, tips, and treatment for this system?
How should I use that speed.slave?

[springycz]

Really no news? WHDload 18.3 didn´t change anything for me.

[Wepl]

It wasn't diagnosed as a WHDLoad problem, so nothing has changed in this direction.
You may try option NoAutoVec if it is caused by NMI interrupts. It the reason is the chip memory speed, then there is no solution.

[springycz]

Ok, thank you! I also have different problem, probably with fast RAM...