My A500+ GVP A530 feels so damn slow

[bloodline]

mr_a500 wrote:

Both the Viper and the A530 are connected to the same bus, just at the opposite ends.

So, you're saying that expansion through the CPU slot would be the same speed as expansion through the side expansion slot?

They are both basically ZorroII

[Amigaz]

patrik wrote:
@AMIGAZ:

Chech if exec.library is in fastmem with Sysinfo - the library list shows in which type of memory each library is allocated. If it is in chipmem, use FastExec to move it to fastmem.

Burst enabled for the datacache might make things slower on a 030. Burst speeds up sequential accesses and punishes non-sequential accesses and data accesses usually arent extremely sequential. This in combination with the 030s small cache..

So skip the "Cpu CACHE BURST" and just let SetPatch enable the caches as usual. It will enable both instruction and datacache for you, but only enable burst for instruction cache (instruction loading tends to be very sequential).

As others have mentioned, the more colours you use, the less chipram bandwidth will be available when say rendering graphics:
Hires/Hires Lace 2 and 4colours: 3.5MB/Sec write (max)
Hires/Hires Lace 8colours: 2.5MB/Sec write
Hires/Hires Lace 16colours: 1.5MB/Sec write

No read results, but I dont have them in my head and the A2000 is tucked away at the moment.


/Patrik

I turned off the burst cache today and it feels a bit quicker and Sysinfo shows that the exec.library is in fastram
I'm running in highres laced at the moment which could explain the big slowness and also using FFS which is the slowest filesystem in earth  :sleep:
Anyway..I'm going to make an AIAB testresult so you guys can see if it's sloow or is it just me.  :-)
Maybe the ECS graphics is the big thief here which makes everything so slow.
i'm also positive that the kickstart is loaded into fast ram since I just commented that command line out of my startup sequence which gave me more chipram after that

[patrik]

@AMIGAZ:

You shouldnt get more chipram when disabling the remapping, but more fastram. If the kickstart is mapped into chipram, it will be even slower than it is originally.

To see if the kickstart really is in fastram, you could use bustest and check that you get more or less the same read-speed for "bustest fast" and "bustest rom".


/Patrik

[Doppie1200]

AMIGAZ wrote:
I'm running in highres laced at the moment which could explain the big slowness and also using FFS which is the slowest filesystem in earth  :sleep:

I don't think the FFS is the bottleneck. The A1200 I have here runs fine with it. It has an 060 inside that is picking it's nose most of the time because of the chipset. Like others said; free up cycles for your CPU by reducing the load on your chipset bus. Or else your CPU has to wait longer before it can use it in case it needs to access something attached to that bus. The A1200 has a wider chipset bus compared to your A500. So the 030 inside the A500 may need more cycles to do the same on that bus. Like a memcpy from fast to chipram.

Anyway I am curious about the testresults.

[Amigaz]

patrik wrote:
@AMIGAZ:

You shouldnt get more chipram when disabling the remapping, but more fastram. If the kickstart is mapped into chipram, it will be even slower than it is originally.

To see if the kickstart really is in fastram, you could use bustest and check that you get more or less the same read-speed for "bustest fast" and "bustest rom".


/Patrik

lol, I meant fastram  :lol:

[srg86]

Agreed, appart from the A530 mounted fast RAM, you have a 32-bit CPU connected to a 16-bit system bus. That's your bottle neck right there IMHO.

The stock A1200, with it's 32-bit bus (and about twice the CPU clock speed) is supposed to be about 4 times faster than a stock A500 with it's 16-bit bus. I expect it is similar for an A500+.

[zipper]

I remember when testing my A500T with Blizzard 2060 and 2 MB Megachip2 the max speeds it could get thru the chipset were between 0.8 MB/s and perhaps 2.1 MB/s and that was when the graphics were handled by a PicassoIV leaving the chipset on very low strain. So with those speeds you can't expect too much. But everything that runs on the processor/memory loop is a different thing. I remember when launching my A500T first time with Blizzard 2060 using just the ECS chipset I was quite disappointed what was happening on the screen. Under two weeks I had acquired a RTG card - what a huge difference!