Hooray! :DancingBanana:

Woohoo!!!!!!!!!!  Imagine playing Total Chaos AGA at 10x the speed! That is faster than WinUAE can do it!  :Excitement: :Shock:

Baby steps seem like leaps here. :P

You've made my day! :D This is what I call an update.

Good luck and may the NatAmi force be with You!

Imagine playing Total Chaos AGA at 20x speed with NatAmi 32bit:D !

Respect

Great work

Comparing with AGA speed is complex to do as Video DMA was 4 times as fast on AGA as Blitter DMA.
Therefore if the LX is 10 times faster than AGA-Video DMA then this means the Blitter is actually 40 times faster than AGA, doesn't it?
 
But whether the 16Lx Bringup Board Video DAM does run as of today with 10times or 20times AGA speed and that Blitter DMA runs 40 times faster than AGA is not really important.

Until Thomas does say that the memory interface runs at full speed and that he finished tweaking the performance - I assume that there is still room for performance improvement. ;-)

I think the 10times/40times performance is just a baby step ...

I'm with the others, please keep us updated this way, it is so exciting to know every little step of the birth of NatAmi!

To use approximately 100% DMA power on AGA one must open a screen of
1280x512x8bpl @50hz
1280x400x8bpl @60hz
640x512x8bpl DBLSCAN @50hz
640x400x8bpl DBLSCAN @60hz

However, the 640x512x8bpl DBL modes are so absurdly slow that nobody in their right mind ever uses them. :loco:  They either view that mode in interlace or they have a hardware flickerfixer/deinterlacer/scandoubler thingamajig.  All us "hardcore Amiga users"(tm) :) have a hardware flickerfixer as it doubles the speed of the AGA chipset while allowing the use of standard cheap PC VGA/SVGA monitors.  The latest hardware flickerfixer to be produced is the Indivision AGA (Hi Jens! :) and is quite popular.

So on my Amiga, 640x512x256 color mode 31Khz doublescan only uses 50% DMA bandwidth.

There is an option to play Total Chaos AGA in the DBLPAL mode but nobody ever uses it, its quite dreadful.  So all the Total Chaos fans are playing it in a 50% DMA Bandwidth mode.

Another troublesome part of the calculation is that since the NATAMI SAGA chipset is doing the doublescanning, this means that on SAGA, 640x512 takes 2x the display bandwidth of 640x256.  On AGA, 640x512 and 640x256 each take the same amount of bandwidth due to the "interlace trick".

Comparing with AGA speed is complex to do as Video DMA was 4 times as fast on AGA as Blitter DMA.   Therefore if the LX is 10 times faster than AGA-Video DMA then this means the Blitter is actually 40 times faster than AGA, doesn't it?

Yes, I agree.  The LX Blitter should currently cruise at 40x the speed of the AGA blitter.

But my asm blitting routines have been up to 72x faster than the AGA blitter for over 10 years, so the 40x faster blitter doesn't help my Total Chaos game coding any.  I expect my blit routines to go anywhere from 4x to 8x faster on Natami.  So I expect Total Chaos blits to go around 300x faster than the AGA blitter could do them.  Zoooooooom!  Natami FTW!
 
Of course the 40x Natami blitter will still greatly help out:

Workbench apps that do their blitting with the blitter
Badly coded AGA games
ALL OCS/ECS games
AMOS games
Blitz Basic games
etc.

But it would need to be a 300x faster blitter to benefit Total Chaos players.

How will Total Chaos benefit from a faster blitter?

This counter also tells only the bus load an the NatAmi bus - which is completely different and absolutely not comparably (for calculations) to the original Amiga bus. I wanted to explain the differences in the "deviation" thread but it got completely off-topic. So the only essence here is that this counter shows that the NatAmi is not at its limit with a common AGA display.

The Blitter is a different thing. Currently I am thinking about the Blitter design concept. I could use the old "unbuffered" Blitter from the C-One design. I am sure this would work with some minor issues which I never fixed. But it will not be much faster than an original Blitter. I am not sure if it would be faster at all because there is at least a 22 clock wait penalty for every 16bit access. And there are 4 accesses necessary for one 16bit word. So I am thinking about a "buffered" Blitter which pipelines the accesses. The buffered Blitter is not  as easy as the bitplanes because it has an ascending and a descending mode and supports self-modifying data.

The Blitter has two memory interfaces, one for read, one for write. The buffered Blitter will be able to read and write "all the time" while it does the calculations pipelined in the background. So either the Blitter will be faster than the CPU or there is a serious flaw in the concept.

Just a question that still lingers.
The buffered blitter version, it does R/W ChipMem and Read FastMem
Or did you real mean that it can read only one type of memory and write the other?

I think i ansered the question already myself but someone please reply.

Team Chaos Leader wrote:

But my asm blitting routines have been up to 72x faster than the AGA blitter for over 10 years, so the 40x faster blitter doesn't help my Total Chaos game coding any.

I think you/we are mixing throughput numbers with algorithm efficiency here.
Even the fastest Cyberstorm 68060 board could at best reach 8 times the throughput of the old OCS Blitter.

The NATAMI Blitter could reach a throughput far higher than a 68060 can reach.

I know you are doing a very unusual case in Blitting for which the normal AMIGA Blitter operation is not optimal.
I think what you look for a some "clever" blitter operation which is more optimal for you special case.

I thought about some blitter optimizations the other day when was coding on 194x. I think those simple optimizations will increase the efficiency by 400% for your usages cases too.

I'll propose the idea on the "dark" side of the forum - we both  only need to twist Thomas arm to implement it.

The buffered Blitter will be able to read and write "all the time"

Yes.  But so do my asm blitting routines.  Gotcha ;)

while it does the calculations pipelined in the background. So either the Blitter will be faster than the CPU or there is a serious flaw in the concept.

Yes there is a serious flaw in the concept of the Amiga blitter.  It has always been so.

The main reason I battled and fought and scratched and clawed my way onto the Natami team was to try to get you to fix the conceptual flaw.  I know I will succeed once I have a working Natami and can do some timing tests to prove things to you.  I tried to explain it to Gunnar in English but he absolutely positively would not believe me.  So I wait.