agranlund wrote: Wed Aug 21, 2024 8:18 am
Those were my expections since I haven't really heard anyone talk much about the W32i.
But here there is a significant difference when compared to the AX card I was using:
This has to be using the hardware blitter, or it has an internal speed difference that is just nuts compared to my other card.
PhilC wrote: Wed Aug 21, 2024 9:21 am
They are still surprisingly quite affordable. Typical I just fixed my Mach 32
That Mach32 is super nice though!
The "only" thing kind of missing is trying to make some code to switch the card between IBM8514 and VGA operating modes at runtime so one can get into 320x200x8bpp resolution for games, and then back again to the desktop.
There's a reason why we ended up getting cards specially optimised for gaming but the Mach32 can be convinced to give it its best shot
(all credits for this fun experiment goes to @Badwolf)
It really starts questioning life when landing on the planet but it's still a valiant effort from what is essentially a clone of an old CAD/GUI accelerator.
Resolution don't seem to matter, it's the same in 1280x1024 as 640x480 - perhaps it's the drawcall count that's slowing it down.
Wonder how much is VDI overhead and how much it could be improved by direct hardware access, then one could let the cpu continue with gamecode and math at the same time as the card is rendering triangles too. One would probably want to double buffer this in vram as well.
PhilC wrote: Wed Aug 21, 2024 9:57 am
Still looks good to me @agranlund. I've got an et4000 handy as well so it's no problem.
What's next? Trident 8900 & 9000 or an Oak chipset?
There's a Cirrus Logic 5426 somewhere in transit to here.. I couldn't resist after seeing the code in svgalib and linuxfb.
I doubt I'll ever make a VDI driver for it a'la Ideks awesome work with the ET4K and Mach32, that's a ton of time and work I'm not ready to put into it -- I just wanted one to have a play around with in some fullscreen type direct-access fashion. At most, possibly EmuTOS support.
frank.lukas wrote: Wed Aug 21, 2024 7:47 am
The W32i chip has a hardware blitter like the MACH32 but the Nova driver doesn't use it I suspect. In my tests the W32i is just as fast as an ET4000ax. The W32i can manage more RAM. There are circuit diagrams online. Perhaps the 1MB in 8x 256kx4 chips can be replaced by 1mx4 chips (4MB) on the card. the W32i has a RAM address line A9...? There are RamDAC chips out there from BT up to 110Mhz.
This surprises me since the drivers support the blitter and the benchmarks, at least in my case, confirm that the W32i is way faster than the AX.
agranlund wrote: Wed Aug 21, 2024 9:50 am
One would probably want to double buffer this in vram as well.
Ah the old VDI flicker! Lovely!
I've always been a little hazy on the purposes of the logical and physical addresses you pass to vsetscreen(). Can the VDI be tricked into off-screen rendering (which natively is not supported) by clever manipulation of these somehow?
BW
DFB1 Open source 50MHz 030 and TT-RAM accelerator for the Falcon Smalliermouse ST-optimised USB mouse adapter based on SmallyMouse2 FrontBench The Frontier: Elite 2 intro as a benchmark
agranlund wrote: Wed Aug 21, 2024 9:50 am
One would probably want to double buffer this in vram as well.
Ah the old VDI flicker! Lovely!
I've always been a little hazy on the purposes of the logical and physical addresses you pass to vsetscreen(). Can the VDI be tricked into off-screen rendering (which natively is not supported) by clever manipulation of these somehow?
BW
I'm sure this is not applicable through the VDI interface, so it's a bit of a tangent, but this guy here does a lot of interesting retro-tech stuff and one of his projects was 3D rendering on a MachX card by means of direct register programming (in MS-DOS)
He ended up doing the hardware triangle rasterisation to outside the visible screen and blitting the result to the visible portion of vram.
I really only guessing but presumably you might be able to hardware "flip" by modifying a vram pointer register too?
frank.lukas wrote: Wed Aug 21, 2024 7:47 am
The W32i chip has a hardware blitter like the MACH32 but the Nova driver doesn't use it I suspect. In my tests the W32i is just as fast as an ET4000ax. The W32i can manage more RAM. There are circuit diagrams online. Perhaps the 1MB in 8x 256kx4 chips can be replaced by 1mx4 chips (4MB) on the card. the W32i has a RAM address line A9...? There are RamDAC chips out there from BT up to 110Mhz.
This surprises me since the drivers support the blitter and the benchmarks, at least in my case, confirm that the W32i is way faster than the AX.
I seem to remember running GEMBENCH on an Atari MegaSTE and the display values were almost the same on both (AX / W32i). Maybe I'm imagining it wrong ...?
edit: I was wrong but I ran the w32i again with Gembench and the ET4AX got 117% for the display values and the w32i got 156%, although the blitting values are the same. But the difference is not that big.
Try Gembench 6.x. GB4 does the blitting between two logical workstations. One of which is located in ST RAM and the W32 blitter can not access ST RAM. GB6 uses only one workstation which is located in VGA RAM.
