AmiQuake new 68k Quake Port for AGA/RTG

[NovaCoder]

Originally Posted by Damion

Just a quick timedemo run with the BFG @ 100 MHz (apologies, this is the Clickboom version):

AGA 320x200: 21.00
Picasso IV 320x240: 21.42
Picasso IV (Z2 jumper): 10.42

BFG @ 50 MHz

Picasso IV 320x240: 11.65

A2000 GVP 2060 @ 50 MHz

Picasso IV 320x240: 9.56

Noice, what a beast

Try AmiQuake RTG too

[grelbfarlk]

Originally Posted by Damion

Just a quick timedemo run with the BFG @ 100 MHz (apologies, this is the Clickboom version):

AGA 320x200: 21.00
Picasso IV 320x240: 21.42
Picasso IV (Z2 jumper): 10.42

BFG @ 50 MHz

Picasso IV 320x240: 11.65

A2000 GVP 2060 @ 50 MHz

Picasso IV 320x240: 9.56

Yeah so the point is that video bandwidth isn't a concern until you pass some threshold of CPU power, say in the couple hundred MHz range.

[Daedalus]

Video bandwidth (in terms of the overall access speed to the graphics card) has a huge impact even at 50MHz. Running a P-IV in Zorro-II versus Zorro-III mode is a very good illustration of that. The figures there are for 100MHz, but it's still an obvious difference at 50MHz.

[grelbfarlk]

Look at the old Amispeed site via archive org. On the bottom are 2D only non HW Accelerated benchmarks.
https://web.archive.org/web/20160120...eed/q***e.html

With a BPPC and a BVPPC it is beaten by the Mediator which has at least half of the bandwidth, probably a lot less. The CVPPC beats the Mediator by maybe 33%. As the resolution increases bandwidth starts to matter more but that's after the CPU is even more bottlenecked, but still compared to the BVPPC with far more bandwidth the Mediator gets farther ahead.

With Novacoder's RTG version I get 14.2FPS on the 060@50Mhz on the CSPPC with a Voodoo. With WarpQuake the 200MHz PPC locks up after a few seconds but looks to be at maybe 30FPS but that's at 320x240.

[James]

Originally Posted by grelbfarlk

With a BPPC and a BVPPC it is beaten by the Mediator which has at least half of the bandwidth, probably a lot less.

BPPC 160 Mhz + BVPPC = 18.5 fps
BPPC 166 Mhz + Mediator = 17.7 fps

[grelbfarlk]

Originally Posted by James

BPPC 160 Mhz + BVPPC = 18.5 fps
BPPC 166 Mhz + Mediator = 17.7 fps

Yeah that's my point, at these low resolutions, it's CPU bottlenecked and the bandwidth from CPU to GPU doesn't matter all that much for Quake.

BPPC 160 Mhz + BVPPC = 18.5 fps
BPPC 166 Mhz + Mediator + Voodoo3 3000 = 17.7 fps
BlizzardPPC: PPC:603e 240 Mhz + G-Rex PCI+ Voodoo3 3000 = 21.4 fps
CyberstormPPC:604e 150 Mhz + Mediator PCI + Voodoo3 3000 = 25.8 fps
CyberstormPPC 604e 233 Mhz + Mediator PCI + Voodoo3 3000 = 29.2 fps

[tomcat666]

Originally Posted by Ras Voja

[ Show youtube player ]

Which Quake1 build is that? To me it looks much better than AmiQuake from aminet... ?

Comparison:



Lighting is MUCH better in the video...

I have never seen these video options in any of the versions:

[Damion]

Originally Posted by Daedalus

Video bandwidth (in terms of the overall access speed to the graphics card) has a huge impact even at 50MHz. Running a P-IV in Zorro-II versus Zorro-III mode is a very good illustration of that. The figures there are for 100MHz, but it's still an obvious difference at 50MHz.

Absolutely. It should probably be noted that the PIV performs a bit poorer in jumpered Z2 compatibility mode in a Z3 machine, vs when used in an actual A2000. The former is 2.4 MB/s bandwidth, vs 3.5 in the latter (bustest), so this is absolute best case scenario really for the A2000. In the A4000 at 50 MHz with the Z2 jumper enabled on the PIV, you get about 7 FPS, vs 9 in the A2000. So the additional 1 MB/s bandwidth is huge, not to mention the BFG has a bit better onboard memory performace compared to the (still very good) TekMagic.

On that note, it seems that the BFG's onboard memory performace amplifies the Zorro bottleneck somewhat more than other accelerators might, as (for example) compared with the Cyberstorm MK2's much slower memory controller, you're almost back to 1:1 again in the timedemo between the 2000 and the 4000 (clock both at 66 MHz, and the Cyberstorm pulls ahead by a mere 1 FPS). Bottom line though, you'd definitely need better than Z2 to get the best from Amiga Quake, even at low-res.

[Daedalus]

Indeed, different accelerator designs can and will have different Zorro transfer characteristics because cycles need to be synchronised. An accelerator optimised for memory performance won't be optimised for Zorro performance, and vice versa.

Even at low resolutions, video bandwidth makes a noticeable difference. You can see that most games where much of the screen drawing is done by the CPU, at low resolution on an A2000 with a P-IV can be beaten by AGA. Quake and Payback are two such examples, and it's simply because AGA bandwidth is higher than Zorro bandwidth.

[trixster]

@tomcat666 that looks like ClickBoom Quake

http://www.clickboom.com/quake/faq.html

[tomcat666]

Originally Posted by trixster

@tomcat666 that looks like ClickBoom Quake

http://www.clickboom.com/quake/faq.html

Interesting, thank you.

[tomcat666]

Must be something else (maybe AGA?), on P96 in the same resolution there is no difference in how lighting works in AmiQuake and ClickBoom Quake ... I do get some 20% boost in performance with ClickBoom quake though

[trixster]

I cannot discern any difference in lighting between AmiQuake and Clickboom quake, so I suspect it is just the way the video has been recorded which makes it look different.

[Reynolds]

Originally Posted by trixster

I cannot discern any difference in lighting between AmiQuake and Clickboom quake, so I suspect it is just the way the video has been recorded which makes it look different.

Yepp, direct video output recording vs video intercom :P

[eXeler0]

Originally Posted by Ras Voja

Does Warp3D MiniGL help?

Not if its a port of a Software-renderer version of Quake.

[eXeler0]

Btw, if someone feels like testing my new Quake map, the download is here:
https://www.exretro.com/downloads/qmaps/qmaps_dl.html
(Its the second one on the page)
It is tested to run on WinQuake and earlier versions of the map were actually tested on Amiga with Novas port.
I get very varying performance on AGA+66Mhz 060, if someone could try it with a Vampire, it would be great to hear how it runs.

[abu_the_monkey]

@eXeler0

I know this is not your issue, but, I can not download your maps from the site you have linked to. am I doing something wrong?

[grelbfarlk]

Originally Posted by Ras Voja

But yes if its some GL or Glide port?

Yes Bromangelo, then it helps a lot, both in speed and visual quality.

For instance:

          Demo  Sonnet 500MHz G3    PrPMC 450MHz G4   Killer NIC 400MHz   Ragnarok 950MHz RAM+
                320x240  640x480   320x240  640x480   320x240  640x480    320x240  640x480
QuakeWOS  (2)    68.3     24.6       93.0     40.3      60.0    25.3      119.3     50.4
glQuake   (1)     x       48.9        x       50.1       x      44.5        x       54.0
Quake II  (1)     x       51.2        x       55.1       x      44.3        x       66.6
Quake III (4)     x       28.3        x       32.0       x       a          x        x

[eXeler0]

Originally Posted by abu_the_monkey

@eXeler0

I know this is not your issue, but, I can not download your maps from the site you have linked to. am I doing something wrong?

Tried right click, ”save link as”..?

[tomcat666]

Originally Posted by eXeler0

Btw, if someone feels like testing my new Quake map, the download is here:
https://www.exretro.com/downloads/qmaps/qmaps_dl.html
(Its the second one on the page)
It is tested to run on WinQuake and earlier versions of the map were actually tested on Amiga with Novas port.
I get very varying performance on AGA+66Mhz 060, if someone could try it with a Vampire, it would be great to hear how it runs.

Here it is running on AmiQuake and pistorm 900 MIPS (rPIzero2W) Amiga 600 in 640x400 resolution. Noticable slowdowns in places, much more than default Q1 maps... unfortunetly.

[ Show youtube player ]