KarlosTKG now on GitHub

[eXeler0]

Originally Posted by Karlos

The shotgun room is gone. The shotgun itself is right there at the end of the hallway right as you come out of your medical bay. Surprised you missed it?

I will just say that unlike the original game, I do try to add some secret areas. Make sure you explore properly. You should have the assault rifle by the time you complete level A.

Well 1st run was just a quick test and keeping my eyes on the FPS counter while toggling stuff. ;-)

Now I played through the whole level.
The framerate in fullscreen 1x1 is for the most part between 5 and 7 fps (on the TF1260 with EC060 at default clock and old firmware)
There are some tight areas where it goes up, but not really on average. On the bridge I get about 4.5 fps, there are places (like next to the right switch at the end of the first long corridor) where it drops below 4fps.
At 2/3 screen it's sort of playable, get maybe 10-12 fps in most places, I think I got like 7.5 fps on the bridge. Some places we get peaks of up to maybe 15 fps.

Another thing i noticed was that the start screen takes much longer before the menus appear.

[Karlos]

I wonder what amipal would get on his machine. He already reviewed an earlier version on it with the standard game.

[Karlos]

@eXeler0

How is the standard T17 game with the same engine, for comparison?

Regarding the station delay, much longer than what? The mod uses a lot more texture data than the T17 game.

[abu_the_monkey]

the delay is probably just loading the overridden wall and floor textures.

and again the TF1260 does suffer with lower than optimal fast->chip mem copy speeds (~5meg/sec) so using it in AGA screen modes will yield lower than expected fps.

[Karlos]

Originally Posted by abu_the_monkey

the delay is probably just loading the overridden wall and floor textures.

True, though they aren't huge as files. I'm not compressing them yet though.

and again the TF1260 does suffer with lower than optimal fast->chip mem copy speeds (~5meg/sec) so using it in AGA screen modes will yield lower than expected fps.

You should look at the frame draw time (ms), this doesn't include the chip ram C2P stage.

[Karlos]

What is the deal with the TF 68060 cards? Other than slower chip ram access, how do they compare to older 060 boards?

[abu_the_monkey]

Originally Posted by Karlos

What is the deal with the TF 68060 cards? Other than slower chip ram access, how do they compare to older 060 boards?

they are a no frills WIP hobby project by Mr Stephen Leary.

they are not designed to be the absolute bleeding edge of performance in every aspect.

[abu_the_monkey]

it would probably be different if we were not using double buffering as the screen would update 'as soon as possible' and not waist time waiting for the screen flip, but would cause graphical artefacts.

[paraj]

I also only get around 5MB/s (~720ns cycle time) on my "period correct" P5 B1260 so it's probably not that much slower. Another advantage is probably that they're available and allow using EC/LC 060's now that the full versions are hard to get/insanely expensive.

[abu_the_monkey]

Originally Posted by paraj

I also only get around 5MB/s (~720ns cycle time) on my "period correct" P5 B1260 so it's probably not that much slower. Another advantage is probably that they're available and allow using EC/LC 060's now that the full versions are hard to get/insanely expensive.

could it just be a bus timing issue? seem to recall raw cpu performance being on point with most other 060 cards.

[eXeler0]

So Im getting better results from my other 2 060 cards but those also use overclocked 060s, so generally I feel its more fair to benchmark on default clock.
My Apollo 1260 runs at 66MHz and my BFG9060 runs at 100MHz.
That being said, the TF1260 could probably perform better with latest firmware, but its a bit of a hassle to update so Ive been putting it off.

[Karlos]

So, is the Lowspec version worth the effort?

[paraj]

Originally Posted by abu_the_monkey

it would probably be different if we were not using double buffering as the screen would update 'as soon as possible' and not waist time waiting for the screen flip, but would cause graphical artefacts.

Triple buffer time! Not a miracle cure, but could save 10ms on average if we really are spending time waiting for the right moment to flip (can't remember how that part works).

Originally Posted by abu_the_monkey

could it just be a bus timing issue? seem to recall raw cpu performance being on point with most other 060 cards.

Going outside my field of expertise, but yes, I guess it's some issue/difficulty in the glue logic interfacing/syncing the chipset with the 060 data bus.

[Karlos]

Triple buffering is a useful trick. You can think of it it as the smallest useful ringbuffer. I've only used it in RTG in the part but equally I've had mixed success. That was back when I used to mess around with hardware rasterisation.

[Karlos]

Back on to the mod for a moment, I've pushed changes to Level B which expands the secret area a bit (as well as giving it a purpose).

[eXeler0]

@Karlos So I finally installed the original TKG (Or actually the 1.1 WHDL version with TurboPatch and FPS counter) on the same TF1260 Amiga.
So TBH, its not running any better than your mod, possibly slightly worse if the FPS counter is to be trusted.
On average I'm getting a few more fps from your LowSpec mod than the original version.

[eXeler0]

Originally Posted by sokolovic

I have the exact same configuration. Original game is a bit difficult to play IMO, especially with textures on the ground and above. But with the Paraj/Karlos/Abu patch it becames very playable now, even in full screen.
I'm still hoping for a whd final version to put it in its dedicated drawer.

Are you saying you are playing TKG in full screen on a 030?
What frame-rates are you getting a 1x1? 2x1?

[Karlos]

I think he means the original T17 game with the current engine.

[Dunny]

FWIW the original T17 version in your engine with a PiStorm is running at a solid 50FPS in RTG mode, AGA mode gets you about 35 to 50 depending on area.

[Karlos]

Regarding the double buffering/vsync, in part this was introduced because the game runs ludicrously fast in UAE. It might make sense to have an option to disable it for slower machines.