BOOSTED GAME COMPARISON

[mikro]

Booster Comparisons.

Hope somebody finds it interesting.

I definitely did. For me this is a really nice proof why accelerating CPU alone doesn't bring any significant performance boost. Unless you have a faster RAM and/or some on-chip cache (and a [patched] game which can use it), there's really no point.

[Steve]

Booster Comparisons.

Hope somebody finds it interesting.

I definitely did. For me this is a really nice proof why accelerating CPU alone doesn't bring any significant performance boost. Unless you have a faster RAM and/or some on-chip cache (and a [patched] game which can use it), there's really no point.

Also remember Agranlunds 'special' firmware for the tf536? I can't remember exactly but didn't it cache ST-RAM (shadowed/mirrored?) into fast ram? (Forgive me if I remember the details wrong) but damn it was ridiculous the speed improvement over a normal tf536! It was like on STeroids! :)

[exxos]

Also remember Agranlunds 'special' firmware for the tf536? I can't remember exactly but didn't it cache ST-RAM (shadowed/mirrored?) into fast ram?

Yeah, I think writes were copied into TTram, so reads could be done at 50mhz basically. Very good project. But there was the problem that DMA devices, such as the blitter and DMA drives couldn't be used as the cache wouldn't have any knowledge of the writes. There was talks on how to fix it, but seems it was just abandoned in the end.

Accelerators are not really useful with graphic intensive things like games because ST RAM is always going to be the bottleneck. This is why 3-D games mostly benefit because the CPU can process parts of code much faster, and if the game can run in TT RAM, then you maximise what the accelerator can do. But you're still ultimately limited by graphics held in STram running at stock speeds. All you can do is maximise the bandwidth available to ST RAM. Aside from if you have huge caches that is.

I've said it before many times, but all these accelerators are, in my opinion are nothing more than dirty hacks to the system. This is why I want to move the Phoenix platform to just have a faster bus, so accelerators are just simply not needed, and you have STram running at 32MHz or whatever, so you simply don't hit all the bottlenecks & problems associated with accelerators.

[rubber_jonnie]

Booster Comparisons.

Hope somebody finds it interesting.

I definitely did. For me this is a really nice proof why accelerating CPU alone doesn't bring any significant performance boost. Unless you have a faster RAM and/or some on-chip cache (and a [patched] game which can use it), there's really no point.

:dualthumbup:

[Darklord]

Part of the intent is to set expectations that with most accelerators you lose compatibility in exchange for some performance improvement.

The video would be to show some examples of where the benefit is greatest.

Ultimately VI shows massive improvement btw.

Awesome! I hadn't actually gotten around to trying it on my 40mhz Pak 68/3 equipped STacy.

On the list... :)

[Darklord]

"I've said it before many times, but all these accelerators are, in my opinion are nothing more than dirty hacks to the system."

Looking at my 40mhz, Pak 68/3 68030 equipped STacy, sulking in the corner... :roll:

:)

[stephen_usher]

The biggest problem with compatibility with PP's game conversions is the way he determines what machine it's running upon and what patches to apply.

Given that since TOS 1.04 we've had the cookie jar and from that you can read such things as what the CPU is and the machine type etc. you would have thought that he may use that info, such as using the 68030 patches if there's a 68030 set in the cookie jar, but no. From what I can see he looks at the TOS version only and then guesses what sort of machine ran that version. So, if you have an STFM with an ST536 installed and running TOS 2.06 it doesn't install the 68030 patches and the game usually doesn't work. (If you use the 512K TOS 3.06 patched for the PAK3 then it thinks it's a TT and usually does work.)

PP's games hardly ever work under EmuTOS I've found.

[stephen_usher]

[Accelerators are not really useful with graphic intensive things like games because ST RAM is always going to be the bottleneck. This is why 3-D games mostly benefit because the CPU can process parts of code much faster, and if the game can run in TT RAM, then you maximise what the accelerator can do. But you're still ultimately limited by graphics held in STram running at stock speeds. All you can do is maximise the bandwidth available to ST RAM. Aside from if you have huge caches that is.

It depends. If there's a lot of calculation behind the "graphics" then it will help a lot if the code is in fast RAM. If it's purely sprites and 2D scrolling linked to the frame rate then it will make zero difference. Better to change the system to be NTSC for more performance at 60Hz.

[coonsgm]

The biggest problem with compatibility with PP's game conversions is the way he determines what machine it's running upon and what patches to apply.

Given that since TOS 1.04 we've had the cookie jar and from that you can read such things as what the CPU is and the machine type etc. you would have thought that he may use that info, such as using the 68030 patches if there's a 68030 set in the cookie jar, but no. From what I can see he looks at the TOS version only and then guesses what sort of machine ran that version. So, if you have an STFM with an ST536 installed and running TOS 2.06 it doesn't install the 68030 patches and the game usually doesn't work. (If you use the 512K TOS 3.06 patched for the PAK3 then it thinks it's a TT and usually does work.)

PP's games hardly ever work under EmuTOS I've found.

To be fair to PP, in the large majority of instances those assumptions are perfectly correct. I would say, I'm lucky because the majority of games I like to play work just fine with an accelerator and the two where I most want improved performance benefit substantially(Ultima VI and Civilization)

I mean a big part of this exercise is to highlight incompatibilities....perhaps they could be fixed by folks more coding capable than I....but to provide transparency that just because you drop $200-$300 on an accelerator doesn't mean you get improved game performance.

With more plug and play accelerators, you start addressing a broader audience that hopefully understands, but I expect a not insignificant subsegment won't. These will be installed semi-permanently so if plugging it in unexpectedly makes someone's favorite most nostalgic game incompatible there will not be happiness.

Even more interesting is the fact that two different computers with their different pedigree may experience differences in compatibility. In the testing I did yesterday, I've already come across a small set of games that work on the STE536 that don't work on my ST536 and vice versa. Not even touching speed benefits.