PC games that made the Amiga look like Shit

[8bitbubsy]

Originally Posted by Promilus

And no, 486 optimized code is not the same for 386 the same way why 030 optimized code won't run just as good on 020 - regardless of core speed difference. If you have cache you and want the best performance you try to make whole timing critical functions to fit there. Cache already opens up a slot for specific optimizations which - when made properly - results even bigger performance difference than from running regular software on 386 and 486 at same speed. Just because if you don't optimize routines to fit cache some of it might run as fast as possible but not all of them and not always. Really playable framerate was only achieved on fastest 486 DX2 which I can hardly qualify that game to be "optimized" for 386

Yup. A 486 is guaranteed to have fast internal L1 cache, while a 386 would have external L1 cache at best (usually no cache at all).
Writing code to be optimal for cache-based systems is a thing.

[Karlos]

Originally Posted by Promilus

Oh, you mean you completed the game with highest details and fairly reasonable screen resolution on 486 25MHz?

I am saying I played Doom 1 SW and Full to completion, full screen on a 486 SX 25 in 1993. It was playable l. I'm not saying it ran at display refresh, my estimation is 15-20 FPS for the most part. There were occasional slowdowns here and there but it was never rendered unplayable, except perhaps to today's twitch reflex generation that can't play anything if it isn't 120fps.

In a repeat, I completed Doom 2016 on my 2.6GHz Q9450 with 8GB DDR3 and GTX780ti, OpenGL 4.5 rendermode (vulkan was glitchy) at 1080p maximum detail settings. The slowest frame rate it achieved in game was around 40fps.

[TCD]

Originally Posted by Promilus

I can say Witcher 3 was absolutely smooth with FX6300 and GTX960 ...

I played it on a Intel i7-4790 with a GeForce 970 and it ran pretty good at (nearly) max details on 1080p. Then again I'm not too fussed about the FPS dropping into the 40s or even 30s as long as it's not stuttering all the time.

[freehand]

I play DCS world in VR at 3 frames a month on my rig it's great

Peoples definition of playable is different to mine, how more silly can it become to win an argument I wonder

[Karlos]

Originally Posted by freehand

I play DCS world in VR at 3 frames a month on my rig it's great

Peoples definition of playable is different to mine, how more silly can it become to win an argument I wonder

Playable has a simple enough definition. You can complete it, it's enjoyable to play and isn't a complete slideshow. For fast paced, competitive multiplayer FPS games you definitely need fast, steady frame rate and low latency, that's a given. For single player games, it's less of an issue if you can't manage 60fps constantly.

[Bruce Abbott]

Originally Posted by Karlos

Playable has a simple enough definition. You can complete it, it's enjoyable to play and isn't a complete slideshow. For fast paced, competitive multiplayer FPS games you definitely need fast, steady frame rate and low latency, that's a given. For single player games, it's less of an issue if you can't manage 60fps constantly.

'Enjoyable to play' is enough for me.

[freehand]

Originally Posted by Karlos

Playable has a simple enough definition. You can complete it, it's enjoyable to play and isn't a complete slideshow. For fast paced, competitive multiplayer FPS games you definitely need fast, steady frame rate and low latency, that's a given. For single player games, it's less of an issue if you can't manage 60fps constantly.

No that is your definition.

Also no one is talking about a constant 60 fps in games your talking about doom in full screen bla bla on 386

[idrougge]

Originally Posted by Bruce Abbott

3 years later the Doom source code was released, and 3 days after that it was ported to the Amiga. Requirements were any Amiga with a 68020+ CPU and 8MB RAM.

And still, it runs rather bad on my A4000 with a graphics card.

[idrougge]

Originally Posted by grond

The mistake most likely was to make the gap that wide and do not provide a recognisable upgrade path to bridge the gap.

The A4000/030 was released in 1993 and was quite cheap in comparison.

That's exactly why I keep saying that Commodore should have left the 020 away from the PCB and put it on a seperate CPU board. For very little extra money (basically the price for the connector on the CPU board) they would have made the base product ready for careful upgrade right from the factory.

The connector — not to mention the carrier board — probably cost as much as the EC020. Keep in mind that the A1200 didn’t even include a SIMM slot for extra memory, which would have improved performance and longevity tremendously. But even plain RAM expansions for the trapdoor were very expensive.

[oscar_ates]

Fortnite, Apex and warzone are the most played games at the moment. Therefore 3d fps games are still lit started with doom in 1993 as genre.

Team deathmatch was an invention of John Romero

[idrougge]

Originally Posted by Bruce Abbott

Compared to the A3000 and A4000 which had more stuff on board, the A1200 was simpler and more reliable. Trying to get those ZIP chips in and out of the A3000 was a pain, and I had endless trouble with the CPU slot after upgrading to an 060 board. The A4000 had cheap plastic SIMM sockets whose retaining clips would snap off after a few insertions, and the 'fast' RAM on the motherboard was much slower than RAM on the accelerator card could have been.

Well, I haven’t seen snapped-off SIMM clips in any A4000 to this day, and I have owned and serviced quite a lot of them. You generally switch out the SIMMs once in the machine’s lifetime.

As for the A3000’s ZIP sockets, I don’t know what Commodore was smoking then. Even shitty Atari had SIMM sockets in their consumer-grade STe by then, Apple had had them for years before the A3000’s release. The ReAmiga 3000 successfully replaces that entire area with two SIMM sockets. There is no discernible reason for Commodore’s reluctance to accept SIMMs for memory expansion, and whatever you might think of the quality of the sockets on the A4000, they are safer than a bank of 16 ZIP sockets.

[idrougge]

Originally Posted by Bruce Abbott

But shareware was never as popular on the Amiga as on the PC

[citation needed]

The Amiga’s shareware scene was one of the biggest. The one thing making the PC scene bigger was that it was concentrated to one big economy, making transaction costs smaller.

[Bruce Abbott]

Originally Posted by 8bitbubsy

Yup. A 486 is guaranteed to have fast internal L1 cache, while a 386 would have external L1 cache at best (usually no cache at all).
Writing code to be optimal for cache-based systems is a thing.

Yep. And so is using asm instead of C. On the Amiga I generally optimize for 68000 because the slowest machine is the one that needs it the most.

All the 386DX systems that I have examined had external cache. I don't know about the one in the comparison video, but in case you forgot I also provided a video of an AMD 386DX/40 with 128k cache and 32MB 60ns RAM overclocked to 50Mhz, which did the Doom timedemo run at 8.8 fps.

All 486 CPUs only had 8k of internal cache, except for the DX4-100 which had 16k. So as far as 'optimizing for cache' is concerned a 486 shouldn't get any more benefit from it than a 386DX with 128k.

How does this relate to the Amiga? Here's a post from amigaworld.net:-

Originally Posted by Cammy

Here are the results from my Amigas (all tests were done in 320x200):

DoomAttack

A1200 030/50Mhz 32Mb (Optimised 020 C2P) - 8481 realtics (8.8 fps)
Amiga CD32 68020/14Mhz 8Mb (Optimised 020 C2P) - 18971 realtics (3.9 fps)
Amiga CD32 68020/14Mhz 8Mb (Optimised Akiko C2P) - 12872 realtics (5.8 fps)

ADoom

A1200 030/50Mhz 32Mb - 9133 realtics (8.2 fps)
Amiga CD32 68020/14Mhz 8Mb - 16294 realtics (4.6 fps)

So there you have it. DoomAttack was exactly the same speed as the 386DX with 128k cache overclocked to 50MHz. ADoom was ~7% slower than DoomAttack, while the 386DX clocked at the normal 40MHz should be ~14% slower than ADoom.

In case you think those figures are anomalous, here's the first post from that AmigaWorld thread:-

Originally Posted by pavlor

Realtks avg.fps...
7549 ------ 9.9 ------ NoteStar NP-740, 80486SX-25 MHz, 4 MB RAM, VGA No SFX
7748 ------ 9.6 ------ Amiga 1200, Blizzard 1230IV-50 MHz, 8 MB Fast, AGA (DblPAL) ADoom 1.3 No SFX

A Blizzard 1230IV-50 is matching the speed of an 80486SX-25!

[idrougge]

Originally Posted by sokolovic

And there is games in 256 colours mode that doesn't have any speed problem. Slamtilt is a good example, even switching between low res and high res on the fly.

I think we need to keep in mind that Slamtilt and Pinball Fantasies are mostly static-screen games. The only things that need updating at 50 fps is the ball, the flippers and possibly a low-bitdepth score display. The rest is handled by the scroll registers.

You can’t apply the same limitations to any freely scrolling action game.

[Bruce Abbott]

Originally Posted by idrougge

[citation needed]

The Amiga’s shareware scene was one of the biggest. The one thing making the PC scene bigger was that it was concentrated to one big economy, making transaction costs smaller.

Note that I said shareware, not public domain.

The most popular Amiga PD/shareware library in the early days was probably the Fred Fish disk collection, which he started distributing on Jan 4, 1986 with 4 disks. On the list of titles in the 1120 disks eventually produced, I counted 573 instances of the word 'shareware'. The average number of programs per disk looks to be about 5, which would mean only about ~20% are shareware. On Aminet I only found 65 packages with the word 'shareware' in the description. But that doesn't tell us anything about how popular each one was.

Amiga shareware games were few, and even fewer were worth playing. A few notable utilities and system extensions such as MUI were very popular, but even if every single Amiga user had a copy it would still be dwarfed by the number of shareware programs installed on PCs - both because there were more than 10 times as many PCs and a higher proportion of 'freely distributable' PC programs were shareware.

The nature of shareware distribution makes it very difficult to accurately estimate the popularity. Just counting the number of titles produced isn't enough. However the shareware fees collected do give some indication. In 1987 Apogee Software was raking in $100-500 a day from their latest game, with a total income of $80-100,000. That's more than many full commercial titles were getting on the Amiga!

[Bruce Abbott]

Originally Posted by idrougge

I think we need to keep in mind that Slamtilt and Pinball Fantasies are mostly static-screen games. The only things that need updating at 50 fps is the ball, the flippers and possibly a low-bitdepth score display. The rest is handled by the scroll registers.

You can’t apply the same limitations to any freely scrolling action game.

So? All games have limitations. Maybe you can get 8 way scrolling at 50 fps but not have large bobs at the same time. Or maybe you have to keep the AI simple to reduce CPU load. Or you might have a window into a larger pre-rendered screen and not have much ChipRAM left for other stuff. There are many ways to make a game run at 50 fps in 256 colors (or more!) if you are willing to compromise in other areas. Game designers always had to plan the game around the capabilities of the hardware. The Amiga was no different.

The reason people recommend using fewer colors on the Amiga is that it is often a poor use of resources. On a PC you don't have much chioce - it's either 16 or 256 colors and nothing in between.

Personally I think 256 colors is often too many anyway because the artists don't know what to do with them. You get extremely pretty graphics that just distract from the gameplay and/or make it harder to see things like bullets and obstacles. Then they call it a feature, not...

[Promilus]

Originally Posted by Bruce Abbott

All 486 CPUs only had 8k of internal cache, except for the DX4-100 which had 16k. So as far as 'optimizing for cache' is concerned a 486 shouldn't get any more benefit from it than a 386DX with 128k.

LoL
I guess you don't really know how different cache behaves. So let me enlighten you. Integrated 8K cache on 486 was internal unified SRAM based cache running at full speed with the core and well integrated with execution unit. It means there's lowest delay between requesting data and receiving data (&instructions - it was unified cache). It's especially important with DX2 and DX4 as they are running at 2 and 3x system clock internally. Now then what about 386 external cache? Well it still does sit on system bus interface and the only good thing is it does access memory faster than from DRAM. And guess what, 486 still had that kind of cache which turns into L2! And so had Pentium (which in case of K6-III led to 3 levels of cache - L1 integrated on die, L2 integrated on die and what used to be external L2 on motherboard became L3 for a minor but visible performance impact).

So - again - writing routines which fits that 8K internal cache of 486 does make the difference because accessing external cache while faster than DRAM is still slower than internal SRAM. That's the whole point of integrating relatively small but fast internal cache. And 68k did that as well without even trying to apply external cache (some really old memory upgrades and turbo cards did try to replicate that external cache but I guess it did introduce some incompatibilities with AOS or apps).

[Bruce Abbott]

Originally Posted by Promilus

So - again - writing routines which fits that 8K internal cache of 486 does make the difference because accessing external cache while faster than DRAM is still slower than internal SRAM. That's the whole point of integrating relatively small but fast internal cache. And 68k did that as well...

This is all so silly. First you say the PC version of Doom wasn't optimized but the Amiga version was, so any comparison is invalid. Then you say it was only optimized for the 486, so the comparison is still invalid. Now you are saying that though the 386 will benefit from caching it won't benefit as much as the 486, so the comparison is still invalid.

Finally you say the 68030's cache provides similar benefit to the 486, which I presume is another reason the comparison is invalid.

But it's all completely irrelevant. The asm code in the PC version had to be changed to work on the Amiga anyway, so if the 68030 makes better use of it than a 386DX so what? That's a plus for the Amiga! You can pontificate as much as you like about how 'optimized' the code was on each platform, but it won't change the fact that in the real world an A1200 with 50MHz 030 has similar performance to a 386DX-40.

BTW just for fun I tried running DoomAttack on my Blizzard 1230IV with caches turned off. Frame rate dropped from 9 fps to 6 fps. In low detail mode it dropped from 12 fps to 7 fps. So the cache does make a big difference. Hooray for 68030!

[Karlos]

Originally Posted by freehand

No that is your definition.

If you're going to play the subjectivity card, then that makes my definition just as good as anyone elses.

Also no one is talking about a constant 60 fps in games your talking about doom in full screen bla bla on 386

486 25MHz. I never tried on a 386, so I have no comment. As I've said more than once, it achieved full screen at an estimated 15-20 FPS.

[ Show youtube player ]

Read the description before you assume the two machines are the wrong way around in the video.

[Promilus]

Originally Posted by Bruce Abbott

This is all so silly. First you say the PC version of Doom wasn't optimized but the Amiga version was, so any comparison is invalid.

You don't even understand simple terms. I did challenge your claim that it was optimized for PC (generally) and I did mention 486 machines there. Right? Right. That means 386 based systems never did receive any particular optimizations. Right? Right. Now did 030 system receive any particular optimizations? Yes? YES! So how's that comparing apples to apples? It is not, that's why I claim such comparison to be ridiculous. Not my fault you're too blind and arrogant to see that.