Genesis demos VS Emulation... Is this a good way to measure accuracy?

Libretro Cores
#1

Because GenesisPlusGX fails miserably here:

It will only show the first screen without the 3D text and stuck there. On the other hand, Picodrive will show half of the demo and freeze later on while BlastEm will show the whole thing.

There’s also the case of this older, well known demo:

Again, GenesisPlusGX fails to display the majority of the sequences correctly. Even the sound glitches after some point. While both Picodrive and BlastEM show everything correctly. Or at least i couldn’t notice any bugs or differences on either of them VS the video, from a fast glance.

Now, i know BlastEM is supposed to be the most accurate Genesis core so there are no surprises here. What does surprise me though is how much better Picodrive seems to be compared to GenesisPlusGX.

This doesn’t take compatibility into account though. An accurate emulator doesn’t mean it’s the most compatible, that has to do with maturity (like how Ares is more accurate than any other N64 emulator but it’s not as complete so it’s also not as compatible as Mupen64plus). In that sense, GenesisPlusGX was better in my mind because i remember having issues in some Genesis games with Picodrive in the past. Though that was sometime ago.

So is Picodrive the better emulator these days?

1 Like
#2

I would like to see differences on real games, not just on benchmarks or demos, before coming to conclusions. Also what about the sound? Genesis Plus GX has different sound engines, such as Mega Drive 1 and 2 models and either nuked or mame engine. Which one is Picodrive using? My point is, there is probably more to the story than just testing demos.

#3

Yeah but those demos are like a stress test for accuracy because they use the hardware in ways that most games don’t.

Not sure about the sound in Picodrive.

1 Like
#4

That is exactly my problem with it. If game don’t use specific features or if the demos “abuse” features, then why does it matter as accuracy for the games? I assume those demos are created with emulators first and then tested on real hardware.

Don’t misunderstand me, I am curious about the topic. And maybe demos are good ways to test emulators. But are the demos perfect too? What if they are not perfectly programmed in example.

#5

Well, the whole purpose of those kind of programs is to test the accuracy of emulation.

If those emulators fail on those programs then most likely the emulators are not perfectly accurate and issues can also happen in real games.

#6

I see. Guess it makes lot of sense to test specific and reproducible cases. Another good way would be to test known bugs in “real games”. Currently doing some research and want to share some resources to this topic (however one is about NES) that go deeper into these kind of subjects: https://www.nesdev.org/wiki/Accuracy and https://emulation.gametechwiki.com/index.php/Emulation_accuracy . Unfortunately they don’t talk too much about Mega Drive / Genesis, but it’s mentioned that a later version of the console fixed a bug in a game. That is something one should have in mind when testing stuff.

#7

I don’t think these demos are made with emulation in mind. They are made by hobbyists who use clever programming and tricks in order to produce impressive effects and sequences that would normally be beyond the supposed capabilities of the target hardware. This started when piracy groups would use such sequences in their “cracktros” back in the C64 days, way before emulators were a thing.

Basically, it’s programmers flexing their skills and digital art.

There are, however, roms that exist to test emulators. These are a different thing though and they are referred as “test roms”.

#8

right, the major difference between demos and test ROMs is that test ROMs are intended to give emu authors information to improve their emulator, while demos are just supposed to look cool and make people who know the hardware say “wow, that shouldn’t even be possible!”

A lot of test ROMs, like blargg’s test suites, use only basic, common behavior, so passing all of the tests means your emu is good enough to play most games, while others test funky edge-case behavior and are intended to push accuracy further. One of the problems I have with the combined single-score accuracy shootouts is that they put the same weight on each test, even though there could be dozens of tests that all hit the same functionality (that is, an emu that implements that one thing could score higher than an otherwise more accurate emu that just misses that one thing).

1 Like