How Accurate is MiSTer FPGA vs Original Hardware?

[@BD12]

It's Ridge Racer! Riiidge Racer!

[@jamesfmackenzie]

😂 If I was there I would have cheered along with Kaz

[@rijjhb9467]

Remember that one? (I do because I'm old, I used to play it on the actual arcade machines. Nonetheless, it's strange that the 2006 public never heard of it, how young were they? RR was still pretty relevant at the end of the 90s, not even 10 years before that E3).

[@musicman7943]

Asking the real questions that I didn’t know I’d ever ask

[@GeorgesChannel]

Great video. I am thinking about to buy a Mister for a while...

[@jamesfmackenzie]

Thanks! I can strongly recommend a MiSTer setup!

Thanks for this video! It answer a question I've had for a while: how really accurate is it to real hardware. Thing is, as we know, FPGA can be identical to real hardware, but it goes as far as the person programming/engineering the FPGA. If there's any mistakes or workarounds, it will have flaws and glitches just like any emulator. I got my Mister not too long ago, but was always wandering if I'm looking at a 1:1 copy of real hardware, or some possibly flawed (but really close) interpretation of the original hardware. Well, your videos seems to prove it is identical (or 99.9999% identical) to real hardware. Awesome to know, and got a sub for that!

[@jamesfmackenzie]

Thanks for your kind message! My take: whist it’s possible for FPGA implementations to be cycle accurate (or even accurate at the gate level etc), this is not always the case. Core developers are often working in a “black box” fashion or using existing software emulators as a reference to create their cores. So it can be *very* accurate, but not always 1:1. One advantage that FPGAs have: once you define components in FPGA, they will interact in parallel with each other, just like with real hardware. If each component is implemented accurately, the FPGA implementation will be very accurate to original hardware. Naively, software emulators operate on a single CPU thread, with the code that emulates each components being executed sequentially - which means you need a *very* fast CPU (compared to the original hardware) to approximate original behaviour. Of course today we have multi GHz CPUs, and also the ability to distribute threads across multiple CPU cores - so with the right programming, it is possible to write very accurate software emulators too - but it’s not easy! 😂 There are some great articles that explain this a lot better than I can - but I hope this helps!

James, is there any chance you can do another comparison? With another PS1 game, and also with another system (like comparing accuracy of slow downs on cores like SNES and Neo Geo, sprite flickering on the Genesis, etc). I say that because a lot of people wanna know if Mister is really accurate, or just a passed on internet legend. 90% accurate, 95%, 99% accurate is not 100% accurate, you know what I mean? You seem to have provided the best comparison on the internet to this day, we'd love to see more of this.

[@jamesfmackenzie]

Thanks! I don’t have my SNES or Genesis hardware with me at the moment, but happy to try another PS1 game :-) Any specific game you’d like to see? 😎

@@jamesfmackenzie Hey James, thanks for passing by :) Your PS1 comparison tells a lot, on so many different things, for example: it tells the rendering is identical (I pasted over both real hardware and Mister images one over the other in photoshop in layers, and it's clear poly wobbling and texture warping is identical). It also tells framerate is identical, and so on. The thing is: we're told Mister is supposed to be the "same" as real hardware, but only comparisons like yours put this to the test. PS1 has a lot of graphical glitches (wobbling, warping, dirthering), so it's kinda hard to software emulate identically. . Your comparison gets us closer to the truth :D No specific games in mind, just some that pushes the hardware further into those glitches, one that has tons of warping and wobbling. Seeing how Mister deals with these glitches show how accurate it really is. Maybe Metal Gear Solid or Alien Ressurection would be great!

[@GreySectoid]

Awesome comparison, could you do other consoles as well (Nintendo, Sega)? Subscribed.

[@chinarut]

lol i can’t tell the difference on first watch - could someone comment on the difference they see?

[@johansjostrand6026]

Pause and look at the yellow car from a medium to close distance and you'll notice that it has light grey edges/outlines on the original PSX whereas it's completely yellow on the Mister core. Not sure if anyone has brought it to Roberts attention.

[@chinarut]

@@johansjostrand6026 I guess the MiSTer is that good - I still can't tell after watching for the yellow car & pausing quite a bit! 😆 is there a time marker (@x:xx) you feel it's most pronounced?

[@johansjostrand6026]

@@chinarut Yeah the Mister is great indeed. Try 48 seconds, set the playback speed to 0.25 or something and hit pause. I'm watching in 1440p resolution in case that matters. Maybe James is running some kind of filter causing this, I believe Robert is all about accuracy.

[@beefquiche]

@@johansjostrand6026 looks like an upscaling issue on the PS1 - PS1 doesn't run at 4k (I know, yep, obvious), so it's likely an upscaling artefact.

[@Zacabeb]

@@johansjostrand6026 That issue is caused by 4:2:0 color subsampling in the video capture rather than the PlayStation itself. 🙂

[@someguy999]

I assume this is just running a demo. Do you think that it would make a difference if the hardware had to process button presses and input?

[@jamesfmackenzie]

It’s a great question! Unfortunately I don’t have a way to synchronise the inputs between the two systems :-(

[@rijjhb9467]

@@jamesfmackenzie It shouldn't be too hard to make a contraption to do that, especially with microcomputers, since thieir joysticks send raw signals to their respective machines. On thing to note is that on the MiSTer, SNAC should be used instead of USB in order to minimize input latency.

[@jamesfmackenzie]

@@rijjhb9467 I agree with that - seems like this would be possible with a microcontroller and some code 😎

[@ragexg]

Can this thing upscale te resolution or is it just 1to1?

[@Trunchisholm]

It's native resolution, but it adds several image options such as wide-screen and 24bit rendering, dithering removal and texture filtering.

[@rijjhb9467]

MiSTer is mostly a 2D affair. It can greatly enhance a 2D picture thanks to its CRT filtersand such, but for 3D stuff I prefer to stick to emulation and crank the internal rendering resolution to 4K. Even a crap CPU/GPU combo can run PS1 games in 4K nowadays.

[@sheimamouff5764]

Pourquoi chipoter? C'est plus proche du hardware,est loin de l'émulation. J'étais un défenseur de l'émulation,mais l'input lag est l'ennemi numéro 1 en réalité,elle est là l'innovation