Retro: With emulation is original hardware worth it?

[Foxi4]

*Snip!*

That's the thing about retro collecting - it requires a lot of love, passion, dedication and often expense. Consoles have to be regularly cleaned, they have to go through proper maintenance (which sometimes includes component replacement), cables need to be kept neat and replaced when needs be, all your gear has to be kept in a dry place that's preferably not exposed to direct sunlight (nobody likes yellowed plastics) etc. - it's almost like owning pets, minus the feeding and walking, plus loads of electronic entertainment. If you want to start collecting, you have to be aware of all this and, preferably, you have to actually enjoy the maintenance. In my eyes "getting things to work" and keeping gear in pristine condition is very rewarding in and out of itself, the genuine experience that comes with it is almost secondary since you can't physically play X number of consoles on a daily basis.

 

[weiff]

It will be a long time before emulation reaches original 1:1 100% across all systems, so much of it is related to the proprietary coding associated with running certain chipsets. Until they can get all that squared away, there will still be a use for hardware even if just for reverse-engineering and coding the new emulators. Beyond that, everytime new hardware comes out someone wants an emulator that does everything again perfect on it... never ending cycle of recreating an old wheel.

 

[Hells Malice]

There's really no difference between emulating and playing the original, unless the emulator doesn't work.

The biggest difference is obviously the controller, but it's actually really easy to get PC adapters for all of them.

I enjoy owning retro games and consoles, but I never delude myself into thinking it's actually better.

 

[FAST6191]

I forgot to mention that about 75% of screens I see are poorly calibrated (and people seem to like TV people having sunburn so they often stay that way), knowing this several devs designed accordingly (and still do to this day in some cases http://www.eurogamer.net/articles/digitalfoundry-vs-hdmi?page=3 about halfway down the page). Audio mastering has known that making it sound good on shitty speakers is a good plan too. Wind in some cheapo hardware/odd decisions (no component and what have you) and the "vision of the devs" becomes far less important for me.

Beyond that, everytime new hardware comes out someone wants an emulator that does everything again perfect on it... never ending cycle of recreating an old wheel.

Assuming you mean something like "Oh wait, someone released android, now I need a SNES emulator for it" then will we not hit the point where the overheads for acceptable emulation for most games* will still be under the min speeds the device runs at? We kind of saw something like it on the PSP where on the GBA emulators were hardcoded to take advantage of basically every cycle and then with the PSP things were more or less just ported over. Equally I reckon coding will continue to get more and more abstract/divorced from the hardware as time goes on (if not done remotely).

*100% transistor accuracy is probably further in the future than not, "100% good enough for 100% of the games most people care about playing, plus the extras emulation affords" is already there in most form factors that matter (TV, handheld and PC).

 

[Foxi4]

I never quite understood the point in 100% transistor accuracy - once you reach 100% software compatibility, your job is done, there's no reason to progress other than showing off your programming brawn. It seems counter-productive - I'll take performance and optimization over accuracy that doesn't improve the experience anytime.

 

[Issac]

I like both. I think nothing beats the original hardware; it's just the feeling, controller and knowing that it's 100% as intended. Then again, emulators are damn handy, with quicksaves and over all convenience.

I'm one of those Vinyl collectors too, so yeah...

 

[FAST6191]

I never quite understood the point in 100% transistor accuracy - once you reach 100% software compatibility, your job is done, there's no reason to progress other than showing off your programming brawn. It seems counter-productive - I'll take performance and optimization over accuracy that doesn't improve the experience anytime.

I reckon it is probably twofold

1) It might well be easier at some level than trying to make cycle accurate code synced between ten different chips, on variable machines too. If you are just throwing more computing grunt at it (or a programmable chip) then fantastic.

2) 100% software is fine if you want 100% released software, however if you wind in potential homebrew code and potential

A bit further down the line it might be nice to use it as some kind of all purpose circuit dev board -- if I do not have to mess around with my oscilloscope for 5 hours to make a component mod on a new system and can rather figure it out by looking at things then fantastic. If it is a FPGA then that changes things again.

Do not get me wrong, I am happy enough with things as they stand now for 16 bit and older systems (give or take quality of debuggers for some of them), and some of the later stuff does well enough for most of my purposes as well, however it is not without purpose beyond coder vanity project.

 

[Jiehfeng]

I've always wondered though, why is emulation still slow when it's run on a PC which is actually faster than the original hardware? Sorry for being a dumbass...

 

[FAST6191]

I've always wondered though, why is emulation still slow when it's run on a PC which is actually faster than the original hardware? Sorry for being a dumbass...

There are two main schools of thought on emulation, three if you count the hardware stuff we have been talking about in this thread.

1) System level replication.

2) Dynamic recompilation.


Processors are very dumb devices and they all speak different languages when it comes to the binary stuff they are decoding.
Ignoring that you are running an operating system on top of your machine where say a SNES might have copied a value into a register you would have to take that value, decode it, convert it into an instruction/series of instructions to do it, do the instruction and then post the results. At the same time you would also be having to emulate the sound chips, reading the cart/disc/whatever, emulate the graphics chips, emulate any non processor stuff (if you want to copy a chunk of cart into memory then it does not have to go through the CPU, usually using something called DMA/direct memory access)......
As the original system, and code for it, all expects this to happen reasonably in step and your little single serial processor in your PC gets to do the job of several chips, probably specialist chips at that*, you end up needing more processing power to do it all in real time.

*the audio chip does audio, it can then probably do a fancy audio technique with its baked in silicon that no x86 processor would ever justifiably include in its instruction set, leaving the emulator author to figure out a way to replicate the effect with basic maths (always doable, just potentially long winded).

Dynamic recompilation makes this easier -- processors are also quite simple devices in terms of the maths they can do where my program might not be. High level languages/compilers aim to help me avoid having to think how to do my crazy maths in small steps. However it will still be rendered into small steps for the CPU to handle, a dynamic recompiler aims to see this, figure out what is actually trying to do, generate some new code and make it so the host system has something a bit more native to it to run and that can be a lot quicker. You still have the potential issues of having to handle many different chips, even more on newer systems, but such a thing is still slower than having a dedicated chip do it.


Not all games need everything to march in lock step like the original system though and not everything the original system did needs to be replicated (the original system does a sinusoidal decay for the audio, go linear and for most it will sound close enough sort of thing all while taking less resources to do). As previously covered most people just want to play a game and that means you can take a shortcut to make said game happen. Developers on old systems and new will often learn the hardware and try to wring every last drop of speed out of it* so these shortcuts can trouble some games. This sort of thing is why the "emulation will never be perfect" mindset can exist and be justifiable in some cases, at least until we have transistor or FPGA driven stuff and we all start arguing about acceptable tolerances were for the original run (consoles are supposed to be a singular platform but you still have allowable tolerances, things like the xbox even having quite different speeds in RAM so they could get it out of the door cheaply).

*this also applies to do the downsides and bugs. For instance consumer grade CRTs were not great and would have a certain amount of blur. Some used it as a crude form of anti aliasing, others even used this blur to blend textures on a wall. http://bogost.com/games/a_television_simulator/ has some

Also have a read of
http://arstechnica.com/gaming/2011/...-3ghz-quest-to-build-a-perfect-snes-emulator/

and if you are up for it have a watch of


FPGA by the way = field programmable gate array. They are chips composed of things you can program to form computing units. Mostly they are used where fabricating a chip would be too expensive but you want something nice and custom done. However as you are quite literally making a logic device using software but rendered in hardware you can replicate existing chips perfectly if your FPGA has enough gates in it to do it.
They are an utter pain in the arse to use well but it is still doable
http://www.eevblog.com/2013/07/20/eevblog-496-what-is-an-fpga/ has more.

 

[Jiehfeng]

--mega-snip--

Well that makes perfect sense for a single processor to emulate many processes of a console, which can make it slow.

Very informative but dumb old me can't understand (or bother to) the latter part. xD (FOR NOW!) Hopefully someone will.
I kinda was surprised to see a long reply to a vague and short question. You seem to know a heck of a lot of stuff, I won't even be surprised if you said that you were one of the crowd in the audience of that presentation.

Also, thanks, and thanks for frying my puny brain cells.

 

[FAST6191]

4 paragraphs and some footnotes is probably not all that long by my standards. I also went slightly off topic to cover a few things that I would probably have included in a later reply to this thread.

On knowing a lot... not really. I just happen to know enough to be dangerous in computers, electronics, materials engineering, video making/restoration and some light literary/entertainment criticism, as such a skillset is almost completely useless around here I then have a lot of free time which allows me to spend my days (and it has been some years now) learning more of the same.

I would love to go to a C3 but when I total it up what it will cost it is usually that or a shiny new tool I have just found out about, or upgrades to them.