RetroUSB AVS - HDMI NES implemented on a FPGA!

[RGB32E]

Update [8/8/2016] - Finally available for pre-order! Standard and LL Editions for $185 and $650.20 respectively.


I had the opportunity to check out Brian's HDMI NES prototype at the Portland Retro Gaming Expo this weekend. Initially this project was to tap the bus lines between the CPU and PPU to generate a signal, which was doubled for 480p HDMI output. I was surprised to learn that the project headed in another direction... a FPGA console built from the ground up! The HDMI NES outputs at 720p (clean or scan lined), and looks great! I inquired about 1080p output and learned that a much faster FPGA would be required for 1080p output and would raise the parts price by $50! The current estimated price is $139.95. The PRGE expo hall was pretty noisy, so I wasn't able to evaluate audio. The case isn't final, and was created via a 3D printer in vinyl material. The console even had a Famicom DB15 connector for expansion controllers and devices! Power was supplied via mini-USB.

Unfortunately this console is at least 6 months out as the online score board system must be completed, and there are no plans to release the console without this feature. Despite the launch window being 6 months out, I was extremely impressed by what was presented!

Picture Album: http://imgur.com/a/BmDIz

[Fudoh]

The dual-slot design for NES and Famicom carts is really neat.

Did you have a chance to evaluate the horizontal scaling once the 1:1 pixel aspect is disabled ? I could imagine that this is about the hardest task in the whole design. If they just have nearest neighbour scaling while true 4:3 aspect ratio is enabled, it would result in rather bad scaling/scrolling.

[RGB32E]

The dual-slot design for NES and Famicom carts is really neat.

It's designed so that you can connect the RAM cart for the Famicom Disk System (FDS).

Did you have a chance to evaluate the horizontal scaling once the 1:1 pixel aspect is disabled ? I could imagine that this is about the hardest task in the whole design. If they just have nearest neighbour scaling while true 4:3 aspect ratio is enabled, it would result in rather bad scaling/scrolling.

The system configuration menu has 3 different settings for aspect ratio - 1:1, 4:3, & 5:3. I set the TV to 1:1 mode (no non-uniform scaling), and didn't notice any horizontal pixel wobbling when scrolling. I believe he kept all of that correct for the different pixel modes - columns of scaled pixels all had the same width.

[Smashbro29]

So it went from 480p to 720p, it has scanlines and has all the abilities of an NES and Famicom? Is it still made from an actual NES or is this like a Retron or something? Does that matter in getting 100% accuracy?

[hermit crab]

Yeah wtf, I thought it was supposed to be a HDMI mod for a real NES but that looks like some kind of famiclone?

It matters.

[eightbitminiboss]

Yeah wtf, I thought it was supposed to be a HDMI mod for a real NES but that looks like some kind of famiclone?

It matters.

Yep, as I recall it was going to be a add-on board for top loaders originally. Apparently not the case now...

[Kyle]

Yep, as I recall it was going to be a add-on board for top loaders originally. Apparently not the case now...

Wow! I trust Brian but this is way different than I expected. He was ripping on the Retron 5 for not being real hardware.

[leonk]

Wow. There's a lot of action in the NES hardware scene these days. Wonder if it's because technology finally caught up with user needs (all 3 use FPGAs), financially it finally makes sense (FPGA and PCB printers/assemblers are dirt cheap while RGB PPUs are expensive now) or a little of both.

I'm not sure which one will "win" but I do see the merits of all 3. Have choice can only be good to keep these retro games alive.

[undamned]

FPGA and PCB printers/assemblers are dirt cheap while RGB PPUs are expensive now

FPGA's are not cheap. Baring an expensive custom case, the FPGA is definitely the most expensive piece of the pie.
-ud

[alamone]

Well correct me if I'm wrong here, but he can still be justified in dissing a Retron,
which is basically just a software emulator with cartridge readers.

On the other hand, a FPGA is a reconfigurable hardware that can mimic
the hardware of the original system. So although it's a different approach,
it may be closer to the original hardware than just a software emulator.

Of course, I'd rather just use the original hardware to begin with,
which is why I'm more interested in the NESRGB. Plus, I'd rather have
straight 15KHz RGB out than have it outputting in HDMI, as I prefer to
play low-res stuff on CRT if possible.

[panzeroceania]

I would rather have a kit that I could install on a real NES/FC, but I am still interested in this project. It doesn't always have to be one or the other. Still would like to see a add on chip but I can see how realistically he'll be able to sell more of these to the everyday gamer.

real world business often leaves us enthusiasts high and dry.

Still a cool looking project. At this point I'm happy to support any development in retro gaming hardware.

[Smashbro29]

Honestly speaking, I'm not too upset about not using original hardware so long as everything functions 10000000% like the original one did. Not to mention it's a great option as it basically saves you buying a Famicom too.

Honestly, it will run 720p with scanlines directly to the TV? Even audio will run via HDMI? That's really impressive.

So let me stop being hype for a second and do the breakdown.

Game genie inside, 4 ports native, will play all NES and Famicom games and accessories in the exact same way as the NES and Famicom (ignoring lightguns of course). It looks nice the suggested price is nice. Screw the new consoles, this is what I'm excited for.

[leonk]

Well correct me if I'm wrong here, but he can still be justified in dissing a Retron,
which is basically just a software emulator with cartridge readers.

On the other hand, a FPGA is a reconfigurable hardware that can mimic
the hardware of the original system. So although it's a different approach,
it may be closer to the original hardware than just a software emulator.

!?!? They're both the exact same thing!! The Retron uses an FPGA to emulate the CPU and so does the HDMI NES. The only difference is the actual FPGA software. I would bet Brian's implementation would be more accurate and have periodic software updates. That being said, they are both emulators.

the RGB nes adapter on the other hand does not emulate the CPU. It emulates some functionality of the PPU and some it lets the original ppu to do (hence why it's still needed) I assume it too might have some corner case glitches but a lot less than the ones that recreate the entire system.

Lastly, there's something to be said about cosmetic aspect of the projects. I prefer my retro consoles to look as original as possible. The nes HDMI might be an engineering marvel but it reminds me too much of NOAC eBay systems that sell for 20$.

[Fudoh]

The Retron uses an FPGA to emulate the CPU

last time I checked the Retron 5 was a generic design based on an ARM CPU with software emulation for all supported systems.

[alamone]

As Fudoh said, I thought that only the older Retron models used FPGA, whereas the Retron 5 just uses a general purpose processor or SoC with a software emulator.

[hermit crab]

LOL. The man is a complete asshole and always talking shit, not just about Retron but about Krikzz and everyone else too. But he did bring us the PowerPak even though people used to say it would be impossible (mappers yadayada). And because of that will have my full respect for that forever (basically he made my childhood dream come true). Not interested in a famiclone though, from him or anyone else. Software emu, FPGA, SoaC, whatever... It's not the real thing! Sure it can be fixed until every known game works, but no code is ever perfect. And even if the code was perfect, who knows if all features of the NES have even been discovered let alone documented yet...

[leonk]

I stand corrected.

From my understanding, if we compare the FPGA based solution to an ARM based one (i.e. put both in a box, and do a double blind test) assuming the software has been correctly ported to the two, then there should be no difference except for performance. So, again, what this boils down to is who has a better software implementation. I would put my money on Brian and the NES HDMI having a better software implementation.

But don't fool yourself. They are both "emulators". One emulates the NES CPU/PPU in software running on ARM, the other emulates the CPU/PPU in hardware. Brian is really good at making emulators (the NES RetroPak is a good example of that) -- but it's not perfect. Many have stepped in to fix bugs in the RetroPak mappers. It's been a few years now that the RetroPak is out, and it's still not 100% perfect.

I assume the same thing will be true with the NES HDMI. 97% of all games will work. But there will be some corner cases that might not. But given his track record, Brian might share the software, which will allow the community to make the product even better for all of us. He shared the source for CopyNES and he shared the source for RetroPak.

[Smashbro29]

I really really hope you're wrong and we get it 1:1 out of the box.

[ApolloBoy]

This actually looks pretty neat, but only having HDMI output kills it for me.

[RGB32E]

This actually looks pretty neat, but only having HDMI output kills it for me.

http://www.hdfury.com/products/hdfury-gamer/

Problem solved!

[Ed Oscuro]

HDMI and no worries about analog signal degradation is the whole reason to get one. I understand the need for RGB systems but this isn't one. There still is at least another person working on an RGB board for the NES, so look at that if you want RGB on a NES. No reason to make this system more expensive than it need be.

They do need to roll out a 1080p version though. 720p is going to add lag on many TVs.

Edit: I realize what follows here is confusing and in some cases might be wrong, so please skip it.

I don't get the apparent criticism that an FPGA is "just" emulation. It's running the native code on the original CPU, reimplemented on the FPGA. It's not as good as a NOAC could be, but let's face it, a NOAC would be expensive to make and also not easily designed to interface with the modern systems touted here. Unless I'm very mistaken, going with the FPGA will allow much better speed (close to native, or even at native NES speeds) than emulation in an ARM box. Basically, the two technologies are considerably different and it's highly misleading to criticize the FPGA implementation as being "just an emulator." That suggests all the features and foibles of emulation and that's not true. It won't have the lag issues of running a PC emulator on a highly powerful PC - but with lag added from the operating system and other sources on a PC, such as graphics compositing; it also won't have the power or inaccuracy issues of translating the opcodes for an underpowered ARM CPU. I am sure there is some downside to the FPGA but it is apparently not the big deal some people think it is.

Hope they make a clear system with that sexy red PCB, but I expect the final to just have a boring gray shell.

[trap15]

Not to offend, but you don't know what you're talking about. An FPGA is literally an ASIC that is field-programmable. This means you can literally create a new ASIC by uploading a new bitstream. This means an FPGA can literally act exactly like whatever ASIC you want to clone (in this such instance, an NES PPU), and work functionally 100% equivalently. There is no such thing as "speed" since it is acting on a per-cycle level, with no translation and no simulation. An FPGA allows for 100% accuracy with no overhead.

The only disadvantages to an FPGA are that average programmers can't develop the bitstreams for them (they're written in their own Hardware Descriptor Language, either VHDL or Verilog usually), and they're typically fairly expensive.

[Ed Oscuro]

I want to stress that I know we all profit from your concise explanations, but I don't want to take the fall for things I haven't actually said, or have the conversation go off the rails and not look at the important things. I think your time would probably be better spent addressing the original point I tried to address - whether FPGA is "just like emulation" (which, clearly, it is not in terms of details - but is an FPGA implementation "just like emulation" as far as the consumer is concerned? My belief here is that a NOAC has significant drawbacks - it's costly to design and manufacture, and is inflexible. Its main benefit, power, might be eroded if implementing a NOAC means that other functions have to be implemented on other chips for purposes of keeping development and production costs down. Obviously I was misleading to present this unstated factor as a benefit - let alone one that consumers would care about. At the same time, I think that people understand that NOAC promises (in theory) to be an exact reproduction of the original system. What they don't know is the schedule of costs and benefits, and whether other approaches really are inferior. This is something that I think needs to be addressed, because it's a bad meme in the classic gaming community that makes people suspicious of things they shouldn't be.

At the same time I had the belief that though a "pure" emulator running directly on an ARM is going to give a better emulation experience than running on a full system with many sources of lag. My question is whether current ARM CPUs can do all the processes of emulation quickly enough to compete with a NES reimplementation running on a FPGA.

I want everybody to disregard what I wrote about "speed" and NOACs in the previous post mainly because I didn't carefully explain what I meant, or whether I thought these presented good arguments for one approach or the other. I would say that I believe these aren't worth mentioning in this discussion and I shouldn't have brought them up.

[ZellSF]

HDMI and no worries about analog signal degradation is the whole reason to get one. I understand the need for RGB systems but this isn't one. There still is at least another person working on an RGB board for the NES, so look at that if you want RGB on a NES. No reason to make this system more expensive than it need be.

They do need to roll out a 1080p version though. 720p is going to add lag on many TVs.

Pretty sure that's very few TVs, not many. I haven't run into any with noticeable more lag in 720p than in 1080p and I'm pretty sure Fudoh and Leo Bodnar has said something similar.

Just saying, so people aren't worried about buying this because of that.

I'm curious how "open" this will be, if future NES clones can learn anything from it, getting clone hardware perfected would be much better than hanging onto really aging hardware. I don't see a need to be upset by this really, for people who want real NES hardware there's alternatives coming (NES RGB board, NES Universal PPU).

[hermit crab]

I'm curious how "open" this will be, if future NES clones can learn anything from it, getting clone hardware perfected would be much better than hanging onto really aging hardware. I don't see a need to be upset by this really, for people who want real NES hardware there's alternatives coming (NES RGB board, NES Universal PPU).

Well it's mainly because it was initially announced as a HDMI mod for a real NES so obviously it turned out a disappointment for people who wanted that. Nothing inherently wrong with developing clones or emulators of course.

Regarding FPGA vs SoaC... Isn't the main difference that FPGA's are general use standard components, can be bought even one at a time and you can update the code as many times as you want, they are just expensive. SoaC's you have to order 10000 (or whatever minimum the factory has) and any bugs will be there forever, but price per unit will be low.

[Ed Oscuro]

Regarding FPGA vs SoaC... Isn't the main difference that FPGA's are general use standard components, can be bought even one at a time and you can update the code as many times as you want, they are just expensive. SoaC's you have to order 10000 (or whatever minimum the factory has) and any bugs will be there forever, but price per unit will be low.

I think that's all right except that SoaC units, being custom ICs, are not going to be a low price per unit - as you said, 10,000 in a batch is too much for a hobbyist project.

There is also the very substantial up-front cost of designing an IC, even before you get to problems - trap15 mentioned the lack of familiarity with using the VDL for FPGAs, but that's far less problematic than trying to design an IC. If you were trying to design a SoaC with HD output, either you're just making clone Nintendo chips (might not be hard to do, if there is a copy of the mask out there somewhere), or you're making that clone chip and also throwing circuitry for the HD video output in there, which doesn't seem simple to do (unless you want to pay somebody money for their design, maybe, and know how to integrate it with the NES CPU). Going the NOAC route you'd probably just end up with another chip dedicated to the HD output anyway. Compare that with the options for FPGA HD video - it looks like there's lots of options pretty much ready-made for getting HD output right from a FPGA, some even might be free. Of course, if the FPGA manufacturer wants, they can even put video for outputting HD video right in the IC with the FPGA (there currently are models that have regular CPUs along with the FPGA). Don't know if there is something like this in the market yet (that is low-cost, at least; there seem to be a few FPGAs out there with HD but intended for industrial machine vision), but IP cores should be fine.

[Ed Oscuro]

HDMI and no worries about analog signal degradation is the whole reason to get one. I understand the need for RGB systems but this isn't one. There still is at least another person working on an RGB board for the NES, so look at that if you want RGB on a NES. No reason to make this system more expensive than it need be.

They do need to roll out a 1080p version though. 720p is going to add lag on many TVs.

Pretty sure that's very few TVs, not many. I haven't run into any with noticeable more lag in 720p than in 1080p and I'm pretty sure Fudoh and Leo Bodnar has said something similar.

Just saying, so people aren't worried about buying this because of that.

I'm curious how "open" this will be, if future NES clones can learn anything from it, getting clone hardware perfected would be much better than hanging onto really aging hardware. I don't see a need to be upset by this really, for people who want real NES hardware there's alternatives coming (NES RGB board, NES Universal PPU).

Good to know - I gotta stop spouting disinformation

Edit: Here's an interesting slideshow (rather old, though, from 2001) about ASIC (i.e. NOAC) versus FPGA. It seems to cover pretty well a lot of the areas we've been talking about. Note the source, though; it's certainly got a certain agenda. Any thoughts on this?

[leonk]

Ed.. you forgot the link

[Ed Oscuro]

Dohh...here it is:
https://www.doc.ic.ac.uk/~wl/teachlocal ... llasic.pdf

But here's a different link that has just the high points:
http://www.xilinx.com/fpga/asic.htm

I think it's in the presentation slides (in the PDF) that it's mentioned that "non-recurring engineering costs" might run into millions of dollars. Would it really be that expensive to make a chip that combined the NES circuitry with HD output? Maybe not but it probably still wouldn't be cheap at all, no matter which route you take.

[leosmendes]

hello ! I would like to thank all the questions and explanations in this topic , I did not know but had some imagination worked as an FPGA , imagined that it was a software - hardware like many here , not as a programmable hardware in order to behave 1:1 as original hardware , it is clear that there is no documentation, but based on trial and error , there is no limit to overcome even if the original hardware , you all know also have glitches or moments Some work requires a " more power " or other resource. but leaving aside the discussion about emulation or fpga , and also suggest ... since it was abandoned the use of the original hardware , then why not also have the option of using a software player not original or better, a card reader roms with all its advantages such as a built GameGenie well as products krikzz ? another thing is the nostalgia of putting the American cartridge like a VCR , but really everyone knows the problems and difficulties and others in this type of insertion . is little wonder that left 2 nes .