MiSTer: MiST on Terasic DE10-nano board.

[Sorgelig]

However, the hardware difference (quality and features) of the DE-10 vs. the Replay (and new Replay2) is quite substantial.

Add almost tripple price and low availablity of FPGA Arcade.

There is 64MB of SDRAM

another 64MB of SDRAM

DE10-nano has 1GB of 800MHz DDR3 512MB available for FPGA usage.

single SD card

is it advantage?

[ijor]

The MiSTer is nice because the DE-10 is inexpensive. However, the hardware difference (quality and features) of the DE-10 vs. the Replay (and new Replay2) is quite substantial.

The Reply is a very nice board, and Mike is a top notch professional. But you can't compare the power of the Reply's FPGA with the one in the DE-10 nano. The DE-10 FPGA has about FOUR times the capacity! Let alone that it is a SOC that includes a dual core Cortex A9 at 800 MHz (plus 1GB DDR3 on the board).

Obviously the Reply has some advantages. But it is impossible to beat Terasic's pricing. Just the FPGA in the DE-10 nano alone costs much more than the whole board

[Sorgelig]

dual DB9 joystick ports

MiSTer has USB for joysticks. Tell me how to connect wireless pad/joystick to DB9

96bit audio

same on MiSTer.

composite/SVideo

OMG...

The daugher-board has Ethernet

on the main board of MiSTer.

multiple USB type A and Mini-B ports

Any HUB can do the same. And if you need to disconnect the board, a single USB connector is more convenient than multiple.

You've forgot another point: MiSTer is fully open source unlike FPGAArcade. If you check the SVN(who use SVN nowadays??) you will see source of several cores only. Minimig on FPGAArcade is not open source for example.
And if you compare amount of cores for both boards...

[Sorgelig]

Probably, Mike could make his FPGAArcade project on top of DE10-nano. I even talked to him and offered cooperation. I think, amount of GPIOs on DE10-nano should be enough for his 68060 expansion. But he still wants to use a fully custom board even now with Replay 2. Well.. Up to him. It will be again very expensive and hard to find board.
And when it becomes expensive, people will think do they really need so expensive board with only few emulators with full of endless expectations or may be it's better to buy Raspberry Pi 3 and get a lot of emulators with it right here right now.

[Newsdee]

-SRAM save support in cores where it makes sense.

This is really important to "finish" some of the cores, since longer games are unplayable without it.

[Sorgelig]

-Adding SD support to the Atari800/5200 emulator so it doesn't require the additional card.

i have plan to virtualize the secondary SD card on main SD. So it will be mounted as image and will work as a read SD card.
Actually only HDL module of SD card is missing. So, if you can make a virtual SD card HDL module, then you will achieve this goal.
There is one virtual SD card module. It's used in some MiST cores. But it's very primitive and doesn't support all SD commands to fully replace a real SD card. So, probably it can be used as an initial implementation.
And then, if virtual image is present for current core, then it will be used instead of real SD card. ARM part is done already.

[JimDrew]

However, the hardware difference (quality and features) of the DE-10 vs. the Replay (and new Replay2) is quite substantial.

Add almost tripple price and low availablity of FPGA Arcade.

The Replay is $299 for the version with Composite/SVideo (a *requirement* in the U.S. retro community as many people use composite monitors). The Replay is not "almost triple the price" just for the DE-10 at $130, let alone the add-on boards (another $70+ when you build single quantity). The Replay2 is being produced in quantity and will retail for around the same price with more features, including the same Cyclone V being used in the DE-10.

There is 64MB of SDRAM

another 64MB of SDRAM

DE10-nano has 1GB of 800MHz DDR3 512MB available for FPGA usage.

Yes, but the RAM is available through a tunnel, so it's slow - the very reason why the SDRAM add-on board was made, correct?

single SD card

is it advantage?

It seems to be according to people I have talked to, and I also agree. Having a single card with folders for each emulation (with sub folders for ROMs, disk images, and hard drive images) is a much easier way to deal with updates. You dont' have to be remembering which card was for what.

You have to remember, people using things like this aren't typically hackers so they probably don't know the difference between a ROM and core.

[JimDrew]

dual DB9 joystick ports

MiSTer has USB for joysticks. Tell me how to connect wireless pad/joystick to DB9

Well, there are wireless joysticks that use DB9 connectors, however, you could plug one into a USB port on Replay.

96bit audio

same on MiSTer.

There is an actual 96 bit DAC with advanced filter controls on the DE-10? There is on the Replay.

composite/SVideo

OMG...

Welcome to the real world. I could barely sell Replay boards without the Composite/SVideo connector. It's not an option for the U.S. market.

multiple USB type A and Mini-B ports

Any HUB can do the same. And if you need to disconnect the board, a single USB connector is more convenient than multiple.

How often do you disconnect a board though?

You've forgot another point: MiSTer is fully open source unlike FPGAArcade. If you check the SVN(who use SVN nowadays??) you will see source of several cores only. Minimig on FPGAArcade is not open source for example.
And if you compare amount of cores for both boards...

Actually, that is not true. Developers have full access to the SVN. There is no "minimig" on the Replay. It's a completely different emulation. Even T68K is being replaced completely right now.

[ijor]

Probably, Mike could make his FPGAArcade project on top of DE10-nano. I even talked to him and offered cooperation. I think, amount of GPIOs on DE10-nano should be enough for his 68060 expansion. But he still wants to use a fully custom board even now with Replay 2.

The DE-10 nano GPIO ports have some limitations. They are not the ideal type of connectors. You probably can't put DDR Ram, as he wants, using those connectors. It might be possible to use some of the other more expensive Terasic boards that have built-in DRAM and an HSMC connector for high speed expansion boards.

But besides the technical side, I'm sure he wants to build his own custom board. And I don' think there is anything wrong about that.

[Sorgelig]

Of course there is nothing wrong about that. Different people have different targets.

MiSTer has a big potential which is not implemented yet. While discrete FPGA solutions like FPGA Arcade requires additional hardware like a real 68060 on daughter board with many high-speed GPIO, hybrid solutions like DE10-nano can beat with tight coupling ARM+FPGA. 68060 can be implemented on ARM side while the whole chipset with RTG graphics will be on FPGA. Communication speed between ARM and FPGA can be very high if high-speed bridge is used (currently unused). And something tells me this combination will be faster and most important FREE, than a real 68060 CPU.

I will create a separate topic about it, so may be it will take attention of interested developers.

[ijor]

-SRAM save support in cores where it makes sense.

This is really important to "finish" some of the cores, since longer games are unplayable without it.

On most cores it is very difficult to implement a "save state" functionality. Saving RAM is not the big problem, but this is only a small part of the state.

[ijor]

The Replay2 is being produced in quantity and will retail for around the same price with more features, including the same Cyclone V being used in the DE-10.

That's not at all what it is published in the website.

[Newsdee]

-SRAM save support in cores where it makes sense.

This is really important to "finish" some of the cores, since longer games are unplayable without it.

On most cores it is very difficult to implement a "save state" functionality. Saving RAM is not the big problem, but this is only a small part of the state.

Ah, this is not about save states but backing up the SRAM used for game saves. In real hardware this is a piece of RAM kept alive by a battery (e.g. Zelda on NES), in emulation we need to store the contents to SD card and read them back on startup.

[Sorgelig]

There is an actual 96 bit DAC with advanced filter controls on the DE-10?

There is HDMI Audio which supports 96KHz audio.
Actually, there is no real audio source on retro systems requiring 96KHz sampling rate.

I could barely sell Replay boards without the Composite/SVideo connector. It's not an option for the U.S. market.

I didn't know people in US are using 20-years old TVs )))))

Actually, that is not true. Developers have full access to the SVN. There is no "minimig" on the Replay. It's a completely different emulation. Even T68K is being replaced completely right now.

Open source is when it's open. If only selected people have access to source it's not open source. So, you just confirm it's not open source platform.
I would think concrete before invest money to closed platform.

including the same Cyclone V being used in the DE-10

Mike doesn't use Altera FPGA. He uses Xilinx.

The Replay is $299 for the version with Composite/SVideo (a *requirement* in the U.S. retro community as many people use composite monitors). The Replay is not "almost triple the price" just for the DE-10 at $130

or $99 if you will buy it as a student/professor which is very easy to do.

let alone the add-on boards (another $70+ when you build single quantity)

Not at all. Also keep in mind you need only SDRAM board for minimal config.
Even with $150 it's twice cheaper with much bigger FPGA + 800MHz ARM + 800MHZ 1GB RAM.
Another advantage - is VIP cores providing a excellent video scaler for HDMI output. Xilinx has no such IP.

[Newsdee]

To be fair the Replay has been in the pipeline for a long time, and since technology has moved on it makes sense to (IMHO) to do a new revision but not necessarily to abandon it. There was a post back in 2013 by MikeJ comparing the MIST and Replay, where he stated:

We are all working to deliver open source cores which are as close to the original hardware as possible.
I designed Replay to be what I want - other people have different ideas and are perhaps more cost sensitive.

http://www.fpgaarcade.com/punbb/viewtopic.php?id=107

And certainly the same ideas remain today. MiSTer is oriented to be a cost-effective and more generic approach (as was the MiST), whereas one could see the Replay as specialized hardware designed for using a real 68060 (for example).

Interestingly in the same thread, Wolfgang recommends:

But a "classic" dev board I'd consider as good "add-on invest" to any dedicated (Minimig, Replay, ...) board. I have myself several generic boards (with different FPGA types, external memory and other features useful for development) and still use them beside the Replay for experimenting and (initial) development of code modules.

Four years later from those posts, the MIST successor is built precisely over such a "classic dev board".

There is a bit of downside to using an industrial/commercial board; Terasic could stop manufacturing them but chances are a successor would appear in that case. What could be a problem is if Terasic decides to abandon the price point (i.e. they don't provide any other boards at a similar price), but even then I'm sure there will be enough cards in the second hand market as alternative. I would say these risks are lower than the risk of a custom card stops being made (assuming the design isn't "open sourced" for anybody to build"). Note that in that regard the MIST is in good shape, as a few clones have started to appear. I've seen at least three, including one that adds composite and S-Video output (check the ZxUno forums for details).

[Sorgelig]

Terasic could stop manufacturing them but chances are a successor would appear in that case.

Terasic still sells DE1 board on Cyclone II. Their business is not to deliver a new boards and replace the older. Probably, as long as Intel will produce this FPGA chip, the board will be produced. And besides the MiSTer, DE10-nano is very interesting in many aspects and i doubt it has risk of sales termination.

At the same time, i would recommend to buy the DE10-nano. Don't wait

[ijor]

I could barely sell Replay boards without the Composite/SVideo connector. It's not an option for the U.S. market.

I didn't know people in US are using 20-years old TVs )))))

Jim is probably exaggerating, but I agree that old school video output is useful. A crisp sharp digital video output is very nice. But many retro fans swear that a retro experience must use a true retro display. And in many senses they are right. The feeling is completely different. In many cases it is not just a matter of feeling, a digital display might miss some color effects produced intentionally on an analog display (yeah, it is possible to emulate them to some degree if you want).

Another advantage - is VIP cores providing a excellent video scaler for HDMI output. Xilinx has no such IP.

Of course it has. Obviously it is not free. But Altera VIP is not free either. It is free only if you have the non-free Quartus version that most of us can't afford anyway.

There is a bit of downside to using an industrial/commercial board; Terasic could stop manufacturing them but chances are a successor would appear in that case.

This is definitely an issue. And it will likely happen sooner or later. Yeah, they probably would came out with something newer, but it won't be compatible with older cores. Ok, as long as you have the sources it might an issue of just a straight recompile, or it might need some porting. But expansion boards might not be compatible.

I think it is the price that you have to pay if you want something that powerful, at that quality, at that price point. As I say all the time, it's impossible to compete with Terasic pricing. You wonder what kind of discount they get from Altera. Conceivable they get a price at manufacturer cost, as Altera might be very interesting in promote those "educational" boards. Remember that some Terasic boards are actually branded as Altera/Intel.

Terasic still sells DE1 board on Cyclone II.

Yes, but the DE1 is an exception. Most other boards were replaced with something newer over the years. Even the big brother of the DE-10 nano, the Cyclone V Sockit is being phased out already.

[Sorgelig]

Yes, but the DE1 is an exception. Most other boards were replaced with something newer over the years. Even the big brother of the DE-10 nano, the Cyclone V Sockit is being phased out already.

Which board has been replaced? Probably some unsuccessful boards. All known boards like DE0, DE1, DE2, DE2-115, etc are all available today. Even DE0-nano is still available and it's direct successor DE0-nano SocKit didn't replace the former one. They both are available. DE10-nano is also direct evolution of DE0-nano SocKit - and again DE0-nano SocKit is still on sales.
So, your theory doesn't work with Terasic case. You've treat Terasic as Apple or Samsung who stop to manufacture the previous models.

And again, The feature/price of DE10-nano is so delicious, that you should buy it now. Even if you are not sure about MiSTer project.

New FPGA models created under Intel ownership is complete rubbish and step to backward direction. So, most likely there won't be a better dev boards than now. And DE10-nano is a golden choice, IMHO.

[JimDrew]

The Replay is open source. You can get the schematics and such for it. The Replay2 will be the same way, and it does in fact have the larger Cyclone V FPGA and high speed ARM, just like the DE-10. The daughter board is something that Amiga people want. Personally, I hate the 060. It was a compatibility disaster, but just like the U.S. market demands composite/SVideo outputs, the European market (primarily) demands to have an 060. I agree that I could write a 040/060 emulator on the ARM side that would rival the real hardware. I have already done that with FUSION-PC, which runs circles even the Vampire core on an I3 or later equipped PC.

[ijor]

Even DE0-nano is still available and it's direct successor DE0-nano SocKit didn't replace the former one.

I said the Cyclone V Sockit (big brother of the DE-10 nano) was phased out, not the DE0-nano SocKit. I also have the DE2-70, and other ones that were discontinued.

Regardless, it might change in the future. But I agree with you, the DE-10 nano price is so low that I don't care.

New FPGA models created under Intel ownership is complete rubbish and step to backward direction.

Some people say Intel is not really interested in the FPGA business per se, but only in applying some of technology to their own products. It would be too bad if they kill the brand

OTOH, they say that FPGA devices are ideal for testing brand new new CMOS fab technologies, something that Intel is very interested. An FPGA is not such sensitive to productions defects because they are designed with a bit of a extra resources that they can compensate after manufacturing testing.

And DE10-nano is a golden choice

To be honest, I prefer to pay a bit more and go for one of the bigger models that include DRAM and many other peripherals. It's only a pity that, for some reason, they include VGA but not HDMI.

[Sorgelig]

It's only a pity that, for some reason, they include VGA but not HDMI.

That's why DE10-nano is a golden choice. Add a miniature size here. And the price difference is huge. SDRAM addon is much cheaper.
For DE10-standard you will have to add HDMI board which will be more expensive than SDRAM board and will add the price on top of already expensive price.
But if price doesn't matter then DE10-standard can be an alternative, but i cannot call it's a better alternative. For project targeted for masses, DE10-nano is still a golden choice.

Actually MiSTer can be easily ported to DE10-Standard. If some one is willing to donate this board to me, i will port MiSTer project to it.

[ijor]

Actually MiSTer can be easily ported to DE10-Standard. If some one is willing to donate this board to me, i will port MiSTer project to it.

Regarding porting to similar boards. Do you have some documentation on what changes you implemented at the U-boot and Preloader in comparison with the default Terasic/Altera build?

[Sorgelig]

I've used other U-boot version. And all changes are tracked on github.

[JimDrew]

If someone was willing to create a complete developer setup that allowed for assembling (compiling I guess?) ARM assembly code, I would port my 68040 x86 assembly core to ARM assembly. I just don't know anything about the developer environment requirements. I would need a setup where I could just unzip an archive, load an example assembly program (that blinks a LED or something) to use as a template, and then I could do the conversion. I don't think I need debugging capabilities as long as I had access to LEDs or a UART.

[Sorgelig]

You can find discussion about build environment here: http://www.atari-forum.com/viewtopic.php?f=117&t=32441
ARM part runs a normal Linux, thus Linux environment for code compilation is more natural and easier to setup. But compiling under windows is also possible.
Although i'm not sure what you are going to do with 68040 emulation alone. It's not about CPU emulation only, it's also about connection ARM and FPGA parts.