MiSTer FPGA Forum

At least the secrecy gives us something to talk about!

My bet is it is a Jaguar core.

DS? He said it wasn't possible before but he is more experienced now and probably has more tricks up his sleeves.

mario64 wrote: ↑Mon Mar 20, 2023 1:17 am

The whole secrecy thing is annoying. He said it’s in case the core isn’t doable but he announced PSX knowing it may not be doable. Makes no sense.

One day, when you learn the personal pressures of having thousands of people waiting to overreact to something you work on, you will then know why he's being secretive about it. He's managing expectations.

You all know he's seen us all talk about N64 and giving it a feasibility test

With how God tier that PSX core wound up, Robert can work on whatever he wants at this point. Still I’m hoping for Jaguar

I'm hoping it's Star Wars Battle Command.

caffeinekid wrote: ↑Tue Mar 21, 2023 7:40 pm

DS? He said it wasn't possible before but he is more experienced now and probably has more tricks up his sleeves.

Definitely not. This post is from page 2 of this very thread, when DS was discussed as a possibility:

viewtopic.php?p=68792#p68792

Most bets are on the Jag. The other possibilities are NGCD, CD-i, VirtualBoy or a big surprise.

It would admittedly be very weird if it was the NGCD considering the WIP core has roughly 80% compatibility at the moment. Then again if it was that means we’d get save states and likely retroactive save states to the Neo Geo core so that’d be a big W.

My vote is on a Saturn core
Only because someone else is working on something has never stopped Robert.
He has his own ideas he wants to achieve and that also includes going beyond what the hardware could do like what he did with the PSX core.

Cebion wrote: ↑Wed Mar 22, 2023 9:25 pm

My vote is on a Saturn core
Only because someone else is working on something has never stopped Robert.
He has his own ideas he wants to achieve and that also includes going beyond what the hardware could do like what he did with the PSX core.

Robert didn't begin working on a PSX core until laxer announced that he wasn't working on it any longer. Then after Robert announced that he was working on it, laxer modified this to a hiatus, and then said he will work on it again.

This is obviously the wrong thread for this, but since it is the Robert Peip thread, I thought it was as good a place as any... I had a thought the other day about ao486. Rather than try to exactly replicate a real CPU (which I know ao486 does not), would it be possible to rip and replace the CPU that is in ao486 with a newly designed super efficient one that works well using the MiSTer's timing contraints and capabilities, etc.? In other words, design a new old-school 386/Dos/Windows/OS2 32-bit compatible CPU optimized for the MiSTer? It would need to be able to run the CPU instruction set, each of which could be newly designed, and possibly a math co-processor, and any modern day techniques could be employed to make this as efficient as possible within the capabilities of the MiSTer. If anyone could do this, it is Robert Peip. It might be amazing how fast that CPU could be, and it might be capable of playing Quake, etc. Then once the CPU is replaced, fix any other remaining issues with DMA, IDE or whatever else might need some tweeking for compatibility. We could end up with a really fast retro PC. This could then be underclocked to somewhat match real world CPU's like the 286, 386 and 486 when necessary for throttling.

I think we need a Mister II for a better retro PC, N64, all pre-2000s arcades and maybe even Dreamcast FPGA. With current Mister everything we get extra now is just a cherry on top.

H6rdc0re wrote: ↑Fri Mar 24, 2023 12:26 am

I think we need a Mister II

The issue is not MiSTer but the FPGA device required.

Some more capable FPGAs already exist but they cost more than a thousand US dollars. I doubt there will be many people willing to invest in that!

thorr wrote: ↑Wed Mar 22, 2023 10:10 pm

I had a thought the other day about ao486. Rather than try to exactly replicate a real CPU (which I know ao486 does not), would it be possible to rip and replace the CPU that is in ao486 with a newly designed super efficient one that works well using the MiSTer's timing contraints and capabilities, etc.? In other words, design a new old-school 386/Dos/Windows/OS2 32-bit compatible CPU optimized for the MiSTer?

Don't expect to see such a complex 32bit CPU on the DE10-Nano with more than 100Mhz.
The AO486 core is not replicating original timings of a 486 CPU anyway and the resulting clock speed isn't that bad either.

Still, the CPU could use some other improvements that are hard to integrate in the current design:

• branch prediction: Would be very useful with such a long pipeline
• real Level 1 caches: The current level 1 cache is behind the (slow) TLB and therefore takes like 5 cycles to deliver a result when the TLB is used(e.g. all windows programs)
• operand forwarding: result registers from a calculation cannot be used for 2 cycles until the next calculation. With so few registers as the 486 has this is a serious issue and can bring down the cpu to 1/3 speed in some situations.

It would be some interesting topic to work on in the future, but honestly, we might just exchange the FPGA technology in future. I bet on a ultrascale+ FPGA the cpu could run at least with 250Mhz.

Thanks! You did an awesome job of polishing a terd with all due respect to ao486 and really made it shine impressively for what it is. It would be awesome to replace the underlying terd with something newly designed to properly maximize performance. The improvements you mentioned all sound like they could make a massive difference. I am familiar with pipelines and branch prediction and for sure that would help and the others sound very impactful as well. When I was able to play Quake II on the PSX core, I thought to myself it should also be possible in ao486. 100 MHz with a performance oriented CPU design would be amazing. The first Pentiums were under 100 MHz and were way faster than 486's clock for clock. The Pentium Pro was another giant leap. Anyway, I don't expect anything but I appreciate your thoughts on the subject and maybe someday something can happen with it.

Here is an interesting comparison with an AMD 486 at 133, twice the frequency of a Pentium 66, which is still far below the maybe possible 100 MHz: https://www.youtube.com/watch?v=tm5lIQ_47Ko

thorr wrote: ↑Fri Mar 24, 2023 7:09 am

When I was able to play Quake II on the PSX core, I thought to myself it should also be possible in ao486.

It's not a fair comparison though; the PSX version was optimized for the hardware and used a GPU, which does not exist in ao486. Not sure if a 3dfx voodoo 1 would work with a 486, but it would be hard to write and fit in the current core.

Still a lot of systems that would fit fine on the DE-10 Nano we don't have before we need to worry about a MiSTer successor. Only a few devs working on cores would even be able to make use of the extra hardware anyway, as the systems that won't fit are so complex Robert is one of the few who could even try take them on.

what can fit on the DE-10 Nano is exactly in the sweet spot of console generation, arcade game era and computer generation that I'm into thankfully.

I would like if it were a GX4000 Amstrad honestly (or a 3DO like second choose)

I think we can safely say based what Robert has teased that he is not working on a GX4000, but this (or an Amstrad CPC+ with GX4000 support) would be a very nice core to get. With the MiSTer project supporting so many systems now, not supporting these Amstrads is looking increasingly like an omission rather than a nice to have. I have a GX4000, is a fun little console, should dig it out and fire it up.

Newsdee wrote: ↑Sat Mar 25, 2023 9:31 am

It's not a fair comparison though; the PSX version was optimized for the hardware and used a GPU, which does not exist in ao486. Not sure if a 3dfx voodoo 1 would work with a 486, but it would be hard to write and fit in the current core.

Yeah a 3D card would help, but I do remember that with the original Quake game, I played that on my Pentium Pro 200 and it was smooth as silk without a 3D card. At half speed, it would still be playable. We had LAN parties and people were playing Quake on 486 100's without 3D cards.

thorr wrote: ↑Thu Mar 30, 2023 4:57 pm

Newsdee wrote: ↑Sat Mar 25, 2023 9:31 am

It's not a fair comparison though; the PSX version was optimized for the hardware and used a GPU, which does not exist in ao486. Not sure if a 3dfx voodoo 1 would work with a 486, but it would be hard to write and fit in the current core.

Yeah a 3D card would help, but I do remember that with the original Quake game, I played that on my Pentium Pro 200 and it was smooth as silk without a 3D card. At half speed, it would still be playable. We had LAN parties and people were playing Quake on 486 100's without 3D cards.

Yes, and idgames open sourced it. So you have it in great quality for modern platforms

Iirc, Games that benefited from 3dfx were Turok, Screamer Rally, ...

Like Robert said, a new x86 PC core could be marginally better than ao486 if built for MiSTer from the ground up, but that would be a lot of work for only a minor improvement. Even if someone managed to implement a Pentium-esque chip on MiSTer it would still be too slow to run late 90s 2D classics like Age of Empires or Rollercoaster Tycoon or Baldur's Gate. There's not enough room on the FPGA for a 32-bit CPU and a GPU, so any 3D accelerated games are right out. Would it be cool to have a better PC core on MiSTer? Yeah. Is it a good use of anyone's time? Not really. There's still so many arcade boards that need to find a home on MiSTer, I'd rather see those before anyone devotes themselves to a new PC core.

Mr. Encyclopedia wrote: ↑Sat Apr 01, 2023 2:07 am

Like Robert said, a new x86 PC core could be marginally better than ao486 if built for MiSTer from the ground up, but that would be a lot of work for only a minor improvement. Even if someone managed to implement a Pentium-esque chip on MiSTer it would still be too slow to run late 90s 2D classics like Age of Empires or Rollercoaster Tycoon or Baldur's Gate. There's not enough room on the FPGA for a 32-bit CPU and a GPU, so any 3D accelerated games are right out. Would it be cool to have a better PC core on MiSTer? Yeah. Is it a good use of anyone's time? Not really. There's still so many arcade boards that need to find a home on MiSTer, I'd rather see those before anyone devotes themselves to a new PC core.

While I personally don't think he'll be doing a 486 or pentium core + GPU, if my understanding is correct, the current ao486 size could be cut down somewhat if a talented developer worked on a new more efficient 486 CPU core. Who knows how much space that could save.

Reed_Solomon wrote: ↑Sat Apr 01, 2023 6:00 am

While I personally don't think he'll be doing a 486 or pentium core + GPU, if my understanding is correct, the current ao486 size could be cut down somewhat if a talented developer worked on a new more efficient 486 CPU core. Who knows how much space that could save.

I respectfully disagree - the reason the core is where it is today is that it is quite efficient because it's based on a software implementation that cut a lot of corners in the interest of performance at the expense of compatibility. A 'proper' 80486 core would only be more complex and slower, but more accurate.

The sound chip support is the largest chunk of what makes ao486 so huge. If everything except for one soundblaster 16 implementation were removed it would probably save a good chunk of space.

H6rdc0re wrote: ↑Fri Mar 24, 2023 12:26 am

I think we need a Mister II for a better retro PC, N64, all pre-2000s arcades and maybe even Dreamcast FPGA. With current Mister everything we get extra now is just a cherry on top.

We may need it but we won't have it because FPGA prices are already high and there aren't any cheap consumer friendly and available FPGAs that both meet those needs and the needs of the MiSTer project. If I am wrong you are free to point me to a chip that is capable, people aren't going to plop down thousands of dollars for a more advanced FPGA.

I don't think it would be marginally better, and I don't think Robert said that. I think it could be 200-400% faster than what we have today and hopefully at the same time more compatible. Also, the cache probably takes up a lot of space. This could be reduced with a better CPU theoretically. The compatibility alone is a reason to work on this. ao486 still has issues.

I be he is finishing the Saturn core by helping srg320 out.