Home Creators Posts Import Register Favorites Logout

Content

Español abajo 日本語は以下

Konami CPU

We are testing the CPU now and we are honestly shocked at how well it is going. If you were here last Autumn you would remember that we were working on the NGP CPU and that progress was slow. Eventually, the freshman engineer who was working on it parted ways with us and I had to look for another person. I found a computer scientist from Venezuela, who had struggled around Europe to find an opportunity and hadn’t been able to find a job where she could grow her skills. She accepted to work with us even though we couldn’t pay much because she wanted to learn. Three months later, she is mesmerized looking at how the CPU she is worked on is moving data up and down. In her own words, when she started she couldn’t understand half of what I was talking about. But now, she has found a new passion and we have found a team member to rely on. We look positively at future CPU developments.

Out Run

There are still some glitches on the left hand side of the screen in some scenes. After investigating them, it looks like the frame buffer is still a bit too slow to handle all scenes. I have tested several ways of improving it but I still have not got a perfect image. The frame buffer is a strategic development needed for more games too, so we won’t give up on this.

CPS 3 Custom Chips

After finishing tracing the first one, we have estimated the effort needed to trace the rest of the chips for which we have die shots. We think it will take about 15 man-months full time to do the tracing. Then there is the circuit extraction. So this is not workable with our current resources. We either look at a different solution for the tracing (some kind of machine learning?) or assume that the reverse-engineered information in MAME is good enough and try to implement a system based on that and the information we got from the schematics.

Español

Procesador de Konami

Estamos probando la CPU ahora y estamos sinceramente sorprendidos de lo bien que va. Si estabas por aquí el otoño pasado, recordaras que estábamos trabajando en la CPU de NGP y que el progreso era lento. Finalmente, el ingeniero novato que estaba trabajando en ella nos dejó y tuve que buscar a otra persona. Encontré a una informática de Venezuela, que había dado tumbos por Europa para encontrar una oportunidad y no había podido encontrar un trabajo en el que pudiera desarrollar sus habilidades. Aceptó trabajar con nosotros a pesar de que no podíamos pagar mucho porque quería aprender. Tres meses después, está fascinada al ver cómo la CPU en la que está trabajando mueve datos de acá para allá. En sus propias palabras, cuando empezó no podía entender la mitad de lo que le decía. Pero ahora ella ha encontrado una nueva pasión y nosotros hemos encontrado un miembro del equipo en quien confiar. Ya no tememos futuros desarrollos de CPUs.

Out Run

Todavía hay algunas fallas en el lado izquierdo de la pantalla en algunas escenas. Después de investigar, parece que el búfer de cuadros todavía es demasiado lento para manejar todas las escenas. He probado varias formas de mejorarlo pero aún no he conseguido una imagen perfecta. El búfer de cuadros es un desarrollo estratégico necesario para más juegos también, por lo que no nos daremos por vencidos con esto.

Chips Propios de CPS3

Tras terminar de trazar el primer chip, hemos estimado el esfuerzo necesario para el resto de los que disponemos de fotos. Creemos que tomará alrededor de 15 meses-hombre a tiempo completo. Luego está la extracción del circuito. Así que esto no es factible con nuestros recursos actuales. O buscamos una solución diferente para el trazado (¿algún tipo de aprendizaje automático?) o asumimos que la información de ingeniería inversa en MAME es lo suficientemente buena e intentamos implementar un sistema basado en eso y en la información que obtuvimos de los esquemas.

日本語

コナミCPU

現在、CPUのテストを行っていますが、正直なところ、その順調さに衝撃を受けています。去年の秋からパトレオンになっている方は覚えていると思いますが、NGPのCPUに取り組んでいて、その進捗は遅かったんです。結局、担当していた1年生のエンジニアが離れてしまったので、別の人を探すことになりました。ベネズエラ出身のコンピューターサイエンティストで、ヨーロッパを回ってチャンスを見つけようと奮闘していたのですが、これまで自分のスキルを伸ばせるような仕事が見つからなかったのです。彼女は「学びたい」という思いから、給料はあまり払えないけれども、私たちと一緒に働くことを受け入れてくれました。3ヵ月経った今、彼女は自分が担当するCPUがデータを上下に動かしている様子を見て、心を奪われています。彼女自身の言葉を借りれば、入社当初は私が話していることの半分も理解できなかったと言います。しかし今、彼女は新たな情熱を見出し、そして私たちは頼りになるチームメンバーを見つけました。今後のCPU開発にも期待したいです。

アウトラン

一部のシーンで、画面左側にまだ不具合があります。それらを調査した結果、すべてのシーンを処理するにはフレームバッファがまだ少し遅すぎるようです。いくつかの改善策を試しましたが、まだ完璧な画像は得られていません。フレームバッファは、より多くのゲームにも必要な戦略的な開発なので、このままあきらめることはないでしょう。

CPS 3 カスタムチップ

最初の1個をトレースし終えた後、ダイショットがある残りのチップをトレースするのに必要な労力を試算してみました。トレースには、フルタイムで約15人月かかると考えています。それから、回路抽出です。そうなると、現在のリソースでは実現不可能です。トレースについては、別のソリューション(機械学習か)を検討するか、または、MAMEのリバースエンジニアリング情報で十分と考え、これと回路図から得た情報をもとにシステムを実装するか、のどちらかです。

Files

Comments

emceebois

My vote, for whatever it's worth, is that you either look into human-validated machine learning, or just take the 15 months it takes. Baking a software emulator into HDL wouldn't make you happy, would it? You're too much of a perfectionist! And that's one of the reasons I'm proud to support you. I also look forward to the return of progress updates on the NGP, if and when you have the resources to return your attention to it

Pixel Cherry Ninja

Thanks for the update. I'm glad the team is doing well. Have a nice weekend.

Jeruro

Creo que deberías hacer una encuesta los patrones de que camino tomar con CPS3. Yo prefiero tener un core de CPS3 mas rápido con la info. de MAME, que aunque no será 100% perfecto se podrá jugar mil veces mejor que en cualquier emulador.

Filip Kindt

Good to hear about the Jotego iniative given the scientist a great learning opportunity!!!

Richard Maxey

Great update Jotego and team! Thank you

Arend Pronk

CPS-3: How about using the reverse engineering information from MAME, and then update this information based upon your teanms findings throughout the months ?

Mat Azel

thanks for the update, as someone that learned very basic VHDL as an EE and owner of a CPS3 board, I'm not surprised at all by that number. I for one would gladly support you for 15months if it means you contributed to reproducing chips on the CPS3 mobo for when mine will fail👍 I really do hope you will consider it.

Brad

I am happy to wait however long it takes for a proper CPS3 implementation, we are all very patient :) Thank you Jotego.

Tonton Kaloun

It's a bit of a shame to take inspiration from the Mame implementation to develop an FPGA core, we lose some of the spirit of the original MiSTer FPGA principle. I vote against it ! :) It doesn't matter if it takes 15 months or if it never happens on our MiSTer. Thank you for everything you've done so far.

Matthew Caron

I'm with several of the other people - I'm fine with a quick and dirty implementation so we can play more games, but I'm behind your work for the long haul and would like to see as accurate a replication of the original hardware as can be done, no matter how long it takes. Meanwhile, I too have been staring at assembly code all week - Cortex-M4 in my case.

Carlo Pedrazzini

Thanks for the update! I’m no expert on the technical side of things but I think extreme accuracy has to be the main driver of MiSTer development. So, no “shortcuts” should be taken. Take all the time it takes, meanwhile we still have software emulation. We trust you’ll do your best.

Andrea Bogazzi

Thanks for the update! Victor ( the yamaha guy that makes a lot of the traces ) was looking into writing a software that could extract the cell structure from the scans

Andrea Bogazzi

I also want to say, there is nothing bad with extracting an implementation from mame and then moving forward from there. All or nothing is a strategy that not always work. A fpga implementation is still better than a noisy pc running mame or a arm board running neofba.

Captain Ginyu

Welcome to the newest team member of the JOTEGO crew! 👋🏻

Rautz

Happy #JTFriday everyone! Great news on how quickly the Konami CPU is getting owned. Seriously great news 😀 I'm looking at building an Outrun cabinet in the next 18 months so hearing there is some minor issues to address is something that comes part and parcel with "different" hardware. That CPS3 update definitely puts a damper on things but seeing mention machine of learning has my attention piqued..... best of luck with this HUGE challenge!! Thanks Jose, the rest of the team and others behind the scenes that continue to work and share your results with us. Churry.

Raziel

I said from the beginning that we should use MAME as the source of truth as they've already been through this exercise of understanding all the custom chips, there's no reason to assume they got it wrong. I always thought it would be very impractical to try and hand-trace every custom chip, that was always going to take a long time. As for machine learning, I happen to know a lot about AI, that's not going to be an option since you will need MANY examples of traced chips to train an AI on (read: thousands, if not tens of thousands), not to mention hardware powerful enough to do the training, and people that know enough about AI to implement an AI that can do it in the first place.

SegaSnatcher

I will just say I have zero issues with you referencing Mame to get a CPS3 core out, if that makes things easier for you. The question is would users know the difference between a high quality core using Mame as a reference vs tracing from an OG PCB? I'm going to say probably not, but from a preservation perspective it might be worthwhile. So it really comes down to what is more important to your Patrons, a quality implementation using Mame as a reference or straight up preservation of original hardware? This sounds poll worthy to me.

davethebulb

In addition to the graphical glitches in OutRun, the music is still off too, with repeated percussion sounds like drums and symbols. Hopefully this will get fixed soon.

John Casey

I'm for taking the time to do CPS3 properly. Yes, 15 months sounds like a long stretch, but other projects you have been working on have ended up stretching out over many months too. How long has Outrun been worked on? Around 9+ months? NGP work started in 2021. Or is the concern that multiple projects you are working on are ending up being dragged out for many months/years? Are you worried that after 15 months of CPS3 tracing there could still be another 15 months putting everything together in a working state?

ER

Thanks for this great update and looking forward to future cores, however long it takes!

Jarbas Silva

Personally, I am fine with waiting 15 months for the tracing. What's a year and 3 months in the grand scheme of things? Not like we have a lack of things to play in the meantime. This Patreon is already pretty atypical since you release new stuff every month, or even every 15 days. Most patreons only see actual releases in months or years time, with only snippets of the work being shared.

Basil The Plant

Presumably you can start with the mame based version and then plop the custom chips in one at a time as and when you do them?

Syro

Thanks for updating the progress. I'm fine with whatever you think works best for you. I personally would have preferred PGM, but I am in no rush, and think that we will all get better results if you work on what's better for your teams work flow. Thank you as always :)

Phily

CPS3 - maybe focus on the reverse-engineered information from MAME and in time (depending on options in the future, crowd effort sourcing, AI etc.) you can target revisit if a certain chip is problematic or slow.