MiSTer FPGA Forum

silverharmonica wrote: ↑Fri Jan 20, 2023 5:21 pm

Nik73 wrote: ↑Thu Jan 19, 2023 4:39 pm

I have the same pre-built mister from ultimate mister (except it has the blisster board) and finally got round to trying the Saturn core this week. Unfortunately I’m having the same exact problem as you.
Did you find any way around it? I’m hoping the sdram isn’t the 4 chip kind that doesn’t work.

I still haven't found a solution, unfortunately. I'm curious if there's a way to identify if we have the Villena sdram that's been known not to work?

I’ve emailed Ricardo today to see if he remembers what were used at the time (Nov 2021). I noticed that ultimatemister SDRAM is currently a ”new” version with 2 chips.
I might just take my mister to bits to check myself.

There is no requirement on any Dev to support non standard 4 Chip Memory. It should never have been made. Why should a dev spend time supporting a non official memory expansion.

Well, amazing service from Ricardo. He replied to my email almost straight away and confirmed that all the SDRAM boards he uses, and has ever used, are two chip official ones so it must be something we’re doing silverharmonica.

I’m going to start again from scratch.

FPGA64 wrote: ↑Sat Jan 21, 2023 5:19 pm

There is no requirement on any Dev to support non standard 4 Chip Memory. It should never have been made. Why should a dev spend time supporting a non official memory expansion.

True, but it is still supported by the Mister and all finish cores are compatible with both configurations. Since it's an open source project, you always have devs happy to come up with solutions for everyone (especially Patreons, but not just) as long as it's actually possible and simple enough. It's really not necessary to get upset, it's not like you have to do it. Reducing e-waste can be a valid reason to actually make it work as well, but again you don't have to lash out. Antonio had good reasons to do the 4 modules in the first place. These modules were cheaper and easier to find for a while.

Neocaron wrote: ↑Sat Jan 21, 2023 6:52 pm

FPGA64 wrote: ↑Sat Jan 21, 2023 5:19 pm

There is no requirement on any Dev to support non standard 4 Chip Memory. It should never have been made. Why should a dev spend time supporting a non official memory expansion.

True, but it is still supported by the Mister and all finish cores are compatible with both configurations. Since it's an open source project, you always have devs happy to come up with solutions for everyone (especially Patreons, but not just) as long as it's actually possible and simple enough. It's really not necessary to get upset, it's not like you have to do it. Reducing e-waste can be a valid reason to actually make it work as well, but again you don't have to lash out. Antonio had good reasons to do the 4 modules in the first place. These modules were cheaper and easier to find for a while.

You're mostly right, but it is isn't supported by the MiSTer. It is a piece of hardware designed to loosely observe existing standards, under the false impression that it would be 100% compatible (hint: it wasn't). This is not the only piece of hardware which was designed, built, and sold by this manufacturer without a comprehensive set of compatibility testing.

Having said that... yes, if it is realistically possible, I'm sure that the core developer will try to make it work. And if not him, I'm pretty sure that somebody else will try. But because the timing characteristics are different on this module, don't automatically assume that this will be easy, or even possible. There are standard RAM designs for a reason - so that the developers have an established standard set of characteristics to build the designs against. When these are treated as flexible, either things stop working, or we end up with a "lowest common denominator" set of reduced design standards which reduces the potential of the machine, and doesn't benefit anyone.

Neocaron wrote: ↑Sat Jan 21, 2023 6:52 pm

FPGA64 wrote: ↑Sat Jan 21, 2023 5:19 pm

There is no requirement on any Dev to support non standard 4 Chip Memory. It should never have been made. Why should a dev spend time supporting a non official memory expansion.

Antonio had good reasons to do the 4 modules in the first place. These modules were cheaper and easier to find for a while.

Yes one good reason is more profit for him, cost shaving of using cheaper chips has never passed onto customer.

Antonio Villena also has designed another memory module XXSD128 in the past these were bad with certain cores due to board routing issues I owned one and had nothing but problems on neo geo core, once faults were known product was withdrawn my advise is stick with official mister project memory.

Nik73 wrote: ↑Sat Jan 21, 2023 5:51 pm

Well, amazing service from Ricardo. He replied to my email almost straight away and confirmed that all the SDRAM boards he uses, and has ever used, are two chip official ones so it must be something we’re doing silverharmonica.

I’m going to start again from scratch.

That’s encouraging, and thanks for saving me the time of disassembling mine! I’ll post an update if I figure out a solution.

Any news on Sergey? Last update on Patron and Twitter was November 16th and last activity on GitHub was December 1st.

silverharmonica wrote: ↑Sun Jan 22, 2023 4:46 pm

Nik73 wrote: ↑Sat Jan 21, 2023 5:51 pm

Well, amazing service from Ricardo. He replied to my email almost straight away and confirmed that all the SDRAM boards he uses, and has ever used, are two chip official ones so it must be something we’re doing silverharmonica.

I’m going to start again from scratch.

That’s encouraging, and thanks for saving me the time of disassembling mine! I’ll post an update if I figure out a solution.

Thanks to some help from Ricardo, I got mine working and figured out what was wrong. As usual it was something simple and my fault. I'm on a Mac and when I was renaming the bios to boot.rom it was keeping the .bin extension in the background. Once I changed that it worked great.
I've found this one works best - Sega Saturn BIOS v1.01 (1995)(Sega)(JP)(M6)[h].bin

Can some one please share the last compile file, please. Thank you

https://github.com/MiSTer-unstable-nigh ... ble-builds

I've never been able to get anything to work with this core, despite countless efforts, so last night I decided to try my 32mb memory. That was even worse That wouldn't even boot the bios - the 128mb I usually use at least gets into a game before hard locking up.

Then I decided to look at the versions of the memory, and the 128mb was 2.5, and the 32mb was 2.2.

With 2.9 being the latest, is this likely the cause of my issues with this core? I dont have any issues with other cores (memtest and jotego's memory test runs fine too).

Also, is there anywhere I can see what the differences are in the versions of the memory sticks?

If 2.2 is worse than 2.5,then probably 2.5 is worse than 2.9 so…
I have 2.9 and core/games works ok.

There's another thread where folks were pointing at the Antonio Villena RAM modules as not working entirely correctly. He went off in a different direction and used 4 chips instead of 2, and apparently some cores don't get along with this arrangement. With the Saturn core being so memory-intensive, it wouldn't shock me at all if it was another unhappy camper on the Villena design.

If your SDRAM board has four chips, that could very well be the problem.

I don't have any other specific knowledge, but apparently this core is marginal on one stick anyway, so if you've got an older, but standard board, that could also be the problem. It seems unlikely that they would have have gone through the pain of a board revision, unless there was an excellent reason for it. If you're handy with a soldering iron, you might be able to transfer your old RAM chips to a new board, which would save some money on the replacement.

About all I know about soldering is not to touch the hot end, so I'd probably just buy a new stick.

Version numbers are not that useful. What is useful is the max stable speed your ram can achieve in Memory Test.
I have run the saturn core both with 2.5 and 2.9. The 2.5 clocks stable at 142 and the v 2,9 clocks stable at 150. Using the 2.9 gave better results but v2.5 if its the official 2 chip design should be fine as long as you can hold > 140 in memtest.

FPGA64 wrote: ↑Mon Jan 23, 2023 7:14 pm

Version numbers are not that useful. What is useful is the max stable speed your ram can achieve in Memory Test.
I have run the saturn core both with 2.5 and 2.9. The 2.5 clocks stable at 142 and the v 2,9 clocks stable at 150. Using the 2.9 gave better results but v2.5 if its the official 2 chip design should be fine as long as you can hold > 140 in memtest.

What memory clock speed does the saturn core require? I think it is not 140mhz so this is just misleading to others.

Fastest listed requirement for a released core is 126mhz

https://github.com/MiSTer-devel/Wiki_Mi ... -use-SDRAM

Looks like it need at least 108mhz

annette wrote: ↑Tue Jan 24, 2023 1:46 pm

FPGA64 wrote: ↑Mon Jan 23, 2023 7:14 pm

Version numbers are not that useful. What is useful is the max stable speed your ram can achieve in Memory Test.
I have run the saturn core both with 2.5 and 2.9. The 2.5 clocks stable at 142 and the v 2,9 clocks stable at 150. Using the 2.9 gave better results but v2.5 if its the official 2 chip design should be fine as long as you can hold > 140 in memtest.

What memory clock speed does the saturn core require? I think it is not 140mhz so this is just misleading to others.

Fastest listed requirement for a released core is 126mhz

https://github.com/MiSTer-devel/Wiki_Mi ... -use-SDRAM

I wasnt giving a speed requirement I was just reporting what worked for me and what clock speed my Ram worked at. Just stating 2.5 or 2,9 isnt useful for debugging purposes since the ram modules can be stable at various speeds

After more than 2 months of silence, Sergey just did a commit on GitHub:
https://github.com/srg320/Saturn_hw/com ... 2023-02-11

Great to see he still has time, will and opportunity to work on this core... it means he's as good as possible, considering the circumstances

That's a relief. Mostly for the fact we no know he's still out there and ok as the circumstances allow. I don't mind lack of core updates as long as he let's the world know he's alright.

Great to see he's ok. I'm assuming the changes have to do with the Saturns CD Block?

Latest update from Srg320 on his Patreon about the Sega Saturn core.

https://www.patreon.com/posts/development-78620033

Great news. Castlevania SotN,one game Interesting for me on Saturn.

Great!

Saturn core improvement are great

Good to hear from him, in such horrendous circumstances.

MrSniffles wrote: ↑Mon Jan 23, 2023 7:57 am

I've never been able to get anything to work with this core, despite countless efforts, so last night I decided to try my 32mb memory. That was even worse That wouldn't even boot the bios - the 128mb I usually use at least gets into a game before hard locking up.

Then I decided to look at the versions of the memory, and the 128mb was 2.5, and the 32mb was 2.2.

With 2.9 being the latest, is this likely the cause of my issues with this core? I dont have any issues with other cores (memtest and jotego's memory test runs fine too).

Also, is there anywhere I can see what the differences are in the versions of the memory sticks?

I couldnt get anything to boot/ properly so I checked and I had a 2.5 128mb.

I just received my 3.0 from Ricardo, installed it, games work instantly

I have mixed results on the Saturn builds with v2.5 (2 chip) passing memtest at 140.
Most of the games work, and I believe I am getting the expected results with perhaps some games which work for others not loading due to sheer luck of the draw. £50 isn't much for me to throw at this, but I really see no reason to be jumping to any conclusions with in development code.
Why should anyone be updating.. Am I missing something?

you shouldn't update. I'm not saying anyone should update.

The core isn't finished so I would just wait it out.