Sega Saturn Core in Development

[dshadoff]

That's not what I recall. v2.9 was made with surface-mount pin header sockets.
However, there was a time when there were questions about which capacitors were to be placed in which positions on the boards for v2.5 (back around 2018, but that got straightened out).

And there were some boards from some makers during that period which were found to have been inconsistent with the standard (Jotego uncovered this a few years back), but I don't recall the details off the top of my head - it's in a thread somewhere on the forum.

[Akuma]

It all depends on His HW setup

I am running a v2.9 ram that clocks at 150 in the memtest so I have a very good ram chip

Also on v2.9 and mine run stable at 147 in memtest after a couple of hours, and no issues here

[Hydrochaeris]

I've been messing with the Saturn core and several games don't run at all (e.g., Radiant Silvergun, Nights), while some don't have music (e.g., Panzer Dragoon), and others have some shimmering white dots on the screen (e.g., Dragon Force II), an issue I haven't seen anyone talk about. I've used both CHD and BIN/CUE files from sets noted here to work, and since some games run fine I doubt it's a BIOS issue.

I have a v2.5 RAM module with two chips (purchased from MisterAddons several years ago). But I just ran memtest and it shows 142 after over two hours of running (and has worked perfectly with every other core). Are my problems with the RAM, and is it because 142 MHz is somehow not enough, or a different design issue as dshadoff mentioned?

I'm thinking about getting a v3.0 stick since I want this core to work properly, but If I get it from MisterAddons it's only guaranteed to test 130+ MHz after 5 minutes, which is worse than my current one. So I'm not sure what to do.

[longnosedgoblin]

I assume the mister multisystem ram is not 2.9?? I get sound issues on most games, and no way to upgrade the RAM. Hopefully ram compatibility can be improved further.

From github "A lot of games have sound issues."

Yeah but these games are ones that 2.9 ram runs fine, so I was questioning what version the multi system uses.

[FPGA64]

I assume the mister multisystem ram is not 2.9?? I get sound issues on most games, and no way to upgrade the RAM. Hopefully ram compatibility can be improved further.

From github "A lot of games have sound issues."

Yeah but these games are ones that 2.9 ram runs fine, so I was questioning what version the multi system uses.

It dosent it uses its own custom version. The important thing is the speed it gets in MemTest. 2.9 and 3.0 come in in the high 140's. 2.5 the low 140's. The higher the speed the betters the saturn core behaves

[longnosedgoblin]

Yeah no idea about that, hopefully optimization will get compatibility better on slower ram.

[Armakuni]

Yeah no idea about that, hopefully optimization will get compatibility better on slower ram.

From what Srg320 has stated on his Patreon speed might not be the only issue why we see some issues on older and 4 chip ram. He claims how is he is using the SDram is a first and is needed for bandwidth purposes. Jotego uses interleaving on his CPS2 core but we dont see similar behaviour so he possibly is doing something different ?

"This is the first core in which the memory chips are interleaved (run in parallel) to obtain the required bandwidth. This is a necessary requirement for the core to work in single sdram mode. Due to longer track lengths and more load on chip ports, it is not possible to guarantee normal work on 4 chip sdram module. Sorry."

[pbsk8]

I think it's all but confirmed now, but just to chime in .... upgrading my single 128MB ram from 2.5 to 3.0 (sourced from Mister Addons) made a ton of Saturn core issues go away. Nights now has music ... that was the first thing I tested

maybe a faulty ram?

Mine is 2.5 and Nights runs with music.

It was discovered with memtest was created that 2.5 and earlier modules had inconsistent speeds. That's the whole reason 2.9 modules were created. So 2.5 ram modules will probably work fine for some.

First thing I did when I got my mister was running the jotego memtest for over 1h and everything was fine. Because 2 years ago the cps1 had problems with some ram modules and it was also related to the speeds. Mine went fine.

[Atohmdiy]

What i find strange with these ram issues things is the change of behavior. I am testing the core for a long time, and for a period audio was working perfectly fine with the few working games i had in a bin/cue format. It's only recently that audio start to glitch, so it looks more like some kind of regression than a problem with my ram stick.
For exemple i was trying duke nukem 3d, audio was perfect. Since the last updates of the core, audio loop / glitch at startup.
Sure my 2.5 ram stick is "old" but since now i never have any issue. I heard people having issue with jotego cores, but no me, so i considered myself "lucky" with my ram stick for whatever reason.

[kerunaru]

Got a v2.9 RAM module from Villena and I can confirm the core works with Villena builds AND v2.9 RAM module

[Akuma]

Got a v2.9 RAM module from Villena and I can confirm the core works with Villena builds AND v2.9 RAM module

Just out of curiosity whats the outcome of the memtest after 1h ?

[Bristles]

My v2.9 hit 146 in memtest, and that was after 1hr. It did hit 148 for 10 mins, but slowly went down to 146, where it stayed.
What annoys me is that almost every v3.0 seller are saying almost exactly the same thing - "Tested at 130Mhz for over an hour", which is poor speeds. Are they just cut 'n pasting the same text from other sellers ?

[Xbytez]

"Tested at 130Mhz for over an hour", which is poor speeds.

The Sega Saturn core requires 107 MHz.

viewtopic.php?p=56296#p56296

[FPGA64]

"Tested at 130Mhz for over an hour", which is poor speeds.

The Sega Saturn core requires 107 MHz.

viewtopic.php?p=56296#p56296

Thats not what has been found though. 107 may be a minimal spec but good compatability only occurs in the high 140's. Its been found that a good 2.9 or above 2 chip design ram gives best compatability and a 2,5 even clocking in the low 140's can give sound issues.

4 Chip design ram is very bad for this core.

[Xbytez]

I agree the 4 chip SDRAM is designed and produced by a third party seller and is not part of the MiSTer projects released hardware.

Newer official MiSTer project memory boards v2.9 and v3.0 have stability and reliability improvements over the v2.5 and v2.4 such as an onboard voltage regulator and more filtering capacitors.

[kerunaru]

Got a v2.9 RAM module from Villena and I can confirm the core works with Villena builds AND v2.9 RAM module

Just out of curiosity whats the outcome of the memtest after 1h ?

144 Mhz after 1 hour

[FPGA64]

Got a v2.9 RAM module from Villena and I can confirm the core works with Villena builds AND v2.9 RAM module

Just out of curiosity whats the outcome of the memtest after 1h ?

144 Mhz after 1 hour

Thats not great for a 2.9 module more what I would expect a 2.5 to give

[DaveGeorge]

Upgraded my 2.5 to a 3.0 from Riccardo and it got rid of that nasty sound issue where the music would just simply cut out after a minute or two.

[16Bittt]

So is there a conclusion re RAM specs required for this core?
it seems there's still conflicting opinions.
my RAM is V2.5 from 2019 and I get 142 mhz on memory test.

[Xbytez]

So is there a conclusion re RAM specs required for this core?
it seems there's still conflicting opinions.
my RAM is V2.5 from 2019 and I get 142 mhz on memory test.

If you experience issues with the Saturn core, which are not a documented issue with the core, you can check GitHub.

Known Saturn core issues are here:
https://github.com/MiSTer-devel/Saturn_MiSTer#status

User reported Saturn core issues are here:
https://github.com/MiSTer-devel/Saturn_MiSTer/issues

It also needs to be appreciated the Saturn core is still in a beta state and development is ongoing, you will probably benefit from upgrading to the latest revision of official MiSTer project SDRAM which is currently v3.0

The last two versions v2.9 and v3.0 of the SDRAM module include improvements to the stability of the SDRAM module, the v2.5 SDRAM memory module amongst other improvements in the later revisions does not have an on board voltage regulator and is not as stable as a newer v3.0 SDRAM module.

[FPGA64]

So is there a conclusion re RAM specs required for this core?
it seems there's still conflicting opinions.
my RAM is V2.5 from 2019 and I get 142 mhz on memory test.

v 3.0 or 2.9 . Only 2 chip design. the faster the better

[AmintaMister]

Is 147 Mhz after 5 hour a good result?

[FPGA64]

Is 147 Mhz after 5 hour a good result?

Should be good enough for the Core.

[marcof]

I've read this post, really nice info, thanks.
I'm not sure about the "4 chip SDRAM" ??? It will never run good with the Saturn core? Should I update it right now?
Or there is a chance, with the core development, the "4 chip SDRAM" will have equal performance as the 2 chip SDRAM?
Thanks.

[FPGA64]

4 chip is a non standard design. The saturn core pushes ram hard. The author of the core may or may not support it but there is no requirement to spend time supporting non standard designs

[Xbytez]

Sergiy Dvodnenko

This is the first core in which the memory chips are interleaved (run in parallel) to obtain the required bandwidth. This is a necessary requirement for the core to work in single sdram mode. Due to longer track lengths and more load on chip ports, it is not possible to guarantee normal work on 4 chip sdram module. Sorry.

The above quote is from srg320's patreon posted on September 22 2023 as a reply to a question in the comments of his original post, this is direct from the developer of the Saturn core about 4 chip SDRAM modules.

[jordi]

do CHDs need to be in separate folder? they don't load anymore

[Hydrochaeris]

The need for the memory to test in the high 140s is incorrect. Something else is going on, though I'm unsure if it's some improvement in the RAM modules' design.

As I mentioned above, my v2.5 RAM tested 142 MHz after a long while and I had many issues with the core. I just received a v.3.0 stick from MisterAddons and after an hour it tests worse than my old one, only 141 Mhz.

Yet all the Saturn-related issues I had are now gone: games that didn't run at all now boot just fine (e.g., Radiant Silvergun, Nights), those without music now have music (e.g., Panzer Dragoon), and the white shimmering effects I noticed in some titles (e.g., Dragon Force II) are now gone. A caveat is that I haven't had time to test the games long term (like playing them a couple hours), but regardless it's an improvement from the old stick.

[dshadoff]

Basically what you are seeing is a different kind of stress on the SDRAM.
Using a 'simple' memory controller (as is included in the memtest core), you can reach one number.
However, srg320's Saturn core uses a different controller, with a different access pattern. This may put different stresses on the chip which aren't demonstrated by the simple controller.

With all this talk of some SDRAM modules being OK while others have issues, it might be useful for the RAMtest core to have srg320's new SDRAM controller included for additional/alternate tests... but since srg320 is busy with the Saturn core, I'm not sure who else would take this on.

[KennyL]

I'm using this spreadsheet to track progress but are there other ones with more granular breakdown like this N64 spreadsheet? That Saturn spreadsheet is too binary and frankly not very useful to track down progress and issues. For example, Daytona USA and other Sega games do run on my setup but music and sound drop out. Is that my issue or is it how it is for the core currently? You can't tell from the spreadsheet. Which games can be completed? No idea.