Final Fantasy III - 128MB SDRAM Issue

[James-F]

https://github.com/MiSTer-devel/SNES_MiSTer/issues/204
Further discussion about this issue here.

I use 32MB SDRAM module and never encountered this issue, though many users with 128MB module did.
The Issue on github was closed because of frustration, but I think it should be re-opened.
MiSTer gods were busy with other huge projects like TGFX16-CD, so do not despair, it will be fixed eventually.

PS.
First post (on the new forum), I feel like a MiSTer noob again.
A fresh start for a brighter future, hurray for the new forum!

[Sorgelig]

Actually i'm looking for such module. It's not a common problem. Only several ppl has such issue.
It's not that issue is forgotten. It's just hard to fix it blindly.
If i will have such module, then i will be able to see mode details. It's also unclear if the same module produces the same problem on other MiSTers.
if someone has such module bought from @rickdangerous (Ricardo Saraiva) then i'm sure he will replace it. You need to tell him the module has exactly this problem, so Ricardo will send it to me.

[keith.f.kelly]

Note that 128MB SDRAM stability can actually depend (surprisingly) on the HDMI connection. If you use a long HDMI cable, or run it through an HDMI switch, it can cause the SDRAM to become unstable at normal speeds. I discovered this the hard way while troubleshooting 128MB SDRAM instability along with Mister Addons. If your repro involves such an HDMI setup, try bypassing any HDMI switch and/or using a different/shorter HDMI cable and see if that magically cures it.

[Sorgelig]

this is weird...
may be hdmi switch draws too much power from HDMI' 5V?
Actually de10-nano is fully 3.3v and doesn't directly use 5V, but if 5V is overloaded, then it may drop bellow the specific voltage where 3.3V dc-dc converter start to fail.

[tenshin2002]

i always ask the same question, why only on FF3 in SNES core?

I have such module that gives garbled graphics in the intro after looping it for 3-5 times, always at the exact same spot. Since i got my I/O board with a fan I looped it for 10 times without issue, so heat is a factor as well? Or maybe the I/O board makes voltages change on the board somehow?

[nagus]

i always ask the same question, why only on FF3 in SNES core?

I have such module that gives garbled graphics in the intro after looping it for 3-5 times, always at the exact same spot. Since i got my I/O board with a fan I looped it for 10 times without issue, so heat is a factor as well? Or maybe the I/O board makes voltages change on the board somehow?

I've had graphical glitches on Super Castlevania IV which ended up being a bad 32MB RAM module.
I opened and closed a Github issue about this:
https://github.com/MiSTer-devel/SNES_MiSTer/issues/122

Note that 128GB SDRAM stability can actually depend (surprisingly) on the HDMI connection. If you use a long HDMI cable, or run it through an HDMI switch, it can cause the SDRAM to become unstable at normal speeds. I discovered this the hard way while troubleshooting 128GB SDRAM instability along with Mister Addons. If your repro involves such an HDMI setup, try bypassing any HDMI switch and/or using a different/shorter HDMI cable and see if that magically cures it.

This post interested me since I noticed that, when plugging in another powered-on HDMI device to my 2-port HDMI switch, my MiSTer's fan would spin even when its main power cable was disconnected. I have two MiSTers. One connected to a 2-port HDMI switch and the other connected to a 4-port switch. I've been running FF3, with its attract mode on loop on both units, for the last 2 hours with no graphical glitches so far. I've also turned on all the other devices connected to the switches. I'll continue to run them overnight and see if there are any issues. I'll report back if there are.

Both running the same ROM file via samba share
MD5SUM: 544311e104805e926083acf29ec664da Final Fantasy III (USA) (Rev 1).sfc

Both DE10-Nano boards are revision C
MiSTer #1 specs:
MiSTerAddons.com's MiSTer IO Board XL v5.6 with Noctua fan
MiSTerAddons.com's MiSTer USB Hub V1.2 with power splitter cable
Zerohimself's 128MB SDRAM XS-D v2.5
Power supply for MiSTer: 5V 6A
HDMI switch: https://www.amazon.com/gp/product/B079JQ9XXV/

MiSTer #2 specs:
MiSTerFPGA.co.uk's Digital IO Board v1.2 with stock fan (using IO board's power port)
MakerSpot Micro USB OTG Hub
Homemade 128MB SDRAM XS-D v2.5
Power supply for MiSTer: 5V 2A
HDMI switch: https://www.amazon.com/gp/product/B07SYHQ2PS/
Power supply for HDMI switch: 5V 1A


Edit: Added power supply specs and specified 128MB SDRAM
Update: 10 hours of looping FF3 attract mode and still no issues. Also, all devices connected to these HDMI switches are 5V devices, mostly USB powered.

[keith.f.kelly]

this is weird...
may be hdmi switch draws too much power from HDMI' 5V?
Actually de10-nano is fully 3.3v and doesn't directly use 5V, but if 5V is overloaded, then it may drop bellow the specific voltage where 3.3V dc-dc converter start to fail.

I think it’s not so much about power draw, but about poor electrical buffering on the DE-10 Nano’s HDMI port. When the MiSTer was powered completely off but simply left plugged into my auto-selecting HDMI switch, it would often randomly switch over to that input for no reason while I was actively using a different device on a different HDMI input, as if the switch was detecting a signal from the MiSTer. And sometimes the MiSTer’s fan would be slowly spinning even though it was powered off. I also had lots of problems with my TV losing video signal/sync from any other HDMI device plugged into the switch while the MiSTer was attached to it.

When I first got my 128MB SDRAM, the first thing I did was run memtest. It wasn’t stable above 120MHz. I actually shipped the SDRAM back to Mister Addons at his request because we couldn’t figure out why it wasn’t stable and we assumed it might be defective or a manufacturing defect. But it checked out fine on his end, perfectly stable at 150MHz. Even so, he cleaned up a couple tiny solder points a bit for good measure and shipped it back. When I tried it again, it still wasn’t stable above 120MHz. Finally, while troubleshooting, I happened to bypass the HDMI switch, and then suddenly it was stable at 140MHz. Then I swapped in a shorter HDMI cable, and it was stable at 150MHz.

Because stable at 140MHz seemed good enough, and I really need to use an HDMI switch in my home theater setup, I bought a different HDMI switch whose auto-select feature can be disabled (manual/remote switching only) and a shorter HDMI cable, and I leave it hooked up that way. With this setup, everything works, although it still takes my AV receiver and TV as long as a second to finally recognize and sync with MiSTer’s video signal after any resolution change, so I often miss the beginning game title screen animations or have to pause the action for a couple seconds whenever the resolution mode changes between a cutscene and the start of a new level.

The next thing I was thinking about trying was some kind of powered external HDMI signal booster. I figure it may act as a proper buffer between the MiSTer and the HDMI switch and cure the remaining signal sync annoyance.

[James-F]

Wow, a lot of people with 128 GB modules,, and here I am with only 24GB of ram in my i7 PC.

[Xzarian]

Ya I'm also one of the few who have this issue with FF3. Sorgelig, I can send you or whoever my module if you need it. I got mine from Porkchop Express if that helps.

As for hdmi switches, I've tried both of the ones I have. I have also tried different cables and different displays with no switch, always the same results. Garbled up FF 3 text.

[Xzarian]

Also, I just tested a 1ft hdmi cable directly connected to a display, same issue, garbled text. I even only had one usb controller connected to the otg cable. So only power draw was from the fan on 5v pin.

[nagus]

keith.f.kelly: Just curious, what are the model numbers of the old and new HDMI switches?

[Xzarian]

Also, not sure if it helps, but I just ran latest memtest for 20 minutes at 140 with no errors.

EDIT: 90 minutes and still going no errors

[Sorgelig]

Ya I'm also one of the few who have this issue with FF3. Sorgelig, I can send you or whoever my module if you need it. I got mine from Porkchop Express if that helps.

As for hdmi switches, I've tried both of the ones I have. I have also tried different cables and different displays with no switch, always the same results. Garbled up FF 3 text.

i'm afraid you will send me this module and it will work well on my MiSTer. Although i have about 5 de10-nano boards - chance is higher.

[zypheratus]

@Xzarian Have you tried running FF3 with a 32MB module with your current setup (IO Board and hdmi switch) ?, is your hdmi switch powered separately ?

[mr-box]

i have the garbled graphics on my d-10 nano too it went away but its back now i just check again not sure what the hell it can be . the problem i have is the title screen screen is corrupt and the lighting strikes turn into yellow blocks lol in text is ok only this games that i have played has this problem

[keith.f.kelly]

Wow, a lot of people with 128 GB modules,, and here I am with only 24GB of ram in my i7 PC.

Hahah! Sorry, brain wasn’t fully engaged. I’ll go correct it.

[keith.f.kelly]

keith.f.kelly: Just curious, what are the model numbers of the old and new HDMI switches?

Sorry, no idea what the old one was as it is long gone.

This the one I am using now:

https://www.amazon.com/gp/aw/d/B06WV5YJ6H

EDIT: I found the one I was using before:

https://www.amazon.com/gp/aw/d/B01IT1WSQU

[ioev]

I actually just got my 128MB ram yesterday, tested out FF3 by letting the intro run through a dozen times or so, and I didn't see any problems.

I'm running DE10 Nano with a heatsink and 12v fan, v2.5 128mb ram module from mister addons, unpowered usb hub running an 8bitdo receiver, bluetooth and usb wireless (not working yet) and hdmi is going through a powered auto-switch I got from aliexpress.

[Xzarian]

@Xzarian Have you tried running FF3 with a 32MB module with your current setup (IO Board and hdmi switch) ?, is your hdmi switch powered separately ?

Ya, it works fine with 32mb sdram switch or no switch. Ya, both switches I use are powered. I don't use an i/o board though.

[Xzarian]

I just ordered another de-10 nano. This is bugging me too much. hahaha

[bootsector]

I have two 128MB SDRAM boards. One of them gives me garbled graphics always at the same place in FF3 (@2:03):

https://www.youtube.com/watch?v=1UKeg4RS0Hk

This same module passes memtest at 140Mhz on two DE-10 Nano boards. It also gives me the same glitch in other DE-10 Nano board as well.

I would send Sorgelig this RAM module, but I've lent it to a friend (along with a full MiSTer setup).

The other 128MB RAM module I have doesn't screw up FF3 intro, but it screws up Secret of Mana after starting a new game or loading a saved one:

https://www.youtube.com/watch?v=xVnuu1ARH1g

Sometimes I can make the SoM glitch to not happen if I press buttons real fast when starting the game and not giving a chance for the intro to run. If intro runs even for a little bit, chances I get corrupted graphics when starting a game increases.

It makes me wonder why this is happening only with Square games.

Also, both games run flawlessly on both 128MB RAM modules with a really old version of the SNES core

[James-F]

Can we make the Memory Tester stricter? Maybe random address smaller chunks?

What is the fundamental difference between the 32MB and 128MB modules that may cause this issue even if both pass full memory write+read at 140MHz?

[mdd45]

I have two 128MB SDRAM boards. One of them gives me garbled graphics always at the same place in FF3 (@2:03):

https://www.youtube.com/watch?v=1UKeg4RS0Hk

This same module passes memtest at 140Mhz on two DE-10 Nano boards. It also gives me the same glitch in other DE-10 Nano board as well.

I would send Sorgelig this RAM module, but I've lent it to a friend (along with a full MiSTer setup).

The other 128MB RAM module I have doesn't screw up FF3 intro, but it screws up Secret of Mana after starting a new game or loading a saved one:

https://www.youtube.com/watch?v=xVnuu1ARH1g

Sometimes I can make the SoM glitch to not happen if I press buttons real fast when starting the game and not giving a chance for the intro to run. If intro runs even for a little bit, chances I get corrupted graphics when starting a game increases.

It makes me wonder why this is happening only with Square games.

Also, both games run flawlessly on both 128MB RAM modules with a really old version of the SNES core

Are you using integer scaling? i have a 128mb module, fan, heatsink and i don't have this glitch...

integer has nothing to do.... Just got the glitch too in third or fourth intro loop

[bootsector]

I have two 128MB SDRAM boards. One of them gives me garbled graphics always at the same place in FF3 (@2:03):

https://www.youtube.com/watch?v=1UKeg4RS0Hk

This same module passes memtest at 140Mhz on two DE-10 Nano boards. It also gives me the same glitch in other DE-10 Nano board as well.

I would send Sorgelig this RAM module, but I've lent it to a friend (along with a full MiSTer setup).

The other 128MB RAM module I have doesn't screw up FF3 intro, but it screws up Secret of Mana after starting a new game or loading a saved one:

https://www.youtube.com/watch?v=xVnuu1ARH1g

Sometimes I can make the SoM glitch to not happen if I press buttons real fast when starting the game and not giving a chance for the intro to run. If intro runs even for a little bit, chances I get corrupted graphics when starting a game increases.

It makes me wonder why this is happening only with Square games.

Also, both games run flawlessly on both 128MB RAM modules with a really old version of the SNES core

Are you using integer scaling? i have a 128mb module, fan, heatsink and i don't have this glitch...

Yes. I do use integer scaling. However, in my case this is not related to scaler or HDMI as it happens when connected to a CRT via IO board as well.

[rickdangerous]

Actually i'm looking for such module. It's not a common problem. Only several ppl has such issue.
It's not that issue is forgotten. It's just hard to fix it blindly.
If i will have such module, then i will be able to see mode details. It's also unclear if the same module produces the same problem on other MiSTers.
if someone has such module bought from @rickdangerous (Ricardo Saraiva) then i'm sure he will replace it. You need to tell him the module has exactly this problem, so Ricardo will send it to me.

Again, If anyone have a module with this problem just let me know if want to trade for a good one. We need one shipped to Sorgelig, and I sent a new one without any failure.

Keep safe, Enjoy MiSTer

[Xzarian]

I just checked some older snes core releases as bootsector pointed out in the github issue. Just like bootsector, the last working version for me is SNES_20190204. Something happened between SNES_20190204 and SNES_20190225 to cause this I assume? As the atari forums are gone, so is the change log that Sorgelig maintained so I can't see what he posted. I took a look at the commits between those two versions and there are a lot of changes suffice to say.

Feb 4, 2019 (This commit happened after the release I think?)
- Reduce audio taps.

Feb 11, 2019
- Fix Vsync length

Feb 12, 2019
- Merge pull request #35 from paulb-nl/vsync

Feb 23, 2019
- Merge pull request #14 from MiSTer-devel/master

Feb 24, 2019
- Rewrite the muxer of chips.
- Minor update.
- sys: update scaler, scandoubler. User port definition. Remove VIP.
- Merge pull request #36 from srg320/master
- SA1 support

Hopefully this helps. I do have a new de-10 nano on the way so I'll report back my findings when it arrives.

[bootsector]

I just checked some older snes core releases as bootsector pointed out in the github issue. Just like bootsector, the last working version for me is SNES_20190204. Something happened between SNES_20190204 and SNES_20190225 to cause this I assume? As the atari forums are gone, so is the change log that Sorgelig maintained so I can't see what he posted. I took a look at the commits between those two versions and there are a lot of changes suffice to say.

Feb 4, 2019 (This commit happened after the release I think?)
- Reduce audio taps.

Feb 11, 2019
- Fix Vsync length

Feb 12, 2019
- Merge pull request #35 from paulb-nl/vsync

Feb 23, 2019
- Merge pull request #14 from MiSTer-devel/master

Feb 24, 2019
- Rewrite the muxer of chips.
- Minor update.
- sys: update scaler, scandoubler. User port definition. Remove VIP.
- Merge pull request #36 from srg320/master
- SA1 support

Hopefully this helps. I do have a new de-10 nano on the way so I'll report back my findings when it arrives.

I've already tested every commit in between the working and the non-working builds. You can see the results here:

https://github.com/MiSTer-devel/SNES_Mi ... -585208380

[Sorgelig]

Those who have intro problem, please test this version:

SNES.zip

(1.8 MiB) Downloaded 357 times

[bootsector]

That didn't fix for Secret of Mana at least. I will ask my friend to test with the other RAM module specifically on FF3 and I will report here.

[Xzarian]

Those who have intro problem, please test this version:
SNES.zip

Unfortunately doesn't fix it for me.