Strange behavior of the NeoGeo Core

[LockeBoss]

Hi,
my De10-Nano has a strange behavior while running the NeoGeo Core. I encountered (atleast i thnk so) no problems in other cores.
In advance i discussed this topic on another forum but we ran out of ideas.

I tested several things now to get rid of this problem (other roms, storage media, cores (versions), (stronger) power bricks and even other SDRAM. With and wthout the USB board and I/O board. I could narrow it down and it seems to be a tempeture thing of the DE10-Nano. There is a heat sink on my DE10-Nano and a noctua fan on my I/O board.

If i do fresh cold boot and go into the NeoGeo Core games crash with a "Exception Error Handling - Illegal Instruction" error message. Some games are crashing every time on the same spot. Like "Nightmare in the Dark" crashes everytime on the button info screen and i hit "A" (or the demo ends on its own) and the gameplay starts.
So i looked at the UniBIOS Gamecard Test and every rom i tested gave me mostly CRC = "NG". Tested several other versions of the same rom and several other UniBIOS versions on different storage Media.

I did test it without the I/O Board (and the Noctua Fan (this seems to be the important part!)). Because if i cold (i mean cold!) boot the De10-Nano and let it run something (does not matter what) for about 5 minutes, start the NeoGeo Core and load some roms, the CRC checks are all "OK". Every single rom i tested was good now. I repeated this several times (atleast 30 times now) to exclude chance.

For my tests i used "Rage of the Dragons" because it does several CRC checks and after one CRC check is "done" (takes about 1:40 mins) i reloaded the rom into the SDRAM via menu (F12).
Even if i let the core run (without fan) and every rom passes the CRC checks, shorty after i plug the fan back in the checks start to fail again after a few minutes. (reloading the rom into SDRAM after every completed CRC check). If i pull the fan plug, the CRC cecks start to be "OK" more and more and finally are all fine. Than it does not matter if i reboot the core or the whole MiSTer (even a cold REboot).

Without fan after a fresh cold boot: [1. round CRC ckeck: 8x"NG"]; [2. : 6x"NG", 2x"OK"]; [3.: 4x"NG", 4x"OK"]; [4.:0x"NG", 8x"OK"]; [5.:0x"NG", 8x"OK"]
Without fan and the MiSTer has done something 5 minutes long (and is warm now) before starting the tests: 5x: [CRC = 8x"OK"]
The MiSTer keeps running and i plug the fan back in: [1.: 8x"OK"]; [2. : 5x"NG", 3x"OK"]; [3.: 8x"NG", 0x"OK"]; [4.: 8x"NG", 0x"OK"]; [5.: 8x"NG", 0x"OK"]

I does not matter if the fan is pluged in and is pointing into another direction (even lies next to the MiSTer) the important part is if its cooling the DE10-Nano or not.

So has anybody an idea whats wrong here?

[jca]

Power supply problem?

[AtomicShroom]

Can you uh... describe/show the issue?

[LockeBoss]

Had to edit a few typos and did not know this would take the post so long to be visible:
"This post is not visible to other users until it has been approved by a moderator."

[bootsector]

Can you please post a picture of memtest after letting it run for about 20 minutes?

[LockeBoss]

I checked this on two different 128MB SDRAM boards (v2.4 and v2.9) both show the exact same behavior on the NeoGeo core.
I let the v2.4 board run memtest on auto for 3 hours but no abnormalities. I also ran the two jotego memtests for 3h each with no problems ( in know the only check the first 32mb).

[aberu]

Can you take a picture of your MiSTer with the fan attached?

[LockeBoss]

Here are two pictures with fan.
I did test it without the USB board and only with an micro USB hub with only the logitech unifying receiver attached.

IMG_20210807_104100.jpg (695.85 KiB) Viewed 7643 times

IMG_20210807_104038.jpg (571.17 KiB) Viewed 7643 times

[blue_chip]

I had exception handling errors too. My RAM Expansion 128 MB was at fault. I bought a new one and had no errors.

[redsteakraw]

Two things unplug the whole Mister from the wall wait a minute and plug it back in. Do a RAM test also. See what it says. If the RAM is bad it might be the case where the other consoles work because their ROMs are too small to end up on a bad block but NEO GEO games are big and fill up some of the RAM chips entirely so if it is bad it will look like a corrupted ROM and cause crashes. If there are any errors get the New Version of the RAM as it is rated for the higher speeds and should work better on the CPS cores.

[MostroW]

if not mistaken then the "bad apple demo" should fill the memory module up far enough?