Are you programming the 2716 EPROMs? What EPROM programmer are you using? What brand of EPROMs are they?

It sounds like you're getting borderline program burns in them. The 5V supply from your bench power supply will be slightly different to your actual games supply, which may be more favorable to the EPROMs programmed cells.

Program each 2716 chip another 3 or so times *without* erasing between programs. i.e. so you top up the program cells.

On the bench you are probably applying the 5V at test point 5. The game inputs the 5V at J4 pins 16 and 17. Try something simple like cleaning any crud off of those pins. Couldn't hurt. Check the 5V and see how much juice the board is getting. Could be the eproms too. I've burned eproms that verify after burning but wont work on the board. Low 5V maybe or bad eproms.

Not true.

You need to make sure your -35 board is jumpered correctly first, locked on LED in addition to other causes, if the game can't see the software in the correct location due to incorrect jumpers for the rom you're using, it will never turn off the LED (It's literally the third and 4th instructions in the game rom)

Does it make any difference if you short CPU pins 39 and 40 together briefly in the game?

As I understand the O.P's opening post, he has two -35 boards jumpered for 2716 that boot on the bench but not in the actual game, so they should be jumpered correctly.

I have seen this 3 or so times on Pinside now and it was due to weakly written EPROMs. Programmer verifies ok but in game with a slightly different 5V level they fail. I've experienced the same thing with the TL866 programmer and 2716 chips.
mongi do you know anyone with an EPROM programmer?

What happens when you try what dothedoo posted?

If the LED goes out your theory is the more likely problem.

If the LED stays lit, it could be a reset/voltage stabilization problem.

Received bad IC's from Marco in the past. Not eproms but other stuff. They did replace them, though. I shop IC's at other places now.

They might make the same mistake burning them at high speed when they need to be programmed at slowest speed.

I recently had a case of two newly burned sets of four 2732 (used like 2716) that would not boot up on a MPU200 yet other another old set of 2716 did. Software works fine on a replacement MPU and the chips verify back OK. Damn ancient EPROMs being a pain. Too bad they don't have one DIP28 ROM socket or any easy way to add one. DIP28 so much more easier to find and burn.

Good advice. I'm getting the best results by changing the GQ-4x4 speed to -2 and burning three times.

I did not take an o scope to the laptop I burn ROMs with but a pretty quick updating DMM indicates the Vpp is not reaching 25vppp when there is a M2732 installed. If I start a burn with no chip in the socket I can see it gets all the way to 25v.

Also if you are buying 2716 and 2732 from China (or from a pinball supplier who is buying them from China) most of them I get are actually 25vpp chip but they say 21vpp. If you burn them at 21vpp on the GQ4X it takes like 3-4x as long to complete the burn as when set to 25vpp.

I did get a batch of 12.5v 2732s marked as 21vpp once. You kill the chip trying to burn them at 21 or 25vpp. If you are unsure start low and step up the voltage.

I wonder if some of these problems are people trying to burn a 25v chip at 21vpp.

I've just been helping someone with some test code on some old hardware which we've been having problems with assuming it was my test code or hardware at fault.
He was programming a 27C256 chip with a GQ-4X programmer and once he selected slow speed the chip programmed properly and the code worked.
So this issue goes beyond 2716/2732 chips with the GQ-4X programmer. 27C256 EPROMs are 12.5Vpp chips.
Something about the high speed algorithm GQ-4X uses is not right. I'd suggest using slow programming speeds to better ensure a deep write.

I wonder if there are any differences in the older gq4x and newer 4x4 versions as I have never really noticed problems with the speed settings. Mine is maybe 8 years old. I normally leave it set in the middle speed for old dip24 chips and fastest for dip28 and bigger. Burned a ton of different chips in that period.

Tms2732 is the only one i consistently had trouble with eproms being corrupted a day after burning.

My GQ4X is the older version - 6 or more years old - I always use the original software provided with the disc that came with it & use original erased eproms either the TMS2532JL, HN462532G or TMS2732, never had a problem with burning these.

The later version does not include a HN462532G option from the drop-down menu which the earlier does, doesn't really matter I guess.

What GQ-4X speed setting do you use for these old EPROMs?

Pity they don't make the speed settings more meaningful as time based options like other programmers so you know how long the program pulses are.

When any of these old Eproms are selected, I leave it at the +2 default speed (speed 1)

Also never had a problem reprogramming original HN462716 Eproms either.

I think I have updated the software a few times. I burn 2532, 2732, and 2716 under the generic profiles. Just make sure I pick the right programming voltage. Looking at the profile setup file it does not appear to do anything different between AM2716, HN462716, or M2716.

For 2532s the HN462532 have had the best luck with. Issues with TMS2532 are I can take a working chip off of a CPU or sound board. UV erase it until it shows blank, but then maybe 50% of the time it will not take another burn. GQ-4X always errors out right away. Sometimes they work and if it takes the burn its good for indefinitely. The Hitachi ones HN462532 usually always work.

I need to change the zif in mine. It is starting to get worn out so I snapped a pic of an older 4x one while it is open. Apparently they are worried about copies as they scratched off a few chip labels. ChongX caps.

I was using a dip 42 adapter before I got the top3000 and I think the fat square posts weakened the spring of the zif contacts.

20210106_121119 (resized).jpg