One thing to consider is the remark factor and revision number of the FM1608. Checking the data sheets you will see as time goes by and revision numbers increased, the rated life of the chips increases by a factor 100 from 1999 to 2004. When buying chips from eBay, they often take tons of the same kind of chip and refurbish them. They straighten the legs, resolder the legs, and they also reprint the top of the chip. If they reprint the top of the FM1608 you don't know if you have FM1608 rated for 10 billion writes (1999) or a FM1608 rated for 1 trilling writes (REV M date codes start around 2004).

All my FM1608 have original imprints, revision M (1 trillion write endruance, 35 years between cycles) and are tested good.

Saddly I cannot compete with Shenzhen sellers. I have squeezed the price point about as low as I can. As long as the batteries are getting off the MPUs is the main thing. About a years ago plus your options for NVRAM was priced up to $40 each. Competition is good the end pinballer.

http://html.alldatasheet.com/html-pdf/80224/RAMTRON/FM1608/245/1/FM1608.html early date code data sheet showing the endurance is only 10billion.

I have lowered my prices since that post too, I need to go back and edit that. You can get a 10 pack of FM1608 for $70 shipped. Plus i have 5101 NVRAM adapters and 6116 NVRAM adapters at good prices too.

Repro boards are in the works. Aux lamp driver is out, Bally -51 sound board coming soon.

http://nvram.weebly.com

That looks like a remark to me. The font and package is wrong. Still probably works, but see my post above about revision number and write endurance. You want 1 trillion write endurance, not 10 billion.

The font used on that package "1" character should not have the horizontal bar across the bottom is what makes me think it is a remark. None of mine do in that package across similar dates.

I thought it was missing a circle shaped divot on the top, but that package does not have them.

Has anyone counted how many write cycles our pinball machines are doing while just sitting in attract mode? Is 10 billion not many writes or plenty?

The new FM16W08 dated 2011+ seem to have a 100 trillion write endurance. Maybe i should start pimping those as new and high write endurance. I paid more than $2 each the FM1608 pulls i have, cant compete with Shenzhen there.

The 6264 RAM is the scratchpad in these later games, no? The lamp matrix i think would be doing lots of writing even during attract mode, maybe not. We would need a software guy to figure that out i suppose.

If the ram was written to every cpu cycle of 1mhz. That is 166 hours to reach 10 billion. 16,666 hours to reach 1 trillion.

And the brand new SOIC-28 FM16W08 would take a whopping 190 years of straight writing at 1mhz to reach 100 trillion write endurance.

The date code and endurance could be important.

That is the surface mount version SOIC-28 of the RAMTRON. You want DIP28 version.

I still think you should be buying these from me =D =D =D I still have tons of them. EEP.

ebay.com link: usr

That user sells good enough dual wipe IC sockets and ships in tubes. They are the MH brand stamp ones that i got. Same ones you usually get from Tayda. One of my test MPUs has probably near 100 cycles on the EPROM spots. While the socket is and easy to pull the chip out after an extreme amount of insertion cycles, it still isnt causing problems.

I think the 100 trillion write endurance of the brand new FM16W08 will make it a moot point. No way a pinball machine will make that many writes if it is powered on writing at 1mhz for the rest of our lives. 10 billion write endurance of a 1999 dated FM1608 is probably a different story.

Something tells me RAMTRON estimated far on the side of caution with the stated write endurance. Makes me wonder if they changed the write endurance rating without changing the actual die of the chip. I guess i could try and find every data sheet release during the revision 'M' die date codes and see if the endurance inflates.

Here is suspected remark comparison from the AliExpress to a real imprint. When they remark these devices, a lot of times they get the font wrong. Main thing i notice is RAMTRON never used a font, at least in my experience, that has the lower bar across the number one character.

Original imprint.
ram real.jpg

Possible remark on AliBaba
fm1608 re.png

For what its worth, remarks tend to work in my experience, but take precaution.

rmy6nuM.jpg

So i looked at the data on quite a few of my FM1608. Most of them appear to be used pulls from copy machines, which that and medical equipment my source told me they parted out the FM1608s i have came from. Looks like specifically the RICOH Afico MP 1800 are on every FM1608 in the last few tubes I opened up. They must have parted out hundreds if not thousands of this specific copy machine.

I have to figure this would be a low write environment for the FRAM. It is just going to storing the settings and what not.

If you want to poke around and look at the data on a FM1608. You can read / write them in GQ-4X with the dallas DS1225Y profile. Might be able to monkey around with audit mode stuff and high scores. Backup your ram if you really wanted to as well.

Untitled.png

90% of the space in each of these RAMs look like they are completely unused and just have 00 written all over them.

I just looked at the contents of a FM1608 from pinball games. In Getaway High Speed 2, there is ASCII in there about ZZ Top(custom message?) and the high score initials.

I found ASCII high scores initials and this in Baywatch's RAM
Untitled.png

6116 nvram adatpers will work in all system 11 games. 11B and 11C can use an off the shelf FM1608 NVRAM with a jumper change which are less expensive.

Oh geeze, here we go machine pin vs dual wipe sockets =D. Probably for another thread.

For what its worth. The dual wipe China sockets labeled with a MH seem to be the best quality ones coming out of China. I have gone through thousands of those sockets with good results. My Bally test MPU EPROM sockets have gone through probably at least one hundred insertion cylcles, while it is loose, there has been no problems.

I have had some China machine pin sockets that where so bad one or two insertion cycles made them very loose, i figure the wrong type of soft metal was used.

If fishtails wants to clear settings at bootup, i would consider that a problem? It would be interesting to see what the NeoLoch ram tester thinks of that FM1608.

I got a response about endurance and data codes from Cypress. Includes an interesting note that a READ counts as an endurance cycle.... copy paste below.....

-=-=-=-=-=-

Hello Andrew,

As per the information from our product experts, we can expect the same data endurance in FRAMs from both 2000 and 2010. Earlier we did not have the data to guarantee 10^12 endurance cycles in older parts, that is why the older datasheets guarantee less endurance cycles. But, please be aware that unlike other non volatile memories read operations in FRAMs also will consume one endurance cycle.

Please let me know, if you need any clarifications.

Thanks and Regards,
Sudheesh

-=-=-=-

This could be a reason to consider the fm16w08 adapters since the endurance is greater by an entire magnitude. Reality also is the endurance rating is a very conservative rating. I would assume that they will likely last a the lifetime of a game.

Also there are some charts that show difference in the VCC supply by only tenths of a volt effect endurance greatly.

I have assumed the data retention was number of years between write/read of a memory location.

I think temperature may effect endurance rating. I will see if I can understand the bar graphs later today.

Looks like voltage definitely effects endurance. The voltage drop across the blocking diode left in games probably helps the FRAM last longer.

Since some of you are buying the FM1608 right from China(boo, buy from me hehe). Is there any interest in DIY kits for 6264/FM1608 RAM to 6116 RAM? I just made the layout and it is super simple because of how similar the footprint is 6264 to 6116

I could sell just the blank adapter board. Adapter board and just the parts to assemble. Or assembled adapter boards and you plug in your own FM1608 into the 6116 adapter.

I have to think about it more. Might be counterproductive for me to do this.

No jumper change needed. I have never seen anything but a 6116 come in a sys11 mpu from the factory. The RAM is hard soldered to the PCB tho.

I think they have to be double sided unless you want to solder machine pins from the top side. That might be difficult for someone with limited practice.

I learned this too when offset staggering pins. You want a little space so you can actually have room to solder. There is a pattern to follow doing these offset pin adapters so you have the most space to get the iron tip in there. You will figure it out after assembling a couple.

I would think scanning of the lamp and switch matrix would be hitting the ram on every scan. Every time the WPC computer increments in the in game clock would be RAM access too. I am not sure if it is counting that stuff by the millisecond or what.

Ramtron/Cypress claims 1 trillion write endurance of the FM1608

A system 11 game runs at 1 MHz (1,000,000). If the CPU read the ram once every cycle it is about 700 days to reach 1 trillion writes.

If you are worried about write endurance, the SOIC-28 version FM16W08 is 100 trillion write endurance.

SOP28 = SOIC-28 surface mount package. FM1608-P have lead in them. FM1608-PG is the "green" version. The package is a different shape with different imprint fonts. That is the difference i see.

The ones that end in -PG are more likely to have a newer date code. Based on what I have seen in my stock the -P went obsolete around 2005/2006 and after that they are all "green" -PG version

I wonder how fast the switch matrix strobes on a WPC game. That would be a good benchmark value to guesstimate how long before 1 trillion read/write cycles happen.

UTsource is as legit as they come from China second source parts. They where about the only place i could find the sound generator for the -51 sound board i am redoing.

FM16L08 is the 3.3v version. I don't think you are supposed to run these at 5v per the data sheet.

I have no idea what FM1608M is. Never seen one, clearly not a remark since the top is scuffed up. Might mean nothing. Revision M is what i on most date codes.

FM16L08 is the 3,3v of FM1608. Might work but over voltage.

FM18L08 is a different IC all together. Would not work as 6264(i assume??)

FM1608M apparantly works fine and makes sense based on part code syntax. Not sure the M suffix means at the end. Military??? That "M" suffix does not show up anywhere in datasheets i have seen.

Yikes that packaging is terrible. Right before i fired EPROMMan for Chinese remark EPROMs. He sent me a bunch of TMS2732 packaged like that and 2/3 of them did not work.

The fact RAMTRON put ESD numbers in the datasheet makes you think it is at least somewhat ESD sensitive. The ESD rating for FM1608 is human body 4kv and "Machine" 300v. I dont really know what that rating means as far as handling, but I am sure wrapping them in stretch wrap stacked on top of each other is not ESD safe.

2732
ebay.com link: itm

2716
ebay.com link: itm

Both are remarks but i have had good success with them. Burned hundreds of each flavor from this seller. They come in washed clean, no label gunk. Pre erased. And i have had almost zero failures from this seller. Note that they sometimes mess up the voltage rating / suffix of the EPROM. My 2716s says 2716A which would be a 21v eprom, but they only burn at 25v. ***EDIT*** I emailed the seller letting him know the imprint voltage was wrong on his device. Looks like they updated the ebay picture and the remark imprint. hah... i guess they at least try and remark it properly.

Considering how long ago these EPROM sizes obsolete, you really don't have many options for EPROMs of this age. EPROMMan was good for a while, but i started getting some REALLY beat up skanky EPROMs from him. They would be all sticky and cleaning up and erasing them was more work than it was worth.

Used EPROMs are going to be sketchy from any source. I am sure EPROMman didnt want to sell bad EPROMs, but how is he going to test them all. Same with the chinsese. They just pull them out of devices, match up like parts, wash clean, erase, straighten and solder bath the legs, rub the original imprint off, remark, resell. I doubt they test them. Who knows if someone burned an eprom with the wrong voltage range 30 years ago.

True NOS 2732 would be pretty rare. ST was probably the last people to make these things considering the Chinese like to use ST logos when remarking. The ones i have i think say 2009 date codes.

They are remarks for sure. Well done enough you can be fooled that it is a truly new M2716.

I have M2716A marked chips that will only burn at 25v. That is a pretty big hint it is remarked. Here is a close up pic from the seller i linked above. That seller excepts best offers btw.

m2716 remark.jpg

another pic i took a while back

http://s27.postimg.org/lnw6xonmb/20141207_115742.jpg

and my notes are...
The M2716-1F1 on the left says 21v right on the chip... yet it only works at 25v. These have been 100% good for me. Used over 100 of them no failures.

The M2732A-2F1 in the middle. Where they normally put The "21V FAST" text at is missing. The A suffix means 21v... these actually burn at 25v. I have gone through 25 or so of my 100. All have worked 100%.

The M2732A-2F1 on the right are 21v chips. I got these from EPROMman and they have about a 25% failure rate right out of the tubes... The bad ones refuse to burn at any voltage. The good ones work fine at 21v.

Interesting... all the ones i got came from Ricoh copiers too, mine came in bulk amounts even. Must be a common source.

I just bit the bullet and ordered 1000pcs more of FM1608-120-P(G). It will take 7 to 10 days for my recycler to pull them off of the copier control boards, restamp the legs, solder both the legs, and test.

Got the impression the FM1608 might be getting close to a limited supply available. Figure i need to strike while the iron is hot if I want to try and stay in the FM1608 game.

Lets dance! I bought a few of these NOS from a board house and i don't intend to get stuck with them.

20151220_124143.jpg

I don't think we have seen the bottom of the NVRAM game just yet.

5101-dip28_(resized).png

I created a new 5101 adapter for people that like to DIY. All through hole components. Jumper out and 7400 installed = dual CE operation. Jumper in and 7400 out = single CE operation. Might offer them up as kits for people buying their own FM1608.

Good call. updated the pic in the post above.

Of course i had a bunch of traces around p11 i had to navigate around. Just added a few VIAs. might clean it up some more before i order.

Got another round of 1000pcs in. These look better than my last batch. So far each one i have read still had Ricoh copier stuff on them.

In my last round of 1000pcs i had maybe 5 bad ones out of the tube and only one report of DOA at a customer. Zero complaints otherwise.

20160108_080048_(resized).jpg
20160108_080137_(resized).jpg

back in stock at nvram.weebly.com

Stacked up like that, one ESD blast could go through numerous devices since they are electrically connected.

If you dont't think it is sitting right in the zif socket, stack on my SIP socket strips on the legs

Here

My best advice is to find a dedicated person who you can trust to shop the Shenzhen market for you.

It is not a specific vendor. It is a middle man that shops shenzhen for me and I pay him for that service. Shenzhen is a "special economic zone". Which I assume means they cheat on the value of their money. If you buy electronic components from anywhere else, you are probably paying too much. You can get brand new in original manufacturer packaging from shenzhen, used pulls, remarks etc... You really need someone who you can trust though.

I give him lists of what I want and he tells me what is available. He is honest with what I get sent as we have an understanding the money stops if I end up getting bad components.

Rub the imprint off of the failed ones and see if anything is below it. See if they are playing around with sending out remarks yet.

i have a few failed FM1608 if ForceFlow doesnt want to send any out.

Those are questionable in a few places and here is my opinion and why....

The batch / date code is the same on every single one. In my experience getting used not remarked pulls from China, it is uncommon to see matching date codes / batch codes in more than a few in a row. All chips having the same date code can be suspect of remarks when buying used chips.

2nd. I think the font is wrong. I believe the "1" character should have a bar at the bottom and your picture does not. I found one of mine with a similar date code and I will include a picture.

It still could be legit though, hard to really say, maybe they changed the font in year 2005 at some point. I can dig through my new stock and look for some around that date and maybe see.

I have had luck sanding off imprints and finding stuff below. Inspecting the bottom side of the chip can tell too sometimes.

20160110_161705_(2)_(resized).jpg

The bottoms would push me more into thinking those could be remarked since how the bottom looks compared to how nice the top is. That being said, the underside imprint looks correct. You should see those two indents one with a country stamp (TH or S i think) the other with a number, I think I see the 07 stamp in the indents when i zoom in. I see that on my -P versions as well.

Just cuz it is remarked, doesn't mean it will not work.

bottom imprints zoom in from the FM1608 i pictured above. One say "07" probably some sort of manufacturing spot or batch number. The other says "TH" which probably stands for Thailand.

_2_20160110_170053[1]_(resized).jpg

_1_20160110_170053[1]_(resized).jpg

Have you seen some fail in that mode? That is VSS to VDD. I know the NeoLoch ram tester will reboot itself if you try and test a chip that is shorted VSS to VDD.

Thanks for the shout out.

I just restocked on FM1608 and I got a better deal this time so i lowered the price per unit around $1 each. The new batch i have is mostly -PG chips 2005+ dated. Quite clean with full length straight legs. Original imprints. Tested and guaranteed good. Ten pack is now $60 shipped from me.

DIY pcbs coming soon so you can use your FM1608s in 6116 and 5101 applications.

http://nvram.weebly.com

This is asked often. Here is the compatibility list break down for NVRAMs.

Sega, Data East, and Stern White Star use a FM1608 / 6264 adapter. Plug and play ready. Original RAM is in a socket.

WMS system 3-4 use a Dual CE 5101 adapter. Not plug and play, original ram is hard soldered to the board.
WMS 6-7 use a Single or Dual CE 5101 adapter. Not plug and play, original ram hard soldered in.
WMS 9-11A use a 6116 adapter. Not plug and play, original hard soldered in.
WMS 11B-11C come with a 6116 installed from the factory, but you can use a FM1608 / 6264 adapter with the RAM jumper change. 6116 NVRAM also works with no jumper change. Not plug and play. Original hard soldered in.
WMS/Bally WPC use a FM1608 / 6264 adapter. Not plug and play, original ram hard soldered.

Gottlieb System 1 is not compatible (typically)
Gottlieb System 80 uses a Dual CE 5101 adapter. Not plug and play, original ram is hard soldered in.
Gottlieb System 3 not FM1608 compatible.

Bally -17, -35, Stern m100, m200. Uses a single or dual CE 5101 adapter. Plug and play, original ram is in a socket from the factory. (m200 needs 2pcs or special combined adapter).
Bally 6803 uses a 6116 RAM and is plug and play ready.

go for it. feel free to clean up my sloppy way of typing. getting stuff out of my brain and onto paper is tough for me at times =D.

anyone know whats up with the color coding of the 3M ZIF sockets? I suspect some of the green 3m ones could be china knock offs.

I have some of the 3M green ones on my test boards, 40 and 28 pins, that i have worn out after only 1000 cycles. I have a black one on my EPROM burner that has probably seen a few thousand cycles and it is still in good shape. The green ones pins start to have memory at the edges and dont fully open up / close.

Are all the green ones dodgy though? I see black ones on the reputable sites, but only the pea green ones on china marts.

I see Tayda removed the 3M branded green zifs and added back their own brand... which is the same as what they had before just no 3M logo.

http://www.taydaelectronics.com/zif-socket-40-pin-0-6.html

When I am home today i will take pics of my worn out "3M" ZIF sockets.

You get what you pay for at times. A legit 3M textools zif is like 10 bucks. The China ones are a buck. It will probably be a pain in the butt to desolder out the 40pin ZIF sockets on my WMS 6 test board which is my 6802 and 6821 tester. The ZIF sockets have really chunky legs.

Ultimately either socket works fine. If you plan on changing the IC in the socket numerous times (like maybe the ERPOMs) i would use the dual wipe lower one.

The machine pin upper one, specially coming from China, tend to wear out very quickly. So these are better if you don't plan on ever removing the IC (RAMTRON FM1608 would fit that description). NVRAM adapter posts are round pin, the machine pin sockets are round. That would be ideal, but then again, it really doesnt matter.

I use this mpu to test 6802 CPUs. This is a picture of a dodgy 3m socket after around maybe max 2000 cycles. I am starting to get false failures with the socket. I have to reseat a few times to make sure the failure isn't zif socket related. From my memory, this came from Tayda electronics. Maybe ebay. No more than a few dollars each.

My eprom burner has a black colored 3m labeled socket that is going strong well past 2k cycles. I even spilt my coffee into it one day a year or so back and had to clean.

20160124_164922_(resized).jpg

What code version is the stand alone RAM tester with a LCD screen at?

I know mine is buggy at times. For example... if i test a 6264 RAM and then test 5101 RAM, the first 5101 RAM test will always report failed. Run the test again and it shows pass.

Also sometimes i get false fails 6264 on the first write / read back pass. Run the 6264 a second time and it will show pass. Even if you power off the tester mid test, turn it back on, it will continually show the first pass as bad until one full test cycle completes, then it works....

Basically if the RAM tester says fail, run the test one time and see if it passes. If it still fails, make sure it is in the ZIF socket well.

I assume a lot of the false failures people are seeing is because of how the FM1608 is recycled. On the sloppier used pulls it looks like they go pin by pin with a soldering iron flattening out the solder after it is pulled from a board. This pushes the flux residue stuff to the edges where it can sit and insulate. ZIF sockets grip on the edges of pins. Another thing done is sometimes the pins are cut short. The pins are typically not too short for a machine pin socket, but too short to mate fully in the zif socket.

As far as the quality of the ZIF socket in the neoloch tester, it appears good, but probably not a legit 3m socket like it is stamped just because of the cost of the ram tester. A $30 zif would eat up a lot of the $50 price tag. I know i have tested 500+ 5101s, 2000 FM1608s, and countless hand soldered adapters. Socket shows no real signs of fatigue. Pins are still going fully close and fully open. I wouldnt worry about the one in the neoloch, even if knock off, seems fine.

http://www.yamamo10.jp/yamamoto/comp/Z80/memory/reference/RAM.pdf
indicates a 62256 RAM from the data sheet.

The all DIP boards finally came today and i am doing some testing. So far so good... Single and dual CE mode working OK for the 5101 adapter. Single CE = no 7400 and jumper in, Dual CE = 7400 and jumper out. I put the 7400 to the right of the FM1608 because i didn't want the adapter hanging off the side of a GTB s80 board which created a new issue... Turns out it is really tight with the LED on a Bally MPU.... dang. Not really a deal breaker as it does fit, but too close for comfort, so i will change it. You have to use a 74LS00 some other 7400 part that will hold logic high when the unused inputs are floating. I have already fixed that for the next revision by connecting unused inputs to other nearby outputs.

The DIP 6116 adapter works and fits a treat.

Kicking around the idea of making these available to order on dirtypcbs.com. I figure some amount of people will do it themselves, some will just want to buy them assembled, but i also don't want to shoot myself in the foot for giving away stuff for free. I am pretty sure DIY boards are going to be doable, i need to sleep on it though. I would probably just sell the PCB and you stuff the parts yourself. The round pin headers are found on aliexpress or ebay. socket for the FM1608 is not required if your pins where not cut short in the recycle process.

Here is some pics i took testing today.... excuse the roached out bally MPU =D
20160203_190457_(resized).jpg20160203_185633_(resized).jpg20160203_182849_(resized).jpg

If it is an 11B or 11C game there is four extra pads and a jumper to change. 11 and 11a only use 24pin 6116 style ram/nvram.

My source has told me the used pull market for FM1608 is getting tapped out. I won't be able to import any more FM1608 after the batch i have now, so I am glad I stocked up when I did. There isn't much left, what is left is pretty rough so they are getting sanded, black topped, and remarked. You can spot remarks because the font is wrong. It is close, but some characters are not correct. You risk getting a different RAM like FM1808 remarked to fm1608 which may not work in all situations. I made a post a while back about spotting remarks.

If i got FM16W08-PG and it didnt come in taped reel package i would be concerned that they are used pulls. Brand new FM16W08 should read blank too. They are probably recycling the SMT parts too.

Mine are $6 each, clean, -PG type, date codes 2005+, original imprints, straight clean legs, in game tested + neoloch tested, ship in tubes with esd protection.

Untitled_(resized).png

The "real time clock" needs a battery or it will not advance when the power is off. With NVRAM the real time clock does not advance when the game is powered off.

Most games this is a who cares, maybe twilight zone is an exception.

WPC computers notoriously keep terrible time in a standard setup, so you are not missing much with NVRAM.

There is factory in China making knock off FM16W08 now. I bought some Aliexrpess "too good to be true price" FM16W08 and they came RAMTRON branded with date codes that should be Cypress. Most of them did work, at least initially, but i did have some dead right out of the tube. I am not using these in my adapters.

I still cannot get anymore FM1608 dip28 package from a source I trust. I am only going to have SMT to DIP28 adapters going forward once what I have left is gone.

Doesnt look like the specific ali item i bought is listed anymore. I found some similar ones though. These are definitely illegitimate in some way or another. The date code its outside of RAMTRON's range and it is missing the dye revision letter.

https://www.aliexpress.com/item/2Pcs-FM16W08-FM16W08-SG-quality-assurance/32576429339.html?spm=2114.01010208.3.18.4LQ5ql&ws_ab_test=searchweb201556_0,searchweb201602_4_10065_10068_10067_10069_418_10017_10060_10061_10062_10057_10056_10055_10037_10054_301_10033_10059_10032_10058_10073_10070_10052_10053_10050_10051,searchweb201603_4&btsid=54545054-0c9b-4fe0-8282-63699618af7b

https://www.aliexpress.com/item/Free-shipping-25pcs-lot-FM16W08-SG-FM16W08-FM16W-SOP-28-Best-quality/32487991322.html?spm=2114.01010208.3.60.1fgjYL&ws_ab_test=searchweb201556_0,searchweb201602_4_10065_10068_10067_10069_418_10017_10060_10061_10062_10057_10056_10055_10037_10054_301_10033_10059_10032_10058_10073_10070_10052_10053_10050_10051,searchweb201603_4&btsid=a7e90e48-4ba8-40c2-a56b-074ceb63ef71

The font looks wrong too

To make 6264 work in place of 6116, you just need lift the two extra address pins and tie them to ground. The /RE and VCC would need routed too. The rest of the pins match up. You could stack sockets the legs floating out and come up with something that worked.

It is possible to have a ram chip that is only bad in one small section. When I read back these NVRAMs that have been used in a game, a good portion of the RAM is still blank or has static data that never changes. As long as the bad section never gets accessed, it may work fine in game.

It is also possible your RAM tester didnt make good connection or it failed it on the first pass due to a glitch of some kind. I use an older neoloch RAM tester and sometimes it reports 6264 NVRAMs as bad. if you test them a 2nd time they read good.

Of my FM1608 used pulls, I have had less than 1% bad out of the tubes. The FM1608 seems to be quite robust and not that static sensitive.

I think i see problems. Highlighted them in red.

damaged track 1 (resized).png
damage track 2 (resized).png

Hakko makes nice flush cutters and they are quite inexpensive. I have two pair I use, one i am more rough on and the other pair i try and be gentle with so the blades stay nice. Even the one i cut things thicker than i should be with them are holding up great with daily use. I have had other recommended brands break on me really quick.

https://www.amazon.com/Hakko-CHP-Micro-Cutter-Clean/dp/B0063GL7UM/ref=sr_1_2

https://www.amazon.com/Hakko-CHP-170-Stand-off-Construction-21-Degree/dp/B017ODDPNO/ref=sr_1_1

get the five pack! hehe

Define cheap =D. Some of the FM16W08 to 6116 adapters can be had for $11 each in 10 packs from different vendors. But no... RAMTRON never made a 2k x 8bit 6116 footprint FeRAM device.

Dallas made some 6116 NVRAMs that use an internal lithium battery, DS1220 You get about 10 years out of one. I wouldn't recommend using the dallas parts either. They are probably more expensive than RAMTRON adapters.

You could do it by stacking IC sockets on top of each other. It might get kind of tedious rerouting the pins manually, but def do able. I made adapters by stacking IC sockets for non jedec pinout ROMs i wanted to dump.

I am still using the older ram only NeoLoch tester. I think it says v1.02 when it starts up.

It gives false failures on the first test of 6264 sometimes. 5114 nvrams the 2nd half of tests shows failed. 5101 rams always show address failure, but the tests are OK. 6116 nvrams always test fine.

My gq4x works fine with windows 10. I had to manually install a special driver if I remember right. Only thing is it gets out of sync when the laptop in the work shop goes into sleep mode. So I keep it unplugged until I am ready to use it.

I want to make socket adapters so I can read / write to all the nvrams with the gq4x.

It is a Dallas / Maxim SRAM with a battery monitor and lithium cell all in one potted package. DS1225 is the 8K x 8 one that would go into a game like T3. They are essentially the circuit Gottlieb system 3 used but all packed inside the dip28 part. I think they still make them, but probably more or as expensive as NVRAM, then they die after about 10-15 years.

https://www.maximintegrated.com/en/products/digital/memory-products/DS1225AB.html

For sure a remark with the 2019 date code but if it works and only $2 that is a steal. With a remark they could be the older RAM chips with the brown out issue where there is less compatibility tho.

Has anyone else noticed the 2005 and earlier chips act up in some systems? In WMS 3-7 and bally -35 if you power cycle the game twice quickly before the 5/12v goes all the way flat the next boot up will get a ram failure and possible corruption.

ones that look like this seem the best. The case and imprint is always 2006+ from what I have seen. These always work in bally / wms 3-7 games.
pasted_image (resized).png

ones that look like this are older and get the failure to boot if you power cycle twice in some systems before the 5/12v collapses all the way in wms 3-7 and bally -35 type mpu boards. They still seem fine in WPC, WhiteStar and DE/Sega boards.
pasted_image (resized).png

There is a third type made in china from early 2000s that looks even different that I can't find a picture of easily. They have the same symptom that the above one does.

I assume it is related to this note in the data sheet but I have had the power cycle twice quickly ram failure with up to a 2005 date code.
Untitled (resized).png

I have not tested system 11, but you may want to try power cycling twice quickly and see if you get a RAM failure with an old date code chip. WMS 3-7 is effected and sys 11 uses the same PS as sys 7 in a lot of games. Stern White Star works fine with the old date code chips and power cycling quickly.

I assume a brown out is when a write or maybe read happens when the Vcc is below the working voltage.

Also noticed that when I use a Bally SDB with a switcher power supply the old date code chips work fine if you power cycle the game quickly. Its only the linear power supplier that get a ram failure and possible corruption when you power cycle twice quickly.

Here is the 3rd FM1608 chip package flavor that also seems to have the power cycle quick issue. I think these are the oldest date codes in this package. Year 2000.
pasted_image (resized).png

Your pictures shows you have all the newer ones which is better since they don't seem to act up in some systems if you power cycle twice quickly. Best I can tell only seems like old bally and WMS games are effected. Most people are probably not going to be trying to use a dip28 nvram in those systems anyways.

I see an FM1608M in there. Anyone know what that M suffix means as it does not seem to show up in the data sheet? It also has a different dye revision letter in front of the date code. Also have one FM1608B in there which are faster but shouldnt matter in any pinball application.

When I read the data off of these used pulls most indicate in readable text that they came out of a Ricoh Aficio copy machine NVRAM modules. Like this thing I found searching in google.

pasted_image (resized).png

FM16W08 surface mount based NVRAM modules have 100 trillion R/W endurance where the FM1608 dip28 only have 1 trillion endurance too. You can measure the /CE frequency and get a good idea how long they should last but on a used pull you do not know how much endurance is left in the chip. I am guessing the endurance figure is a conservative number and they would not get much of workout in those Ricoh NVRAM modules but you just don't know.

datasheets say the letter before date code is the dye revision. The FM1608M says revsion A where the FM1608 all usually have revision M.

Untitled (resized).png

Measured 31000 Hz in a bally game attract mode. Assuming worst case where every access is in the same row (which more than likely is not) FM1608 endurance would be about a year straight powered on 24/7. FM16W08 would be 100 years.