Sega Dreamcast System RAM 16MB+ & JTAG hack

hello here I leave some codes of memory chip 86pin I have seen in several Dreamcast.

NEC D4564323G5 de 9ns
Hyunday HY57V653220B TC-10
Samsung K4S643232C-TC80

Anyone know if these memory chip of 86 pins also existed but twice the capacity of the Dreamcast is 16 megabytes each that would be amazing to look in some applications such as graphics cards or RAM modules


I found this on the NAOMI 2 years after image search system board'll inspect the type of memory you use and see if it is possible O_O

As you can see the Hitachi SH-4 CPU supports the 32 megs quietly but can also be seen that there are 4 chips each of which is 8 Mb words are identical to the difference between the Dreamcast and NAOMI DC is the NAOMI PCB comes in ready to host the 4 chip and the DC no reason I see very difficult to add another 2 modules takeoff unless the CPU and manufactures bridge and the bus for those two extra chip. Still another possibility is that these 86pin chip they come out in versions 16 megs of being so serious issues to find an application which used as graphics cards or RAM is a posibildad truth that can not be ruled out but it's fucking look hard chip may not exist maybe if god only knows ...

http://www.system16.com/hardware.php?id=721

HI RES BOARD NAOMI






PD: Sorry for my ingrish xD TANKS Google Traslate

 

Hello and welcome to the forums.

Farnell has 86TSOPII in 32MB each or higher for a total of 64MB installed RAM, but then to use them you need 9 bit address which means even more pins need to be remapped because of the address multiplexing.

2x 16MB each seems to be the limit for 8 bit addresses for total of 32MB and this requires the least effort, but it still means remapping pins 21-23 and drilling into the back of SH-4 W6.

I haven't been able to take my project further yet. I bought an Atten 858D+ SMD rework station, but I still need a new soldering iron and spare parts, but at the moment I do not have the money to buy them as I have some expensive items in the mail to pay for

I didn't forget about this project, it is still high on my to do list, but as I only have a small budget it will take me some time to complete it.

 

About Farnell chips what you mention can not give me the code od these for google it and try to find some applications where were used like graphics cards or RAM , would help me a lot to know the codes, I did not know existed for more capacity.

 

The specific chip model code doesn't matter too much. Any 86TSOPII SDRAM will solder on, you have to tailor it to the SH-4 pins / address multiplexing depending on the size, bus width and address bits regardless.

All that matters is you have 2 x (128 x 32) x 8b 86TSOPII SDRAM for a total of 32MB SDRAM on a 64 bit bus with 8 bit adresses as the pinout *should* be generic. If I am right remapping pins 21-23 should work, but again, I do not have the money to complete the experiment any time soon.

I need to work hard to pay off my debts at present. I'll be poor until the new year. In the meantime other people are more than welcome to test and improve upon the theory.

 

Now that would be a really cool thing. Though I always thought of the lag as "part" of the game, and used it to my advantage, would be fun to see how far my score would go down without the lag. Personally I always thought that was more of a processor issue, much like an overclock mod to a genesis/megadrive can make it so when Sonic looses 100+ rings from getting hit, that it does not slow down. But looking at purely the CPU speed of both the NAOMI and the DC just now it seems they were both at 200mhz. Someone correct me if I am wrong.

 

Can't you get ram chips with same pinout?
It would be something i would like to do.

 

Pinout isnt the issue, its lack of address lines available.

 

Address lines isn t a problem either, its the traces.

You can fit whatever size you want upto 4GiB using either 32 bit or 64 bit bus and between 8 and 12 address bits but if you deviate too much, you basically have to remap everything into a great big wiry mess which probably wouldn't work for anything else.

The dreamcast has 64MB mapped per area, the standard bus width is 64 bit and the address is 8 bits, under these circumstances I think 32MB (2x 16MB) is max and it needs minimap remapping of pins to the address lines to work in theory. Keeping it simple FTW

 

Available address lines is an issue, otherwise you wouldnt be drilling into a chip.

The address lines are not available on the board, which is why all the hacking is required.

 

The address lines are available if you follow the traces.

The only reason it needs to be drilled is because to match the address multiplexing for the particular format and size, there is no trace but we still need acess.

There are forms which would go onto the SH-$ with little to no drilling because the lines are available, but thyey would mean using a different bus width or size which is less likely to be compatible.

 

So.... the required address line isnt available on the board (as you need to drill to get to it)... which is an address line issue, not a pinout issue - like i said?

Point being, bart was implying its just a case of matching the pin out. To double the ram, you need an extra address line on the ram chip. Its not about the pinout, but getting that address line to the chip.

 

It is certianly not unavailable. The address line available whether you drill or not. It is just difficult to access.

But yes, it isn't just a case of matching pinout. Matching the pinout , bus width and adress bits you will end up replacing the 16MB RAM with another 16MB. If you alter the size you need to match the approapriate adress line multiplex for the other addresses to be made available.

 

Being under a BGA chip, with no broken out trace/pad is inaccessible/unavailable by most peoples definition. Obviously I wasnt saying it didnt exist - else you wouldnt be attempting this hack, just that its not available at the ram chip pads nor any other trace/test point.

Which is why I said its an address line issue, not a pinout issue. Which is correct, I dont understand why you decided to jump on my reply to bart.

 

Then again you are missing the point as per usual.

99% of the problems you are invovled in are to do with your use of words.

 

Funny how the issue only arises with you, id look into that.

Also, its kinda funny you are claiming I am the one difficult to understand, given your typical post.

100% of the problems I am "involved in" are actually to do with you - no one else.

 

I don't dispute the fact people may struggle to understand my posts, I never did.

Its funny how you forget the issue arises with other people and again it is because of your use of words lol.

Lets not forget you are adding more detail each time you retread the same ground. If you have covered all of the bases from the outset...

But anyhow, enough of that.

Not every address pin one needs is unavailable. When you can go in through a trace, test point, or even through the front the address lines are not unavailable. Drilling through the back is only for ease of access.

As I said already, there are address multiplexes for memory sizes / formats which require little to no modification. There are some which require heavy modification and the one with the least modification is probably the most compatible.

 

So.... drilling a pcb? hard? are schematics availeble or you sure you dont drill a ground plate and/or destoy another important line?
:witless:

 

There are pics in the OP and more information can be found in the SH-4 documentation and the leaked Dreamcast schematics. The what where when why and how are covered in the remainder of the thread

Now that I have an SMD rework station I may try my luck sliding a fluxed wire through the front to W6 whilst heating up the SH-4 and hoping it takes to it. This is something I can easily try without connecting up the pins to anything functional and without having to spend any more money, but I am much too busy to do it this weeekend plus without the other stuff I would have no way to test the connection any time soon.

 

I pressume you'll pre-tin the wire before anything else? if so there shouldn't be a problem.

surely theres an avaliable trace you can tap into without having to drectly hook up to the SH4?... if not, it doesn't exactly inspire me with confidence for my plans with holly. Just out of interest how is your particuar brand of SMD rework station, performance wise? i was thinking of going with a Saike, but if your having a good experience with yours i may reconsidor.

 

If there was a trace I wouldn't have this problem There are only traces for the pins the Dreamcast uses by default, but of course the SH-4 comes equipped with all of the other pins and they are balled too, just some are grounded, the others are left floating.

I have an Atten 858D+. It is by no means amazing b ut it seems to get the job done and it was cheap. It seems to get some good reviews anyway.