RAM 320XE

[ctirad]

I ordered two packets of 20pcs for now. The cost was the same as if I would order 100 and the small packages easily pass the customs vithout additional VAT and other fees.

[+wood_jl]

I would like 2 as well! Looking forward to this like I did RAM 320XL. Excellent!

[Roydea6]

If this thread is becoming an order thread. I would like to order (1) 130XE.

 

Thank You for doing this for the Atari Community..

[BillC]

Ctirad already posted about 2 weeks ago that he will start a separate order thread when he has some devices ready.

http://atariage.com/forums/topic/178885-ram-320xe/?view=findpost&p=2827142

[Caterpiggle]

Ctirad,

 

Is possible for you to increased from 320k to 1 mb on the same board card ?

[ctirad]

No, the maximum is 320kB plug&play or 576k with a mod inside atari.

But it is funny, you asking right now, because the new XL version, on which I'm currently working, will have either half meg or 1 MB fitted. Of course you will allways have to mod your atari to make anything over 320kB possible.

[Caterpiggle]

Hello ctirad,

 

Can you be kind to tell me how much cost to buy 1 ? In euro of course.

[Marius]

I still disagree about that 320K limit. Could you please post an example of software (a test program is also ok) that proves that more than 320K without mod is not possible.

 

I have no problems at all with my 512K in my Turbo Freezer. No issue at all. Not with selftest, not with any ramdisk program, not with any demo. It simply works. Unmodified atari.

[+Larry]

Just wondering if Ctirad is referring to the PORTB issue such that Basic must be disabled (or enabled with a hardware switch)? AFAIK, you can do 512K, but not 1 meg without a mod. BTW, I'm far from adamant about that! :-)

 

I wonder if Hias could offer some info as to how the Freezer does 512K -- maybe some bank-switching within the freezer, itself?

 

-Larry

[Marius]

Well Ctirad repeats every time that 320K is max, while TurboFreezer offers a 512K XE compatible RAM expansion (576K in total) and I did not experience any incompatibility at all so far.

 

But… since I might not be in the circumstance to get the issue, because I do never create that circumstance, my question now is: please provide a program, or describe a situation, where one would get the incompatibility he is talking about. Then I could conform (or not) that the Freezer indeed does not solve that issue.

[HiassofT]

Just wondering if Ctirad is referring to the PORTB issue such that Basic must be disabled (or enabled with a hardware switch)? AFAIK, you can do 512K, but not 1 meg without a mod. BTW, I'm far from adamant about that! :-)

 

I wonder if Hias could offer some info as to how the Freezer does 512K -- maybe some bank-switching within the freezer, itself?

I don't know how the RAM 320XL/XE is actually implemented, so I'm speculating a bit. Maybe ctirad can give some definitive answers.

 

My guess is that the RAM 320 uses the EXTSEL line to disable internal RAM at $4000 and map in it's RAM banks. Doing it this way you can only use PORTB bits 2-6. Bits 0 and 1 are out of question to use (same for the Freezer) because they'd also map BASIC and the OS in/out - certainly not what you want to have. You also can't use bit 7 because that would map in 2k selftest ROM at $5000 and you can't disable it using EXTSEL (EXTSEL can only disable RAM, not ROM). So I _think_ this is why the RAM 320 only supports 256k of extended RAM.

 

The Freezer uses a different approach. It doesn't use EXTSEL but disables internal Atari memory by using the refresh (REF) line, which works both for RAM and ROM. So we can also use PORTB bit 7 and have up to 512k of extended RAM.

 

so long,

 

Hias

[Marius]

Thanks for the explanation, that's what I needed to read.

 

So the Freezer proves that it is possible to have more external XE-compatible ram without modding an Atari, but that's because EXTSEL line isn't used. Aha... well that's interesting.

 

Thanks again!

M.

[ctirad]

My guess is that the RAM 320 uses the EXTSEL line to disable internal RAM at $4000 and map in it's RAM banks. Doing it this way you can only use PORTB bits 2-6. Bits 0 and 1 are out of question to use (same for the Freezer) because they'd also map BASIC and the OS in/out - certainly not what you want to have. You also can't use bit 7 because that would map in 2k selftest ROM at $5000 and you can't disable it using EXTSEL (EXTSEL can only disable RAM, not ROM).

Exactly. But am afraid it is even more complicated, because there are some sings that the selftest ROM is not mapped, but copied to the $5000 area and thus the PB7 can be reused safely without any additional mod. The selfetest would be probably not correctly handled in some cases, but who cares.

 

In fact, the RAM320XE in 576kB mode seems to work raliably without any mod. But there are so few programs which actually use the 576kB and the memory tests showed me 100% working memory even in times there were still a lot of bugs in my MMU implementation, so it is not a bullet proof for now. This is also the answer for the ProWizard.

 

The Freezer uses a different approach. It doesn't use EXTSEL but disables internal Atari memory by using the refresh (REF) line, which works both for RAM and ROM.

This is somewhat clever solution, but /REF is ofically an output signal. Are you sure the ANTIC /REF output is open collector? There is no pullup on my schematics so I would be careful.

Edited September 27, 2014 by ctirad

[HiassofT]

But am afraid it is even more complicated, because there are some sings that the selftest ROM is not mapped, but copied to the $5000 area and thus the PB7 can be reused safely without any additional mod.

Could you please elaborate on this topic?

 

If you look at the Atari MMU logic you'll see that the selftest ROM is just remapped from $D000 to $5000 if PB7 is low. That's it.

 

The selfetest would be probably not correctly handled in some cases, but who cares.

Problems in this area are actually quite easy to find. If you have an error in your logic controlling the presence of the selftest ROM your Atari won't power up. Upon boot the OS does a ROM checksum check, including the selftest area, and if the checksums aren't correct it jumps to the selftest. Then, if you don't have the selftest ROM present, the Atari just locks up.

 

This is somewhat clever solution, but /REF is ofically an output signal. Are you sure the ANTIC /REF output is open collector? There is no pullup on my schematics so I would be careful.

REF is not an open collector output, but it's still safe to pull it low using another open collector output (it would not be safe to pull REF high, though).

 

Actually, that's also how the chip works internally. In the NMOS manufacturing process you don't have push-pull drivers like on CMOS devices, but a transistor on the low side coupled with a rather weak pull up. This is also the reason why you don't see any sharp rising edges on your scope when probing NMOS outputs.

 

This trick was first used in the original Turbo Freezer by Bernhard Engl back in the late 1980s. Bernhard works as a professional chip designer, so I'd say he knew what he was doing

 

so long,

 

Hias

[ctirad]

Could you please elaborate on this topic?

I don't remember where I read it. It was most probably misleading info, as the schematics are clear.

 

Problems in this area are actually quite easy to find. If you have an error in your logic controlling the presence of the selftest ROM your Atari won't power up.

At power up the PORTB is filled by default values and thus there is not any problem until some application would't to touch PB7 for some other season than run selftest.

 

Actually, that's also how the chip works internally. In the NMOS manufacturing process you don't have push-pull drivers like on CMOS devices, but a transistor on the low side coupled with a rather weak pull up.

I see. As long as the all existing ANTICs are made by the same technology, it is safe. But it's still a nasty hack.

[+CharlieChaplin]

In fact, the RAM320XE in 576kB mode seems to work raliably without any mod. But there are so few programs which actually use the 576kB and the memory tests showed me 100% working memory even in times there were still a lot of bugs in my MMU implementation, so it is not a bullet proof for now. This is also the answer for the ProWizard.

 

Well,

 

it depends on what you are using the 576k RAM for. If you search for games, you will not find many programs (a few that utilize 1MB RAM like Atari/GTIA Blast, etc.), but if you search for tools and demos you will find a lot of programs that can utilize 576k RAM.

 

Next thing, how do you "setup" the 576k RAM, is it a) Rambo mode only with blocks 8ACE (each with 8 banks of 16k RAM) or b) Compyshop mode only with blocks 26AE (each with 8 banks of 16k RAM) or c) combined Rambo+Compyshop mode with blocks 02468ACE (each with four banks of 16k RAM). I personally prefer option c), because with this setup every program that requires either Rambo or Compyshop mode will work, most older DOS versions that support up to 320k RAM (like e.g. Super-DOS, Turbo-DOS, Bibo-DOS, etc.) will also work fine.

 

Looks like some A8 software is limited to a max. of four banks per block (e.g. Turbo-DOS, Bibo-DOS, Super-DOS and others) and thus does not work with U1MB in 576k or 1MB mode (where 8 banks are used per block). It is easy to guess why, because the original 130XE had only one block (block E) with four banks, 128k XRAM offered two blocks (blocks AE) again with four banks per block and 256k XRAM (Rambo, Compyshop and others) offered four blocks (26AE or 8ACE) again with four banks per block - thats why some software does not care (or does not allow) for more than four banks per block. Because of that my U1MB is almost always in 320k mode, I only switch to 576k or 1MB mode when required.

 

I also have two 800XL equipped with 576k SRAM enhancement by mega-hz. I did not install a switch, thus 576k RAM is always available. It supports the above mode c) with Rambo and Compyshop blocks available at the same time (and a maximum of four rambanks per block). All A8 software seems to work fine with it and I must say that I prefer that SRAM enhancement over the U1MB because of its high software compatibility. Only programs that require 1MB RAM or separate Antic access do not work with this SRAM enhancement, but thats it so far. (And I have tested many thousand A8 programs in over ten years with that enhancement; I even had 512k XRAM before, since a good guy extended my 256k Compyshop compatible XRAM to 512k XRAM: http://www.b-pahl.de/atari8bit/RAM-Disk/doku/Upgrade_1/upgrade_1.html the enhancement you see in the picture was mine!)

 

What setup will your 576k mode for RAM-XL or RAM-XE use - mode a) Rambo only (4 blocks x 8 banks), mode b) Compyshop only (4 blocks x 8 banks) or c) combined Rambo+Compyshop (8 blocks x 4 banks) or d) a completely different mode ?!?

Edited September 27, 2014 by CharlieChaplin

[Mathy]

Hello ctirad

 

But there are so few programs which actually use the 576kB ....

 

That's what every hardware developer says.

 

Every software developer says there aren't many upgrades over 320kB.

 

Is this a vicious circle of a classic case of the chicken and the egg?

 

Sincerely

 

Mathy

[+CharlieChaplin]

Well,

 

576k XRAM is good if setup correctly, because 1) you do not have to switch between Compyshop and Rambo mode anymore (both can be available at the same time, software will simply use the mode it requires), 2) there are demos and tools that can utilize 576k and even more RAM...

 

Thus 576k is an advantage, even when only 320k is used by software (no need to switch between CS/Rambo modes).

[ctirad]

c) combined Rambo+Compyshop mode with blocks 02468ACE (each with four banks of 16k RAM).

This one.

 

I also have two 800XL equipped with 576k SRAM enhancement by mega-hz. I did not install a switch, thus 576k RAM is always available. It supports the above mode c) with Rambo and Compyshop blocks available at the same time (and a maximum of four rambanks per block). All A8 software seems to work fine with it and I must say that I prefer that SRAM enhancement over the U1MB because of its high software compatibility.

I can say the same for the basic RAM 320XL/XE mode. Even it is not either RAMBO or COMPY SHOP, the compatibility is almost 100% with any demos, games or utils I tried.

[ctirad]

That's what every hardware developer says.

Every software developer says there aren't many upgrades over 320kB.

Never heard that before. Actually, I saw the 320kB only in 130XE machines, because it was easy to made just by adding RAM chips and three 74xx gate chips. The most common expansion around me was 576k.

[ctirad]

there are demos and tools that can utilize 576k and even more RAM.

I'm a lot into demos and I know only one, maybe two of them. And no such game at all.

[+CharlieChaplin]

Well,

 

regarding Games:

 

- Atari/GTIA Blast topic: http://atariage.com/forums/topic/220575-gtia-blast-update/page-3

post Nr. 55 and 75 contain the packed 1MB version of Atari Blast

 

- Space Harrier topic: http://atariage.com/forums/topic/184410-space-harrier-released/?hl=%20space%20%20harrier

if someone would convert this from cart (BIN/CAR/ROM) into an executable file (COM/XEX), it would require 1 MB XRAM...

 

- there are some A8 games that do support up to 320k RAM (e.g. Kaiser II, The Brundles and some others), but as of yet I do not know any A8 game that supports or requires 576k RAM

 

 

regarding demos:

 

attached you will find most (all?) of my 1 MB demos, some tip-animator demos were created by me

 

 

regarding tools:

 

attached you will find some tools that support up to 1MB, most of them require a proper DOS however (MyDOS or SpartaDOS)

 

regarding 576k and 320k demos and animations, well I have dozens of them, but they are sorted from A-Z and not in the 320k or 576k directory right now (this is some work I still have to do)...

Demos 1MB.zip

Tools 1MB.zip

Edited September 28, 2014 by CharlieChaplin

[576XE]

Ctirad,

Just whistle and I'll bye your creation!

Please, if possible, by PayPal.

130XE

Awaiting your PM.

[+Larry]

Being a practical guy (I think), I'm sure that Ctirad will sell plenty at 320K. At 576K, he could sell maybe a few more. If the price is close, then sure, lots of folks will opt for the 576K. The one and only program that I use that requires (even) 320K is the terrific Mean 18 golf game ported from the PC/64.

 

@ Mathy -- did you buy a Freezer with it's 512K?

 

-Larry

[ctirad]

post Nr. 55 and 75 contain the packed 1MB version of Atari Blast

So we have one unfinished game that requires 1MB just because there is no loader routine present btween the levels yet.

 

 

attached you will find most (all?) of my 1 MB demos, some tip-animator demos were created by me

Only the Prototype is the real demo (a subaverage one). All other is just a lot of precalculated crap which can actually fill any amout of memory.

 

BTW, the Prototype exists only because Poison wanted something that would utilize his new U1MB. Fandal (author of the code) said that with a little effort it could squeeze it maybe to 256kB, but it don't worth it. I fully agree with him.