Memory upgrades, hard drives and device DMA (St/e alike)

[carmel_andrews]

Following on from the 1gb a8 upgrade thread on the hardware forum, i was just wondering if this size memory upgrade was either possible or appropriate for the A8, if not, what would be the max size of memory upgrade

 

Also, given that some people are thinking about moving away from the traditional D301 (xl/e) and d800 (4/800) bank switch, would future upgrades be able to go beyond the 16k bankswitxh limit

 

And given the increased memory size that a 'potential' 1gb upgrade would give, would that command better storage capabilities from HD's/DVD's and removable memory (i.e sd cards or mem sticks), got me wondering...does the a8 do 'device dma' like the st/e etc machines do for high spead data xfer

 

And also a dos that removes the 16/48 meg partition stroage limit as well as number of devices limit (8 or 9 last time i recall)

[MEtalGuy66]

Following on from the 1gb a8 upgrade thread on the hardware forum, i was just wondering if this size memory upgrade was either possible or appropriate for the A8, if not, what would be the max size of memory upgrade

 

Also, given that some people are thinking about moving away from the traditional D301 (xl/e) and d800 (4/800) bank switch, would future upgrades be able to go beyond the 16k bankswitxh limit

 

And given the increased memory size that a 'potential' 1gb upgrade would give, would that command better storage capabilities from HD's/DVD's and removable memory (i.e sd cards or mem sticks), got me wondering...does the a8 do 'device dma' like the st/e etc machines do for high spead data xfer

 

And also a dos that removes the 16/48 meg partition stroage limit as well as number of devices limit (8 or 9 last time i recall)

 

Get a 16bit machine, carmel..

[_The Doctor__]

SpartaDosX 4.42 does 512byte blocks on storage and 15 drives. Works pretty well too! That shattered the rules we all came to accept. In fact borrowing from that concept some have upgraded firmware on I/O devices to address the larger blocks as well. So the max became much more than we hoped.

 

Those two changes address your storage concerns greatly. As for memory... There are a great deal of upgrades out there. But basically 2 root standards, and one sub-variant. There was talk of a third upgrade type/method but I fear it is just that.

 

It is conceivable that the same kind of thing could happen, Someone could conceivably build on the memory concepts proven on a modified devices and possibly apply it to the 8 bit directly.

 

I hope that helps you in some small way.

[MEtalGuy66]

The possibilities for memory upgrades are infinite.. On any platform, as long as your talking about bank switching.. (which in the case of the ATARI extended ram, we are.) As far as the bank size goes, 16k is pretty big for a machine whose adress space is only 64k.. Thats 1/4 the memory map.. You could go bigger, but why? As far as the total size of a memory expansion.. heh.. The sky's the limit..

 

For instance, one additional 8-bit register of bank switching, on a stock 130XE gives you 256 times as much extended ram.. (16megs)...

two additional 8-bit registers of bank-switching would give you 65536 times as much extended ram.. (4gigs)...

three additional 8-bit registers would give you 16.8 million times more extended ram.. Which comes to just about 1 terrabyte..

 

Do I need to go further?

 

So you see, the issue is not how much can we use.. But how much physical ram can we actually wire into an atari, and what standard do we use as far as bank-switch registers....

[danwinslow]

Carmel, I think that thread about 1GB upgrade was a joke.

[MEtalGuy66]

Carmel, I think that thread about 1GB upgrade was a joke.

It really could be done, Dan... Think about this.. 72pin SIMMs go up to 128megs in size, and still support fast-page mode.. You could do a gig with just a few more bits of bank select.. With the availability of cheap programmable logic, its not at all out of the question..

 

coming up with a useful application for it might be quite a challenge..

Edited February 5, 2009 by MEtalGuy66

[danwinslow]

Well, I suppose it could be done, but I still think the original thread was a joke.../shrug lol maybe it wasn't.

Edited February 5, 2009 by danwinslow

[spookt]

Apologies if this is a stupid comment but wouldn't it take a long time to load 1Gb of data?

[MEtalGuy66]

Apologies if this is a stupid comment but wouldn't it take a long time to load 1Gb of data?

 

Thats not a stupid comment at all.. heheh. It takes the atari like 10 minutes to load 16megs from hard-disk to memory.. Imagine how long it would take to load a gig...

[ClausB]

As far as the bank size goes, 16k is pretty big for a machine whose adress space is only 64k.. Thats 1/4 the memory map.. You could go bigger, but why?

My original 256K XL upgrade used 8 32K banks but I quickly changed it to 16 16K banks after the 130XE came out. Now, in retrospect, I wonder if 32K might have been better. Each 32K bank had its own page 0 and stack page, so task switching would have been easier. The upper, non-banked, 16K could have held a task switching kernal and DOS and maybe common screen RAM. You could have had up to 8 applications running in their own 32K spaces taking time slices of the CPU. Something like that for the Atari in 1985 could have been a big deal.

[danwinslow]

Yes, I have a tasking routine that I use that has to copy the stack page each context switch...having a banked stack would be killer for that kind of thing.

[Rybags]

Task switching would be cool... but the nature of how most hardware registers work (ie read or write only) makes things somewhat harder.

 

But if you have programs which still allow the OS to do it's shadowing, it eliminates much of that problem.

For a short time, I worked on a task-switcher for the 130XE... I devised some methods to at least detect if DLIs are being requested. Detecting if they're enabled would need a custom OS.

 

As for devices doing their own independant DMA. IIRC, it's not possible via the PBI and would need modifying the machine itself.

It would also require the device to be able to know what cycle within a scanline we are on.

The other barrier is that there aren't very many cycles at all within a scanline that are guaranteed never to have DMA happening from Antic, once you take all the possibilities of DMA-Width and HSCROL into account.

[danwinslow]

My tasking routines are strictly self contained, so if there are any DLI's I already know about them

 

Doing a general case tasking supervisor for external programs would be very hard I would think.

[+bob1200xl]

Actually, I think a 130XE has the HALT line on the PBI. With that, I think you could do DMA.

 

but, don't tell Ken - he'll just kill some more 1200XLs...

 

Bob

 

 

 

 

Task switching would be cool... but the nature of how most hardware registers work (ie read or write only) makes things somewhat harder.

 

But if you have programs which still allow the OS to do it's shadowing, it eliminates much of that problem.

For a short time, I worked on a task-switcher for the 130XE... I devised some methods to at least detect if DLIs are being requested. Detecting if they're enabled would need a custom OS.

 

As for devices doing their own independant DMA. IIRC, it's not possible via the PBI and would need modifying the machine itself.

It would also require the device to be able to know what cycle within a scanline we are on.

The other barrier is that there aren't very many cycles at all within a scanline that are guaranteed never to have DMA happening from Antic, once you take all the possibilities of DMA-Width and HSCROL into account.

[Rybags]

PMSL.

 

But does the Halt output occur early enough such that a) a secondary DMA source knows not to touch the bus in time, and b) in the event it can do its own DMA, there's still enough time for it to do so without affecting system stability.

[+bob1200xl]

Hey, Claus!

 

How goes things?

 

Bob

 

 

 

As far as the bank size goes, 16k is pretty big for a machine whose adress space is only 64k.. Thats 1/4 the memory map.. You could go bigger, but why?

My original 256K XL upgrade used 8 32K banks but I quickly changed it to 16 16K banks after the 130XE came out. Now, in retrospect, I wonder if 32K might have been better. Each 32K bank had its own page 0 and stack page, so task switching would have been easier. The upper, non-banked, 16K could have held a task switching kernal and DOS and maybe common screen RAM. You could have had up to 8 applications running in their own 32K spaces taking time slices of the CPU. Something like that for the Atari in 1985 could have been a big deal.

[+bob1200xl]

HALT occurs in the middle of the cycle before DMA takes place. Plenty of time...

 

Looks like HALT is not open-collector, though. You would have to split the buss externally. Wouldn't need HALT if you did that.

 

Bob

 

 

 

PMSL.

 

But does the Halt output occur early enough such that a) a secondary DMA source knows not to touch the bus in time, and b) in the event it can do its own DMA, there's still enough time for it to do so without affecting system stability.

[mos6507]

So you see, the issue is not how much can we use.. But how much physical ram can we actually wire into an atari, and what standard do we use as far as bank-switch registers....

 

IMHO, the issue is not how much memory can be wired in, but whether any software other than RAMdisks will ever be written to utilize it. History has shown that none ever will be.

[drac030]

IMHO, the issue is not how much memory can be wired in, but whether any software other than RAMdisks will ever be written to utilize it. History has shown that none ever will be.

 

Do you mean that there is no software other than ramdisks, that uses the memory extensions? That's not true.

[MEtalGuy66]

So you see, the issue is not how much can we use.. But how much physical ram can we actually wire into an atari, and what standard do we use as far as bank-switch registers....

 

IMHO, the issue is not how much memory can be wired in, but whether any software other than RAMdisks will ever be written to utilize it. History has shown that none ever will be.

 

Well, in the interest of keeping additional DRAM control logic to a minimum, I'd say that 72pin SIMMS are about the largest denominator for easy use.. Thats all I was referring to.. Certainly, if your willing to implement the schemes used on newer, faster devices, the sky is the limit.. What is the largest DIMM module they make now? 8gigs?

[MEtalGuy66]

Actually, I think a 130XE has the HALT line on the PBI. With that, I think you could do DMA.

 

but, don't tell Ken - he'll just kill some more 1200XLs...

 

Bob

 

 

To be honest, Bob, the reason I initially retrofitted a 130xe motherboard into a 1200XL was because someone ASKED me to do it to their machine.. This is after I refused to do your PBI mod to it, because I could not think of an easy way to do it without producing something like what beetle has in his "1200XLD"... This guy was really concerned with "nice work" and chose me based on some work I had done for someone else that looked really nice inside. So since I couldnt do it "nicely," I came up with the idea of putting in the 130xe motherboard.. At the time I had several that I had already completley stripped and "rebuilt" to very nice standards..

 

Anywayze.. I ended up liking the machine so much, that I was reluctant to let him have it back after it was all done... So I built myself one... Put pics on atariage, etc. etc... It wasnt long before someone made an offer on that one that I'd have been a fool to turn down, so I ended up building two more... One for myself, and one just in case I get another offer like that..

 

And that't really the truth of the matter.. I did not actually set out to kill 1200XLs.. But that said, the 1200XL with the 130xe board in it is by far my machine of preference.. It works absolutley the best for the most applications, and of course, the keyboard is beyond comparisson to any of the other XL/XE line..

[kurtm]

IMHO, the issue is not how much memory can be wired in, but whether any software other than RAMdisks will ever be written to utilize it. History has shown that none ever will be.

 

Do you mean that there is no software other than ramdisks, that uses the memory extensions? That's not true.

 

You are mistaken drac030. The Last Word, MAE, various games and other utilities do not exist. You're just imagining them.

 

[drac030]

You are mistaken drac030. The Last Word, MAE, various games and other utilities do not exist. You're just imagining them.

 

Damn. I always suspected something like this.

[ClausB]

How goes things?

OK, thanks. Too much going on though. No progress on the video thing, sad to say.

 

How are things in the sunny West?

 

You could have had up to 8 applications running in their own 32K spaces taking time slices of the CPU.

Oops. Only up to 7 apps.

[mos6507]

Do you mean that there is no software other than ramdisks, that uses the memory extensions? That's not true.

 

There are a select few. Not much. And most are app software that nobody uses anymore like Paperclip. I'm not aware of any games that use more than 128K of RAM.