What do you do with a 1-MB upgrade?

[+Larry]

I have several. I like them because they are neat, but I have yet to find a very practical use for them. Like the mountain, maybe "because it is there?" Not trying to be negative here -- just curious.

 

I've read before that with 512-byte sectors, that writing to a drive or card, it is faster than writing to the internal ram disk. (?) In the future, when we go faster than 1.79 MHz, then that won't be true, but even with 256-byte sectors, the MyDos menu or MEM.SAV loads essentially as fast from an "external" drive as from the internal ramdisk.

 

But still, what do you use your 1 MB upgrade for?

 

-Larry

[Rybags]

Not a lot, yet. But Ramdisk was about the last thing on my mind when I upgraded. Also have my SDX on IDE+2 configured to use media rather than Ram for mem-sav.

 

Get enough upgrades out there, bigger games and demos will become the norm.

[mimo]

Laugh at people without 1meg upgrades

[flashjazzcat]

Why should RAMdIsks suddenly operate as fast as hard disks if the CPU is faster? The limiting factor of RAMdisks is the fact they're buffered, not the speed of the processor. And a faster processor would mean greater HDD bandwidth, since we're only using a small fraction of the IDE bandwidth at the moment.

 

[Rybags]

Yep, the scaling will always be in favour of IDE unless you get the CPU going like 6 times or more the rate that PIO could handle.

 

Ramdisk has the disadvantage of the intermediate buffering or at the least the fact that PortB has to be changed twice for each byte read. IDE you can just burst direct to where the data is.

[+bob1200xl]

PBI interfaces are not inherently faster than ramdisks but there is generally more code space for a PBI device than a ramdisk driver. This allows you to use inefficient, but fast code.

 

In most all PBI devices, however, the code sits in a slow memory IC that cannot be run at high speed. A ramdisk driver sits in main memory where it can be run at maximum speed. In theory, the ramdisk will run 6x faster at 14mhz than it does at 1.79mhz.

 

In addition, most IDE based storage devices should not be run at speeds much over 3.58mhz.

 

Ramdisks are OK for legacy software, but high speed CPUs will have lots of linear memory - use that, instead.

 

Bob

[flashjazzcat]

PBI interfaces are not inherently faster than ramdisks but there is generally more code space for a PBI device than a ramdisk driver. This allows you to use inefficient, but fast code.

 

As Gary says, you can't "burst read" from a RAMdisk because the destination buffer might be in the same banked region. So you might use an intermediate buffer in low RAM, which would remove the need to bankswitch for every byte read. But it's still double the work of reading the IDE buffer directly into the destination. This is one fundamental advantage of hard disks as things stand at the moment. No amount of loop unrolling, etc, will make the buffered RAMdisk as fast.

 

In most all PBI devices, however, the code sits in a slow memory IC that cannot be run at high speed. A ramdisk driver sits in main memory where it can be run at maximum speed. In theory, the ramdisk will run 6x faster at 14mhz than it does at 1.79mhz.

 

Is that why results on the 65816 1200XL with the (RAM-based) MyIDE soft-driver were so impressive, Bob?

Edited April 27, 2012 by flashjazzcat

[+bob1200xl]

I didn't do the MyIDE tests, but I would say that running in RAM would make a huge difference - yes.

 

Bob

[+Larry]

Gary, Jon...

 

For the record, I am assuming two things:

1) External PBI transmissions will still take place at current speeds.

2) Accelerated (expanded) memory will operate at least double that.

 

I remember the same logic as to why the MIO ramdisk should be faster than the internal ramdisk, but I could never get more than a whit of difference (MyDos).

 

Why don't you show some test numbers?

 

BTW, this wasn't the intent of the thread, but since the issue has been raised...

 

-Larry

[kenjennings]

As slow as it may be compared to an IDE or other kind of device, here the Ramdisk makes makes moving between DOS and a cartridge with memsav on nearly instant. A lot faster than any floppy or even SIO2PC. So, I guess the comparison is between nearly instant and instantly instant

[Gunstar]

We definately need more programs that use 1MB Memory. I've never really used a ram disk, and always thought it pointless since it's gone when the computer is shut off, which can happen very often on a non-multi-tasking Atari 8-bit. Sure there are some now with battery backup, but really, I don't see what the hurry is to get a 64k or less program loaded that only takes a minute at normal SIO speeds anyway. I've always found 3x SIO speed to be fine and anything faster is nice, but I don't really care; I still load the majority of my software off of floppy disks, and MyIDE is plenty fast too.

 

I already have 512K memory, and I don't think I've ever used more than 320K of that, and that with demos and a couple games like Commando+ and Yie ar Kung Fu (yeah, I think I spelled it wrong). So do I need 1MB memory? I don't see much of a reason yet, until some killer apps come out that use more than 512K.

 

I expanded my memory not just for games, but for Word Processing, and I may decide I need 1MB memory to fit more of my novel into memory, but right now, 512K is enough, and more than likely the novel will be split up into files per chapter, so more than 512K at a time probably isn't necessary anyway.

 

If I ever upgrade an Atari to 1MB, it won't be my 512K 1200XL, I'll do it with another computer. But I have no plans to go beyond 512K since I can't even use all of it aside from a demo or two.

 

But, if you haven't already expanded the memory, if a 1MB upgrade doesn't cost much more that a 256/320/512/576K upgrades, then you might as well go all the way and do 1MB.

Edited April 28, 2012 by Gunstar

[+DarkLord]

Waiting for Flashjazzcat's bright, new and shiny GUI to be released!

[flashjazzcat]

Gary, Jon...

 

For the record, I am assuming two things:

1) External PBI transmissions will still take place at current speeds.

2) Accelerated (expanded) memory will operate at least double that.

 

I remember the same logic as to why the MIO ramdisk should be faster than the internal ramdisk, but I could never get more than a whit of difference (MyDos).

 

If you're talking about accelerated RAM in the context of faster CPU speeds - I know nothing about that other than what Bob's said: that code in ROM is limited by the speed of the IC. The only test results I've seen were from a 7Mhz 1200XL running a MyIDE cart using my soft-driver. If the comparisons are limited to ROM-based PBI code (which was not necessarily a given), then my assumptions about increasing RAMdisk / HDD speeds may have been incorrect.

 

Why don't you show some test numbers?

 

I can show test numbers from stock hardware when I get time if it's of interest. Trying to stay focused on the FAT32 / ATR code for Ultimate at the moment. It's urgent.

 

Waiting for Flashjazzcat's bright, new and shiny GUI to be released!

 

Well, the GUI will use it for sure, especially if you have a word processed document containing a hundred 24-point fonts.

[+Larry]

Hi Jon-

 

I certainly understand about your time constraints. In the meantime, I've got a lull here, so I'll revisit this I/O topic. I'll even get out my SDX cart!

 

The MyIDE test in the XL7 was unique (at least at that time), because I saw a post from you that your MyIDE SDX driver was ram based. The XL7 had no PBI code, so there was the only chance to get a hard drive working with the XL7. IIRC, the RWTEST score was something like 160,000 B/sec.

 

KMK ultimately wrote a ramdisk for SDX that went in the linear memory. So there are several comparisons to be made, as well as for some existing hardware.

 

-Larry

[flashjazzcat]

The MyIDE test in the XL7 was unique (at least at that time), because I saw a post from you that your MyIDE SDX driver was ram based. The XL7 had no PBI code, so there was the only chance to get a hard drive working with the XL7.

 

Yeah - recollections are a little hazy about the XL7 tests, but the driver was indeed RAM-based. I had perhaps naively failed to consider the issue of slow ROM chips. I do recall being told that it was the first HDD driver which had worked at all with the XL7. As far as PBI hardware goes, it appears that one or two popular interfaces have design "issues" which might prevent them from working with a faster clocked CPU at all.

[Rybags]

Another little explored area is multitasking.

 

It can be done although it'd be a case of keeping several machine state images in extended Ram that you could switch among e.g. Shift+Reset then choose from a menu.

[flashjazzcat]

Another little explored area is multitasking.

 

It can be done although it'd be a case of keeping several machine state images in extended Ram that you could switch among e.g. Shift+Reset then choose from a menu.

 

I've thought about multitasking in the context of the GUI but given the dreadful performance one would likely experience with 2 or 3 processes running at the same time, task-switching (more akin to what you describe) strikes me as a more practical and useful proposition. In this case, only the application with the focus would be running at any given time (along with any interrupt or timer services).

[+CharlieChaplin]

Well,

 

what would I do with a 1MB RAM upgrade ?!?

 

1) Programming? No, I am not a programmer, not at all, so this is no option for me. But there is T.Karwoth`s XE Macro-Assembler 2.1 available that supports up to 1MB XRAM via MyDOS, see attachment here...

 

2) Gaming ? Hmm, I do not know any games that require 1MB XRAM (Space Harrier therefore is available on cart.), nor if that would make sense - loading time from tape some hours, from disk still 10-20 minutes, from hardisk also quite some time (more than a minute I guess)...

 

3) watching Demos ? YES. Thats all I WOULD do with it, a) watching graphic+sound demos and b) listening to sounds (mostly by using some tools to playback sampled sound, MOD, NEO/converted MOD, etc.). As said before, loading from disk takes some 10-20 minutes, but anyways, here are some 1MB Demos and sounds/soundplayers:

 

a) Demos that require 1MB:

 

- 1MB Raytracing Demo by Solocoder of Ace (Abbuc), originally running only on K.P. 1MB Megaram 3 upgrades, but thanks to patches by Fandal now running on almost every 1MB upgrade:

http://www.atari.fan...p?files_id=5905

 

- 1MB Raytracing Animation Megademo by Lizard of BBSL (also requires an IDE interface+harddisk, a KMK-IDE or IDE+ I guess, have not tested this); see attachment here... ANIM1MB.zip

 

- Brullwurfel by EPI+PIN / Tristesse, 124 seconds of 22khz, 8Bit stereo-sample (supports Pokey and Covox): http://www.atari.fan...p?files_id=2519

http://epi.atari8.in...brullwurfel.zip

 

- Cityscape by Andreas Koch (thats me!), created with Tip-Animator 2xB by converting an AVI / MPG / MOV to the A8: http://atarionline.p...yscap_1088k.xex

 

- Dr. Snuggles Intro by Andreas Koch, again created with Tip-Animator 2xB by TeBe and converting a youtube video to the A8: http://atarionline.p...rsnug_1088k.xex

 

- Pitfall (Atari Jaguar) Intro by Andreas Koch, created with Tip-Animator 2xB by TeBe and converting a youtube video, see attachment here... PITFALL_1MB.zip

 

- Prototype by Poison, PG and Fandal, a converted "Karateka" GIF-Animation:

http://www.atari.fan...p?files_id=6277

 

- Willy Fog Intro by Andreas Koch, created with Tip-Animator 2xB by TeBe and converting a youtube video: http://atarionline.p...lyfog_1088k.xex

 

Maybe there are some more demos...?!?

 

b) Tools that support 1MB:

 

- Boss X by Mirko Sobe, a GUI in TB XL available for MyDOS and sooner or later also for BeweDOS:

http://www.abbuc.de/...bossx1035de.zip (german version)

http://www.abbuc.de/...bossx1035en.zip (english version)

 

- (the) CTH-Browser for SpartaDOS 3.2x by Tom Hunt, see attachment here... BROWSER.zip

(Note: on the disk you will find Bewe-DOS to avoid any possible copyright conflicts, alas, the CTH-Browser does not work under Bewe-DOS)

 

- Inertia 4.x by Tebe, a MOD-player that supports Pokey, Covox, mono or stereo and up to 1MB XRAM: http://madteam.atari...uzytki/it45.zip

 

- NEO-Tracker 1.7 and 1.8 by EPI/Tristesse, plays/converts to/edits NEO sounds on the A8, supports Pokey and Covox, mono or stereo and up to 1MB:

http://epi.atari8.in...tracker-1.7.zip

 

- (the) Small Printery by Walter Lojek (Abbuc), see attachment here... SMALPRT.zip

 

- (the) Sound-Utility by CTH/Tom Hunt, a *.WAV and *.DIG (digital sample) player, see attachment here... WAVPLAY.zip

 

[- Tip-Animator 2.9 and 2xB by TeBe, a PC tool that can create tip-animations for Atari XL/XE with up to 1MB XRAM; see TIP-Animator topic here at atari-age...]

 

- XE Macro-Assembler 2.1 by T. Karwoth, running under MyDOS, see attachment here... TK_ASS_21.zip

 

Maybe there are some more tools that support up to 1MB RAM...?!?

 

Well, I said I "would do", since I have not installed a 1MB RAM upgrade yet into my XL`s. I have four U1MB, but I refuse to cut lines/traces on any of the XL motherboards... Greetings, Andreas Koch.

ANIM_1MB.zip

browser.zip

SMALPRT.zip

WAVPLAY.zip

TK_ASS_21.zip

PITFALL.zip

Edited April 28, 2012 by CharlieChaplin

[+DarkLord]

 

I've thought about multitasking in the context of the GUI but given the dreadful performance one would likely experience with 2 or 3 processes running at the same time, task-switching (more akin to what you describe) strikes me as a more practical and useful proposition. In this case, only the application with the focus would be running at any given time (along with any interrupt or timer services).

 

That happened early on with the ST's GEM/TOS desktop as well. There were a couple of apps that let you accomplish it. One commercial example is "The Juggler". It wasn't actual multitasking, but it was useful. Here are a couple of pics:

 

 

 

[+Larry]

Hi Andreas-

 

Thanks for the very thoughtful and "plentiful" answer. More stuff than I ever imagined.

 

I know what you are saying about cutting traces -- but I did get one of Dropcheck's adapters, so now I don't really have any good excuse! Perhaps your list of uses will motivate me.

 

-Larry

[flashjazzcat]

I've thought about multitasking in the context of the GUI but given the dreadful performance one would likely experience with 2 or 3 processes running at the same time, task-switching (more akin to what you describe) strikes me as a more practical and useful proposition. In this case, only the application with the focus would be running at any given time (along with any interrupt or timer services).

 

That happened early on with the ST's GEM/TOS desktop as well. There were a couple of apps that let you accomplish it. One commercial example is "The Juggler". It wasn't actual multitasking, but it was useful.

 

...

 

 

Thanks for posting that. True multi-tasking is a very attractive ideal, but we have to offset the gains against the work involved. I think being able to click into notepad, copy some text, then click into the word processor and paste it is a far more useful feature than being able to listen to music in the background while hammering away at an unresponsive text editor or file browser in the foreground. I think simple task swicthing would be very nice, and very doable.

[+Larry]

Still more RWTEST data to be generated, but here's a starter:

 

IDE+2, MyDos 4.55 DD, Transcend Industrial CF Card:

WR 20777

RD 28476

Ave 24617

 

IDE+2, MyDos 4.55 DD, Hitachi 1.8" drive

WR 28251

RD 29305

Ave 28788

 

Note: a "real" drive always turns in slightly higher scores than any CF card I've tested. These Industrial cards turn in poorer Writes, presumably due to the error-checking procedures.

 

800XL Internal ramdisk, MyDos 4.55

WR 44624

RD 36700

Ave 40622

 

Note: Nearly all MyDos internal ramdisk test values are in line with these numbers. However I did get two runs of 52K WR and 50K RD. I have no idea how. The MyDos ramdisk is always SD. Looking back at RWTEST results from several yeas ago, these results are very consistent.

 

MIO Ramdisk, MyDos DD

WR 26533

RD 28874

Ave 27704

 

Note: These numbers look slower than the numbers with the "old" firmware from several years ago. (?)

 

For comparison, here is the MyIDE cartridge, MyBios firmware 4.7 Beta-4, using Transcend Industrial Module, MyDos 4.55 DD

WR 19062

RD 23739

Ave 21431

 

Note: if you look through the older numbers, the current results are vastly improved!

 

EDIT: results following added 29 Apr 12

 

Now for some impressive and interesting SDX results:

 

MyIDE SDX 4.45 (FJC Driver), DD

WR 40071

RD 50345

Ave 45208

 

MyIDE SDX 4.45 (FJC Driver), 512 BPS

WR 55308

RD 64376

Ave 59342

 

MyIDE SDX 4.45 Ramdisk (RAMDISK.SYS 8,8 )

WR 22186

RD 24391

Ave 23288

 

Note: So very interesting that under MyDos, the internal ramdisk is by far the fastest device. Under SDX, it is so very slow.

Jon or KMK, can you explain this behavior -- at least on the SDX side? I think that I remember that Drac030 mentioned one time that the SDX ramdisk was slow, but I don't remember any discussion of "why."

 

Have not been able to get the newer SDX drivers working with Bob1200XL's XL7 but for here were the RWTEST results from 2010:

 

SDX 4.42 (?) Flashjazzcat's ram-based MyIDE SDX driver, DD

WR 53066

RD 122716

Ave 87891

 

Note: "You ain't seen nuthin' yet..."

 

SDX 4.42 (?) Flashjazzcat's ram-based MyIDE SDX driver, 512 BPS

WR 157077

RD 163622

Ave 160349

 

Note: WOW!!!

 

EDIT: results following added 13 May 12

 

Now for some more down-to-earth results, this time using Bob1200XL's "WIDE" 65816 system with integrated IDE flash card hard drives:

 

Dos 2.0s (in single density mode)

WR 16227

RD 18700

Ave 17463

 

Dos 2.0D (DD)

WR 29976

RD 26355

Ave 28166

 

MyDos 4.5 (single density)

WR 13974

RD 14332

Ave 14153

 

MyDos 3.07 (DD)

WR 30679

RD 26533

Ave 28606

 

Note: The major difference between MyDos 3.X and 4.X is that the 4.X versions added "big disk" support. 3.X is still a "small disk dos" that has support for 720K DS-DD disks. "Big disk" support evidently slowed MyDos down 15-20%.

 

-Larry

[Gunstar]

Another little explored area is multitasking.

 

It can be done although it'd be a case of keeping several machine state images in extended Ram that you could switch among e.g. Shift+Reset then choose from a menu.

 

I've thought about multitasking in the context of the GUI but given the dreadful performance one would likely experience with 2 or 3 processes running at the same time, task-switching (more akin to what you describe) strikes me as a more practical and useful proposition. In this case, only the application with the focus would be running at any given time (along with any interrupt or timer services).

 

There was a program released through AIM (Atari Interface magazine) back in the early 90's called 'Snap-shot' that allowed two or more programs in extended memory and switch between them at a push of a button. Matbe you should look into that for some ideas. I'll see if I can locate that program anywhere...

[flashjazzcat]

There was a program released through AIM (Atari Interface magazine) back in the early 90's called 'Snap-shot' that allowed two or more programs in extended memory and switch between them at a push of a button. Matbe you should look into that for some ideas. I'll see if I can locate that program anywhere...

 

Thanks. In the GUI all we have to do to switch tasks is switch banks and copy the stack and page zero, since all applications are in the $4000-$7FFF area anyway (although code and resources may occupy more than a single bank). I imagine the snapshot program's job was a little more involved, since it had hardware registers and display lists, etc, to worry about.

[+CharlieChaplin]

Another little explored area is multitasking.

 

It can be done although it'd be a case of keeping several machine state images in extended Ram that you could switch among e.g. Shift+Reset then choose from a menu.

 

I've thought about multitasking in the context of the GUI but given the dreadful performance one would likely experience with 2 or 3 processes running at the same time, task-switching (more akin to what you describe) strikes me as a more practical and useful proposition. In this case, only the application with the focus would be running at any given time (along with any interrupt or timer services).

 

There was a program released through AIM (Atari Interface magazine) back in the early 90's called 'Snap-shot' that allowed two or more programs in extended memory and switch between them at a push of a button. Matbe you should look into that for some ideas. I'll see if I can locate that program anywhere...

 

And even more

 

programs, like MTOS by Tom Hunt, XL2 by John K.Picken, MOS, Programmers PAL and probably some more task-switching programs... -Andreas Koch.