Jump to content
IGNORED

Calling on 1200XL extended memory experts for help!


Gunstar

Recommended Posts

I have an AtariMax Rambo compatible 512K upgrade board in my 1200XL. It seems to have a bad bank of memory or something wrong to make it look like a memory bank is bad, because I've swapped out the ram chips with the exact same result in memory tests.

 

I'm using two different extended memory testing programs, one is Mem-test V 1.0 by Jaku-B and the other is +Memory Tester by Bloomer/Pentagram. Both test all other memory banks before and after to be good. Mem-test seems to indicate bank 15 if I am reading it correctly and +memory stops at bank 16 and says it's dead and I have to push start to continue through the test. But maybe it stops on 16 but is really saying the previous bank 15 is dead.

 

Can this be anything other than a problem with a ram chip? If it were anything other than a ram chip, wouldn't all memory above the bank testing as bad also test as bad or non-existant?

 

As to the memory upgrade in general, I have already swapped out 2 other PIA chips that were known good, just in case, I swapped out 74LS158 chip that is one of two, the other being replaced by the Rambo board. And I even swapped out the GAL chip on the Rambo Board and all the ram chips. And of course quadruple checked all connections, jumpers, etc.

 

So I am stumped. unless I managed to replace the bad chips with other bad chips in the exact same location, more than once...

 

One of the memory test programs says it's bank 16 that is bad, the other shows a spot that I guess would be called 87 if it wasn't dashed out, which, if I count down left to right in the horizontal rows, is the 15th bank of memory. Pictures below.

 

post-149-0-00114900-1473538695_thumb.jpg

post-149-0-38360200-1473538756_thumb.jpg

post-149-0-20601400-1473538811_thumb.jpg

Link to comment
Share on other sites

I think I've figured out this is an issue of mirrored base ram being ignored with these programs, and to move along as there's nothing to see here...still, 31 banks is odd, they should be showing either 28 banks of extended ram, or 32 banks of total ram.

Link to comment
Share on other sites

The very first banking RAM, the Axlon, also mirrored base RAM, the so-called home bank. So did my 256K XL, and that provided advantages. You could access the upper 16K, normally under the OS, in the banked region without disabling the OS ROM. And the 2K normally hidden under the I/O registers were available as well. It's a shame those test writers didn't get it.

 

Back to the OP: I'm not sure of the 512K banking details, but it's likely that your RAM is fine, and the suspect banks are mirrors of the base 48K, which the test programs can't handle.

  • Like 1
Link to comment
Share on other sites

The very first banking RAM, the Axlon, also mirrored base RAM, the so-called home bank. So did my 256K XL, and that provided advantages. You could access the upper 16K, normally under the OS, in the banked region without disabling the OS ROM. And the 2K normally hidden under the I/O registers were available as well. It's a shame those test writers didn't get it.

 

Back to the OP: I'm not sure of the 512K banking details, but it's likely that your RAM is fine, and the suspect banks are mirrors of the base 48K, which the test programs can't handle.

 

Thanks. Yes, this upgrade is based on Rambo, which we all know was based on your 256K XL upgrade, then extends it to 512K, and yes, that's total memory. This is a 1200XL with 16 piggy-backed pair of ram (32 chips total) 8 original base memory sockets.

 

The whole 64K+extended memory came about with the 130XE that had two rows of ram instead of one, and you could leave the base memory alone and use 8x256K chips in the second row, or even piggy-back 8x256K's for a total of 320K or 576K respectively.

 

It seems all the memory testing programs, so far, are based on the XE assumption and don't work right for Axlon or 256 XL original extended memory upgrades.

Edited by Gunstar
Link to comment
Share on other sites

OK, thanks. I still seem to have a memory issue, but I'm not sure it's the ram, or at least apparent ram issues might only be a side-effect of a problem in the circuit somewhere and other tests aren't made for it.

 

Everything I load and test/use up to at least 128K seems to work fine, but when I attempt to load/run program files such as Commando+ (320XE) and other super-cart-to-file programs they crash once started.

 

I'm currently thinking I make have all or most of the first 256K memory (base and extended) working and an issue is occurring somewhere with the second set of 256K ram banks. The other memory test programs are showing 31 banks of good memory and 1 bank of bad memory, but the bad memory is supposedly located at $83 and/or $87 mirrored banks.

 

This is the 3rd Atari I've had with a Rambo or compatible upgrade in it, the second with 512K, and all the programs I'm having issues with worked fine on my other 512K Rambo.

 

Could be lumpy test code. Would you mind testing with my tester and show a screen shot?

attachicon.gifEXTEST.zip

Edited by Gunstar
Link to comment
Share on other sites

I understand, I was just hoping it might be a side-effect of a common problem that someone like you could say "oh, yeah, it does that if..." I'm not sure that there are any GAL equations or schematics (for the Atarimax Rambo board) available to the public. But since I replaced the GAL chip I'm thinking it's elsewhere beside that Rambo board. This isn't just a 1200XL with a Rambo upgrade, it's totally maxed out on fixes, mods and upgrades, aside from accelerators and graphic boards. So it could be something isn't playing nice with another upgrade. I'm sure I will eventually figure it out, I was just hoping for some expert's insight.

Link to comment
Share on other sites

Your test shows 31 banks instead of 32 (32 total banks, or 28 extended is what it should be), like all the other memory tests I've used, but, your test is showing bit 4 of Port B working where it was shown inactive with all the other tests.

post-149-0-89860800-1473995205_thumb.jpg

Link to comment
Share on other sites

Thanks.

 

The bits used part doesn't account for anything like bit 4 being needed to be zero in order to

operate the extended memory in the first place. Other tests are masking it's use to show the

above fact so it's really not being used as a banking bit per se here either, I wanted to see

every bit used even once, it's just a collective result of the banking bytes. No

reinterpretations allowed in other words. If it ever is a zero, it gets counted by this one.

 

87 is missing as reported by other tests. Something smells in Denmark, but I have no idea why

that particular arrangement of bits would be doing this on that machine. They are working in

other combinations. It's as if the upgrade were written to exclude this byte on purpose. Not

very likely that is the case, I'm tending to agree that some other upgrade is interfering at

this point. How easy would it be to start disabling them one by one to find the guilty party?

 

Seeing a table of 31 bytes is a first for me, it would be an even 32 if bank

87h were there and that would be the standard display from a 512K machine.

 

Your test shows 31 banks instead of 32 (32 total banks, or 28 extended is what it should be), like all the other memory tests I've used, but, your test is showing bit 4 of Port B working where it was shown inactive with all the other tests.

Have to disagree with 28 count. You don't have 4 extended memory banks on a

1200 to start with like might be found on an XE. There, 4 extended memory banks

are often replaced outright and those banks are not added to, they are just

missing and not available thru portb at all. It COULD be that they could be

made available, I've never seen it done that way. To then account for them is

a useless exercise. I don't pay any attention to the description of a memory

upgrade beyond the ballpark size figure, because so many call it this when it

never was possible and it winds up being that instead with the very best

intentions too. 4, 8, 16, 32, and 64 are what I'm used to seeing be it an XE

or an XL. Home bank is stock base ram and is not counted with this tester.

 

Do let us know if you find out anything about this issue. Is your other 512K

also an atarimax rambo like this one? I may have to get one of these upgrades

and do some snooping up close, thanks again.

 

I did some work on my EXTEST and it's slightly changed from before

although it won't show any differences because of that work. It was

mainly to clean it up and remove some unnecessary stuff, a bit of

knowledgeable debugging. And source and assembly listing too so

anyone might see how this one works with the added comments in it.

 

Extest2.zip

Edited by 1050
  • Like 1
Link to comment
Share on other sites

1050: "Have to disagree with 28 count. You don't have 4 extended memory banks on a

1200 to start with like might be found on an XE."

 

No,but I'm not referring to 4 banks of extended memory like an XE, but the 4 banks of BASE memory. When you replace the 8x64K chips with 8x256K chips, on an XL, four of those banks are mirrors of base memory, so if there was a PROPER memory test program that tested ALL memory, base and extended, then it should show up as a machine with 4 banks of base memory, 64K, and 28 banks of extended memory, (448K+64K=512K: XL's with either 256K or 512K upgrades have 64K base and 192K/448K extended memory in the 256/512K ram chips ). So a properly working mem-test on an XL should either report a full 16 or 32 banks of base and extended ram, or, 12 and 28 banks of extended memory. None of the current mem-test programs do either. They all seem to assume a 130XE and don't account for XL's at all.

 

So if anyone ever makes a proper test that doesn't just assume 64K base memory and actually tests ALL banks, then 28 should come up as extended memory on a 512K XL,upgrade. So far, all memory test programs have been based on the assumption of an XE, and not XL's, so some of the XL ram banks are ignored as base memory, $83 and $87 are mirrors of base memory in an XL memory upgrade.

 

Here's how it was explained to me on another site by Brenski:

 

"XRam021 and XRam019 WILL report only 15 of 16 banks of ram (for a 256K upgrade) - that's normal, as being an extended ram test your "board ram" is excluded from the check.

so it should report 192k. Fact that you're getting 240k on XRam test is a little strange - unless i misunderstand you?


The "Rambo XL" using 16 banks of 16K - works like this:
Utilizing Port B bits to control memory configuration, Bits 0, 1 and 7, deal with ROM/RAM switching.

In RamboXL, bit 4 enables banked CPU & ANTIC RAM. Banked RAM substituting for stock RAM at addresses $4000-$7FFF. Bits 2,3,5 and 6 access one of sixteen 16K banks. Four banks will mirror stock 64K 800XL RAM, and four others replicate extended 130XE RAM.

I found this: - does it help at all?


$FF none normal
$83 0 mirror of normal $0000-$3FFF (page 0, stack, OS variables, DOS, etc. = dangerous bank selection)
$87 1 mirror of normal $4000-$7FFF (address = same bank area = useless for bank selection)
$8B 2 mirror of normal $8000-$BFFF (screen RAM, not much other use for bank selection)
$8F 3 mirror of normal $C000-$FFFF (RAM below OS ROM = useful)
$A3 4, $A7 5, $AB 6, $AF 7, $C3 8, $C7 9, $CB 10, $CF 11 extended
$E3 12, $E7 13, $EB 14, $EF 15 = XE compatible extended


XRam Programs do not address $87 because it mirros stock RAM. Any crashes caused access to $83 bare due to it trying to write to slots occupied by DOS and OS.

 

this is because - in the example of a 256k ram upgrade:

in 800xl/1200xl - (as you say) first 64 is board ram, so is a "given".
the complication now occurs where XRam021 or XRam019 progs test the extended ram and try to access $83 and $87 - which on Rambo boards - are mirrors of actual base system memory.
XRam is effectively seeing banks of memory with flags that say "system Ram" - so it ignores these"

 

Me: This should be the same as a 512K upgrade but with an additional 16 banks of ram. With XL's, you are replacing the base memory chips in the same locations with chips that will be BOTH base AND extended memory, since there is no second row of chips for extended memory like with an 130XE. Which is why I asked for experts of 1200XL EXTENDED MEMORY, and not just experts of extended memory on ANY Atari. Apparently, neither you or I am experts on XL extended memory and how it works.

Edited by Gunstar
  • Like 1
Link to comment
Share on other sites

I'm not a programmer or I would make my own memory test, I'm just a sophomoric hardware guy. But it would be nice if a programmer made a memory test program that actually covered ALL Atari's and included base and extended memory. This test would include the Axlon standard for 400/800's, the Rambo standard for XL's and the 130XE(Freddie, CPU/ANTIC) standard and include base and extended memory for all machines. So far only the XE standard has been done, and even they just assume the first 64K is working too. At least there should be one that covers both XL and XE memory, since no one ever makes use of 800 Axlon extended memory anymore...

Edited by Gunstar
Link to comment
Share on other sites

The thread title is then not quite on target then. What you want is a

Rambo/Wizztronic upgrade expert. While I have a wizz board, I've never

gotten around to populate it to then give it a run thru. I'm sure if

I had done that, I would have noticed some of those odd things about it.

It's exactly what I love to do with ramdisks in fact, get under the

hood, and do some snooping around. I would have noticed base ram in

there, I'm quite sure of that. And then would have found ways to

keep it isolated from the ramdisk banks proper.

 

So I have attempted to include some code that should operate with that

type of memory upgrade and it should automatically reserve those when

they do show up as mirrored bank windows. But I can't test it since

I suspect the Rambo choice in the emulator is a bit fudged in this

regard. I think their main concern was in getting the extended

memory under that selection to use the Rambo bytes, duplicating the

base memory into the extended was just not done perhaps. And I can't

know without a real rambo or wizz, I've now lost the board too so then

there's that issue. It's not happening here anytime soon is the

takeaway.

 

Just this year Axlon users have 4 megs for a ramdisk, so I consider your

Axlon summery to be misspeak, I don't believe the end game for the Axlon is

even in play. Very few numbers in that camp for sure, but that fact never

meant much to me anyway. You can't be much of a power user with just

4 banks.

 

So let me know what this version does on your machine, I may have to clean

house and build the wizz after all. I need a screen shot as a minimum.

And a memory dump afterwards of 0x3000 -- 0x34FF would help debug it

if it's got some flaky to it on the real machine. Emulator use shows no

change for all sizes except stock 64K, had never tried that one before,

didn't like what I saw on other versions. So yet another fix, what else

could be new for a programmer?

 

Extest3.zip

  • Like 1
Link to comment
Share on other sites

I have reduced my Rambo compatible 512K back to 256K, still getting errors with other memory tests, so I have come to the conclusion that none of them work right with AtariMax Rambo board. But I know all 256K is working properly by loading programs requiring 192K extended memory. Then I saw the new version of your Extest you posted here, and tried it, and it does properly identify 12 banks of extended memory now. Thanks! Now I just have to test some more 256x8 chips to find another 8 good chips, then I will redo the Rambo to 512K and try your test again. Thanks!

  • Like 1
Link to comment
Share on other sites

Thanks guys, please post screen shots if you can or otherwise just post the list of

bytes shown. That's the only feedback I have.

 

Interesting side note that just came to mind concerning the missing bank 87 with

my old tester and the other testers too. They are properly marking it as home

bank and then NOT including it in the total count, it's the other three banks

that aren't getting the same treatment so they are counted as genuine banks.

 

The tester was working all along, we just have 'troubles' understanding what it's

trying to say with a missing bank - I was blaming the hardware GAL for example.

Edited by 1050
Link to comment
Share on other sites

Looks like it came up with the right banks but it ran very quickly. Does it actually run any tests on the RAM? Also the PORTB bits aren't quite right. This upgrade uses bits 2-6 including bit 5.

 

attachicon.gifIMG_20160921_173959.jpg

I got the exact same as you with the AtariMax Rambo. All the other testers would show 2-6 excluding 4. It doesn't seem to be doing a full ram test though, only half the ram testers I've tried seem to run through ram in entirety.

Link to comment
Share on other sites

Looks like it came up with the right banks but it ran very quickly. Does it actually run any tests on the RAM? Also the PORTB bits aren't quite right. This upgrade uses bits 2-6 including bit 5.

Thanks Claus,

Only testing done is of the reconnoiter of working banks, calling it a tester is then a

matter of language convenience truth be told. It tests one byte out of 16K so not much

of a test. A proper testing of ram would be far more time consuming as you've noticed.

 

 

Bit use again needs to be interpreted. What I see is that all four bit 5 banks just

happen to be base ram banks and so they were not included in the bit report, so the

bit report minus 8F, 8B, 87, and 83 is correct after all. Not that there isn't a

better way to do the bit report mind you, I prefer to keep it this very simple way

and deal with the results we might not like so much with an explanation of just

how simple the bit report is made. It's taking the lowest value banking bit and

reporting those bits only -- turns out to be correct as far as it goes. KISS is

the acronym often applied and the stupid part is this bit report, dumb as a box

of rocks but it doesn't know how to lie. This level of purity really appeals

to me, it will be a hard sell convincing me it needs to be changed. You have

a unique report screen from other upgrades other than the wizz copy of course.

We had a unique report with one missing byte from before too. Home base bank 87

mirrored to FF was properly MIA and I thought it was a defect in the upgrade

instead of the correct (as far as it went) results should be. Now it's also

removing all the other mirrored to FF base banks.

 

A3 is the lowest value banking byte allowed into the list, expressed in binary then

10100011 inversed is

01011100 and this is the bit pattern of the bit report.

 

83 from the earlier test is then

10000011 inversed is

01111100 and that was the bit pattern there.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...