#More on BAD CHIP RAM
29 messages in this thread
Well, I guess it may not be a bad (RAM) chip, but its still INDIGESTION.
Per my previous message It seems that the vertical-striping that I was
seeing is in some way related to overscan. SceneGenerator seems to be
overscanning but I can also now get DPAINt III to pretty reliably plop up
the same vertical stripe(s) when I jump around in screen modes and page
sizes enabling OVERSCAN – I sometimes also get wierd 'striping' from the
arrow pointer like its "dripping" its image vertically over the full
(top-bottom) screen.
I suppose I'll open up the beast and carefully re-seat averything, but I am
starting to worry that some undocumented feature of my now departed Insider
(the $C00000 memory ?) is the cause of the problem… Do I need such memory
to get overscan ?
PS – Monitor is Panasonic DT-M140 but I've had it for years so it is not a
variable factor with this new problem (unless it is failing).
-help – da – via Whap!
Daryl,
Ok…is there an easily accessible width adjustment on your
monitor? If so, try narrowing it. You may simply be seeing the right or
left edge wrapping around or "folded" as it's known in the TV industry.
Don
Don,
Thanks for your thoughts. I did try all of the things you suggested –
here's the next iteration…
It only seems to happen in Overscan mode, always in SceneGenerator,
sometimes in Dpaint. The double stripe vertical bar is almost dead center
in the screen. Monitor problems seem ruled out because of the
pointer/sprite smearing. When I run DPAINT and SceneGenerator I often have
more peculiar problems like the pointer/sprite (in DPAINT it seems to be a
smeared version of all the "specialized DPAINT pointers with "PICK" and
"SIZE" and the arrow all smeared north-south…
I have seen NO other symptoms of any kind on my machine over the week since
I removed the Insider. The problem didn't occur back when I had the Insider
plus the Xetec with 4MB, only since I removed the Insider.
If I take the Xetec controller/ram unit off the expansion port
SceneGenerator runs FINE on my 512K config… but even if I run NoFastMem
with the 4MB Xetec on the bus SceneGenerator and DPAINT overscan have the
vertical bar again.
Is there ANY linkage between Overscan and the $C00000 memory that the
Insider gave me ? I certainly hope I haven't boshed my Agnus, Denise, Paula
or 68K but I can't see how a loose or marginal socket fitting on the 68K
would give such consistent but limited effects…
Is there any diagnostic I can run against Agnus or Denise to see if they
are OK ?
-thanks – da – via Whap!
Daryl,
I know of no linkage between the C00000 memory and overscan other
than C00000 memory is on the CHIP bus, but not accessible (outside the
address range) by the custom chips.
But, I'm not a hardware expert..so could be wrong.
However, I've got an idea…try this. Go into Preferences, move
the screen centering gadget in Preferences so it's against somewhat to the
left of the right edge of the "screen" in preferences. See if that cures
things. If not, try moving it all the way to the right and try again. Let
me know how that works out.
Don
C00000H RAM is only on the chip bus in some A500s and A2000s. There never
was C00000H RAM that was non-fast RAM on the A1000.
-Greg
Greg,
I realize that there wasn't any C00000 ram on the A1000 as it came
from the factory. You had to add a board that was addressed there, but if
you did…it showed up just fine during autoconfig (Dean said it wasn't
true autoconfig…just that memory at that address was found
automatically).
And, as you know, it showed up as "half-fast" RAM.
Thus it showed up on the CHIP bus but not accessible by the custom
chips…yes/no?
Don
Nope. Think about it. Most of those A1000 memory boards fit into the CPU
socket and not anywhere near the custom chip bus. CBM just decoded the
extra 512K at C00000H in the A500 and A2000. The ROM checks for chip
registers or RAM. If it finds ram at c00000H, it adds it to the system
list. Therefore no autoconfig registers are needed. It also gave priority
to that ram, which is why the 512Kfast/512K chip A2000s and A500s were
slower than the A1000 by a few percent. It was an easy way to get the 1 meg
machine's extra 512K recognized. There is no manditory tie-in to the chip
ram bus.
-Greg
Greg, this is Greg 😉
I think what Don was referring to was what is known as the Amazing
Computing Magazine 512K upgrade along with Dave Haynie's auto-config mod.
This consisted of 16 64K x 4 chips piggybacked on the original 8
motherboard memory chips along with some decoding that placed the new
memory at 80000H. Addmem was used to bring the memory online. There was
also a somewhat-kludge using the drive LED to trigger an auto-config
sequence. Later Dave Haynie came out with a 2 chip mod on the daughter
board that relocated the memory to C00000H so that it configured
automagically.
This memory rode on the motherboard chip memory and ran at the same speed
as chip. However since it, of course was not accessible as chip, it came to
known as Half-FAST memory.
Greg (The other one)
Greg White / MicroSearch 72230,1617
We listen to our customers..carefully
Greg,
Nope…I was thinking of things like the Insider.
Don
Sorry, after reading his answer to you about the memory being on the chip
bus, I assumed you were talking about the piggyback mod.
^^^^^^^
Greg
Greg White / MicroSearch 72230,1617
We listen to our customers..carefully
The memory board in question is an Insider; it is true fast ram. The
c00000H RAM being autorecognized was implemented due to the need for it in
the Amiga with Zorro I slots and 68010 processor that was never released.
The fact is you can decode any type of memory to any location, provided the
PALs allow you to (if not, make your own)
The OS just happens to find and use that memory. The piggyback hack is
just another instance of someone taking advantage of this feature.
In your example, the piggybacked RAM lives on the chip ram bus and is
indeed half-fast, but that's not pertinent to this argument.
-Greg
Sorry, I assumed ya'll were talking about 2 different things.
Greg
Greg White / MicroSearch 72230,1617
We listen to our customers..carefully
Greg,
I'll bow to your expertise.
It's not the way it's been explained to me before, but I'm not a
hardware expert…you are.
Don
Just remember that memory at C00000H is found because the OS is set up that
way. You can have any type of read/writable memory decoded at any speed
(virtual paper tape?) and the OS will use it.
My Rejuvenator design moves all three custom chips onto their own bus,
separate from the old motherboard chip ram bus. When two jumpers are set
properly, the motherboard 512K is decoded at C00000H and it is found
automagically. Pull the jumpers, due to a memory conflict (say with the
aforementioned Insider), and you have to addmem it back in at whatever
address the user set the jumpers to.
I have run tests with Dpaint and a memory speed test (RAMSPEED from Fish
31) and have proven that a 16 color hires screen causes the ram benchmark
to slow only a very small amount versus just using chip ram alone. The
speed difference was on the order of 28.5 s seconds with the fast ram, 72
seconds with the 16 color hires screen (DPaint doing nothing).
-Greg
Don,
I see that Greg already answered, but I'll confirm it. The $C00000
memory placed in the A1000 by the Insider (II) was true FAST ram. There is
no contention with video.
The way the A2000/500 was designed is the only reason $C00000 memory is
known as half FAST ram. The reasoning I see was twofold. First, it was
cheaper since the Agnus could be used to control the memory. Second, doing
things that way allowed the memory to be converted to CHIP ram with an
Agnus upgrade.
Dean
DKB Software
Dean,
Ok. You and Greg are certainly better at hardware than I am. I
obviously misunderstood/forgot/misread/etc. the original explanation.
Sorry for the error.
Don
Don,
HOW IN THE WORLD did you come up with that ?!
I was really bummed at the thought that my new 4MB/AutoBoot Xetec might be
at fault so I tried to think of what ELSE had changed – well natch to test
with it off the bus I had had to boot from floppy WB, so I tried this again
but WITH the Xetec ram and scsi on the bus – just no autoboot, NO
PROBLEMS… but I noticed right away the different prefs colors (default)
and remembered that I had been fiddling with interlace, colors, monitor
width/height, horiz centering AND the Prefs Centering gadget…
Went back to BOOT: system-config and reset prefs to default and the whole
mysterious mess DISAPPEARED !?
Now I'm a happy camper but am certainly NAGGED by not understanding WHAT
the heck is happening here ? I can see how jiggering the screen-centering
might interact with my monitor's capabilities to get the bars (? can I ?)
but the "sprite splattering" seems like it could only be derived from
problems on the Amiga End ? Or is it not the monitor ? Does the centering
setting somehow trash things in Overscan ? 1.3 ?
I'd be interested in what led you to your correct guess.
PS – Upside of all this sweat and bother (on my end) is that I did crack
open my old (1986) Amiga Hardware manual and start re-examining Agnus,
Paula and Denise… fascinating stuff, definitely neat for 1986 but I
expected more changes than their (and my) widening girth over the past 4
years…
THANKS
– da – via Whap!
Screen centering affects how much "free time" there is available at the
beginning and end of each scan line. Sprites are moved and displayed during
those time periods (end of line). When you re-center, you can easily reduce
the amount of free time at the end of the scan line where the sprites are
handled by the custom hardware such that there is not enough to do it
right. As a result, the hardware starts the job, gets partway in, and then
loses it's marbles as far as that sprite goes. Display of bitplane (normal
screen data) has a much higher hardware priority, and so the sprite
generator is clobbered, and bitplane display begins.
That's it in a nutshell.
Ben
Amateur Radio Callsign is A A 7 A S
Ben,
Thanks – tho I would have thought moving the screen/center, e.g. to the
right, would reduce end-of-line at the right, but add to it on the left ?!
Not being argumentative, since I saw the problem only in Overscan you've
clearly hit the nail on the head – just wondering.
-tnx – da – via Whap!
Think of the end of one line being part of the beginning of the next…
they occur one after the other with no video in bewteen. For exact sprite
clot placement, refer to the Rom Kernal Manuals.
Ben
Amateur Radio Callsign is A A 7 A S
Daryl,
Ben explained it in his message to you. It's a known "feature" of
the OS and Dean saying it sounded like a mangled sprite tweaked my memory
enough to remember the Preferences centering trick to "cure" it.
Don
Daryl,
The Insider is unlikely to have anything to do with your problem. What
it sounds like is a problem with the Denise chip. Your description sounds
like a 'splattered sprite'. If not the Denise, then the next most likely is
Agnus.
I would guess that you may have zapped something when you had the
machine open.
Dean
DKB Software
Dean,
Glad you're here and I too feel its unlikely that the removal of the
Insider is a factor, except that the problem only occurs since that removal
– natch I hope I didn't bosh Agnes or Denise… how would I see such
limited effect if so ? I have been and am using my system actively and
normally excepting this problem with SceneGen and Overscan…
When I removed the Insider I just removed the two clips and the board then
removed and re-socketed the 68000… its been 4 years since I installed it
so… WUZ there anything else I needed to reverse to get back to my vanilla
A1000 ?
Is there, to your knowledge, any connection between the $C00000 memory I
was getting with the Insider (and don't have now) and the Overscan video?
I DO see that "splattered sprite" syndrome if I run SceneGenerator plus
DPAINT III in overscan, not always, but usually something screws up and I
get sort of a conglomeration of DPAINT pointer/sprites (e.g. the "SIZE" and
"PICK" and the arrow, all smeared out top-to-bottom…
I am not anxious to start desoldering $90 Agnus and Denise chips (might be
eventual incentive to Rejuvenate) so is there any way to specifically
isolate which chip, which address line ? Diagnostics ? It does seem that a
totally scootched Agnus or Denise would leave me with an unworking machine
– so I'm hoping to establish whats special about Overscan and mebbe even
what, specifically, would have to happen to what Copper address line (or
whatever) to give me the vertical bar in the middle of the overscan screen
?!
-thanks for your further thoughts…
– da – via Whap!
Daryl,
Removing the two clips, and the Insider is all that nees to be done to
reverse the Insider installation. (gotta move the 68K too, but you did
that)
$C00000 memory will have no effect upon the splattered sprite problem.
It is possible that the Xetec memory board might have a problem, though if
the problem shows up the same way regardless of what order you load the
programs, I doubt its a memory chip problem.
Both the Agnus and the Denise are socketed, so temporaily replacing them
to test them isn't a real difficult problem. I would suggest taking the
machine to your local dealer to see if they can test the machine for you.
It is entirely possible to trash just a portion of the functionality of a
chip, so zapping the machine can show just about any concieveable symptom.
Dean
DKB Software
Dean,
Thanks – I found the problem to be related to me re-setting the
preferences/screen center gadget (See my reaponse to Don) – 'course I still
don't see how its happening, but resetting it to default makes the problem
go away… Thanks for the tip on P, A and D.
And, if you're the guy to say it to, thanks for the Insider – even tho I've
sold mine now to pay for the 4MB Xetec it was (is) great stuff, well made,
fairly priced and timely to market for the A1000, you should team up with
Greg T. and give us the A-10000 !
– da – via Whap!
Daryl,
Thanks for the kind words.
Regarding the reason that the screen centering caused your problem, dig
a bit deeper into the Hardware Reference manual. Look in the section
describing the video DMA channels. The explanation is too involved to try
to describe here, but the manual should at least give some pointers.
Dean
DKB Software
Dean,
Looking at my A1000 motherboard with the 68000 on the right, which is Agnus
and which is Denise ?
-tnx
– da – via Whap!
Look at the silkscreen. The custom chips are labeled P A D. So the custom
chip closest to the disk drive is PAula, the middle chip Agnus, and the one
closest to the power supply is Denise.
-Greg
Daryl,
Look at the front of the machine right behind the 256K memory port.
There are three 48 pin chips, the 'D', 'A', and 'P' in white next to the
chips tell you which is which.
Dean
DKB Software