Faicuai Posted May 21, 2021 Share Posted May 21, 2021 1 hour ago, drpeter said: With @Philsan's 800XL it took only 8 mins from cold for the effect to develop... Ok, I retimed the whole thing again. In a nutshell: The test in post #3216 loses its blue stripe within TWO (2) minutes from power-up. The REVISED multi-column test in post #3222: begins losing each individual stripe within TWO (2) minutes from power-up, starting with column 1, then 4, then 2 and then 3 (in that order). By minute #5 after power-up, everything has disappeared, leaving only the far-left reference column visible. That's about it. 1 Quote Link to comment Share on other sites More sharing options...
drpeter Posted May 21, 2021 Share Posted May 21, 2021 6 minutes ago, Faicuai said: The test in post #3216 loses its blue stripe within TWO (2) minutes from power-up. The REVISED multi-column test in post #3222: begins losing each individual stripe within TWO (2) minutes from power-up, starting with column 1, then 4, then 2 and then 3 (in that order). By minute #5 after power-up, everything has disappeared, leaving only the far-left reference column visible. Very interesting. So even (some of?) the oldest GTIAs are subject to a degree of thermal variability in this regard... (I'm asssuming it's the GTIA, but without applying a heatsink to the chip might be difficult to prove...) Quote Link to comment Share on other sites More sharing options...
drpeter Posted May 21, 2021 Share Posted May 21, 2021 13 minutes ago, Faicuai said: By minute #5 after power-up, everything has disappeared, leaving only the far-left reference column visible. And in the case of your 800, all players are affected. Does it show a 'preference' between players? In other words, do rectangles of a particular colour tend to disappear within the vertical bands before others, or is it the same across all colours? Quote Link to comment Share on other sites More sharing options...
Faicuai Posted May 21, 2021 Share Posted May 21, 2021 2 hours ago, drpeter said: I'm asssuming it's the GTIA, but without applying a heatsink to the chip might be difficult to prove... It is worth mentioning a couple of things on my test bed, here: The machine is an 800, Week-4/1980, originally fitted with a CTIA. I extracted (carefully) the CTIA and stored it safely (so we don't damage it), and installed a GTIA, manufactured around 1981 or 1982, which was supposed to be "new", and part of my spare-parts inventory. (NOTE: I did test that GTIA extensively with multiple diagnostics, including Acid800 v1.2 Beta) It would be possible to install a functional heat-sink on GTIA, because we could ease the transfer of heat from the chip into the massive RF shield-frame of the 800, instead of simply radiating it inside of the CPU-card chamber provisioned in the shield-frame. We could definitely explore this, but I would need to find a suitable heat-sink for this purpose (have not seen one, yet). 1 Quote Link to comment Share on other sites More sharing options...
drpeter Posted May 21, 2021 Share Posted May 21, 2021 (edited) 24 minutes ago, Faicuai said: By minute #5 after power-up, everything has disappeared I'm assuming from this that if you run Artefact_1 from absolute cold, the bands start at 5 then one-by-one jump up to 6? Edited May 21, 2021 by drpeter Quote Link to comment Share on other sites More sharing options...
Faicuai Posted May 21, 2021 Share Posted May 21, 2021 2 minutes ago, drpeter said: Does it show a 'preference' between players? ONLY in terms of the ORDER in which columns begin to disappear. I will try to record this for you, and post it here... About 3 mins. of net footage (showing actual changes)... Quote Link to comment Share on other sites More sharing options...
drpeter Posted May 21, 2021 Share Posted May 21, 2021 2 minutes ago, Faicuai said: installed a GTIA in place of around 1981 or 9182, Ah, so not totally vintage then Quote Link to comment Share on other sites More sharing options...
+Philsan Posted May 21, 2021 Author Share Posted May 21, 2021 1 hour ago, drpeter said: With @Philsan's 800XL it took only 8 mins from cold for the effect to develop... And on a 130XE 20 minutes with Spidey, if I recall correctly. Quote Link to comment Share on other sites More sharing options...
drpeter Posted May 21, 2021 Share Posted May 21, 2021 (edited) It's amazing that we're uncovering new facts about this hardware after 40 years! (assuming this effect has never been previously documented anywhere) Edited May 21, 2021 by drpeter 2 Quote Link to comment Share on other sites More sharing options...
drpeter Posted May 21, 2021 Share Posted May 21, 2021 (edited) 15 minutes ago, Faicuai said: It would be possible to install a functional heat-sink on GTIA, because we could ease the transfer of heat from the chip into the massive RF shield-frame of the 800 I guess even a small heatsink would slow down or limit the development of the effect, even if it was not entirely prevented (as a proof of concept) Edited May 21, 2021 by drpeter Quote Link to comment Share on other sites More sharing options...
Faicuai Posted May 21, 2021 Share Posted May 21, 2021 8 minutes ago, drpeter said: It's amazing that we're uncovering new facts about this hardware after 40 years! Here it is (should show correctly on, at least, Windows/10 & Edge browser. If not, just download the clip first, and playback on your end): Please, ignore magnified saturation of colors: IMG_2390.mp4 Quote Link to comment Share on other sites More sharing options...
Beeblebrox Posted May 21, 2021 Share Posted May 21, 2021 13 minutes ago, drpeter said: It's amazing that we're uncovering new facts about this hardware after 40 years! (assuming this effect has never been previously documented anywhere) Who'd have thought the need for installing heatsinks in these computers would ever arise?! Very interesting deductions over the last few days - well done. Would be good to hear if anyone else has ever retrofitted heatsinks to any original components on the A8's to alleviate any specific issue. (That reminds me - back in the late 90's I got into overclocking my then AMD K6-2 233mhz based tower PC and made my own water cooled heatsink. (Seriously - ran an aquarium tank and pump into the PC directly into a heatsink of my own design. Seems insane when I look back but it worked surprisingly well at the time! Not planning on doing the same for the A8 mind - bit overkill - although...........heh heh ) Quote Link to comment Share on other sites More sharing options...
drpeter Posted May 21, 2021 Share Posted May 21, 2021 5 minutes ago, Beeblebrox said: need for installing heatsinks I'm not sure I'd go so far as to call it a need! The fact that this is such an edge-case in usage presumably largely explains why it's taken 40 years to come out of the woodwork. I wouldn't so much label this as a fault in the GTIA as a previously-undocumented variation in behaviour. The Altirra Hardware Reference Manual already documents at least one other known one. 1 Quote Link to comment Share on other sites More sharing options...
drpeter Posted May 21, 2021 Share Posted May 21, 2021 (edited) 25 minutes ago, Faicuai said: Here it is Astonishing footage! With respect to the players, the effect develops in the order: Player 3 (blue)->Player 1 (pink)->Player 2 (mauve)->Player 0 (green/yellow) Vertically, there is a subtle top->bottom progression Horizontally, it's as you said- Column 1->Column 4->Column 2->Column 3 All of these overlap somewhat. But the end result is that the last rectangle standing is the bottom green/yellow in Column 3! Edited May 21, 2021 by drpeter Quote Link to comment Share on other sites More sharing options...
Beeblebrox Posted May 21, 2021 Share Posted May 21, 2021 8 minutes ago, drpeter said: I'm not sure I'd go so far as to call it a need! The fact that this is such an edge-case in usage presumably largely explains why it's taken 40 years to come out of the woodwork. I wouldn't so much label this as a fault in the GTIA as a previously-undocumented variation in behaviour. The Altirra Hardware Reference Manual already documents at least one other known one. sure, fair points. Should have some more images posted tmrw btw. Quote Link to comment Share on other sites More sharing options...
Faicuai Posted May 21, 2021 Share Posted May 21, 2021 11 minutes ago, drpeter said: All of these overlap somewhat. But the end result is that the last rectangle standing is the bottom green/yellow in Column 3! IMPORTANT update: This all got me curious, so I fired up one of my 800XLs (they are IDENTICAL, just a few serial #s apart). The result is that fading begins almost EXACTLY at the same point as my 800 (minute #2, from power up), but progresses instead for an additional 6 minutes (going all the way to a total of 8 minutes). The fading process and sequence is very similar to the 800, and it ALSO ends in a complete disappearance of all columns, with the only notable exception being that the bottom stripe of Column three is NOT where or how it ends on the 800XL. Instead, the fading takes in groups of equal sizes on that column, and no single stripe is left alone, at the end. They all die together. 8-) 1 Quote Link to comment Share on other sites More sharing options...
drpeter Posted May 21, 2021 Share Posted May 21, 2021 (edited) 1 hour ago, drpeter said: Astonishing footage! With respect to the players, the effect develops in the order: Player 3 (blue)->Player 1 (pink)->Player 2 (mauve)->Player 0 (green/yellow) Vertically, there is a subtle top->bottom progression Horizontally, it's as you said- Column 1->Column 4->Column 2->Column 3 All of these overlap somewhat. But the end result is that the last rectangle standing is the bottom green/yellow in Column 3! Actually, that's not all entirely true, as bizarrely for Columns 1 & 2 (only) the player order is: Player 1 (pink)->Player 3 (blue)->Player 2 (mauve)->Player 0 (green/yellow) Edited May 21, 2021 by drpeter Quote Link to comment Share on other sites More sharing options...
a8isa1 Posted May 21, 2021 Share Posted May 21, 2021 (edited) 11 hours ago, drpeter said: OK, this is the corrected version.... drpeter_Artefact_Test_3.1.xex 21.51 kB · 13 downloads a8isa1's 1981 800 <--click me Takes only 1:15 for all the test segments to black out from a cold start. I needed to flash a cartridge with drpeter_Artefact_Test_3.1.xex otherwise the machine would be mostly warmed up before the test started. Edited May 21, 2021 by a8isa1 1 Quote Link to comment Share on other sites More sharing options...
drpeter Posted May 21, 2021 Share Posted May 21, 2021 9 minutes ago, a8isa1 said: Takes only 1:15 for all the test segments to black out from a cold start Does it show the same progression as @Faicuai's 800 (see video above)? Quote Link to comment Share on other sites More sharing options...
a8isa1 Posted May 21, 2021 Share Posted May 21, 2021 4 minutes ago, drpeter said: Does it show the same progression as @Faicuai's 800 (see video above)? You can see what happens if you click my link. I don't know how to embed videos. Sorry. 1 Quote Link to comment Share on other sites More sharing options...
Faicuai Posted May 21, 2021 Share Posted May 21, 2021 6 minutes ago, drpeter said: the same progression No. Fairly different, in fact. Beginning with 1st column, which disappears completely before all of the rest. Interesting, to say the least. Quote Link to comment Share on other sites More sharing options...
drpeter Posted May 21, 2021 Share Posted May 21, 2021 (edited) 13 minutes ago, a8isa1 said: You can see what happens if you click my link. Sorry, missed that. Definitely subtly different, but I think fits in with the emerging general pattern, but with some significant variations on the theme Player 3 (blue)->Player 1 (pink)->Player 2 (mauve)->Player 0 (green/yellow) Vertically, top->bottom Horizontally, Column 1->Column 4->Column 2->Column 3 Edited May 21, 2021 by drpeter Quote Link to comment Share on other sites More sharing options...
drpeter Posted May 21, 2021 Share Posted May 21, 2021 (edited) @Faicuai , @a8isa1 still interested to see what happens with drpeter_Artefact_1 if run on these 800s from a definite cold start Edited May 21, 2021 by drpeter Quote Link to comment Share on other sites More sharing options...
ivop Posted May 21, 2021 Share Posted May 21, 2021 Do you know how to fix this in RastaConverter in the first place? Perhaps one can detect the specific loads and stores within a 6 cycle window (not too many combinations I think), and then evaluate that particular kernel line as pretty bad in the late hill climbing sense? Quote Link to comment Share on other sites More sharing options...
drpeter Posted May 21, 2021 Share Posted May 21, 2021 (edited) 17 minutes ago, ivop said: Do you know how to fix this in RastaConverter in the first place? Perhaps one can detect the specific loads and stores within a 6 cycle window (not too many combinations I think), and then evaluate that particular kernel line as pretty bad in the late hill climbing sense? the author @ilmenit, and @phaeron who also subsequently worked on it, have indicated on multiple occasions that they don't have the time to go back to RastaConverter and fix the longer-recognised issue with <=5 cycle windows, so I think for the moment we are left with patching the RastaConverter files themselves when glitches occur. I sadly lack the coding experience to tackle RastaConverter, although it is open source. Edited May 21, 2021 by drpeter 1 Quote Link to comment Share on other sites More sharing options...
Recommended Posts
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.