Photon's A1700

[Photon]

Nah, it looks quite ok. And of course I've watched those demos before, and a lot more these last weeks, and played 3D games, and...

[Kitty]

Photon what is that joystick on the left in your picture? I'm gonna get me one of those if they are still available to buy somewhere?

[Jope]

Originally Posted by Kitty

Photon what is that joystick on the left in your picture? I'm gonna get me one of those if they are still available to buy somewhere?

That's a black Suncom TAC-2.. Very easy to find.

[Photon]

Yep, with a blade switched key from a 70s terminal keyboard as left fire button. Normally the left one looks like the right one. This gives several benefits even if the original button is great for many games, better than most joysticks of its era actually.

[Jope]

The TAC-2's buttons were always its achilles heel.

As a kid I spent lots of time filing down the bronze rings to get rid of the oxide and have them make proper contact. Then I saw the light, bought a competition pro and never looked back. ;-)

But recently I heard a good trick that keeps the TAC-2 buttons working for longer. Take some copper paste or arctic silver heat sink compound, and smear a thin layer of that on the bronze rings. Problem no more and you can actually fire reliably with the TAC-2. :-D

[chiark]

Nice Job! I might have to look at doing that if I ever do go for another 1260... The lack of RAM on the desktop version frustrated me immensely as I couldn't get many 060 demos to work with 32MB.

[Photon]

Originally Posted by Jope

The TAC-2's buttons were always its achilles heel.

As a kid I spent lots of time filing down the bronze rings to get rid of the oxide and have them make proper contact. Then I saw the light, bought a competition pro and never looked back. ;-)

But recently I heard a good trick that keeps the TAC-2 buttons working for longer. Take some copper paste or arctic silver heat sink compound, and smear a thin layer of that on the bronze rings. Problem no more and you can actually fire reliably with the TAC-2. :-D

Well, they're of course tighter when they do have good contact , less 'key travel' and all that... but what usually happens is you can't precision-tap to make sure 100/100 presses are hits. It's usually very good enough for when fire needs to be 'held' rather than tapped, such as in karate games or whatever (cos the joystick is just as tight and nice). But shooters will often leave you unsatisfied and cramped. Another reason is I had a keyboard key that fit snugly in the hole and was also excellent

Originally Posted by chiark

Nice Job! I might have to look at doing that if I ever do go for another 1260... The lack of RAM on the desktop version frustrated me immensely as I couldn't get many 060 demos to work with 32MB.

Why do you think I did it? Another reason was making max use of the turbo card, since it doesn't allow just the one 64MB stick.

[Photon]

OK, after another go at 100 MHz, really going all out and giving it all I've got, it turns out overclocking Amiga is very similar to overclocking a PC.

I cleaned the edge connector, and everything socketed on the Apollo including SIMMs and their sockets. Went through 4 ATX PSUs to get it to even boot, a 500 Watter with a 5V reading of 5.00V or higher under load did it. In retrospect, it really was quite linear. Enough over-amperage (~18-20A) on the 3.3V to make the CPU boot without RAMs, and a good enough 5V for it to work with RAMs. After that, it's all about the cooling to make the sensitive CPU like it for long enough.

Everything is nicely set up inside, with two thick leads for 5V and 3.3V. Swapped the 47uF cap on the board and added one same over 3.3V. There's really nothing left to improve as far as the hookup is concerned.

Result: 1.5 minutes of full CPU load. Set CPU60 options to "worst" (slowest) and it runs forever. This was after swapping the cooler for a larger, brand new pure copper cooler for PC gfx cards.

Potential issues ruled out by testing:

1. Edge connector (if you get black or yellow screen and no boot or guru, it's not on right. you need to ram it in straight, never 'rock it in').
2. Memory not coping (with the ApolloBusTest I coded vs. change of cooler, better cooler made the difference)
3. Memory, Mach chips, or other chips overheating. Doesn't mean all Mach chip versions can take running @ 100/2=50 MHz (as in, that's not ruled out yet but I wouldn't assume it to be a problem - without proof.)
4. CPU gets very hot (it doesn't, it's just very sensitive when it draws so many Amps)

So basically this rig now has an ATX connector at left and I moved the CF drive to the floppy slit, and I can swap oscillator and PSU at whim for further testing.

>> Gallery here <<

The basic conclusion is that my specific 68060R6 is the same as all the other such CPUs and would take 100MHz with a huge cooler that fits in a tower but not in an Amiga case. And so that closes it for me, as this will be another BYOC demoparty Amiga.

Anything up to 95 MHz should be fine in my A1700, 96+ is where it gets sensitive and magical whether it works or not. (Ie. gets sensitive and must not get one degree over room temp, etc.)

Just plugged in the 120W PicoPSU and turned the cooler to minimum RPM, and it's fast and almost inaudible. That's the way, uh-huh, uh-huh, I like it...

Still, you can bet I'll try to get a 95 MHz oscillator asap and give that a go...

-~-
This really would have taken much longer without Roy Bates's experiences and help, so a big thanks to him!
-~-

Also made a nice little stand for the 13" LCD TV for portability, and I won't need to bring the 12V PSU for it obviously.

"That'll do nicely." (c) Han Solo.

[cosmiq]

insane.... :P

[Paul_s]

hows about some water cooling?

[Photon]

It has been suggested Paul_s

Well, I haven't really done anything, just did a lot of tests and pa pic update. The tape bits inside were mostly for quick fastening of cables. I don't mind it, better than trying to do a too-tidy wiring scheme and have to redo it when you get another idea. But the white tape on the Apollo is gone, was just there to protect the 3.3V/5V during testing. I'm all about solid builds using techniques at first glance considered hacky and messy.

Just take the ATX connector, for example. It will sit there forever and fits perfectly (well nip off 2mm of one RF shield stand) because of how I hotglued it. Why make a neatly filed hole in the back and order some rare ATX female that can be screwed to the case, or some craaaazy sh*t like that? It won't ever get done, it will be worse, and look better.

These are the PSUs I tested in 100 MHz mode and the results (in bootability order):

PSU		Limit**		3.3V current	5V acc	Result with ram		Result without ram
120W (Pico)	49.8W		6A		4.95V	No boot			No boot*
250W other	120W		20A		4.82V	No boot			Boots fine
500W Fortron	?		24A		4.85V	No boot			Not tested, likely fine
250W Dell	150W		18A		5.05V	Almost to "devs:pref..."	Boots fine
145W Fortron	83W?		10A		5.08V	Past "devs:preferences"	Boots fine
600W Loudmofo	?		20A		5.13V	1.5 minutes stable	Not tested, likely fine

*One boot to just after devs:Preferences but then No Boot every time.

** Watt load limit shared for both voltages 3.3V and 5V. This affects max Ampere on both.

Note: 5V acc is measured over memory chips while running. 
3.3V current is subject to the Watt load limit and may not be as high if the 5V draw is big, 
so take those with a grain of salt.

Thoughts on the table:

• REAL amperage on the 3.3V needs to be large enough (~10A?) for it to boot without RAM at all.
• 5V UNDER LOAD needs to be 5.00 or better to boot with RAM at all.

as for why the 600W fared better than the 145W I'll let you explain Roy I'm still not sure. Might be that every single PSU shares a load limit between 5V and 3.3V and 350W and up makes the 3.3V amperage go over 10 real amps or something.

And the 1.5 minutes was with the CPU running under 100% with Best settings. And that's nothing, of course. It needs to do that day in and day out.

I've ordered 2x92 MHz and 2x95 MHz crystals Roy, let's see if EuroQuartz takes such small orders (and sends them to Sweden!!)

[roy bates]

ok this isent going to be easy to understand,but ill try to explain how i see it.


when the psu is turned on, in those first few miliseconds all the rails get power.
then the cpu draws more than normal,so the 3.3 volt detect senses more load,and outputs more amps on the 3.3 volt rail,(this is fed back to the psu)
at this point the 5 volt rail drops by 5% and the 3.3 volt rail goes back to normal.(there is a spike at startup)
this is where it gets interesting,the supply does not stabalise properly and the 5 volt rail stays 5% under voltage=no boot or instable system.
as the memory is connected to the 5 volt rail,it fails at startup because its at its lower end of the parts tollarences.
(the supply only detects highs not lows.)

ok this is different to how it works on mine,i have re-wired mine for use as a at supply,so at startup,the psu puts out max power at a constant voltage all the time(mine has no sense wire i removed it when i converted the supply.)
plus mine is 400watts.

also to confuse matters even more this revision 060 works between 95 and 105mhz,this is confirmed.

also because of the shrink in mask it also fails at a lower temprature,if youve had a revision 1 you know how hot it gets at 50 mhz,so this makes sense as the later masks run cooler they also fail at a cooler temprature.
it only starts to get hot if its fpu is in use or its on for long periods of time.
(active cooling is a must at these frequencys,heatsink and fan,12volts i suggest something like a 486 type cooler,vga type should work,but not passive,it wont work for more than 15 mins tops and thats if it runs at these speeds.)

i hope ive made enough sense for everyone to understand.

[Photon]

Agree that some PSUs go strangely low on 5V under load, maybe more while having a bigger load on 3.3V.

The CPU is really not a problem as you say, and agree it seems to have a more sensitive temp range even when you feel it should be able to handle "finger warm"... the CPU is not a problem because you just add a good active CPU/gfx PC cooler. It's only a problem for me because I can't fit such a cooler in my case

(I think what you're looking for is actually something like water cooling: small plate (big flange won't dissipate heat fast enough), copper pipes on plate leading the heat away to bigflange (tm) which has thin fingers that a good fan blows on. A big metal clump just doesn't get the heat away fast enough. After the heat is moved to another place you can cool it off at a moderate pace. A Peltier should actually be optimal for this kind of low temp sensitivity, but see this.)

About the spike: every time I've not gotten a boot-up, I've always Ctrl-A-A'd. Is there a second spike on reset?? Sounds weird. But yes, since most (all?) PSUs seem to share a limit for 5V+3.3V there is something going on there but I just can't explain why the reset doesn't work.

Be happy to mod a PSU according to instruction to test the spike thing. Just cut the brown wire going into one of the orange +3.3V pins?

BTW, got word back from Swe reseller of Euroquartz, they want $22 per oscillator. Hm... maybe I'll order one 95MHz if the PSU mod doesn't work out.

[roy bates]

Originally Posted by Photon

Agree that some PSUs go strangely low on 5V under load, maybe more while having a bigger load on 3.3V.

The CPU is really not a problem as you say, and agree it seems to have a more sensitive temp range even when you feel it should be able to handle "finger warm"... the CPU is not a problem because you just add a good active CPU/gfx PC cooler. It's only a problem for me because I can't fit such a cooler in my case

true,thats why my second one cant have the lid back on.(1200D)apart from the fact it has two siim sockets.(im quite fond of the idea of putting it in an a500 case)

Originally Posted by Photon

(I think what you're looking for is actually something like water cooling: small plate (big flange won't dissipate heat fast enough), copper pipes on plate leading the heat away to bigflange (tm) which has thin fingers that a good fan blows on. A big metal clump just doesn't get the heat away fast enough. After the heat is moved to another place you can cool it off at a moderate pace. A Peltier should actually be optimal for this kind of low temp sensitivity, but see this.)

the only problem with peltier"s is they have to get very hot,to get very cold.

Originally Posted by Photon

About the spike: every time I've not gotten a boot-up, I've always Ctrl-A-A'd. Is there a second spike on reset?? Sounds weird. But yes, since most (all?) PSUs seem to share a limit for 5V+3.3V there is something going on there but I just can't explain why the reset doesn't work.

this is where peaple get confused,you will not be able to do a "soft "reset
if the 5 volt rail drops below 4.85 volts,wouldnt you agree.
but then again if your getting a green dim power light thats something else normaly.

Originally Posted by Photon

Be happy to mod a PSU according to instruction to test the spike thing. Just cut the brown wire going into one of the orange +3.3V pins?

one would assume yes,but it is a conlusion drawn from the supplys i was using,im assuming that not all supplys are made equel.especialy silent ones.

Originally Posted by Photon

BTW, got word back from Swe reseller of Euroquartz, they want $22 per oscillator. Hm... maybe I'll order one 95MHz if the PSU mod doesn't work out.

you might not need to mod it, as the problems only occure at around 96-97mhz range,for some strange reason.

i must admit that the supplier i used years ago(for xtals) doesent seem to exist anymore(i suspect they closed shop),but if you dont mind paying the money it doesent matter as you only want one.its a shame as they where reletively cheap,and built them to any frequency.(plus they where freindly over the phone and very helpfull)

[Photon]

Actually I made both leds red to match the case, looks royal

At $22 a pop, I ordered a 95MHz one from the Swe retailer, let's see if they take the order.

Oh, I thought Peltier worked like fridges, ie. you give them juice and they lower the temp. Hmkay. I guess that's out then.

Anyway I'll cut the wire and see if there's a difference. That should make the 5V 5.00V+ stable then? I'm thinking it still won't work as I had 4 PSUs at > 5.00V after boot and it still wouldn't go fully into startup-sequence.

[zipper]

Originally Posted by Photon

Oh, I thought Peltier worked like fridges, ie. you give them juice and they lower the temp. Hmkay. I guess that's out then.

They do - but you have to move the heat somewhere...(as "bonus" you get bad efficiency with Peltiers, so still more heat)

[musojon74]

Yeah they heat exchange dont they? Feel around the back of your fridge...

[roy bates]

yes but they also generate even more heat in the transfer(upto THREE times)and use even more power,thats why to use them effeciently you would need water cooling.
as a side effect they also make condensation when they are not running.(straight after use)


hmm,i might make a water block and radiator,i can make one pretty easely.(just to play around with)

[Photon]

Since I just repaired some Joytech screens, I ditched the BYOC13S ('bring your own crap 13" screen' :P) idea and made the A1700 a portable.

Well, luggable. But it's only a bit wider than the A600 portable

At 0:45AM, CLOSED & All Cats Accounted For:


Supposed to show DPaint with Earth.hireslace in antialiased-interlace thunderstable 640x480... C1084S looks over its shoulder, like an envious bigger brother...

And >> Gallery pics

They're all as bad as this one but...

I simply had a thin but rigid Alu L-bracket that I painted matte black to match the screen. (Tried a silver Joytech LCD first and it looked really sharp on its own. But the black LCD just matched the matte black line on the A500 case so well that it just had to be a black one.)

I lined it up with 2 screw holes already present in the case, made some marks and drilled holes. Cut the 'overlap flange' of the lower case half and filed the top half and stuck the bracket between the halves.

Fitted the screen so it would line up and twisted the PS2 Slim attachment screws of the LCD against the bracket to make marks and drilled 2 more holes.

Glued 2 M3 nuts to the holes, just to hold them in place enough so I could use pliars and a screwdriver to fasten the LCD to the bracket, and done.

So now this baby is waiting for a 95 MHz oscillator (no, they never got back to me about the order of 1xoscillator...) and I'm also going to internalise some useful stuff like my Nokia speakers and maybe a headphone jack - just because I have the space, and to remove the need for bringing anything but A1700p + PSU to South Meet after Xmas.

[roy bates]

they dident get back to you?,have you paid for it online,if you have you should give them a ring and ask if it been dispatched.all i know is they have a minimum order.