Q:Commodore A3640 tinkering - Watch out I have boing balls!

  • Thread starter Thread starter arnljot
  • Start date Start date
  • Replies Replies 83
  • Views Views 13073
Re: Q:Commodore A3640 tinkering - Watch out I have boing balls!

arnljot said:
Okay, I think I'm catching up now. Put me down if I'm not :-D

  1. When increasing the frequency of a Commodore A3640 it's preferable to underclock a higher rated CPU than overclocking a cpu(running a 040@40 @33mhz is preferable to running a 040@25 @33mhz both in terms of heat and performance). I get the heat bit, but not the performance bit.
    MIPS = Million Instructions Per Second. Only underclock in this instance cos it's cheap enough to do so. Thermal Heat dissipation will be down to the Revision of CPU used.[/*:m:14hv18pc]
  2. Say what ever you want about Commodore engineers but the A3640 is a piece of shit engineering?
    Yes[/*:m:14hv18pc]
  3. The card is the bottle neck when frequency is increased. If it runs at 40mhz, the motherboard will generate extra heat too (SuperBuster, Ramsey, Alice, etc. jfr da9000@a.org)
    No[/*:m:14hv18pc]
  4. If the card runs at 40mhz it will die faster (caps or other) Yes[/*:m:14hv18pc]


:-) Now, my boing balls are in line. I agree, lets do the 33mhz frequency.

Just let me know who I have to send the bits to. We might need to re-evaluate pricing too as I have a rather sexy Blue LED Fan/Heatsink combo which I can supply with the CPU & X-Tal.
Tbh, I'd just solder the X-Tal straight onto the A3640. Mine is not sweet enough to be used in a socket. It's been on & off things more times than a wh0res drawers have been up & down! :|

Kin
 
Re: Q:Commodore A3640 tinkering - Watch out I have boing balls!

I have inspected the caps (105, 106 and 107) closely now. None of them are leaking, but all three are mounted the wrong way around.

I've also asked Anthony Hoffman if he has a v3.2 u209 (pre-programmed Lattice GAL22V10D-10LJ GAL).

C102B and C102E are hard to inspect since they are beneath the heatsink. But they appear to be fine too.
 
Re: Q:Commodore A3640 tinkering - Watch out I have boing balls!

Kin Hell said:
Also, The Mobo will be fine because the timing is sent to the card & as da9000 already concurs, the architecture of the A3640 is very dodgy at these kind of speeds.

No, no. AFAIK, the EXT jumpers say: get me the clock for the motherboard from an external source. The source SENDS the clock to the mobo. That source happens to be the A3640. So if that card is running with a 66Mhz crystal oscillator, the motherboard is getting 33Mhz to drive the custom chips. Which means they'll be overclocked, so the motherboard won't be "fine", unless those chips are cooled.

@arnljot:
To make this brief: if you're going to overclock it, which I think is OK as long as it's done right, then do it properly:
1) cool the CPU
2) cool the A3640 (chips on it get hot), also replace SMD caps
3) cool the custom chips and any other motherboard chips that get hot, I'd also replace the SMD caps

Don't expect miracles, but a ~37Mhz 040'ed A3640 will allow you to play Doom nicely ;-)

And anything that doesn't need high memory bandwidth (3D games and such do, mp3 player does not), will run quite happily, while at 25Mhz it wouldn't have.
 
Re: Q:Commodore A3640 tinkering - Watch out I have boing balls!

@ da9000

'Put the gun down' - 'Put the gun down'. :mrgreen:

:Doh: - Someone smoke me a Kipper & I'll be back for breakfast.........

The Mobo always takes it's speed from the Crystal on the Mobo.

The INT/EXT jumpers merely tell the Amiga which Speed crystal to use for the CPU/FPU, NOT the Chipset! :wink:

Kin
 
Re: Q:Commodore A3640 tinkering - Watch out I have boing balls!

Kin Hell said:
The Mobo always takes it's speed from the Crystal on the Mobo.

The INT/EXT jumpers merely tell the Amiga which Speed crystal to use for the CPU/FPU, NOT the Chipset! :wink:

Kin

I absolutely agree with Kin.

The Mobo conatains 2 Oscillators
1 - 50MHz which is used in "INT" mode to drive the 3620/30 CPU card which has no oscillator on board or "EXT" which means the relevent CPU Slot pins will hold the clock generated by the CPU card's on board oscilator, for ex, the A3640.

2 - 28.37516MHz (for PAL) this Oscillator is used for the chipset timing & is independant of the CPU oscillator.

I think the speed limitations of the A3640 are mainly down to the card's board design & the speed that some of the devices can operate at, the delay line does not help either as it fixes the range but by altering the delay line (as rkauer suggests) higher speeds can be attempted but failure or stability maybe compromised due to crosstalk & noise problems due to the cards track layout..etc.

TC :mrgreen:
 
Re: Q:Commodore A3640 tinkering - Watch out I have boing balls!

Also, the (again) MC68040 will not produce any heat (true: cooler than a 060!). The XC versions are another matter.

The NTSC versions have a motherboard clock of 28.8383 (NTSC clock). Good for people who converts a NTSC unit and reuse the crystal on some old GVP 030 @25MHz :wink:

Any "overclock" over 33MHz (even "under clocks"!) on the A3640 must have the mod made, even to run at 34Mhz.

But in a better world you will swap the PAL chips on this pure crap board to GAL units. I have more details on how to make it, not pertaining to the thread.

The defence rest, M'lord.
 
Re: Q:Commodore A3640 tinkering - Watch out I have boing balls!

My head is spinning and I have vertigo.

This is why we at my job normally don´t let the engineers correspond per email with the client on CC. We usually gather in a room with a whiteboard and then have the thing hammered out within 0.5h

:lol:

This is poetic justice I guess, or pathetic. Either way, I´m totally lost :-D
 
Re: Q:Commodore A3640 tinkering - Watch out I have boing balls!

Where is the A3640 card atm?
 
Re: Q:Commodore A3640 tinkering - Watch out I have boing balls!

Ah, someone assertive - I respond well to that.

It is in an antistatic bag, inside a box. In the living room. Next to a huge pile of Amiga gear.

In prioritized order, this is what it needs done:
  1. Changed and corrected caps[/*:m:3qhn5reg]
  2. Upgraded to v3.2 Anthony Hoffman has given me a good deal on that[/*:m:3qhn5reg]
  3. Underclock and activly cool. 33mhz as my boing balls have shrunk.[/*:m:3qhn5reg]

Now, how this will be done I have no idea atm.
 
Re: Q:Commodore A3640 tinkering - Watch out I have boing balls!

Okay Arnljot. Is the card slik Screened as a 3.1 (Along the CPU connector edge) with GAL chip U209 ending - 02? (Top Left area of card with CPU Connector nearest you)

For reference Pic, see the Pic here on TC's Thread....

Linky

In the Pic, U209 Gal chip is the one in the bottom right corner of pic. That ends -01 & is normal for a Rev 3.0 A3640.
A Rev 3.1 A3640 is silk Screened 3.1 by the CPU Connector but U209 will end -02.
A Rev 3.2 A3640 is Silk Screened 3.1, but U209 will end -03!

The other GAL Chips might also be different revisions depending on the overall revison of the A3640 & Anthony Hoffman is probably your best bet for the Tech! :wink:

Kin
 
Re: Q:Commodore A3640 tinkering - Watch out I have boing balls!

We've covered this, it's a v3.1 card most likely.

arnljot said:
I have no idea what version it is. It has no sticker on U209.

It just reads:
TI
22V10-10CFN
40307XWAF

Is there any software way to tell?

Also here:
arnljot said:
rkauer said:
If you find a MC68040 (not the fry-an-egg XC68040), you can overclock even the 25MHz over 40MHz without issues (on 3.1 and 3.2 boards).

Ah, crap. I've only found an XC - I'll continue to scout for an MC then.

rkauer said:
Your board don't have any sticker?
U401, U208, U207, U213, U203 and U400 have stickers. They all say "*-01" and stuff like "exlat", "term" "mapper", "rst", "lebus" and "oebus".

rkauer said:
Also there is the version printed on the board, near the socket.

Aha! I do have a 3.1 board :)

So: For some reason the sticker on my U209 has been removed. So the text on it not indicative of the version. But I think it's safe to assume it's a v3.1 card. It could be version 3.2 but that is unlikely. Just as it's unlikely it's v3.0, since it's silkscreened v3.1
 
Re: Q:Commodore A3640 tinkering - Watch out I have boing balls!

Great stuff Arnljot. The label probably came off with Age & heat. They literally curl up with age!

I guess you dont want my 40Mhz XC 040 to run it cooler @ 33Mhz! :?:

Heres a Pic of a 3.2 - TC owns this now! - Any Help for you?

You're only concerned with the numbers after the dash. (-) :wink:

& as rkauer so elequantly puts it, GAL are better than PAL even though my old Rev 3.2 has PAL chips fitted! :lol:

Kin

*edit*

The Pic rkauer shows with a jumper wire is NOT needed, though I'm sure someone with the tech. knowledge can explain why?
 

Attachments

  • IMG_0026_resize.JPG
    IMG_0026_resize.JPG
    111.4 KB · Views: 3
Re: Q:Commodore A3640 tinkering - Watch out I have boing balls!

I still want it Kin Hell. No changes on that part of the plan. We are going to rise it from 25mhz to 33mhz this board. But I do want a 40mhz piece then.

But I need to know from Zetr0 if he also can do caps and a new GAL for this. If not I'll have Anthony Hoffman do it. It just changes logistics.
 
Re: Q:Commodore A3640 tinkering - Watch out I have boing balls!

Kin Hell said:
<SNIP>

& as rkauer so elequantly puts it, GAL are better than PAL even though my old Rev 3.2 has PAL chips fitted! :lol:

Kin

*edit*

The Pic rkauer shows with a jumper wire is NOT needed, though I'm sure someone with the tech. knowledge can explain why?

OK, more technical info: the delay line acts as a filter, but impends the boards to run in higher frequencies than the original 50MHz bus. The A3640 design uses delay lines for part of it's clock circuitry, tying up frequencies. So eliminating this specific delay will free the board for some time constrains.

Notice that some 3640 will not work with at 33MHz (66MHz crystal) if the cut-and-join hack is not done. But those boards will work at 30MHz if a 60MHz crystal is fitted, instead (no hack this time).
 
Re: Q:Commodore A3640 tinkering - Watch out I have boing balls!

@rkauer

When you say some, do you mean regardless of revision or do you mean some pre v3.2?

Also, can one determine which ones by inspecting them or is it Lotto?
 
Re: Q:Commodore A3640 tinkering - Watch out I have boing balls!

Butting in, my 1992 Rev 3.1 had a 66Mhz Crystal Mod without rkauers explanation of the Wire mod & was 101% Rock Solid. Active cooling was added though! ;)

Kin
 
Re: Q:Commodore A3640 tinkering - Watch out I have boing balls!

arnljot said:
@rkauer

When you say some, do you mean regardless of revision or do you mean some pre v3.2?

Also, can one determine which ones by inspecting them or is it Lotto?

The mileage may vary... Some 3.1 boards accept a 66MHz oscillator without the hack (like Kin's one), others only go up to 60MHz without hack. With the wire bridge-and-cut job even 37MHz (depending on the 040 version) and some rare boards can go up to 40MHz (80MHz crystal!).
 
Re: Q:Commodore A3640 tinkering - Watch out I have boing balls!

I'll be doing the "safe thing" then. I'll be going for 33mhz. And it's going to be done by Anthony Hoffman in NZ. He'll also upgrade it to v3.2
 
Re: Q:Commodore A3640 tinkering - Watch out I have boing balls!

@ rkauer

Thanks for the Technical info there m8, really apprecited! :thumbsup:

@ Arnljot

Last PM from you was kinda joking/hinting @ the Cooler I have here for a few bucks more & we haven't discussed further!
I expect it should pull off the 060 CPU okay but I dont have a working socket to fit your 040 in other than a knackered CSMKIII that has a reworked and rebuilt socket array. Whilst any CPU in there does still get warm, it's an absolute sod of a job getting any CPU into the rebuilt socket due to the slightly irregular pin socket alignment. I did swap out 060's to take the CPU being dead out of the equation but it was a 2 hour event! :wooha:

If I pull the cooler off the 060, it leaves a sticky Thermal Pad on the Heatsink which I can not protect once removed. If I could drop the 040 into my dead CSMKII, I'd have to desolder the Voltage regulator Resistor on the CSMKIII in the home built socket in order to Juice the 040, thereby getting the Thermal pad warm in order to get it stuckk on for real. All these bloody new Icons & not a single "Flogging a Dead Horse" smiley anywhere! :stick:

I might be able to find a bit of plastic somewhere to press on for protection for shipping to Anthony, but it could all be in vain anyway. I really need to fire the 060 in a socket to warm it for removal which could make the whole frigging event a 4 hour job! :tired:

Whatchya wanna do? - Buy some thermal Pads from somewhere? - I'm fresh out & wanna do the right thing for you! :Doh:

Beer time! :beer:

Charlie
 
Re: Q:Commodore A3640 tinkering - Watch out I have boing balls!

@Kin

I've PM'ed you re the cooler and pad.

Also, I'm seriously starting to consider your CR card. I'm going to look in previous correspondance to see if it also has a battery. :)
 
Back
Top Bottom