Full 68040 for Quadra 605

[Fizzbinn]

Thinking about the best way to upgrade my Quadra 605 cpu to a non-LC version that will support running at 33Mhz (overclocking seems like a not to difficult jumper re-soldering project)

I see see there are “XC” and “MC” versions of the full 68040 rated at 25 and 33Mhz:

http://www.cpu-world.com/info/id/Motorola-identification.html#68020__68030__68040__68060__68881_and_68882

The XC, “Pilot production device”, seems to be more common on eBay, the stock LC chip I currently have is an XC, should I not worry about that?

Are chips rated at/marked 33Mhz a nice to have or a must to modify the motherboard to run the CPU at 33Mhz? I wonder if this might be related to the XC vs. MC question? Perhaps MC parts are better suited to over clocking?

I’ve seen comments seeming to imply questionable authenticity and/or problems with Motorola 68k family chips on eBay sold from Asia, does anyone have any experience (positive or negative) or advice on this? Some examples of what I’m considering:

Motorola XC68040RC25B 09D50D, early date 9144, 68040, vintage CPU, GOLD

https://rover.ebay.com/rover/0/0/0?mpre=https%3A%2F%2Fwww.ebay.com%2Fulk%2Fitm%2F254183698130

Motorola XC68040RC25A, 04D50D, 68040, vintage CPU, GOLD

https://rover.ebay.com/rover/0/0/0?mpre=https%3A%2F%2Fwww.ebay.com%2Fulk%2Fitm%2F254220262657

Manufacturer:MOTOROLA MPN:MC68040RC33 EncapsulationGA

https://rover.ebay.com/rover/0/0/0?mpre=https%3A%2F%2Fwww.ebay.com%2Fulk%2Fitm%2F202379069510

1PCS MC68040RC33 MOTOROLA EncapsulationGA

https://rover.ebay.com/rover/0/0/0?mpre=https%3A%2F%2Fwww.ebay.com%2Fulk%2Fitm%2F113180285890

Thanks!

 

[johnklos]

Most m68040 CPUs in Macs are XC because Motorola took so darned long to "qualify" the '040 to make an official MC part:

Official MC mask announcement

Original XC68040s (D43B, D50D, D98D, E31F) run hot and are .8um masks, but will run fine at 33 MHz with a heat sink.

The E26A is an XC part with a .65um mask, which might be a little cooler.

Starting with the E42K mask, they are fully qualified MC68040 parts with .65um masks.

Then K63H reduced the mask to .57um.

Finally, the latest / last version is the L88M, which is also .57um, but runs cooler than the K63H according to some sources. This version is usually made by Freescale and has their logo in place of the Motorola M.

Some eBay sellers scrub the markings off of older m68040s and print newer markings on them to sell them for more money. Unlike the m68060 which has bits in a status register which identify clearly the revision of the CPU, there's no easy way to find out whether you have a real, proper, newer mask. If you have a good thermometer and/or can measure your 5 volt current draw, you can compare the old CPU with the replacement, but other than writing some tricky code to look for specific errata in certain models, there's no straightforward way in software to see what you've got, so try to go with sellers that have high seller ratings.

 

[johnklos]

BTW - all m68040s, with the only possible exception being 1990 engineering samples, were made from the same batches that made 40 MHz parts, so pretty much all of them can run at 40 MHz with good cooling, and all of them can certainly run with no issues at 33 MHz with a heat sink. Running a Quadra 605 or Performa 475 system at 40 MHz, though, requires a 20 MHz crystal oscillator and a wee bit more soldering.

 

[trag]

Running a Quadra 605 or Performa 475 system at 40 MHz, though, requires a 20 MHz crystal oscillator and a wee bit more soldering.

John, maybe you can answer this.  I've never understood it.   Why does the Q605 upgrade to 40MHz require the addition of a 20MHz oscillator.  Why is it not sufficient to just set the resistor pattern for the clock generator to the 20/40MHz setting?

 

[johnklos]

If I remember correctly, stuff (no idea what) won't run at the full 40 MHz. Don't know if memory is too slow, or if some of the peripheral chips won't run at 40 MHz, or what. So the idea is to run the motherboard at 33 MHz, then remove a surface mount resistor to disconnect a clock going to the CPU's clock generator (the MC88920), then wire the output of the 20 MHz oscillator to the pad which leads to the CPU's clock generator. It's outlined here:

Quadra 605 overclock to 40 MHz using crystal oscillator

I've done this on many Q605 boards, and they typically don't run at 40 MHz, according to Clockometer, but they run close - usually around 38 MHz. I don't know how this works this way, but it does

 

[Fizzbinn]

@johnklos Great info, just what I was looking for, thanks!

 

[trag]

If I remember correctly, stuff (no idea what) won't run at the full 40 MHz. Don't know if memory is too slow, or if some of the peripheral chips won't run at 40 MHz, or what. So the idea is to run the motherboard at 33 MHz, then remove a surface mount resistor to disconnect a clock going to the CPU's clock generator (the MC88920), then wire the output of the 20 MHz oscillator to the pad which leads to the CPU's clock generator. It's outlined here:

Quadra 605 overclock to 40 MHz using crystal oscillator

I've done this on many Q605 boards, and they typically don't run at 40 MHz, according to Clockometer, but they run close - usually around 38 MHz. I don't know how this works this way, but it does

Thank you for the wayback link.  I was looking for that page the other day and it's gone in the original.   I also have used that method successfully.   I just always wondered why the straight resistor swap failed.

Here's a thought.  I wonder if the original author of that article ever tried the resistor method of overclocking, after replacing the MC88920  with an MC88916DW.   It's possible that his original MC88920 wouldn't support distributing 40MHz, but did support 36 or 38.

I think this bears some experimentation.   My hypothesis is that the simple resistor method actually works, if the clock buffer will run at 20/40.  The clock buffer is the MC88920 or the MC88916DW, not to be confused with the clock generator.   The clock generator chip ( 343S1135 ) is the one that hte resistor swaps affect.  As different resistors are installed, it outputs different frequencies.    Presumably, R93 is in line with the output pin of the clock generator chip.

But that's all a hypothesis.   It needs testing.

John, have you done the clock buffer swap?   I ask, because I'm wondering if there's much difference between the MC88916DW70 and MC88916DW80 in practice.    I have a small supply of the latter.  I also have a reel (996 pieces) of MC88916DW55 I picked up for next to nothing.   It would be nice if they were like the 68040 and the speed markings weren't really all that limiting.

 

[johnklos]

John, have you done the clock buffer swap?   I ask, because I'm wondering if there's much difference between the MC88916DW70 and MC88916DW80 in practice.    I have a small supply of the latter.  I also have a reel (996 pieces) of MC88916DW55 I picked up for next to nothing.   It would be nice if they were like the 68040 and the speed markings weren't really all that limiting.

You know, one of my Q605 motherboards has a MC88916DW80. When I find that specific motherboard, I'll try that out. I bought, I think, half a dozen MC88916DW80 and only used one, so this would be good to know.

I find it interesting that because of the amount of time spent using the bus, my QuadDoubled machines running at 50 MHz with a 25 MHz memory bus are, for many uses, slower than a 40 MHz overclocked Q605. Good, solid, true 40 MHz overclocks would be very nice, particularly when I start to build my Q605 cluster

 

[Fizzbinn]

Fingers crossed one of the 2 68040’s I got here will work:

https://rover.ebay.com/rover/0/0/0?mpre=https%3A%2F%2Fwww.ebay.com%2Fulk%2Fitm%2F132803371629

XC68040RC25B

O50D 

MC68040RC33A

L88M

Sold as-is but will be a good deal if one (especially the 2nd one) actually works. 

 

[Elfen]

Love the subject here. My only concern is Overheating and how to deal with it. the LC case does not give enough space for a good properly sized heat sink to be put into place, and the LC Fan is something least to be desired.

 

[johnklos]

Hey, Fernando! Yes, heat is an issue if you're running at 40 MHz, but if you have an L88M or K63H mask, it's already cool enough. For an early mask (any XC part) at 33 MHz, you need a heat sink or you need to make sure the fan in the case works, and for an early mask at 40 MHz, you really need both the heat sink and fan in order to make sure the CPU doesn't overheat.

Most '040 sockets have clip mounts on the sides, and the heat sinks and clips that one can get from Quadra 650 / Quadra 800 machines fit fine inside of a Q605 / LC475 case. Or, when the socket doesn't have clip mounts, or when you don't have the right clip and heat sink, you can always use something like Arctic Alumina thermal adhesive epoxy to hold on a proper heat sink. It's pretty permanent (I've never tried to get a heat sink off after adhering it), but works well.

 

[Fizzbinn]

Fingers crossed one of the 2 68040’s I got here will work:

https://rover.ebay.com/rover/0/0/0?mpre=https%3A%2F%2Fwww.ebay.com%2Fulk%2Fitm%2F132803371629

XC68040RC25B

O50D 

MC68040RC33A

L88M

Sold as-is but will be a good deal if one (especially the 2nd one) actually works. 

Both work, not bad for $35! 

Both arrived with a bunch of bent pins, on the L88M (of course it would that one) one corner was basically flattened. They were packed well, likely why they were sold as-is. I heated them up with a heat gun and slowly bent them back up using tweezers and a mechanical pencil. With no lead the mechanical pencil tip slides right over the pins.  

I still plan to use a heatsink, even though with the LM88 I could probably get away without one at 33Mhz. I’m thinking one of these 45x45x10mm guys:

https://www.amazon.com/gp/product/B07PGNS2TX

with some thermal grease in the middle and some thermal double sided tape:

https://www.amazon.com/gp/product/B019MU1DTK

on on the edges, the epoxy seems a bit too permanent and I couldn’t find a heat sink with a mechanic clip that looked like it would work on a 68040.

 

[Fizzbinn]

Love the subject here. My only concern is Overheating and how to deal with it. the LC case does not give enough space for a good properly sized heat sink to be put into place, and the LC Fan is something least to be desired.

The fans in my LC III and Quadra 605 are really loud/noisy, with a cooler chip now I’m thinking about following the advice from this page:

http://quadra605.mcdonnelltech.com/upgrades/

“Installing a 150 ohm 1/4 watt resistor on the positive (red) lead of the case fan will slow the speed of the fan. This will result in considerably less noise while maintaining sufficient cooling”

 

[Byrd]

If it's a crappy fan, it'll still be crappy at lower RPM.  Perform a deep clean on the fan - blast out dust (ideally with air compressor), relube with good oil - if no better replace with another 60mm low RPM unit.

 

[johnklos]

Is it a 5 volt or 12 volt fan? I forget. Noctua has 5 volt and 12 volt available at that size:

https://smile.amazon.com/Noctua-NF-A6x25-5V-Computer/dp/B01K4HRLRI/

https://smile.amazon.com/Noctua-60x25mm-Blades-Bearing-Premium/dp/B009NQMESS/

 

[Forrest]

The power supply fan on my 660AV was making lots of noise. I opened the power supply case (3 screws) and attached a new: 80 x 80 x 20 mm, axial, ball bearing fan, 12V DC. I bought an NBM brand fan from www.mouser.com for just under $10. It's very quiet - no resistor needed.

The 660AV is the same size as the 610, and I'm guessing the power supply shares the same size fan.

 

[Fizzbinn]

The Quadra 605 internal layout is the same as the LCII/III (and Performa 405/450/475), the fan is 12v 100mA 60 x 60 x 15 mm. When pulled from the case clips mine doesn't sound all that loud, it seems to be resonating when held up tight against the metal shielding/bottom of the case by the two plastic clips. Hoping it would do so less if it was slowed down some.  May look at buying a replacement fan too if I can find one in the right size that claims to be quiet...

 

[Cory5412]

Thought: How much space is between the case and the fan? Could you add a rubber band or, like, some blue painter tape to close the gap a little bit?

A newer, quieter fan also couldn't hurt.

 

[Fizzbinn]

It’s really tight, I tired the painters tape idea and it didn’t seem to help. The plastic clips built into the case really hold the fan in tight on the sides too. Ended up buying some new fans to try:

Delta AFB0612LB-TXNS Brushless Tubeaxial Fan, 11.55 cfm

https://www.amazon.com/dp/B00R5C9XDQ/ref=cm_sw_r_cp_tai_236kDbEZHT69C

Gelid Solutions Silent 6 Computer Case Fan, 16 cfm https://www.amazon.com/dp/B008S1HNPS/ref=cm_sw_r_cp_tai_Ma7kDb83VYP2E

After some fun finding a connector that would work (https://rover.ebay.com/rover/0/0/0?mpre=https%3A%2F%2Fwww.ebay.com%2Fulk%2Fitm%2F182330584197) I was able to test them out. 

The Delta seems to blow more air but made the same level of noise installed as the original fan, the Gelid one seemed in between air flow wise (very scientific hold hand above the fan while running with the case top off “test”) and it was a bit quieter... but still not as quiet as I’d like. 

Next up the resistor in-line. With the Delta fan the airflow now seems less than the original but still decent, however the noise seems more than cut in half! Closing it up there is still some noise but much more inline pitch and volume wise more modern tech.  My spouse, who is usually nonplused by computer noises and would comment on the old noise level, agrees.