Overclocking 604 Processor Cards

[Phipli]

I thought I'd just post this info because I've been looking at it this morning and while I have the datasheet out and heatsinks off...

The following table shows the bus multiplier settings for the 604 and 604e. Note "PLL_CFG" 0 through to 3.



These PLLs are represented by 4 pairs of resistor pads on the processor cards. The correspond as follows (shown for two card variants, older first, then newer).




The first card shows a 3x, the second is 5x.

R1, R2, R3 and R9 represent a '1' in the table. R6, R7, R8 and R13 represent a '0'.

The first board shown (dark green) is a 150MHz 604, with a 50MHz bus. To bump it up to 175MHz, you would move R7 and R8 to the positions R2 and R3.

To change the bus speed you need to replace or override the big tin can clock on the board. This is easy enough to do as they usually have an output disable pin, that you can use to stop the soldered clock, before injecting your own clock.

Sadly these machines sometimes won't run much past 50MHz. I understand some even struggle with 50MHz.

 

[Snial]

I thought I'd just post this info because I've been looking at it this morning and while I have the datasheet out and heatsinks off...

The following table shows the bus multiplier settings for the 604 and 604e. Note "PLL_CFG" 0 through to 3.

View attachment 59584

These PLLs are represented by 4 pairs of resistor pads on the processor cards. The correspond as follows (shown for two card variants, older first, then newer).

View attachment 59585
View attachment 59586

The first card shows a 3x, the second is 5x.

R1, R2, R3 and R9 represent a '1' in the table. R6, R7, R8 and R13 represent a '0'.

The first board shown (dark green) is a 150MHz 604, with a 50MHz bus. To bump it up to 175MHz, you would move R7 and R8 to the positions R2 and R3.

To change the bus speed you need to replace or override the big tin can clock on the board. This is easy enough to do as they usually have an output disable pin, that you can use to stop the soldered clock, before injecting your own clock.

Sadly these machines sometimes won't run much past 50MHz. I understand some even struggle with 50MHz.

Curious. Can the same thing be done with a standard 117MHz 603e on a PB1400, perhaps to hike it to a cacheless 133MHz 603e?

Also, I think you posted a link to a massive compendium of technical manuals for Classic and later Macs recently? I meant to download it, but couldn't think of the right term to search for it.

 

[Phipli]

Curious. Can the same thing be done with a standard 117MHz 603e on a PB1400, perhaps to hike it to a cacheless 133MHz 603e?

Probably, I haven't looked how they're set up. Have to admit I don't tend to overclock laptops because of heat issues.

Also, I think you posted a link to a massive compendium of technical manuals for Classic and later Macs recently? I meant to download it, but couldn't think of the right term to search for it.

Do you remember what topic was being discussed? Was it just the Inside Macintosh Manuals or something else?
https://vintageapple.org/inside_o/

 

[Snial]

Probably, I haven't looked how they're set up. Have to admit I don't tend to overclock laptops because of heat issues.

Sure, but in the case of the PB1400, the 133MHz (and 166MHz I believe) CPU upgrade fits on the same connectors with the same heatsink, with the same external bus speeds (37.5MHz?), so I would have thought it'd work out. Actually, it should be less heat than, say a proper 133MHz CPU, because there's no L2 cache. Admittedly, it probably wouldn't get a lot faster: 133/117.5=1.13x, so maybe you'd only get an 7% improvement? 166/117.5 = 1.41, so maybe you'd only get a 20% improvement.

Do you remember what topic was being discussed? Was it just the Inside Macintosh Manuals or something else?
https://vintageapple.org/inside_o/

Oh, I remember you saying it gave all the pinouts for various connectors.

 

[Phipli]

Sure, but in the case of the PB1400, the 133MHz (and 166MHz I believe) CPU upgrade fits on the same connectors with the same heatsink, with the same external bus speeds (37.5MHz?), so I would have thought it'd work out. Actually, it should be less heat than, say a proper 133MHz CPU, because there's no L2 cache. Admittedly, it probably wouldn't get a lot faster: 133/117.5=1.13x, so maybe you'd only get an 7% improvement? 166/117.5 = 1.41, so maybe you'd only get a 20% improvement.

Its doesn't usually work like that - an overclocked chip usually produces more heat than a chip rated for that speed.

Oh, I remember you saying it gave all the pinouts for various connectors.

Thats a specific book

https://vintageapple.org/inside_o/pdf/Guide_to_Macintosh_Family_Hardware_2nd_Edition_1990.pdf

 

[cobalt60]

Awesome info. Do you know how many different card variants Apple used?
I like the 604 series. I was even thinking about building my 7300 into a 604ev system, and overclocking a bit. Would be fun trying to meet the higher power and cooling demands in the smaller case.

Sadly these machines sometimes won't run much past 50MHz. I understand some even struggle with 50MHz.

I remember reading something like this years (decades?) ago, but recently @Coloruser replied to me here with the following quote:

Checking my notes from the lab days indicates that 80% of the TNTs ran stable at 60mhz while less than 50% of the Tsunamis did that.

So is it Tsunami boards you were referring to, or is your experience that also the TNT boards don't go much past 50MHz? Have you noticed differences among the various TNT/Nitro/Kansas boards?

 

[Phipli]

Awesome info. Do you know how many different card variants Apple used?

At least 13 speed variants. Way more part numbers.

I like the 604 series. I was even thinking about building my 7300 into a 604ev system, and overclocking a bit.

Careful - the Apple manufactured 604ev cards use a different pinout to the older 604 and 604e cards. They shouldn't be used with a 7300. They're only for the later Kansas 8600s and 9600s (250MHz and above).

There are some third party 604ev cards that do work in older machines though.

So is it Tsunami boards you were referring to, or is your experience that also the TNT boards don't go much past 50MHz? Have you noticed differences among the various TNT/Nitro/Kansas boards?

I don't have any clones. I don't tend to overclock to the limit because it leads to instability on hot days or after long use. Personally I'd not put a machine past 50MHz. If I did I'd be putting all fast EDO memory in it and doing some serious testing. I'd push until it crashed, and then clock it back down a load to make sure it stayed stable.

There is no fun in a fast machine that randomly crashes.

 

[cobalt60]

At least 13 speed variants.

I meant like PCB configurations. You showed 2, and then I believe the 604ev are different. So that's at least 3. Basically wondering how many different PCBs I would ever need to know the locations of the PLL CFG jumpers.

Careful - the Apple manufactured 604ev cards use a different pinout to the older 604 and 604e cards. They shouldn't be used with a 7300.

Yep, I've got an 8600/250 (604ev) board that looks like a direct swap (ignoring PSU anyway) into the 7300.

 

[Phipli]

Yep, I've got an 8600/250 (604ev) board that looks like a direct swap (ignoring PSU anyway) into the 7300.

Don't!

Just because something physically fits doesn't mean it is electrically compatible - I said the pin out is different, not the connector is physically different.

Do not plug that processor into your 7300!

 

[cobalt60]

I mean I have an entire 8600 Kansas motherboard.

 

[Phipli]

I mean I have an entire 8600 Kansas motherboard.

That makes more sense - I thought you meant a processor card. They're both boards.

 

[cobalt60]

That makes more sense - I thought you meant a processor card. They're both boards.

Yes, I should have just added those few extra letters so there'd be no ambiguity.

I think it'd be fun to build oddball systems into 7500 type cases. Another that'd be cool is a dual 604e (again, would need PSU and cooling considerations). But I really want to do a 604ev, as I've heard even the 604e was faster than the G3 in certain tasks, and with the 604ev having a cache setup more like (or identical to?) a G3, I'd really like to do some benchmark comparisons there.

 

[Phipli]

But I really want to do a 604ev, as I've heard even the 604e was faster than the G3 in certain tasks

The G3 outperforms the 604ev clock for clock by quite a bit in general use. Its why Apple dropped the 604 line suddenly, and development of a more advanced 604 based chip was dropped too. Apparently, a 266MHz G3 is nominally the same or faster than a 350MHz 604ev (the fastest Apple sold). I should have benchmark data somewhere to check.

The G3 has impressively fast integer performance.

, and with the 604ev having a cache setup more like (or identical to?) a G3, I'd really like to do some benchmark comparisons there.

I believe... the 604ev had more execution units, more advanced FPU and supported Multiple processors.

The L2 cache was not like the G3 cache. The G3 introduced the "backside cache" configuration with the cache running at about half the CPU speed in Apple designs.

The cache on 604ev designs was still at bus speeds I believe, even though they moved it onto the processor card.

 

[Coloruser]

Kansas used double the bus speed for the cache. Thus the cache rus at 100mhz vs. Tsunami‘s 50 Mhz. And yes, despite the FPU, the G3 was overall faster than the 604ev In real world Mac OS use.

 

[Phipli]

Kansas used double the bus speed for the cache. Thus the cache rus at 100mhz vs. Tsunami‘s 50 Mhz

Ah excellent, thanks for the correction - I should have looked it up instead of trying to remember. That would be quite an improvement, although not G3 backside cache level improvement.

If the cache fails on a G3 the difference it makes is shockingly huge, it in the region of halves performance.

 

[cobalt60]

The first card shows a 3x, the second is 5x.

The image you posted is for a 603e. I believe the highest multiplier for the 604e is 4x. I bought a 240/60 604e card hoping to run it at 250/50, but sadly it seems the max multiplier is 4x.

 

[Phipli]

The image you posted is for a 603e. I believe the highest multiplier for the 604e is 4x. I bought a 240/60 604e card hoping to run it at 250/50, but sadly it seems the max multiplier is 4x.

View attachment 65633

I uploaded the table from the wrong datasheet sorry, I was cross referencing the 603 and 604 at the time. Here is a 604e equivalent. The multiplier table is identical where they overlap.


Not sure why your document says only 4x. Guess it was older.

The 9600/233 has a 46.6MHz bus and a 233MHz CPU, so 5x multiplier. That is what my second photo of resistors shows.



Note the clock and CPU speeds, plus the multiplier of 1011. This is a 604e (i.e. not a 604ev). This is the fastest variant before apple switched to the 604ev with the 250MHz 8600 and 9600, which used a revised logic board, I believe for increased power delivery to the CPU card.

 

[cobalt60]

The 9600/233 has a 46.6MHz bus and a 233MHz CPU, so 5x multiplier. That is what my second photo of resistors shows.

Oh wow, well, I guess I'll give modifying my Power Computing unit a try. I already have the crystal oscillator. Are you familiar with Power Computing 604e modules?

Would you mind sharing a link to that datasheet?

 

[cobalt60]

If I had to guess, Id say its R1 to R4, corresponding to CFG0 to CFG3

 

[Phipli]

View attachment 65673If I had to guess, Id say its R1 to R4, corresponding to CFG0 to CFG3

Just bridge R2 with solder, leave the clock alone and it should start at 270MHz. It's only a 12.5% overclock, so will likely work fine.