68040 mhz
- Thread starter konrad
- Start date
I would expect a late mask 040 to do 50mhz with ease but make sure you stick a good heatsink on it.
Newertech's Clockometer, part of the Gauge series.
I've gotten E42K and 02E31F mask chips up to 57mhz on an accelerator. L88M/K63H around the same, 55mhz being the max I recalled. Earlier masks aren't reliable much past 33mhz under ideal conditions.
That said at these speeds the 040's bus drivers are well out of spec, there's no way you will attain anything close to these speeds on a complete system bus. If you have a normal fan and heatsink on the 040, that should be enough cooling for any speed it's able to do. Accelerators mostly have small buffered high speed busses so they can go a bit further without blowing timings. If you strapped the 040 into large buffer mode (greater drive strength) rather than the default small buffer mode you might get a hair more out of it, or removing extra RAM/ROM modules to reduce bus load.
Realistically, IMO, if you can manage a stable 45mhz on an 040 Mac I'd call it good.
For 030s, much past 60mhz gets wobbly on an unbuffered accelerator and 67mhz is the utter limit on a buffered accelerator. Same issues as the 040.
That said at these speeds the 040's bus drivers are well out of spec, there's no way you will attain anything close to these speeds on a complete system bus. If you have a normal fan and heatsink on the 040, that should be enough cooling for any speed it's able to do. Accelerators mostly have small buffered high speed busses so they can go a bit further without blowing timings. If you strapped the 040 into large buffer mode (greater drive strength) rather than the default small buffer mode you might get a hair more out of it, or removing extra RAM/ROM modules to reduce bus load.
Realistically, IMO, if you can manage a stable 45mhz on an 040 Mac I'd call it good.
For 030s, much past 60mhz gets wobbly on an unbuffered accelerator and 67mhz is the utter limit on a buffered accelerator. Same issues as the 040.
Has anyone ever achieved 50 MHz or more?
@zigzagjoe just said that he’d successfully tested up to 55-57mhz.
I was looking for information on the internet and I found this. http://centek.online.fr/atari/phenix/p_tech2.htm
According to this publication, the 040 cannot reach 50 MHz, is that true? Regards
False? I can’t see where in the article it says that. The article just talks about Motorola’s marketing scam, advertising speeds with PCLK rather than BCLK. Many of us do run 040 chips up to and in excess of 50MHz - for example I am runnng my LC475 at 50MHz.
Last edited:
Yup. And I overclocked my DayStar 68040 to 50 MHz.
Occasionally, DayStar 68K accelerators come up for auction on eBay. Sometimes the 50 MHz PowerCache 68030 card asking price will be higher than the sold price of some Turbo040 cards. Other times, you'll find both cards, but they'll be the slower versions. This post provides details on swapping out parts for a higher speed, and the overall performance of both cards at different frequencies.
I purchased a IIsi that included a DayStar P33 PowerCache 68030 card. The card was only at 40 MHz (rather than 50 MHz) and did not include the 68882 coprocessor.
There are three versions...
I purchased a IIsi that included a DayStar P33 PowerCache 68030 card. The card was only at 40 MHz (rather than 50 MHz) and did not include the 68882 coprocessor.
There are three versions...
IMO it's a little more nuanced than "scam". My observations are that the 68040 behaves like a clock doubled 68030. Oversimplifying - but still -overall performance lines up disturbingly well with the 030 if you look at it that way, and it fits nicely with the common reasoning that an 040 is about twice as fast as an 030 clock-for-clock.
The caveat with that argument is to run an 030 at peak performance demands too much of the rest of the system with the technology of the time. A segmented (buffered) bus with L2 cache is mandatory, and you *still* get hosed on writes unless you do something exotic with your DRAM. Aka, take a look at the IIfx and the resulting mess. The massively improved write back cache of the 040 allowed keeping the CPU appropriately fed with a more sedate (economical) system design. Of course, there were other improvements - namely the bus interface and FPU - but the CPU core itself I don't think was much improved beyond that.
My guess is it was more of a political discussion. At the time (1990), bus clock was synonymous with CPU clock speed.... while using the PCLK for advertising gives a nice "big number = better" it also means the IPC isn't much improved over the 030. If your "50mhz" 68040 is going head to head with a 25mhz 486 ... that's really not a good look. Think of Pentium 4's ridiculous clock frequencies but relatively poor real-world performance. Using the lower number implies that the IPC (and design) is much improved and perhaps also suggests that the chip might scale frequency further over the lifetime of the design than it otherwise might due to limitations. Just splitballing, though.
That article seems to be dissecting motorola's advertised instruction timings and I do think something weird was done with those, both with 040 and 060. In the 68060 thread I've a post disussing that. I'm still not entirely sure I didn't make a mistake somewhere, as these are complex chips with complex timings, but I'm pretty sure motorola was playing games with that data. Weither it was intentional or "simply" obfuscated by complexity is perhaps the question.
What I don't understand is that Motorola was manipulating data against the competition. I saw the image; it says approximately 50 MHz. Is that tool reliable? Thanks for listening.
Yes
