Good OSes for your 68k Mac

[CC_333]

Need to remember that the 601 is better than the 603, clock for clock. So a 117MHz 603 is less good than a 100MHz, perhaps even 80MHz or lower, 601.

Yeah, but see, at least in the context of the Power Macs, that was largely because the 603 lacked sufficient cache (8k vs the 601's 32k) for Apple's 68k emulator to run efficiently, and since the initial PPC releases of System 7 were mostly 68k code running in that emulator (indeed, for the initial release (7.1.2) I believe the emulator itself was the only thing that wasn't purely 68k), overall performance ended up being pretty mediocre, especially on the 603.

The 603e rectified this by doubling the L1 to 16k.

c

 

[Phipli]

Yeah, but see, at least in the context of the Power Macs, that was largely because the 603 lacked sufficient cache (8k vs the 601's 32k) for Apple's 68k emulator to run efficiently, and since the initial PPC releases of System 7 were mostly 68k code running in that emulator (indeed, for the initial release (7.1.2) I believe the emulator itself was the only thing that wasn't purely 68k), overall performance ended up being pretty mediocre, especially on the 603.

The 603e rectified this by doubling the L1 to 16k.

c

Hum, not entirely - I meant the 603 family. Compare benchmarks for a 200MHz 603e with a 100MHz 601 scaled to the same MHz.

I wasn't talking about the 68k thing. I don't own a 6200.

I'm not criticising the 603 family, it is just the bottom end vs the 601 and 604, and the post I responded to suggested that the 601 was somehow less good than the 603.

 

[Phipli]

Yeah, but see, at least in the context of the Power Macs, that was largely because the 603 lacked sufficient cache (8k vs the 601's 32k) for Apple's 68k emulator to run efficiently, and since the initial PPC releases of System 7 were mostly 68k code running in that emulator (indeed, for the initial release (7.1.2) I believe the emulator itself was the only thing that wasn't purely 68k), overall performance ended up being pretty mediocre, especially on the 603.

The 603e rectified this by doubling the L1 to 16k.

c

Here is an odd comparison, but one I have to hand on my phone. The MDK software graphics benchmark. This is PPC native, raw pixel pushing. Not Quickdraw accelerated, or anything else accelerated.



If we convert all scores to 100MHz, we get..
601 : 35
604e : 55.2
603e : 33.5

This particular test the 601 just edges out the 603e, but in other workloads, the 601 has even more of a margin. Especially floating point I anticipate.

Sorry I don’t have more figures on hand. I just woke up and have somewhere to be.

 

[pizzigri]

well I had I actually thought that the whole point was performance per megahertz per power consumption ratio - i mean, there are many examples in the history of microprocessors in which basically a newer CPU is less efficient than the older but at the same time has a huge increase in megahertz and better power handling and gives more performance per watt, and therefore you now have a vastly more powerful cpu.
what happened with, like, well the g5.... which could not go further than that.... vs intel

 

[CC_333]

Ah, I see. I guess I misunderstood.

c

 

[Snial]

well I had I actually thought that the whole point was performance per megahertz per power consumption ratio - i mean, there are many examples in the history of microprocessors in which basically a newer CPU is less efficient than the older but at the same time has a huge increase in megahertz and better power handling and gives more performance per watt, and therefore you now have a vastly more powerful cpu.
what happened with, like, well the g5.... which could not go further than that.... vs intel

Typically ARM measured their processors in terms of DMIPS (Dhrystone MIPs) / Watt. Because ARM CPUs were so efficient, that always made them look good against both 8-bit CPUs and CISC 32-bit CPUs.

I'm not sure the G5 couldn't go further than that, as I see it, it's really just that Intel had a huge development team compared with the AIM alliance so they were able to out-compete Apple though sheer brute force. The same applies to the new Apple Silicon Macs (and I'm typing this on a MacBook Air M2): Apple probably has fewer people working in the design teams than Intel, but it's easier to scale ARM architectures, so they're winning.

If we convert all scores to 100MHz, we get..
601 : 35
604e : 55.2
603e : 33.5

This particular test the 601 just edges out the 603e, but in other workloads, the 601 has even more of a margin. Especially floating point I anticipate.

I think this is also reflected in SpecInt95, SpecFP95 scores:
601 @66MHz 1.69 or 1.82,
603e @166MHz = 3.62 or 3.94 (1.43 or 1.56 at 66MHz)
604e @ 200MHz = 6.91 or 7.22 (2.28 or 2.38 at 66MHz).
But also, clearly the 603 architecture had a lot going for it as it became the basis for both the G3 and G4 CPUs.

 

[Phipli]

But also, clearly the 603 architecture had a lot going for it as it became the basis for both the G3 and G4 CPUs.

Oh, absolutely it had really good power characteristics - perfect for laptops, and the G3 with its backside cache made desktops that lightly sipped power and ran incredibly fast.

But, the 603s had lower IOPS than the 601 was my only point.

 

[CC_333]

Why wasn't the G3 based on the 604 instead? At least as far as IOPS go, the 604 was clearly better! And it was multiprocessor capable, not that it mattered much at the time, seeing as only a select few applications – not even the OS! – actually utilized the second processor.

Had Apple bought BeOS and used it as the basis for what eventually became Mac OS X instead of NeXTStep, things would've been much more interesting in this regard, as BeOS's support for multiprocessing was very robust, and from what I've read, it ran very well on 604-based Power Macs, particularly those equipped with dual CPU cards....

c

 

[Snial]

Oh, absolutely it had really good power characteristics - perfect for laptops, and the G3 with its backside cache made desktops that lightly sipped power and ran incredibly fast.

But, the 603s had lower IOPS than the 601 was my only point.

Why wasn't the G3 based on the 604 instead? At least as far as IOPS go, the 604 was clearly better! And it was multiprocessor capable, not that it mattered much at the time, seeing as only a select few applications – not even the OS! – actually utilized the second processor.

It's a good point. The more I look at it (via the https://arstechnica.com/features/2004/10/ppc-2/ articles), the more I think that the 750 was a good choice in the short-run, but a poor choice longer-term. It looks like it took a lot from the 603e (i.e. basic pipeline) a few things from the 604 (BHT and Branch each) and addd an extra simple Integer unit.

Hypothesising on the rationale:

• Apple wanted a single successor that would work for desktops and laptops, so the 603e had to be the choice (much cheaper to fund one CPU than 2).
• The 603e had more scope for improvement, because it was a 4-stage pipeline (diminishing returns already with the 604's 6 stage pipeline).
• The 603e was simpler to build on, so it provided a quicker timer to market.

Having said that, yes, the 604(e) was much more elegant. Admittedly, I've never used a 601 / 604(e) Mac (hmmm, maybe I had access to an 8500 at one point), merely the 603(e), 750 and my iMac G5. Of all of them, I'm more of a 603e fan for the relative simplicity of the architecture.

Had Apple bought BeOS and used it as the basis for what eventually became Mac OS X instead of NeXTStep, things would've been much more interesting in this regard, as BeOS's support for multiprocessing was very robust, and from what I've read, it ran very well on 604-based Power Macs, particularly those equipped with dual CPU cards....

I think we were all v excited by BeOS, more so than Danger-man Steve Jobs at the time!

 

[werdna]

Has anyone ever released a non-Apple OS that works with 68000-based Macs?

 

[cheesestraws]

Has anyone ever released a non-Apple OS that works with 68000-based Macs?

NetBSD.

 

[werdna]

NetBSD

Was NetBSD ever released for the original 68000? It seems like it requires at least an 030.

By 68000 I just mean 68000 - the Plus, SE, Classic, etc.

 

[twelvetone12]

I used to run Debian Linux on my quadra 700, with X and all.

 

[stepleton]

By 68000 I just mean 68000 - the Plus, SE, Classic, etc.

If you count the Macintosh XL (i.e. the Lisa) in that list of 68000-powered Macs, there were several OSs: at least two Unixes, the Lisa Office System and Workshop of course, and the Monitor.

 

[cheesestraws]

Was NetBSD ever released for the original 68000? It seems like it requires at least an 030.

Oh, I misread your original post, sorry!

Someone did get EmuTOS working, though I don't know if it was tested on real hardware or just in emulation.

 

[werdna]

EmuTOS working

Looks like just barely in Mini vMac!

If you count the Macintosh XL (i.e. the Lisa) in that list of 68000-powered Macs, there were several OSs: at least two Unixes, the Lisa Office System and Workshop of course, and the Monitor.

Yes, of course; don't forget GEMDOS, which if I remember correctly runs on the Lisa with about as much hardware support as EmuTOS has on the Plus (almost none, lol). I mean't not to count the Lisa. Aside from EmuTOS, apparently, the closest thing I've ever seen to another OS for '000 Macs is Macintosh Servant, which obviously doesn't count as it's just a Finder replacement.

 

[cheesestraws]

GEMDOS, which if I remember correctly runs on the Lisa with about as much hardware support as EmuTOS has on the Plus (almost none, lol).

GEMDOS on the Lisa supports quite a bit of the hardware. Are you thinking of EmuTOS on the Lisa?

 

[werdna]

GEMDOS on the Lisa supports quite a bit of the hardware. Are you thinking of EmuTOS on the Lisa?

Maybe I'm getting mixed up then, it's been a while since I've looked into it

 

[cheesestraws]

Maybe I'm getting mixed up then, it's been a while since I've looked into it

I'm moderately confident in talking about GEMDOS' hardware support on the Lisa.

Unwinding the stack a level, fairly sure that MacMinix supports 68000 machines, although this may not count as it uses MacOS for I/O

 

[ArbysTPossum]

I've done a lot of tests on my own machines. 030 and 040 machines use 7.5, 603 machines use 8.6.
I forget what my G3 uses, but G4 is dual boot 9.2 and 10.4/10.5