Metropolitan legend or esoteric spec?

[Framiga]

Hi all,

i've heard some time ago, a guy (known as a technician) told me that a 68060@50 Mhz CPU, works internally at 100 Mhz so like a 68040@25 Mhz works internally at 50 Mhz.

So we all know, that a 060 has a clock oscillator at 1:1 ratio while a 040 has 1:2 one.

Any comment?

Ciao

[Framiga]

Wow! a lot of answers!!!

In datail:

A couple of years ago, i've downloaded from Aminet, a program (don't remember wich).

In the readme file, there was the config of the author that said:

A2000 with A2060@100 Mhz.

From this info, i've done the question in the previous post.

Have you ever heard something similar?

Ciao e grazie

[Dan]

I knew that the 040 works with double speed internally but I don´t know about the 060.
I went looking for a amigamagazine that had an article about the 060 when it was new but i couldn´t find the magazine.

Does aybody know?

[Framiga]

Take a look here:

http://www.emulation.com/catalog/off-the-shelf_solutions/production-test_adapters/upgrade_motorola/

$ 296.40 ea. is a little bit expensive!

Ciao

-edit- Yes, you were right!

All MC68040 are clocked doubled processors. Thus a advertised 25MHz 68040 processor has actually an internal clock speed of 50MHz and an external clock speed of 25Mhz.

[Elektro]

what do you mean internally / externally?

[Framiga]

The MC68040-25 works inside itself (internally), at 50 Mhz if  externally clocked with a 50 Mhz oscillator.

Externally (to RAM and so) works at 25 Mhz.

No info about tha MC68060 one.

Ciao

[Floid]

Well, the 060 is superscalar with two pipelines, so... it's probably two 50MHz pipelines.*  About good as having a 100MHz CPU in the best case.  In x86-land, the Pentium was the first 'iteration' of the family to have two integer units, though NexGen's Nx586 may have beaten them to the punch (I'm not sure how the releases were timed, the original Nx586 didn't *have* floating point units, and that design team was acquired by AMD, resulting in the K5 and K6-* series of processors)... Practically everything since then (for the 'desktop' market, anyway) has used the same techniques; the PPro, which begat the P-II, was the first to have it throughout the chip.  (Some else will have to chime in on the PowerPC interations.)  Modern chips can/have had things like 4 or more pipelines, though I don't remember the exact numbers for Athlon/Opteron/PPC970/P4 off the top of my head.

But to drag this up to the things you see discussed on ArsTechnica and the other PC sites today...
The Pentium 4 has some units that run at double clock ("Simple ALUs," apparently), but other things can still take more than one tick to execute, and thus 'drain' the pipeline, negating some of the benefit.  Intel's been trying a different take than other manufacturers, going for a sort of 'waterslide' approach (I think their marketing buzzword is "Netburst," and the common description is 'long and skinny'), versus, say, having a couple sets of stairs ('short and wide').  Sounds great until some kid gets stuck halfway down. ;-)  

The PPC970 seems like a bit of a compromise between extremes...  When Motorola was pushing G3s and G4s, they were really 'wide' and 'short' versus the x86 lineup... so you get more instructions per clock, and more resistance to 'pipeline bubbles,' but you can also have a harder time clocking the unit up to competetive levels, since there are more units to keep in synch.  In practice, it sort of evens out, and you get relatively comparable performance for the same number of transistors spent.. and processor lines 'tuned' better to one or another application.  (G4 AltiVec is a big win for some things, and the rest of the chip has proved pretty decent for general-purpose computing; the P4's design has, in turn, proved pretty decent for stuff like video and audio work, where there's more math to crunch in between conditional branches, and AMD's managed to keep their stuff goodly fast all-around, usually with a price/performance advantage.)

*If you want to know the specific details of the chips, find the technical docs, or someone who actually wrote assembler for them...  I just have the 'big picture' view of anyone who's spent a week reading Ars and the britnews sites.

[Framiga]

Thanks for the useful reply!

By the way the guy that has reported on its program readme, an A2060@100 Mhz, was "in theory" right? or was jokeing?

 

[tonyw]

Joking, I think. Or maybe just exaggerating, in a Salesperson sort of way.

I remember when the '040s were released about the same time as the Intel 486s, and people were comparing the performance of the processors Vs their clock speeds. Someone pointed out (in AF or CUA?) that the Intel CPUs used single-edged clocks while the Motorola CPUs used dual-edged. So a 68040 at 25 MHz was equivalent (in operations/sec) to an 80486 at 50 MHz. Certainly the comparison was consistent with the observed performance.

That was before the days of intelligent cores. You can't do a simple compare like that on more modern CPUs.

tony

[Elektro]

I find this a bit weird... I heard of f.e. fpu units runnig at half speeds in some cpus (VIA I think) but this externally / internally sounds to me like comparing fsb and inner cpu clock... IMO

[Dr_Righteous]

Framiga wrote:
Take a look here:

http://www.emulation.com/catalog/off-the-shelf_solutions/production-test_adapters/upgrade_motorola/

$ 296.40 ea. is a little bit expensive!

Ciao

Ahh sweet! I was looking for one of those...