[geeks] OSX 10.4.2, FireFox 1.0.5 out

[Charles Shannon Hendrix]

Thu, 14 Jul 2005 @ 09:36 -0400, Kurt Huhn said:

> On Jul 14, 2005, at 4:28 AM, Jonathan Groll wrote:
> > Seriously not intended as a troll. Granted, many, many factors come  
> > into
> > play when doing comparisons, only one of which is clock speed. However,
> > many of us, and for too long now have blithely accepted the apple  
> > gospel
> > that PPC is the inherently "superior" architecture. Surely, though,
> > there must come a time when a processor with enough pure 'grunt' can
> > outperform one that is slower but more able?
> >
> 
> MHz != speed.  

I don't believe that's what he is saying.

He's saying that the clock speeds of Intel/AMD CPUs are being boosted so
much that they are making up for architectural issues.

AMD is achieving more performance at lower clock speeds, much like other
CPUs.  From what I read in the K7 papers, the AMD CPUs are *NOT* Intel
architecture CPUs.  They are a different beast under the covers, and the
Intel ISA and other bits are more like a layer of emulation.

It seems AMD could design and ship a competitive non-Intel CPU if they
were not tied to the Intel-compatible market.

Clock-for-clock, I don't see the G5 spanking the AMD like it does the
Intel.  It seems to me that the AMD64 especially, despite its Intel ISA
ball-and-chain, is quite competitive.

I have a theory about why Apple is going with Intel instead of AMD, but
I'm going to post that in another thread.

> Other improvments may bring the Intel processors closer  
> to being acceptable (faster pipelining, faster RAM, switched bus, etc)  
> but "speed" as measured in MHz is not one of them.  I will grant that  
> AMD and Intel are bringing their Desktop processors closer to  
> acceptable, but they will never get there unless they're willing to  
> dump the legacy crap that surrounds them and start designing them to be  
> put into elegant and well-designed systems.

I think AMD is doing that.  They are hampered by the need to be
compatible with Intel.

If they were free of that, I think they'd really shine.

> I see nothing there indicating that the developer macs are  
> outperforming G5s.  I see a vague reference to "as little as 10  
> seconds" to boot to the desktop.  Now there's a reliable benchmark!!   
> Not.

Hey, Windows users put quite a lot of value into quick reboots... :)

> If we're going to claim that the intel systems from Apple are faster  
> than anything else, why don't we compare them side by side, under equal  
> environments and conditions, and take real measurements, with real  
> applications, under heavy load, and performing functions that people  
> who buy high-end workstations need them for?  Until I see such tests,  
> I'll not even consider "it feels faster" with the smallest grain of  
> salt.

Of course, to be completely fair you need to test headless, because the
G5 pays a heavy price for the GUI.

Things like image manipulators, sound encoders, and stuff like that
would make useful test of code that people actually use.

Maybe put Darwin on PPC up against an AMD64 running something like *BSD
or Linux?  As long as you avoid system calls, I think that would be
pretty good.  No GUI, just some batch jobs with compute intensive tasks
to avoid comparing Mach with a monolithic UNIX.


-- 
shannon "AT" widomaker.com -- ["And in billows of might swell the Saxons
before her,-- Unite, oh unite!	Or the billows burst o'er her!" -- Downfall
of the Gael]