G5/Opteron comparison...

[archiea]

Beam, consider this my Xmas present to you... :p

http://www.barefeats.com/g5op.html

Its interesting what they both have in common... the OS's aren;t full 64bit and not allof the apps are optimised...

[Beamtracer]

Hi Archie, I noticed that story last time you mentioned barefeats.

The G5 won some of the tests. The Opteron won some.

At the moment, Windows on the Opteron is only 32-bit. Only Linux on Opteron will allow it to run 64-bit.

The current Mac OS has some 64-bit features, and some 32-bit features. The OS can address 8 gigs of RAM, for example.

I'd like to see Apple go pure 64-bit (which I'm sure they will in the next couple of years) but doing so will probably break compatibility with today's apps.

[Ade]

I dont really trust barefeats though...

I also want to see Lightwave tests.

[Beamtracer]

Their benchmarks comparing the Apple G5 to Intel Pentiums showed the G5 winning convincingly.

http://www.barefeats.com/pentium4.html

[Aegis]

Their benchmarks comparing the Apple G5 to Intel Pentiums showed the G5 winning convincingly.

Hmm... I know for sure that ATM the G5's still lag behind the Xeon's for LightWave rendering - I'd like to see a full set of benchmarks for LightWave running on the dual 2.0Ghz G5 but at the moment www.blanos.com only has partials.

For the record, I've got a dual 3Ghz Xeon and all of the blanos benchmarks for the dual G5 are beaten by this (but the Mac isn't that far behind) - I'm looking forward to seeing performance comparisions as OSX makes the transition to full 64-bit (and as the CPU speeds get ramped up).

Personally, I think the G5 is a very attractive machine for graphics work but at the moment still doesn't represent value for money (for me anyway).

[Ade]

Price performance xeons are expensive... BUt i feel its all about software optimising and currently noone trouches intel sse2 for lightwave.

I dont even know any special instyructions NT has done for g5 lw?

[Beamtracer]

Originally posted by Aegis
For the record, I've got a dual 3Ghz Xeon and all of the blanos benchmarks for the dual G5 are beaten by this (but the Mac isn't that far behind) - I'm looking forward to seeing performance comparisions as OSX makes the transition to full 64-bit (and as the CPU speeds get ramped up).

Hi Aegis,

While 64-bit applications would be cool, I think the most important thing affecting speed of the G5 is optimization.

The G5 is a new processor. There are many new features in the IBM driven "G5" compared to the Motorola "G4". Of course, Lightwave 7.5 came out long before the G5, so it takes advantage of none of these new features. No optimization at all.

The real question is... How much G5 optimization will Lightwave 8 have? This is what will make the speed difference.

From what I gather (by reading between the lines of Newtek statements) is that LW8 will bring compatibility with Apple's new Panther OS. I also gather that some optimization will be done, though a lot more could be done.

My speculation is that Newtek is aiming to do more substantial optimizations in later releases of Lightwave, ie. Lightwave 8.5 and beyond.

But I'm still hoping for some improvements this time 'round.
The Windows version of Lightwave has long had massive optimizations for the Intel processors.

Optimization is more important than how good your processor is.

[Zarathustra]

I read somewhere that IBM was supposed to release a new compiler, which I guess is suposed to make optimising for G5 easier?

[silvergun]

Is that the xlc compiler? I heard ibm was gonna release it some time early in 2004 but theyre trying to incorporate it into GCC. They claim it can speed up the processor by 50% which sounds pretty good. If anyones gonna take the G5 seriously, newtek, luxology, alias and the like really should start putting more effort into their mac products.

[tallscot]

After Effects screams on the G5 compared to the dual Opteron and the dual Xeon. I don't know why.

The dual 2 Ghz G5 is, by far, the fastest platform for AE.

Lightwave? The G5 is close to the Xeon, as mentioned. It will be interesting to see if Intel can keep up, considering the G5 will be at 2.5/2.6 Ghz in a month or two, and at 3 Ghz this summer (2004) with the 90 nm IBM 980 processor.

And Longhorn isn't due out till 2006/2007. Jeez, by the time Longhorn comes out with OS X features like Quartz and the side pane, OS X will be updated another two or three times. Why is it taking Microsoft so long to update their OS when a company like Apple with a much smaller R&D budget is pumping out really great innovative features every year. Exposé rocks my world.

Use what you like, though.

[Beamtracer]

Originally posted by tallscot
Why is it taking Microsoft so long to update their OS when a company like Apple with a much smaller R&D budget is pumping out really great innovative features every year.
Apple is faster because:[list=1]
Apple is a smaller more nimble company with more creative talent than Microsoft
MS needs to see what Apple does first to know what the next Windows features will be
Windows has more problems anyway (ie security and other OS issues)
Open Source. Apple's OS is based on the open source BSD UNIX which shares the development burden.
[/list=1]

[Draven]

It should be pointed out that it is possible to use the 64-bit instruction capability of the Opteron (and Athlon64) without having the operating system even aware of the 64-biut instructions, in much the same way that software can make use of MMX and SSE and SSE2 instructions when the operating system precedes the existence of these instructions and has no allowance for them... for instance, LW's P4 optimizations (well, SSE2 optimizations) work just fine under Win98, which came out well before SSE2 did.