How fast are the G5s, anyway?

How do you know they are not as fast as everyone thinks? Have you used one? Just like with the P4, you won't see it's true potential until software gets optimized for it.

Welcome to the boards, but some friendly advice. This topic is NOT Future Hardware, it is Current Hardware. The FH forum gets choked as it is, so please put any future threads in their correct place. Members will take them more seriously that way.

Back on topic, I think the Dual 2.0 will blow away any Pentium based machine in real world apps. Of course getting people to realize that is all that matters is a different story.

Welcome aboard.

The most objective answer is that at this point, we don't know. When Panther (OS X 10.3) and the new gcc come out, and apps are updated, we'll have a better idea. Still, I think that your basic hunch is right, and what the 970 does is put Apple hardware back in the race. There are (already) some tasks in which it will excel, and some in which it will lag a bit behind (especially in single processor configurations vs. the Pentium 4 where its application is compiled with gcc and the Pentium's with Intel's highly optimized ICC).

But there are other factors to consider: The system bandwidth is phenomenal, the capabilities for multiprocessing are excellent, and the 64-bit nature of the beast means that, for example, the Wolfram Research guys had to come up with a bakeoff script in Mathematica that didn't crash the Xeon that the dual G5 was pitted against.

Right now, the dust is still settling, and the fastest machine hasn't even shipped yet, so it's a bit early to draw any hard and fast conclusions.

I sugget that people read Apple's G5 Performance Primer before they start comparing Apple's and Oranges. There are differences in the architecture of the G5 that has impacts when running G4 optimized code.

It's interesting that people just repond to BS and believe what others say without even doing any individual research into the subject. These speed tests are as bad as the megahertz myth...

I agree with you. They will help us close the Gap but it will be a while before we'll have a definitive speed advantage. Apple and IBM will have to kick butt on this. I'm sure the G5 will get faster as Compilers are smarter about utilizing Altivec and the immense FPU power of the G5

My view is that for software that isn't optimized for AltiVec a dual 2 gig G5 will be at the same level as a dual 3 gig Xeon. For software that is optimized for AltiVec the dual G5 will exceed the dual Xeon, by as much as 2x-3x in cases where the AltiVec optimizations are as agressive as possible.

I think you are probably right. Due to the diffences in architechture they will probably be faster at some things, slower at other. The argument will rage and there will be no clear cut winner.

But in the end, just getting into the ballpark is all that's needed. Apple is still a better experience--but the cost in speed was getting just too big for the better experience to be worth it. As long as the speed isn't a problem, it's Apple's interface and elegance that will make the sale.

. <----- This is your brain

O <----- This is your brain on a G5

Any questions ?

What happens with a dual G5

$$$$$ <------- This is your wallet

$0 <------- This is your wallet after a dual G5


Any questions ?

with apologies to the orignal poster ...



As a long time IT professional I recommend Apple G5 all the time. Ill go into meetings where people are just crying for me to help them gain synergy by decreasing their TCO while at the same time increasing their ROI, yet these people look like a deer caught in the headlights when I flat out tell them that the only way to do that is by looking at taking the next step to the next level by integrating their asset management supply side relationships into leveraged content delivery paradigms, with an eye towards aligning their collaborative relationship initiatives towards common goals and the first step in that direction is to move to an OS that has Olog(n)performance, namely Apple G5.

Ive been able to do this in the past with a a few Fortune 500 companies by implementing a strict B2C affinity marketing plan which relies heavily on E-mediation performance metrics, something that not everyone is willing to go through.

In short, dont even come to me with questions about your Value chain collaborative commerce unless youre willing to pay the piper and upgrade to Apple G5 because this is not your daddys economy and youll get nowhere by running legacy operating systems. Times have changed and unless youre willing to change with them youll be left behind wondering what the hell happened to all your profits.

Just a quick note. I'm slammed here at work.

I think the thing to keep in mind here is that the main wait that people are talking about here is a matter of weeks. Once the G5 is in peoples hands we will see how it does as a starting point.

Other that that, programs need to be updated.
For example: In regards to low scores in Xbench. (XLR8 )


Other things to keep in mind are

• The newer PIV after they get there die shrink to clock up but will be running over 100 watts.
• IBM has designed the G5 to clock up in clock speed. We should actually be relatively close in clock speed in the next 12 months.
• IBM is focusing on high end chips. They are fabing not only G5's and their Power4's but also chips like the AMD Opteron and Nvidea GPU's.
• Unlike the launch of the G4 where Mot. was unable to actually ship the clock speed they announced, IBM launched the G5 at speeds exceeding those they announced.

The main thing is that we have a chip today that has a fantastic all round design, a nice roadmap for the future, an incredible front side bus design, and that is running with reasonable power use characteristics without shrinking in die size yet.

yadda, yadda, yadda.

-tink

madmax559



Next time post the Google translation please!

-tink

It's deja vu because this happens every single time there's a significant change in processor architecture. It happened to the Pentium 4 on its launch as well.

And Panther will not be "pure 64 bit," which is good, because a pure 64 bit OS couldn't run any 32 bit code. That would be bad. Panther will allow any interested application to take full advantage of the 970's 64 bit capabilities, which is an excellent start.


You'll have to wait until IBM's management is as dysfunctional as Motorola's first. Waiting for support for a processor architecture is different; it didn't take more than a couple of months for applications to adapt to the 7450 G4, for example.


Hmmm? You don't have to wait for anything to get blazing performance out of a G5. We're back in the hunt right now, and even ahead in some areas that are important to Apple customers right now. All I'm asking for is the short wait for things to catch up to the 970 before we start making claims about what its potential is - just as things had to catch up to the 7450 and the Pentium 4 and many other architectures.

All that said, I can never counsel against a distrust of marketing hype. Your questions are good questions, and I hope I've answered them.

on a Dual G5: OO

on a maxed-out Xseve Blade: OOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO OOOOOOOOOOOOOO

I have said it before and I'll say it again, then again, I don't know how any of this stuff works so it's all I can say, thus, without further ado, I'll say it:

The only speed comparo you can do between platforms that means anything is to take an app and run it on both platforms through a series of tests, then see which one gets done first and by how much. That's all. You use the "system" and that involves the hardware and the software and how they work or don't work together. If an app isn't available for one or the other of the platforms, then you take the two most comparable apps to do the desired task and see where you get. That's really all that matters, how fast they do, for lack of a more meaningful term, "stuff." Not benchmarky stuff, but real stuff.

If an app is better optimized on one platform or the other and runs faster because of it, then you have to say that said system is faster for that purpose. The laggard may get optimized and spring ahead, but the machine that gets the task done first is, practically speaking, fastest.

Does the way in which it is achieved really matter? This was true when pro wintels were ahead, and it's true now that pro macs may be ahead.

The only testing proceedure that matters is this:

Set down a dollar value.

See what it gets you from either camp.

Do the "stuff" you're going to do in real life

See which one finishes first.

Easy as pie. Xbench, Spec, whatever, means nothing, really and truly. Take the G4, not a desktop beast by any means, but according to benchmarks it would be 6-7X slower than a desktop X86, we know (from timed tests) that it can be up to twice as slow in the worst cases, but not 6X as slow, no way.

Matsu,

I agree with your reasoning, and it might explain why some people are getting upset about the Cinebench scores, myself included.

Cinebench2003 is basically an automated Cinema4d V8 (minus the tools). It uses the same rendering engine as the app, and uses real Cinema4d 3d scenes. Its about as real world from a benchmark as you are ever likely too find.

Agree, things like X-Bench etc are just basically pissing contests, and mean nothing in reality.

Of course, we've heard directly from Maxon that CB does not work well with G5's, and both CB and its real world equivalent will be updated ASAP.

I have been involved in Systems Analysis for a number of years and the one thing that keeps coming to mind is that it really doesn't matter how "fast" the machine is as much as how "fast" a person is able to complete a task on it. In this regard, I find Macs more efficient and thus "faster" than the competition.

I am (with baited breath) anticipating the arrival of my dual G5. I look forward to a "fast" piece of equipment with an "efficient" interface which will make my tasks more pleasant to complete.

One thing that seems really hard for non-technical people to understand is that sometimes very small things can cause major performance issues in new hardware. Performance is "fragile". This is true of all new hardware, not just Apple's. Fortunately in this case most of the problems can be fixed very easily -- first by removing a couple of rarely used instruction (which only appear in highly performance optimized code), and second by recompiling with a new compiler switch. A new minor patch of the software can deliver a very substantial performance improvement without any serious work on the part of the developer. If the developer then uses Apple's new tools the can optimize performance even further.

This is the best 1st post I've ever seen. You have summarized in a few sentences what millions of Windows users somehow cannot understand.

Welcome aboard.

Tell this to some musician running/using realtime software synthesizers/processors.

For some tasks, performance really matters.

i feel exactly the same way!
it's us audio/video people who really need these.
the one exception is talent....that is something that speed can't really help!

His statement is not absolutely true, but it's generally true and absolutely worth stating.

The overwhelming majority of desktops and notebooks will never attempt to process 2,000 7.1 24-bit tracks, each with 128-band EQ and 10,000 high-quality effects, while rendering the Big Bang in FORTRAN and the next Pixar film in Maya. And surfing pr0n sites.

For the bulk of people, a good interface is a better productivity booster than a fast processor. Even when it's not that clear, it's still a valuable advantage - if you're working in real time, the last thing you need is an interface getting in your way! A creative person can always work with fewer sounds and fewer effects, but anybody has to be able to do what they want to do when they want to do it.

what do you expect from someone who will reach his ultimate combined physical and intellectual efficiency in about 3 years time.

maybe he is fast and he already did.

"IBM launched the G5 at speeds exceeding those they announced."

Good point.

So when Steve Jobs said: '3 gig' then I'm quietly hoping for another surprise...

While people may query the current speed grades of G5, it IS early days, software from vendors is unoptimised, Apple's OS X is unoptimised, benchmarks are unoptimised and few in number, the amount of tests are few and far from diverse/comprehensive. The machines are only shipping. The only machine we have had benched is a 1.6 which is a 'Yikes' G5? No, not really. But it IS the entry machine and even unoptimised, it's giving the dual G4s a real headache... Not bad when it's giving 1 and a half gig of mhz sway to the previous fastest Apple machine.

All things adding up, people are complaining that the entry level is slightly faster than a dual G4. Sorry, but the irony isn't lost on me.

Still, for those people who still aren't convinced now, I think Apple will probably bump these machines relatively quickly. In Jan' 2004, it will have been over six months since announcement. Time for a bump. I think we'll see 2.5 gig G5. Dual of course! And for each 500 mhz we get 970? We get 1000mhz of G4 fpu power! My point? The dual G5 2.5 is, relatively speaking, close by (compared to the G5 wait/debacle!) It is this dual G5 2.5 that will give the Intels out there a real headache. I just don't see Intel reaching 4 gig by January. IBM and Apple will, therefore, be piling on the pressure and really closing the gap. It will, in my view, silence any doubters.

As for the AMD64. Hmmm. I'm not sure AMD can ratchet up 500 mhz between now and Jan'...given the mhz problems they've had with both the Athlon and Hammer. I think 'our' 2.5 G5 will prove a match for them. By Jan' 2004, we should see plenty of 'updates' which are 'optimised for G5' by then...and of course, the ace in the pack?

Panther.

In the meantime, the point about a dual 2 gigger and a productive interface says it all (now that we have a G5.) Given hardware of parity or exceeding PC performance...then it will be interesting to see how many creative pros who defected to Wintel 'come home'. All things being equal, the benefits of 'X' should really give Apple the edge they've been looking for since their tower sales began to *slowly* bleed to death

A figure of 60% of workstation users moving to Mac has been quoted. And I'm not surprised considering that even an unoptimised dual g5 hangs with Xeons costing heaps more. You get 'X' and iapps thrown in for 'free' with a fantastic choice of high end Apple software...and now Mac looks like a real value added solution, a productive platform.

If I was on the PC side (which I kinda am/am not-I have an Athlon tower but am typing this on my wife's iBook, the PC tower got taken out by 'So big' and I can't be arsed to purge the virus at the moment...) then I'd be tempted. I'll be going G5 within five months.

Lemon Bon Bon

...fast..really fast..but not as fast as you can imagine...

OK quick update; reason for the recall of G5's is the side door not closing properly so CPU throttled as if the door is open - this explains some of the lousy benchmarks vs. G4 boxes published. QED.

...was refering to those few G5 1.6 boxes out there that have been recalled to Apple manufacturing.