[Eug Wanker]

From this AnandTech thread.

As for the machine, I got to use one! This was cool. I wandered off into the testing lab and, though it was specifically for having Apple reps help you with your code, since there was an open seat, they let me sit down and try out the G5 for a bit. I didn't get to go crazy on it because, well, I didn't wanna look like some dumb kid that's just playing with the machine, but then I noticed the guy next to me writing Email, so I didn't feel so bad!

It's true. The machine flies. I used the new Apple industry standard "bounce" test to judge start times and the like. When I got to the machine, it was open physically so I could be shown the insides. The cooling system is a thing of beauty and the machine was very quiet, especially compared to G4. The rep closed it up, plugged it in, and it took about 13 seconds or so, give or take, for the entire machine and OS to be up and ready to go. Oh, this was the dual 2.0GHz machine. Now, there wasn't a whole ton in the way of software on this machine, of course, it being set up for coding, but all of Apple's software was, including iMovie, which is a beefy app. I launched every single app on this machine and none of them took more than 1 bounce. In fact, they all took less than 1. The machine just felt so incredibly slick and powerful, it was really something to see.

[The Placid Casual]

Great Scott!

I was putting the purchase off,but if it answers the OS X responsiveness 'issue' by that much... wow... I may just buckle...

[DeathMan]

[slider]

Hehehehehehe Ahhhhh yes!

[businezguy]

Wow! That sounds like an awesome system.

I'd like to add to that, in that I know somebody who got Panther and is trying it out.

They are running a G4 733 MHZ, and they say that programs like iMovie are opening in one bounce. Now imagine Panther combined with the new G5. Now, THAT would be impressive.

By the way, he also says it is very buggy, and doesn't run all of his applications, but that is to be expected.

[Eug Wanker]

I'd try to get that XBench on there, but there's this little issue of I don't like being banned from conferences!

Anyway, one thing I noticed is, the G5 appears to pick up steam. It has a energy saving capability that scales the dual 2.0 down to a single 1.3 while doing things like word processing and email checking. When the CPU load increases, it throttles the CPU's back up to the dual 2.0 in a millasecond, so you don't feel any hit or lose any processing time. Pretty slick. It makes the machine use up a fraction of the power and keeps it very cool.

[Skywalkers new Hand]

Originally posted by Eug Wanker:
I'd try to get that XBench on there, but there's this little issue of I don't like being banned from conferences!

Anyway, one thing I noticed is, the G5 appears to pick up steam. It has a energy saving capability that scales the dual 2.0 down to a single 1.3 while doing things like word processing and email checking. When the CPU load increases, it throttles the CPU's back up to the dual 2.0 in a millasecond, so you don't feel any hit or lose any processing time. Pretty slick. It makes the machine use up a fraction of the power and keeps it very cool.

Wow that is totally cool! Wonder how it figures it out.

[Eug Wanker]

Originally posted by Skywalkers new Hand:
Wow that is totally cool! Wonder how it figures it out.

It sounds just like Intel's SpeedStep, which they used in their laptops.

[Socially Awkward Solo]

Originally posted by Eug Wanker:
It sounds just like Intel's SpeedStep, which they used in their laptops.

Apple has has prosessors that slow down since 1993 or so.

So does intels know what you are doing and slow it down or does it has to be a user command?

[Scotttheking]

Originally posted by Socially Awkward Solo:
Apple has has prosessors that slow down since 1993 or so.

So does intels know what you are doing and slow it down or does it has to be a user command?

Intel procs can do it automagically.
The P-M 1.7GHz has been observed to go as low as 300MHz.

[chrisutley]

That's one snappy SOB, eh? I can't wait!!!!!!!!!!!!!!!!!!!!!!!!!!!!!

[jcadam]

Originally posted by Eug Wanker:
I'd try to get that XBench on there, but there's this little issue of I don't like being banned from conferences!

Anyway, one thing I noticed is, the G5 appears to pick up steam. It has a energy saving capability that scales the dual 2.0 down to a single 1.3 while doing things like word processing and email checking. When the CPU load increases, it throttles the CPU's back up to the dual 2.0 in a millasecond, so you don't feel any hit or lose any processing time. Pretty slick. It makes the machine use up a fraction of the power and keeps it very cool.

I'll be disabling that feature when I finally get a G5. No reason not to run a DESKTOP at full throttle all the time. It's not like the cooling system can't handle it.

[Socially Awkward Solo]

Originally posted by jcadam:
I'll be disabling that feature when I finally get a G5. No reason not to run a DESKTOP at full throttle all the time. It's not like the cooling system can't handle it.

WHY do you need it at full speed to look at porn exactly?

Yes the cooling can handle it but you sure as hell will hear it working.

[mac freak]

And we all know computer parts last longer when they run with less heat .

[Pierre B.]

Originally posted by The Placid Casual:
I was putting the purchase off,but if it answers the OS X responsiveness 'issue' by that much... wow... I may just buckle...

And I think the G5 runs right now under 10.2.7. Just wait and see when Panther arrives.

[Eug Wanker]

ThinkSecret has some screen grabs of this 2 GHz "Neo".

Particularly interesting is this bench:



This bench is normalized to a 1 GHz G4, and only tests one processor. To confirm that 1 GHz G4 = 100, I ran this test on my TiBook SD.

Code:
Test              Dual 2.0 G5            TiBook 1 GHz
-----------------------------------------------------
Integer               172                     100
Floating Point        270                     100
Vector                208                     100
-----------------------------------------------------
Average              216.7                    100
Estimate @ 4 GHz     433.3                    400

So the PPC 970 G5 overall is not a screamer in terms of this benchmark when comparing overall MHz for MHz performance, being slower at integer and averaging only 8% faster overall than the G4. Of course, AFAIK it doesn't take into account CPU cache, and the hugely increased bus speeds. The G4 is very hobbled by the SDR 167 MHz memory bus.

Also, the PPC 970 G5 is a screamer at floating point however, as everyone had predicted all along. It runs 35% faster than the G4, MHz for MHz. This is important because while most apps are integer-based, many of those aren't really hindered by CPU speed that much. ie. Who cares if Keynote is integer? It runs fine on my TiBook so of course it'd run great on a computer 72% faster. OTOH, a lot of the hard core number crunchers depend heavily on floating point, and that is where the PPC G5 shines. Plus, they're the ones often needing the huge memory bandwidth to shuffle all that data around. Correct me if I'm wrong...

Also, we can look at it this way: If we are to consider theoretical raw computational power, the FP speed of the dual 2.0 GHz PPC Neo is 5.4X the speed of my TiBook, even though the integer speed is "only" 3.4X. Plus it's got a superfast cache and huge memory bandwidth.

Any way you cut it, it's gonna fly!

[jasong]

I'll be disabling that feature when I finally get a G5.

And we all know that Apple has no idea what it is doing when it comes to these machines . . . it's not like they designed them . . .

-- Jason

P.S. Are the G5s on site prototypes or shipping versions? I think they are prototypes (based on what Joswiack said about the changes made to the G5s used for the benchmarks), so this "feature" may not exist on the shipping version.

[Eug Wanker]

Originally posted by jasong:
And we all know that Apple has no idea what it is doing when it comes to these machines . . . it's not like they designed them

Nothing wrong with disabling this feature, assuming you can stand the fan noise (and don't care about your electricity bills).

[MacNStein]

Also, the PPC 970 G5 is a screamer at floating point however, as everyone had predicted all along. It runs 35% faster than the G4, MHz for MHz. This is important because while most apps are integer-based, many of those aren't really hindered by CPU speed that much. ie. Who cares if Keynote is integer? It runs fine on my TiBook. OTOH, a lot of the hard core number crunchers depend heavily on floating point, and that is where the PPC G5 shines. Plus, they're the ones often needing the huge memory bandwidth to shuffle all that data around. Correct me if I'm wrong...

Also, we can look at it this way: If we are to consider theoretical raw computational power, the FP speed of the dual 2.0 GHz PPC Neo is 5.4X the speed of my TiBook, even though the integer speed is "only" 3.4X. Plus it's got a superfast cache and huge memory bandwidth.

Any way you cut it, it's gonna fly!

You're spot on with that assessment. I've often wondered why people feel like they need a 2GHz+ speed machine to run the vast majority of applications that the majority of users use. Most people could easily "get by" with a 700MHz eMac. Let's face it, the speed bump represented by the G5 is for appearances when it comes to the needs of most users. However, I'm not bitching, if getting into the GHz battle is the only way to make the Mac platform a more viable contender, then I', all for it.

BTW, Eug, regarding that Dual 2GHz G5 that guy was working with, was it running Jaguar or Panther? If on Panther, then I know where some of that speed comes from and it's no mistake that Apple is releasing Panther after the G5 is released. From reports I've read and from my own experience, Panther alone will bring about quite a substantial improvement in performance in most OS functions.

[Eug Wanker]

regarding that Dual 2GHz G5 that guy was working with, was it running Jaguar or Panther?

He didn't say, but I'm assuming Smeagol X.2.7 (G5 variant, build 6S45), because he didn't make any comments about new OS features, that's what the WWDC pix are showing, that's what the machines are shipping with, and Panther is supposed to be too buggy anyway for real demo machines.

[Eug Wanker]

As yet unsubstantiated, but interesting benches nonetheless:

Hey everyone, I've been at Apple's developer conference and had a chance to install and try out After Effects on a new G5. I ran the Night Flight file that has come to be the standard for AE benchmarking. Since I didn't want to sit there and watch it render for hours, I ran just the first 10 interlaced frames from the project's pre-set render queue...

http://www.aefreemart.com/tutorials/...ghtflight.html

Here are my results for this test on the three computers I have available to me:

1 x 1.0 GHz G4 PowerBook 17" - ~30 minutes (3 min/frame)
2 x 2.66 GHz Pentium Xeon from Boxx - 11 min, 39 sec (1.2 min/frame)
2 x 2.0 GHz PowerMac G5 - 6 min, 1 sec (0.6 min/frame)

I ran the Xeon test on a couple different identical machines to make sure mine wasn't just running slowly, but got identical results. Of course my Mac bias is well-documented, but I'm sure many people here can vouch for me as an honest person. If the results had gone the other way, I'd just keep my mouth shut and let someone else break the bad news.

Other observations about this test that may ultimately work in the Mac's favor:

1) The machine was not running 64-bit Panther, but only a tweaked version of 32-bit Jaguar. Likewise, AE is obviously not yet compiled to take advantage of the G5 chip in any way. Both or these situations will automatically be rectified in the future.

2) Night Flight is very CPU-intensive, but not very disk I/O intensive. I think the 1 GHz system bus and other details on the G5 will provide greater gains for typical projects that rely more heavily on I/O.

[The Placid Casual]

Over at AI, the concensus is that the SkidMarks 'bench test' app only looks at ONE processor...

Anyone know if this is true?

[Eug Wanker]

Originally posted by The Placid Casual:
Over at AI, the concensus is that the SkidMarks 'bench test' app only looks at ONE processor...

Anyone know if this is true?

Yes, and my table and comment already takes this into account.

Code:
Test              Dual 2.0 G5            TiBook 1 GHz
-----------------------------------------------------
Integer               172                     100
Floating Point        270                     100
Vector                208                     100
-----------------------------------------------------
Average              216.7                    100
Estimate @ 4 GHz     433.3                    400

"If we are to consider theoretical raw computational power, the FP speed of the dual 2.0 GHz PPC Neo is 5.4X the speed of my TiBook, even though the integer speed is "only" 3.4X."

[Deestar]

Originally posted by The Placid Casual:
Over at AI, the concensus is that the SkidMarks 'bench test' app only looks at ONE processor...

Anyone know if this is true?

It seems to be a pretty old but this from xlr8yourmac says that "it does not take advantage of dual processors at this time"

http://www.xlr8yourmac.com/OSX/os_x_...ark_tools.html

[clarkgoble]

I understand that the XBench test was off for unknown reasons with the G5. I forget what site was discussing this. (Ars I think)

I also suspect that the combination of an unfinished OS and probably more optimization work in gcc 3.3 will entail things only getting faster.

The altivec is one I'm really excited for. Doesn't optimizing altivec for the G5 over the G4 require certain changes?

[kupan787]

Originally posted by clarkgoble:
The altivec is one I'm really excited for. Doesn't optimizing altivec for the G5 over the G4 require certain changes?

The altivec engine on the G5 is the equivalent to the 7400. The altivec engine on the 7455 is much improved (don't recall all the details, but it is different in what it can handle). However, because of the much better bus on the G5/970, altivec wpont be starved, and would beat out a similallry MHzed 7455.

Now if the 970 had a altivec engine equal to the 7455 (and in a future revision in might), that would scream!

[Catfish_Man]

Originally posted by kupan787:
The altivec engine on the G5 is the equivalent to the 7400. The altivec engine on the 7455 is much improved (don't recall all the details, but it is different in what it can handle). However, because of the much better bus on the G5/970, altivec wpont be starved, and would beat out a similallry MHzed 7455.

Now if the 970 had a altivec engine equal to the 7455 (and in a future revision in might), that would scream!

7400 style vector allows for 1 permute instruction and 1 other every cycle. 745x style allows for any 2 instructions. The 970 also has much longer vector latencies than the G4, but makes up for it (mostly, at least) with large amounts of rename registers, extensive out of order execution, hardware prefetch streams, a beefy vector load store unit, and insane memory bandwidth. So, yeah, optimization is a bit different, but it won't matter for most apps.