PowerMac 1.6GHz Benchmarks

[jason166]

Well said... don't make quick assumptions

After seeing all sorts of people complaining about the performance on the benchmarks and people saying they will be re thinking the purchase of a G5 I was gonna say something to the tune of

These programs are not good indicators of G5 performance

stingerman has explained this so I don't have to.

Well said.


And no, this does not mean that every program will have to be recompiled to get decent performance. Only apps that make use of procedures with DST Altivec calls and those that have code that is highly optimized for the G4 scheduler.

It is important to note that all the libraries in 10.2.7 have been recompiled to work well with the G5 so apps that make standard API calls will benifit allready.

Jason

 

[sparkplug]

Nasa bench 2ghz Dual

Here is another benchmark result, this one from nasa on the 2ghz dual running fluid dynamics simulations in JET3D

Nasa test dual 2ghz G5


Here is the bottom line

Based on raw scalar floating point performance in Jet3D, a 2GHz G5 system can match a 2.66GHz P4 system, and this is a dramatic improvement from earlier tests where G4 systems lagged behind higher clock speed P4 systems. Based on an extrapolation of current P4 results, the 2GHz G5 would lag newly announced 3.2GHz P4 systems in Jet3D scalar floating point performance by about 20%, but this kind of comparison is best deferred until G5-aware compiler tools become available (since a 20% performance gain is well within the potential of compiler optimization).

The G5 is a nice machine, the term "fastest" however is an embarresment.

 

[aje]

i'm desperate to leave the world of windows but as a flash developer macs totally suck at compiling and rendering flash movies. 1000% slower generally on the g4's i've been playing with (not QS).

i would love to know if any flash developers who share my pain have had a chance to render/compile a decent size flash application/movie on the new G5's - and if so was it at the least as fast as a decent pc?

i suspect from my google searches that its more of a architecture problem than anything, so i'm wondering if the new g5's architecture will speed up the calculations for flash movies??

 

[BrandonRP0123]

Originally posted by garymm
does this mean that every app has to be recompiled in a special G5 version to take advantage of it? the PS update isn't a complete recompile, yet it still adds tonnes of speed, supposedly.

To my knowledge, no. We must realize that between Mhz, the ceiling gets higher.

Between generations of architecture (think the ``G'' in G5) the ceiling not only gets higher, the room gets wider.

For a wider room, unless you have a wider ladder (or move the ladder) you won't be able to touch the ceiling in every spot of the room

 

[halabar]

Originally posted by aje
i'm desperate to leave the world of windows but as a flash developer macs totally suck at compiling and rendering flash movies. 1000% slower generally on the g4's i've been playing with (not QS).

That sounds like the Flash "compiler" is not optimized for the Mac at all.

We'll never get around that. If the software is written poorly, performance will always suck.

-- Karl

 

[stingerman]

Originally posted by garymm
does this mean that every app has to be recompiled in a special G5 version to take advantage of it? the PS update isn't a complete recompile, yet it still adds tonnes of speed, supposedly.

Re-compiling for the G5 is fast and easy. That was one of the points that was illustrated during the WWDC keynote. The photoshop demo during WWDC was of a simple re-compile and yet it blew away the dual XEON.

As far as this PS update notice what Adobe says:

• Modifies many Photoshop operations to fully utilize the G5 processor capabilities
• Replaces the Adobe Color Engine component (ACECarbonLib) with a new version that is designed for the G5 processor

ACE effects every part of PS. The Plugin does a lot more than you would think a plugin would do.

 

[Midiplaya]

So, I do quite a bit of audio work and run Virtual PC for a variety of apps. Virtual PC performance is pathetic on a 1GHz machine with 1GB Ram. I am anxiously hoping that the G5 machine will run Virtual PC fast enough to "feel" like a decent PC. If not, I am going to have to switch back.

 

[BrandonRP0123]

Originally posted by Midiplaya
So, I do quite a bit of audio work and run Virtual PC for a variety of apps. Virtual PC performance is pathetic on a 1GHz machine with 1GB Ram. I am anxiously hoping that the G5 machine will run Virtual PC fast enough to "feel" like a decent PC. If not, I am going to have to switch back.

Flip it to 16-bit color in the VPC. You should notice a vast jump in GUI performance. I've also found out that having a plain windoze desktop also speeds things along better (save the fancy wall paper for your REAL computer (Mac)

 

[Longey Nowze]

Re: Nasa bench 2ghz Dual

Originally posted by sparkplug
Here is another benchmark result, this one from nasa on the 2ghz dual running fluid dynamics simulations in JET3D

Nasa test dual 2ghz G5


Here is the bottom line




The G5 is a nice machine, the term "fastest" however is an embarresment.

if you notice, they only used one of those 2GHz CPUs, and it still beat the 2.66GHz P4, read that again please.

Thank You
MaT

edit: from the NASA tests

Though dual processor benchmarks are not presented in detail here, it is worth noting that the G5 system benchmarked at 498 MFLOPS and 0.125 MFLOPS/MHz for scalar Jet3D performance when two processors were used.

the singe 2GHz G5 had 254 MFLOPS and 0.125 MFLOPS/MHz the P42.66GHz had 255 MFLOPS and 0.096 MFLOPS/MHz... you do the math

edit: one more very important thing to remember that test was done on pre-release machines runing BETA software the results of the final products should be much better.

 

[stingerman]

One more thing to keep in mind why new benchmark software is needed and you should ignore these benchmarks: The new ASIC Controller actually steps down the CPU speed and then steps it back up slowly depending upon the current load. So those benchmarks may actually be showing results of the processor running at a slower average clock. The benchmarks need to be coded to tell the system that these are benchmarks and give us everything you got, as if you were under a heavy load. This is a major advance of the G5 architecture.

The dual G5's actually step down to 1.3GHz under light loads, using 1/6th the power. The fans don't even need to spin at all to keep them cool, though Apple keeps them spinning at a trickle in readiness. As the load on the processors increases the ASIC slowly steps up the clock, monitoring each processor independently as far as heat goes, each fan spinning for its own CPU's needs and at a possibly different rpm. If there is a fan failure or the door isn't closed right, the CPU's will run only at their slowest speeds which will make sure the do not melt.

I hope this helps put the benchmarks in a better light.

 

[BrandonRP0123]

Re: Re: Nasa bench 2ghz Dual

Originally posted by Longey Nowze
if you notice, they only used one of those 2GHz CPUs, and it still beat the 2.66GHz P4, read that again please.

Thank You
MaT

Exactly. Plus the P4 is 660Mhz faster than the G5. While Mhz myth, sure - the fact remains that the G5 would have proven itself if they ran it against a 2.0Ghz P4

 

[BrandonRP0123]

Originally posted by stingerman
One more thing to keep in mind why new benchmark software is needed and you should ignore these benchmarks: The new ASIC Controller actually steps down the CPU speed and then steps it back up slowly depending upon the current load. So those benchmarks may actually be showing results of the processor running at a slower average clock. The benchmarks need to be coded to tell the system that these are benchmarks and give us everything you got, as if you were under a heavy load. This is a major advance of the G5 architecture.

The dual G5's actually step down to 1.3GHz under light loads, using 1/6th the power. The fans don't even need to spin at all to keep them cool, though Apple keeps them spinning at a trickle in readiness. As the load on the processors increases the ASIC slowly steps up the clock, monitoring each processor independently as far as heat goes, each fan spinning for its own CPU's needs and at a possibly different rpm. If there is a fan failure or the door isn't closed right, the CPU's will run only at their slowest speeds which will make sure the do not melt.

I'd love to see a Dell do that. Or a Compaq. Or an IBM. Heck, even our brand new $8,000 Dell PowerEdge in the NOC doesn't do that (or even come close).

 

[Midiplaya]

$1590 !

Turning the color down to 16bit doesn't help. I am doing things where I need to open and close a lot of files fast. I also find that anything CPU intensive takes 3-4X's as long as a regular PC. I still prefer using OS X as my normal desktop and putting up with the performance as necesary. I really hope that I can get better performance though and snappiness out of the PC desktop.

That aside, if you can get the educational discount and ditch the modem and DVD-burner for a DVD-ROM, then you can get a 1.6 for $1590 or a 1.8 for $1900.

That is a bargain.

 

[BrandonRP0123]

Re: $1590 !

Originally posted by Midiplaya
Turning the color down to 16bit doesn't help. I am doing things where I need to open and close a lot of files fast. I also find that anything CPU intensive takes 3-4X's as long as a regular PC. I still prefer using OS X as my normal desktop and putting up with the performance as necesary. I really hope that I can get better performance though and snappiness out of the PC desktop.

With the price of RAM these days, and the amounts of it - I wonder when RAM drives will be fashionable/ever implemented again. Imagine sticking your VPC image into one of those bad boys on a G5

 

[Powerbook G5]

VPC is not going to fly no matter what you do, it's just a fact of life trying to emulate a whole computer enviroment. It's about the convenience than performance with it, which it does have.

 

[MrMacMan]

Yeah that is nice.

I'm sure more info will be flowing in soon.

And when my Apple store gets them, I think I might have to stea... i mean *borrow* one... forever.



So far so good.

 

[wizard]

Re: Re: Re: Re: a Photoshop test is what I need....

If your going to be doing a serious amount of programming get the dual machine. Both Java and the gcc environment can take advantage of duals. Gcc (on linux at least) can, via make, take advantage of two processors when building a project.

Thanks
Dave

Originally posted by scem0
yes, a lot of surfing.

But a lot of programming too (the more the better ). I know C++, Im learning basic and java. I hope to be using all 3 of those languages a lot.

And with the amount of time I spend on my computer even if all I did was surf it wouldnt be overkill. If I save a hundredth of a second on every webpage I went to I'd save a BUNCH of time.

scem0

 

[wizard]

This won't help my perception of the machine, especially the 1.6 GHz machine, as I honestly believe that performanc of the hardware is going to be found to be lacking in many areas. That doesn't mean we won't see good results in specific areas, but the online information that I've seen and the write up on Ars simple indicates that the 970 is a processor that will just keep parity with the current Intel hardware.

But that isn't what I wanted to comment on anyways, what is interesting is that you described an almost ideal chip set for a LAPTOP. If the description is accurate this is a bit of a surprise in a desktop. It does appear that these could be very energy effiecent machines. So to the people receiving the machines; if anybody has the abiltiy to measure energy draw on these machines please do so and post it.

Thanks
dave

Originally posted by stingerman
One more thing to keep in mind why new benchmark software is needed and you should ignore these benchmarks: The new ASIC Controller actually steps down the CPU speed and then steps it back up slowly depending upon the current load. So those benchmarks may actually be showing results of the processor running at a slower average clock. The benchmarks need to be coded to tell the system that these are benchmarks and give us everything you got, as if you were under a heavy load. This is a major advance of the G5 architecture.

The dual G5's actually step down to 1.3GHz under light loads, using 1/6th the power. The fans don't even need to spin at all to keep them cool, though Apple keeps them spinning at a trickle in readiness. As the load on the processors increases the ASIC slowly steps up the clock, monitoring each processor independently as far as heat goes, each fan spinning for its own CPU's needs and at a possibly different rpm. If there is a fan failure or the door isn't closed right, the CPU's will run only at their slowest speeds which will make sure the do not melt.

I hope this helps put the benchmarks in a better light.

 

[HopefulSwitcher]

For all you people out there that are contemplating getting a mac or a dell... get a mac...

For those thinking between dell and another pc... get a mac

It's not war veterans that make me proud to be living in this country.. its companies like Apple lol XD

I love you G5... you sexay thang you!

Now... Steve Jobs... I command you to release the new powerbooks now!

 

[nagromme]

Originally posted by aje
i'm desperate to leave the world of windows but as a flash developer macs totally suck at compiling and rendering flash movies.

I don't care as much about compile speed--it's never bothered me n my G4--but Flash playback isn't up to speed. Macromedia has acknowledged their failing in this regard, and said to expect big Mac speed optimizations in future versions. (The commend was made specifically in regard to the player.)

There's no flaw in Macs that makes them unsuitable for vector animation--Macromedia simply has an issue to address.

 

[mim]

Originally posted by HopefulSwitcher
Now... Steve Jobs... I command you to release the new powerbooks now!

Heh! Nice try.

<<Cut to dusty Biblical scene. Steve's in a toga, surrounded by nubile young ladies not wearing that much. He surveys the crowd from his podium>>

Steve: Who shall we welease?
Crowd: Welease the powewbooks! (giggling)

Steve confers with advisers: What? There is no Powewbook?
Steve to crowd: There is no "Powewbook". Who shall we welease then?
Crowd: Welease Woderwick!

Steve: Welease Woderwick!

<<Cut to classic white Apple slick add>>

Powerbook style voice over: Woderwick. The next big - and small - thing - because that's all there is - from Apple.

 

[hvfsl]

Originally posted by mim
Heh! Nice try.

<<Cut to dusty Biblical scene. Steve's in a toga, surrounded by nubile young ladies not wearing that much. He surveys the crowd from his podium>>

Steve: Who shall we welease?
Crowd: Welease the powewbooks! (giggling)

Steve confers with advisers: What? There is no Powewbook?
Steve to crowd: There is no "Powewbook". Who shall we welease then?
Crowd: Welease Woderwick!

Steve: Welease Woderwick!

<<Cut to classic white Apple slick add>>

Powerbook style voice over: Woderwick. The next big - and small - thing - because that's all there is - from Apple.

More monty python than biblical.

 

[kaos_de_moria]

That doesn't mean we won't see good results in specific areas, but the online information that I've seen and the write up on Ars simple indicates that the 970 is a processor that will just keep parity with the current Intel hardware.

wait... that was exactly what everyone expected from apple... right?

ppl seem to forget fast what they expected before the G5 was announced!

 

[mainstreetmark]

Geezz... bitch bitch bitch...

if the G5 isn't the fastest computer overall, it's certainly the fastest in a few big categories. The new PCs may be faster at some things, but by a very small fraction now.

The G5 is good news all around. (Yes, it would have been nicer without the 'World's Fastest' claim, and that poor kid getting blown through his house)

 

[sparkplug]

Longey Nowze writes

if you notice, they only used one of those 2GHz CPUs, and it still beat the 2.66GHz P4, read that again please

Yes I did notice this, how exactly is this relevant? I allso noticed that they did not test a dual athlon, opteron or xeon machine, just a single g4, single g5 and single p4, does this somehow invalidate the results? no.


By all means elaborate as to how 254 for the G5 "beat" 255 for the P4 ? you yourself quoted these mflop scores in closing. Unless you are referring to the scalar performance, ie mflops per clock, Indeed the G5 is faster per clock, but this does not mean it is "faster" certainly not "faster" in any usefull terms. IE actual purchasable today performance. Seeing as one cant actually purchase a 2.6ghz G5, nor indeed a 3.25 for that matter.

The nasa bench quite clearly shows that the 2ghz G5 chip does about the same work (1 point slower) that a 2.6ghz P4 does running their real word code. Maybe you should read it again.