G5 Speed Tests

M

mindbend

Registered
Here's a couple of links of early speed tests:

http://www.macworld.com/2003/09/reviews/macworldlabfirstg5testresults/

http://www.macaddict.com/news/news_007.html

I'm sure there are others starting to trickle in and I'd love to see them.

Cynical View:
After doing some quick math, it looks like the G5 is only providing about a 10% advantage in terms of architecture relative to CPU clock speed. In other words, if you prorate the G4 to the 2 Gig clock speed, you're only about 10% behind the G5 at 2 Gig (give or take depending on the test, etc.)

This indicates that, so far, the G5 versus G4 benefit is around 10%. For a machine that claims to be the fastest in the world, my bet is that a top model PC could easily match these test results.

Optimistic View:
Panther isn't here yet. Optimized apps aren't here yet. 64-bit hasn't been ironed out yet. It can only get better.

-----

Does anyone have other speed test links? Any thoughts on just how good the G5 will REALLY be (if you can pull yourself out of the reality distortion field for one moment). :) Any reactions to these and other speed tests?
 
In the two years you've been posting have you ever had a non-cynical view? ;)

And what type of math are you doing that can prorate a G4 with on chip 256k L2 and off chip 2 MB L3 cache? And then to think you can compare those results with a G5 with on chip 512k L2 and (currently, though the POWER4 can access up to 32 MB) no L3 cache. You have way too many factors that are outside a simple cycle-for-cycle calculation.

Apple has known for quite some time that software that is Altivec enabled and dual processor aware runs faster on a dual G4/1.42 then the same software would on the fastest Pentium 4 system. But not everything is Altivec enabled and dual processor aware. Mac OS X does a good job of helping with the multiple processors, but nothing matches code written to use all of a system's advantages. But the problem has been getting Motorola to deliver. Motorola had the hardest time just getting past 500 MHz, what makes you think they are going to reach 2 GHz any time soon?

Cycle-for-cycle calculations are interesting though, specially for seeing the progress of chip design. If we had a G4 with on chip 512k L2 and no L3 cache to work with do you think you would end up with just 10%? The thing is, the PowerPC line has been increasing with cycle-for-cycle performance with each new line. I think that is a good thing. Try doing a cycle-for-cycle calculation with the Pentium II, Pentium III and Pentium 4 processors. I think you'll find that the Pentium II is still the speed champ that way. Intel knows this and has been trying to find a way out of their MHz trap for quite some time. They give up performance in a rush to over clock their processors. Had they been able to keep equal performance with cycle-for-cycle with the Pentium II and still made it to 3.0 GHz, they would have had an incredibly fast chip on their hands, but they needed the cycles more.

For the best benchmarks I've seen so far, I suggest Craig Hunter's Page. He has since provided the following benchmarks

  • dual G4-1GHz Xserve (single CPU only): 105
    dual G4-1GHz Xserve (both CPUs): 207
    dual G4-1.25GHz PowerMac (single CPU only): 129
    dual G4-1.25GHz PowerMac (both CPUs): 256
    dual G5-2GHz PowerMac (single CPU only): 254
    dual G5-2GHz PowerMac (both CPUs): 498
    single P4 2GHz: 192
    single P4 2.66GHz: 255
    single P4 3.2GHz (extrapolated): 307

By the way, in Craig's cycle-for-cycle comparison between the G4 and G5, he came up with a 22% increase in performance for the G5 over the G4.
 
I'm not cynical or optimistic about the G5... I believe that we all have to wait a bit more to clearly see what this beast can really do! I just hope that we will not have to wait more than 2-3 months :rolleyes:

I for one, I will not be able to use my Dual G5/2GHz until perhaps the end of October when they will deliver it to me here in Greece :(

I also believe that the RAM thing that Mac Addict posted is closer to the truth about the G5 than any other site or person out there so far :p
 
The G5 is much more revolutionary than just a speed bump for 32-bit apps. Yes it runs 32-bit apps faster, but that's not the point. It's built for 64-bit, it's built for multi-processor, it's build for high speed memory,...
 
French readers will like to check
http://forum.macbidouille.com/viewtopic.php?t=43952&highlight=cpuclass
and
http://www.macbidouille.com/niouzcontenu.php?date=2003-09-14#6623
It shows that there is an internal file in Final Cut Pro 4 that compares the different CPU speed (for the current version of the software of course).

fcp4cpu.jpg
 
Well, you can have your laughs at my expense, that is fine. I can take it. ;)

But here's the deal. The reality is soon upon us and it will be irefutable. It will be undeniable. It will be factual.

And that reality is the simple fact that the G5 either will or will not match or beat the best PCs (or G4 Macs for that matter) at a given task. You can rationalize it all you want. You can talk Cache, you can talk megahertz myth. You can talk architecture. You can talk 64-bit. You can talk optimized. You can talk whatever you want. In the end, it either is faster or it isn't. (Of course the G5 is faster across the board than the G4, but is it faster relative to other speed gains in the industry...).

As for optimizing as we go. I'm not holding my breath. Yeah, Apple will optimize their OS and their Apps, but Altivec proved that developers are in no rush to take optimizing for Apple hardware seriously. Several years of Altivec netted virtually no gains from thrid party developers. Same for DP aware apps. Apps like Media Cleaner are a freaking joke. Same goes for Lightwave. And the Photoshop plugin for Altivec. Nice try. Thanks for nothing. Shall I go on? Nah...

Here's a nugget. Altivec didn't help help Apple one bit to catch up with the speed gap.

RacerX:
Yes, I am highly cynical after dumping thousands of dollars into woefully inadequate G4 chips and sluggish GUIs. I make a living off performance and the G4 didn't cut it. The only reason I stuck around (aside from untold investments in software) was because of that classic Apple dream, the hope for something better, lots better. The G5 may in fact be it, but it's got a ways to go to really knock my socks off. That's all I'm saying.

As for the prorate example. Yeah, it's not the best approach, but it's just an example. Just something to work with for conversation. I think it's very safe to say that a 2 Gig DP G4, were it to exist, would compare admirably to the current DP 2 Gig G5. That concerns me. I was expecting a wider gap in my theoretical model. In a few more months or a year or two, if the 64-bit dream and other optimizations don't kick in, then I'll really be concerned.

Craig Hunter's Specs:
Cool. Thanks for the link. Too bad I have no concept of what to do with that information. Where's the Photoshop, Lightwave, Quake, etc. tests? Also, Craig himself helps prove my point :" Thus, the increased raw vector performance of the G5 is largely due to its higher clock speed.", regarding vector performance.

And really, Jet3D? I hate to be a jerk, but that test does not qualify for anything that the average G5 will be used for. That is a completely obscure test that obviously highlights the G5's stronger points. C'mon! ;) Craig's tests actually made me even more concerned. He flat out says that the best G5 still lags the best PCs. And you better believe 3.2 gig PCs will be replaced very soon.

I'd be curious to see truly comparative charts using all processors from all manufacturers and OS's. I've always wanted to start a site that allowed people to submit their own data using files provided by the site (Photoshop files, 3D models for rendering, 2D animations for AE, demo games for Frame Rates tests, etc.) Too lazy to do it I guess. Yeah, the G5 is faster. Great. But how does is compare to the best PCs today and how will it compare in six months or two years when the G5 has had time to mature and unfortunately the Intels have had time to grow yet again.

Macbidouille:
Thanks for the graph and link. That chart (after having translated its supporting text through Sherlock) again reaffirms my prorate theory. That chart is indicating how efficiently FCP performs its RT functions relative to other machines. Result? 30% benefit of DP G5 2 gig versus DP G4 1.42. Exactly in line with my prorate example.

Footnote: I alway sending up feeling the need to defend my position. I'm a huge Mac fan. I have all the Mac history books and have read them multiple times. I have distant conenctions to Stevo (he has no idea I exist, however). Apple changed my life with the Apple II and the fist Mac. Literally changed my life. I love Apple. I'm just not blind anymore. Stats are stats. I'll probably continue to use my beloved Macs until I die, but I'm definitely going to complain if they don't deliver every step along the way.
 
Originally posted by mindbend
I'm just not blind anymore. Stats are stats. I'll probably continue to use my beloved Macs until I die, but I'm definitely going to complain if they don't deliver every step along the way.
Click to expand...

And exactly who are you going to complain to? Apple can't force people to optimize for their systems. Most Mac software ends up being an after thought, so it is not using the environment even close to it's best abilities. How is this Apple's fault?

The reason the test done by Hunter are the best I've seen is that they are not stacked for or against any processor.

That is a completely obscure test that obviously highlights the G5's stronger points.
Click to expand...

Actually it is a test that doesn't play to any processors strong points. It doesn't play to any bias on the part of software makers. It is only looking at raw performance. It is actually even better than Apple's test as Hunter didn't limit himself to any given compiler.

Also, Craig himself helps prove my point :" Thus, the increased raw vector performance of the G5 is largely due to its higher clock speed.", regarding vector performance.
Click to expand...

I would hope that would be true, otherwise (as can be seen from the data) the G5 is only 22% faster in a cycle-for-cycle comparison. His data showed an increase of 71% in raw performance while comparing the 1.25 GHz G4 with a 2 GHz G5.

What was your point? ::sleepy::

You sure didn't seem to have one that I saw (that had anything to do with your quote from Hunter). A 20%+ performance gain per cycle in a platform is nearly unheard of in the computing industry. Compounded with the actual increase in over all clock speed, it is simply incredible.

By any measure, this was beyond what any other processor platform has had in the way of improvements in years (decades actually). The only way to not be amazed at the jump from the G4 to the G5 is by knowing that the G5 is actually following the POWER3, POWER3-II and POWER4 lines and is not just the next processor developed after the G4 (even then, it still shines).

Where's the Photoshop, Lightwave, Quake, etc. tests?
Click to expand...

No where, where they should be when benchmarking a processor. Those are proprietary software that can be weighted by the makers (Adobe crippled Premier for the Mac to prove their "PC Preferred" argument). The Jet3D test is using a program developed for NASA that can be optimized to run on any processor... without bias.

If all you wanted was a Photoshop test, MacAddict's test with 2 GB of RAM was fine. It doesn't tell you about the processor. I'm sure that with 4 GB of allocated memory Photoshop is going to run even faster.

He flat out says that the best G5 still lags the best PCs. And you better believe 3.2 gig PCs will be replaced very soon.
Click to expand...

His test were normalized to a single processor. Yes, a single Pentium 4 at 3.2 GHz (307) beat the single G5 at 2 GHz (254). Lucky for us Apple sells the 2 GHz system with two processors (which scored 498 when both processors were used).

I'm sure that the 3.2 GHz Pentium 4 is going to be replaced very soon, but until it reaches 5.2 GHz, single Pentium 4 processor systems are going to be slower than the current dual processor G5s at 2 GHz.

But this is raw performance, how willing a software company is to making their products take advantage of a system is outside what Apple or IBM can do with their processors and systems.

The only thing I see as sad in all this is Apple's policy of not making the mid-range models out pace the former high end models. They really could have made the 1.8 GHz G5 faster (by adding a second processor) if they wanted, but that would have made the dual 1.42 GHz G4 systems seem completely out of date. At least in the current configuration, apps that are dual processor aware are going to be faster on the dual G4 than on the single G5.

You need to spend a little more time studying the data if you plan on making your cynical spin stick. Best of luck though. ::ha::
 
It's too bad (for Windows users) that Intel doesn't allow dual Pentium-based processors, since those would probably kill. But since they're to Microsoftian to allow multiple processors, Apple's dual machines are still here.
 
Mindbend:
If you really are THAT concerned of Apple acting stupid then you should really switch sides...

There I'm sure you will find performance, quality, EVERYTHING actually that you are dreaming or wishing of Apple delivering... ::ha::

In the beginning it will be tough and maybe costly but in the long run you will have the best thing in the world:
Peace of mind(bend) that only the Wintel/Amd side can offer you :p
 
Originally posted by arden
It's too bad (for Windows users) that Intel doesn't allow dual Pentium-based processors, since those would probably kill. But since they're to Microsoftian to allow multiple processors, Apple's dual machines are still here.
Click to expand...

Actually, Intel allows dual Pentium4 based computers... They called XEON! :D

You know, like the Dull one running @3GHz that Steve, Phil, et al trashed with the Dual G5 in WWDC 2003 ;)

Intel promotes XEON for Workstations and Servers and P4 for PCs ::ha::
 
Thanks for the responses and clarifications. I really do appreciate the information. What threw me off on the Jet3D charts was that I now see he's only factoring in one CPU. That 's good info. Congrats to the millions of Jet3D users everywhere! [end sarcasm] (Just having some fun).

My main point of this thread was to find as many speed tests for the G5 as possible. We got a few here, so that's good.

I still, personally, find the Spec and Jet3D tests completely useless for the simple reason that they don't show us any real world usage information. For that matter, the Final Cut chart is poor as well, I just happen to like it since I use FCP a lot.

As for going to the dark side. I'm already there. We have multiple PCs in our shop. I don't like them. I'm a Mac user. But I'm also a businessman. As I've said a million times, I've run multiple tests of PCs against our Macs and the Macs get destroyed every time. That is a fact. It's not up for debate [using my tests and machines anyway]. These tests obviously don't factor in things like usability, stability, enjoyment factor, interface, GUI, etc. Just pure render, I/O tests.

We do not have any G5s yet, nor do we have the latest PCs, so I can't run any modern tests now. My business partner will be getting a top end PC in January and I will be getting a top end G5. We'll be running the old battery of tests and I'll post them here.

My prediction is simply that the two machines will basically be even, give or take a test here and there. And that's good place to be for a Mac user. Better place than we've been in ages. I'm going to hesitate from further speculation since I don't have access to those machines.

Please, feel free to post any other speed tests. I'd love to see them.
 
dear mindbend.
go read THE ECONOMIST or somethin and let us worry bout the G5.

or maybe u can open a business and sell DELLS.. im sure ud convince any loser to by one... buh bye :)
 
I know only one speed test in which Macs and PC's score equally well, no matter what conditions you use or how many times you test them.
 
tsizKeik,

Attacking the person does not nullify their claims I have not seen the G5 to be faster, but I do not have a G5. I can see where mindbend is coming from because this is NOT a mac world. The majority is the PC, and despite how Apple and AMD try to make out that mhz doesn't matter, it does. Depending on the architecture, and other limitations, it matters less, but it still matters. I love my mac, but until Apple can come out with some serious ERP software, you will be living in a Windows world.
 
G5 optimizations are very, very important, because some relatively common PPC instructions choke the G5.

I believe most of the benchmarks we're seeing are using unoptimized code. G5 performance will increase with time.

It's an amazing chip. A miracle, really, for Apple.

(Really, weren't you all afraid Apple was doomed, as poorly as Motorola was performing? I was. Before Apple pulled this rabbit out of its hat.)

Doug
 
Relax people! Relax!

Mindbend if you use Final Cut Pro here is something that you will like:
http://www.apple.com/uk/hotnews/articles/ibc/gallery/gallery5.html

The piece that may interest you:
"...Highlights included a sneak preview of a pre-release version of Final Cut Pro 4 which had been optimised for the Power Mac G5..."

If anyone thinks that the G5 is *ONLY* 10% faster than the G4, or even slower, he/she is pessimistic and NOT realistic or factual because EVERYTHING that the G5 based Macs brought us is A LOT faster than the G4 based Macs but it will take us some months before we will understand the real benefits let alone experience them...

dktrickey I totally agree with you! Also, does anyone here remembers how "bad" OS X 10.0 (or even 10.1) was performing compared to 10.2? Expect at least the same things with the same G5 hardware when optimized apps will arrive...

-Are Desktop Wintel/Amd faster than the Dual G5? NOPE but still, WE SHOULD NOT CARE!
-Are Workstation/Server Wintel/Amd faster than the Dual G5? Some YES, some NO!
-Is the G5 an amazing future proof architecture for the Mac platform? Heck YES! :D

Some people judge also Steve Jobs for spreading BS for the G5... The same goes for Apple in general :( Then I bet that Adobe, Mathematica, et al people that CLEARLY supported Apple/Steve Jobs claims in WWDC 2003 are spreading BS too! :p

I bet that others are full of BS and Reality Distortion Fields... :rolleyes:

Let me get this straight:
Moto G3 = the Past
IBM G3 = the Past and the Present (perhaps the future too :p)
Moto G4 = the Present
IBM G5 = the Present and the Future

On the Wintel world I think that Apple must start worrying about Amd and not Intel... Intel in the months ahead offers only marketing while Amd offers Weapons of Choice ;) like Amd64 and Opteron... Still, those things for at least 2 years down the road will have to use Windows XP + SP1/2 :eek: What a wonderful computing the Dark Side offers :rolleyes:
 
You must log in or register to reply here.
Back
Top