View Full Version : XBench Scores in Rosetta (Can someone post their 1.67 G4 PB Marks?)

Jan 17, 2006, 11:10 AM
Results 28.29
System Info
Xbench Version 1.2
System Version 10.4.4 (8G1165)
Physical RAM 512 MB
Model iMac4,1
Drive Type Maxtor 6L160M0
CPU Test 14.50
GCD Loop 152.44 8.04 Mops/sec
Floating Point Basic 14.12 335.51 Mflop/sec
vecLib FFT 6.55 216.23 Mflop/sec
Floating Point Library 21.81 3.80 Mops/sec
Thread Test 43.70
Computation 150.09 3.04 Mops/sec, 4 threads
Lock Contention 25.57 1.10 Mlocks/sec, 4 threads
Memory Test 35.16
System 24.40
Allocate 10.97 40.27 Kalloc/sec
Fill 46.21 2246.74 MB/sec
Copy 98.92 2043.17 MB/sec
Stream 62.92
Copy 72.95 1506.81 MB/sec [altivec]
Scale 60.15 1242.72 MB/sec [altivec]
Add 60.84 1296.00 MB/sec [altivec]
Triad 59.49 1272.70 MB/sec [altivec]
Quartz Graphics Test 47.60
Line 57.27 3.81 Klines/sec [50% alpha]
Rectangle 45.63 13.62 Krects/sec [50% alpha]
Circle 48.33 3.94 Kcircles/sec [50% alpha]
Bezier 55.32 1.40 Kbeziers/sec [50% alpha]
Text 37.19 2.33 Kchars/sec
OpenGL Graphics Test 93.84
Spinning Squares 93.84 119.04 frames/sec
User Interface Test 11.78
Elements 11.78 54.08 refresh/sec
Disk Test 93.66
Sequential 119.59
Uncached Write 117.84 72.35 MB/sec [4K blocks]
Uncached Write 113.18 64.04 MB/sec [256K blocks]
Uncached Read 125.71 36.79 MB/sec [4K blocks]
Uncached Read 122.38 61.51 MB/sec [256K blocks]
Random 76.97
Uncached Write 42.49 4.50 MB/sec [4K blocks]
Uncached Write 103.43 33.11 MB/sec [256K blocks]
Uncached Read 92.79 0.66 MB/sec [4K blocks]
Uncached Read 125.16 23.22 MB/sec [256K blocks]

Above is from an ArsTechnica Thread of XBench scores (http://episteme.arstechnica.com/groupee/forums?a=tpc&s=50009562&f=174096756&m=396006927731&r=266001037731#266001037731) in Rosetta. Can someone run X-Bench in a 1.67Ghz PPC PowerBook so we can get an idea of how crappy performance is going to be or won't be. Note that this is on a 1.83Ghz iMac not a MacBook so we aren't talking Apple's to Apple's (well we are but well you know what I'm talking about. :p

Jan 17, 2006, 01:38 PM

Just took one off their website...



System Info
Xbench Version 1.2
System Version 10.4.4 (8G32)
Physical RAM 1024 MB Model
PowerBook5,8 Processor
PowerPC G4 @ 1.67 GHz L1 Cache
32K (instruction), 32K (data) L2 Cache
512K @ 1.67 GHz Bus Frequency
167 MHz Video Card
ATY,RV360M11 Drive Type

CPU Test 63.97
GCD Loop 154.37 8.14 Mops/sec
Floating Point Basic 40.67 966.25 Mflop/sec
vecLib FFT 81.02 2.67 Gflop/sec
Floating Point Library 52.30 9.11 Mops/sec

Thread Test 74.28
Computation 70.38 1.43 Mops/sec, 4 threads
Lock Contention 78.65 3.38 Mlocks/sec, 4 threads

Memory Test 27.25
System 31.48
Allocate 127.53 468.31 Kalloc/sec
Fill 36.46 1772.96 MB/sec
Copy 16.65 343.98 MB/sec

Stream 24.02
Copy 24.60 508.15 MB/sec [altivec]
Scale 23.85 492.76 MB/sec [altivec]
Add 23.63 503.44 MB/sec [altivec]
Triad 24.02 513.93 MB/sec [altivec]

Quartz Graphics Test 86.75
Line 128.05 8.53 Klines/sec [50% alpha]
Rectangle 75.01 22.39 Krects/sec [50% alpha]
Circle 103.13 8.41 Kcircles/sec [50% alpha]
Bezier 84.48 2.13 Kbeziers/sec [50% alpha]
Text 66.84 4.18 Kchars/sec

OpenGL Graphics Test 88.23
Spinning Squares 88.23 111.92 frames/sec

User Interface Test 53.03
Elements 53.03 243.40 refresh/sec

Disk Test 29.79
Sequential 63.98
Uncached Write62.44 38.34 MB/sec [4K blocks]
Uncached Write 55.14 31.20 MB/sec [256K blocks]
Uncached Read 83.95 24.57 MB/sec [4K blocks]
Uncached Read 60.75 30.53 MB/sec [256K blocks]

Random 19.42
Uncached Write 6.35 0.67 MB/sec [4K blocks]
Uncached Write 45.64 14.61 MB/sec [256K blocks]
Uncached Read 66.95 0.47 MB/sec [4K blocks]
Uncached Read 86.33 16.02 MB/sec [256K blocks]

Jan 17, 2006, 01:46 PM
28.29?? :eek: Holy crap... Rosetta is, well, crap!

Jan 17, 2006, 10:59 PM
I'm actually surprised Ars Technica even used this benchmarking software. Mods here have already noted it's unreliability, as has barefeats, and XLR8yourmac. Take the benchmark, put it on your mac, then compare the score with an identically configured system--scores drift for no apparent reason. Even running the app on the same machine has given me quite different scores.

Bottom line: real-world applications are the real-world benchmark tool. Xbench doesn't compare.

Jan 18, 2006, 04:21 AM
Surely there must be a better tool to benchmark macs with?;) Give us the photoshop tests.

Jan 18, 2006, 10:03 AM
macbook pro + rosetta emulation = ibook ppc power!

or so Xbench would have us believe.

My xbench scores drift less than 10 percent. So, its still a valid knock at early adopting pros... reality will show, emulation not a great way to start a creative app.


Jan 18, 2006, 11:12 AM
Did anyone actually expect Rosetta to be anywhere near as fast as native performance? It is, after all, emulation (or JIT translation, depending on semantics).

I think you have to see Rosetta for what (I believe) it is: a stop-gap measure to make MacBooks basically viable for early adopters until key apps are ported to intel / universal. I had no expectation of good performance, just adequate performance for basic apps like Office. Frankly if it performs as well as [a 1ghz] ibook, that's better than I expected, and I think it very likely meets Apple's intent (as a stop-gap). Certainly a 1ghz ibook runs Office adequately. Frankly, when I played with it at the Apple store, office ran as faster _or faster_ than it does on my 1.25ghz pb.

2006 will be the year of Rosetta, but I think by the second half of the year a very large number of applications will be universal or intel native. By 07 only a die-hard minority will still be talking about Rosetta (perhaps the same folks that are still clinging to Classic).

There will be a great gold-rush to get apps universal, espcially when it becomes clear that apple is selling 50% more PCs by year end, including ~5-9m Intelmacs this year alone. Think about it - if you're early out of the gate with your universal app, you'll get on the "list of universal apps available now" that is sure to be all over these mac websites, and you're quite likely to see a huge number of initial sales. So I'm sure every developer is racing to get there. Even if you're a large developer, like adobe, the prospect of many millions of "crossgraders" waiting to insatiably try your app on their shiny new Intelmacs will be too great to resist.

There will be laggards, sure, and it sounds like Office will be slow to come out as Universal, but that's ok if it runs that well.

So, bottom line is that I can buy a MacBook now that runs _almost_ as fast as my current 1.25Ghz PC in Rosetta, and way faster for native. Apple pro native apps are just 30 days beyond my MacBook, and I'm sure hundreds of native/universal apps are just 30-90 days behind that. I think Apple hit exactly the mark they were trying to hit.