Which benchmarks would I run on an Intel Mac?

[jsw]

If I decide to open this little box of mini goodness™ in front of me, where would I find suitable benchmarks to run? Assume I have nothing installed except for the default apps - no 3rd party games, apps, etc.

Any suggestions? At the very least, I'll run them on my 17" Intellimac to ensure they run and get some baselines, then I'll decide about opening the box.

[badmofo9000]

Xbench is now a universal app.

[gnasher729]

Xbench is now a universal app.

XBench is also the lamest excuse for a benchmark ever seen.

Try this: Download Handbrake. Take a DVD and convert it to H.264, using two pass conversion for optimum quality. That is a real benchmark: It measures the time for something that people actually do, and they care if it runs faster.

Or import two hours of digital video into iMovie, then convert it to a single sided DVD. That's something that takes hours and hours and people have to wait for it.

Just make sure that you can reproduce exactly what you are doing; H.264 encoding for example depends extremely on what exactly you do; converting digital video to DVD depends on the amount that you convert, above one hour iDVD has to work really hard to fit things on a DVD.

[Spanky Deluxe]

Windows Vista!!

{s******s} - Edit: Woah, that wasn't a swear word! Ok {snickers} then!

But no, seriously, gnasher's on to something with Handbrake.

[plinden]

Or import two hours of digital video into iMovie, then convert it to a single sided DVD. That's something that takes hours and hours and people have to wait for it.
This looks like a good one to me. I noticed this used both cores (up to about 150%) on my iMac. (By the way, iMovie and iDVD are excellent consumer apps - the newer themes aren't too cute and can look quite professional. I used "reflection - white" for my Christmas 2005 DVD.)

[jsw]

Try this: Download Handbrake.
Would the 512MB in the stock mini "invalidate" the test, or is Handbrake less memory-intensive?

[godbout]

This is not all that fair though because Handbrake is using the GPU to a VERY large extent and GPUs with h.264 decoding built in (X1600) will dominate without showing you how fast (all around) your comupter is. You should do mutliple benchmarks if you want to be fair like an xcode build or something... I guess the important thing is that if you are seeing how fast the intel mac is in optimal conditions or under normal use... if you are seeking the latter than maybe a THE photoshop test or iTunes encoding as well... just some suggestions...

[tjwett]

Xbench is now a universal app.

it is indeed but just a heads-up on this...

i recently ran XBench on the Core Duo iMacs at the Apple store (all had 128MB graphics and 2GB RAM) and they all scored around 50. that's about half what my dual G5 1.8 w/1.25 GB RAM and lame 64MB video consistently scored. so i don't know how much i trust the overall XBench with the Intels. they felt twice as snappy as my G5 but scored half in XBench. perhaps it's just one portion of the test that is way off and can be disabled?

[jsw]

it is indeed but just a heads-up on this...

i recently ran XBench on the Core Duo iMacs at the Apple store (all had 128MB graphics and 2GB RAM) and they all scored around 50. that's about half what my dual G5 1.8 w/1.25 GB RAM and lame 64MB video consistently scored. so i don't know how much i trust the overall XBench with the Intels. they felt twice as snappy as my G5 but scored half in XBench. perhaps it's just one portion of the test that is way off and can be disabled?
Odd. Results from my 17"/128MB/2gb Intel iMac:
Results 92.64
System Info
Xbench Version 1.2
System Version 10.4.5 (8G1454)
Physical RAM 2048 MB
Model iMac4,1
Drive Type Maxtor 6L160M0
CPU Test 70.91
GCD Loop 238.28 12.56 Mops/sec
Floating Point Basic 83.35 1.98 Gflop/sec
vecLib FFT 41.42 1.37 Gflop/sec
Floating Point Library 62.24 10.84 Mops/sec
Thread Test 174.45
Computation 152.11 3.08 Mops/sec, 4 threads
Lock Contention 204.49 8.80 Mlocks/sec, 4 threads
Memory Test 99.53
System 96.80
Allocate 92.77 340.67 Kalloc/sec
Fill 97.45 4738.22 MB/sec
Copy 100.49 2075.55 MB/sec
Stream 102.43
Copy 84.63 1747.97 MB/sec
Scale 89.64 1851.94 MB/sec
Add 125.81 2680.07 MB/sec
Triad 122.97 2630.69 MB/sec
Quartz Graphics Test 68.13
Line 63.74 4.24 Klines/sec [50% alpha]
Rectangle 57.85 17.27 Krects/sec [50% alpha]
Circle 58.05 4.73 Kcircles/sec [50% alpha]
Bezier 82.50 2.08 Kbeziers/sec [50% alpha]
Text 90.39 5.65 Kchars/sec
OpenGL Graphics Test 124.16
Spinning Squares 124.16 157.50 frames/sec
Disk Test 82.26
Sequential 94.25
Uncached Write 106.45 65.36 MB/sec [4K blocks]
Uncached Write 94.76 53.61 MB/sec [256K blocks]
Uncached Read 117.48 34.38 MB/sec [4K blocks]
Uncached Read 71.53 35.95 MB/sec [256K blocks]
Random 72.97
Uncached Write 40.53 4.29 MB/sec [4K blocks]
Uncached Write 94.91 30.39 MB/sec [256K blocks]
Uncached Read 90.86 0.64 MB/sec [4K blocks]
Uncached Read 116.31 21.58 MB/sec [256K blocks]

[plinden]

it is indeed but just a heads-up on this...

i recently ran XBench on the Core Duo iMacs at the Apple store (all had 128MB graphics and 2GB RAM) and they all scored around 50. that's about half what my dual G5 1.8 w/1.25 GB RAM and lame 64MB video consistently scored. so i don't know how much i trust the overall XBench with the Intels. they felt twice as snappy as my G5 but scored half in XBench. perhaps it's just one portion of the test that is way off and can be disabled?

See my thread about this here (http://forums.macrumors.com/showthread.php?t=181325). This describes how to improve your Intel Mac XBench results with a simple tweak to the Quartz settings.

XBench is almost certainly badly written if I can improve my performance from 55 to 90 by changing one configuration.

[jsw]

See my thread about this here (http://forums.macrumors.com/showthread.php?t=181325). This describes how to improve your Intel Mac XBench results with a simple tweak to the Quartz settings.

XBench is almost certainly badly written if I can improve my performance from 55 to 90 by changing one configuration.
Interesting... I didn't change anything, and got the score above.

Gotta love pseudo-random benchmarking apps.

[plinden]

Odd. Results from my 17"/128MB/2gb Intel iMac:
Results 92.64

Do you have "beam synchronization" turned off?

Edit: ok, you don't - did you download XBench today? I wonder if they updated it. Although it appears to be the same version.

[jsw]

Do you have "beam synchronization" turned off?
No (see above), but I can try it.

Or maybe I should say "no, as far as I know." I play with the dev tools at times, so maybe I do... but I don't think so.

Edit: yes, I just got it today, but it does seem to be the same version.

[tjwett]

See my thread about this here (http://forums.macrumors.com/showthread.php?t=181325). This describes how to improve your Intel Mac XBench results with a simple tweak to the Quartz settings.

XBench is almost certainly badly written if I can improve my performance from 55 to 90 by changing one configuration.

good to know.

[plinden]

Interesting... I didn't change anything, and got the score above.

Gotta love pseudo-random benchmarking apps.
You also get Random Uncached Write 5x my result. All my results were in line with every other XBench result I've seen for the Intel iMac.

[jsw]

You also get Random Uncached Write 5x my result. All my results were in line with every other XBench result I've seen for the Intel iMac.
Maybe they did change something recently. Dunno.

Result using "Disable Beam Synchronization" (apparently, it had been set at "automatic" before):
Results 110.72
System Info
Xbench Version 1.2
System Version 10.4.5 (8G1454)
Physical RAM 2048 MB
Model iMac4,1
Drive Type Maxtor 6L160M0
CPU Test 70.91
GCD Loop 238.49 12.57 Mops/sec
Floating Point Basic 83.34 1.98 Gflop/sec
vecLib FFT 41.39 1.37 Gflop/sec
Floating Point Library 62.28 10.85 Mops/sec
Thread Test 174.72
Computation 155.99 3.16 Mops/sec, 4 threads
Lock Contention 198.57 8.54 Mlocks/sec, 4 threads
Memory Test 97.20
System 92.16
Allocate 89.75 329.59 Kalloc/sec
Fill 97.29 4730.50 MB/sec
Copy 89.85 1855.80 MB/sec
Stream 102.81
Copy 84.16 1738.28 MB/sec
Scale 89.13 1841.30 MB/sec
Add 127.90 2724.64 MB/sec
Triad 125.22 2678.71 MB/sec
Quartz Graphics Test 114.31
Line 102.57 6.83 Klines/sec [50% alpha]
Rectangle 113.48 33.88 Krects/sec [50% alpha]
Circle 112.50 9.17 Kcircles/sec [50% alpha]
Bezier 110.88 2.80 Kbeziers/sec [50% alpha]
Text 137.55 8.60 Kchars/sec
OpenGL Graphics Test 122.68
Spinning Squares 122.68 155.63 frames/sec
User Interface Test 239.28
Elements 239.28 1.10 Krefresh/sec
Disk Test 83.12
Sequential 96.81
Uncached Write 107.11 65.76 MB/sec [4K blocks]
Uncached Write 94.78 53.63 MB/sec [256K blocks]
Uncached Read 119.71 35.03 MB/sec [4K blocks]
Uncached Read 76.47 38.43 MB/sec [256K blocks]
Random 72.82
Uncached Write 40.65 4.30 MB/sec [4K blocks]
Uncached Write 95.38 30.53 MB/sec [256K blocks]
Uncached Read 90.04 0.64 MB/sec [4K blocks]
Uncached Read 114.44 21.23 MB/sec [256K blocks]

[plinden]

No (see above), but I can try it.

Or maybe I should say "no, as far as I know." I play with the dev tools at times, so maybe I do... but I don't think so.

Edit: yes, I just got it today, but it does seem to be the same version.
Ok, I compared your results to my 2 week old results, yours don't have the User Interface Test. All the other tests except for the Disk tests are the same as mine. I'm guessing they've done a silent update to XBench by removing the UI test rather than fixing any code.

I'm at work so can't test this on my own iMac. That'll have to wait till to get home in 4 or 5 hours. What's your hard drive? I'll check that too when I get home.

[jsw]

What's your hard drive? I'll check that too when I get home.
Stock 160GB model:
Maxtor 6L160M0:

Capacity: 152.67 GB
Model: Maxtor 6L160M0
Revision: BACE1GE0

[plinden]

Maybe they did change something recently. Dunno.

Result using "Disable Beam Synchronization" (apparently, it had been set at "automatic" before):
Results 110.72
...
User Interface Test 239.28
Elements 239.28 1.10 Krefresh/sec
...

Ok, this is confusing. UI test shows up now. So they've modified the code to do the UI test if beam synchronization is off? And possibly modified the disk tests. Agghh, why do I have to work?

Edit: shows how much you can trust xbench ...

[plinden]

Stock 160GB model:
Maxtor 6L160M0:

Capacity: 152.67 GB
Model: Maxtor 6L160M0
Revision: BACE1GE0

So, I removed xbench from my iMac, and downloaded it again from the website, but got the same results as before.

Ostensibly, you and I have the same iMac (17", 1.83 GHz, 160 GB HD, 2 GB RAM), but your hard disk is a Maxtor. Mine's a Seagate, ST360023AS.

Anyway, I've wasted enough time with xbench. It's obviously not worth using, if modifying a single config item increases the score on my iMac from 55 to 90.

[risc]

Play a 1080p H.264 video via QuickTime and tell us if it drops any frames since Apple are saying these are home theatre boxes.

[Eidorian]

This is not all that fair though because Handbrake is using the GPU to a VERY large extent and GPUs with h.264 decoding built in (X1600) will dominate without showing you how fast (all around) your comupter is. You should do mutliple benchmarks if you want to be fair like an xcode build or something... I guess the important thing is that if you are seeing how fast the intel mac is in optimal conditions or under normal use... if you are seeking the latter than maybe a THE photoshop test or iTunes encoding as well... just some suggestions...Is HandBrake GPU intensive at all? I can't find documentation on it's GPU requirements. :confused: I know you need a G4 for the stable official releases.

I'd test iSquint, Handbrake, iMovie/Quicktime encoding on it. Then I'd follow up with 1080p trailers for fun.