Panther benchmarks repository
People here have been claiming how so much faster Panther is compared to Jaguar/Puma. We all want to see some Panther benchmarks comparing it to Jaguar(or Puma). Since we're comparing the OSs, please report results from both Panther and Jaguar/Puma.
You may choose your own favorite apps to benchmark, but here are some suggestions:
1. Let1kWindowsBloom -- Probably the defacto standard today in measuring the "snappyness" of the window system. Simply and compatibile from OS 8.6 through Panther.
2. XBench
3. Skidmarks suit aka. Basher benchmarks (found on Apple developer tools CD)
4. CineBench2003
5. Boot times between similarly configured Jaguar/Puma and Panther installs
6. Journaling performance on/off in Panther versus Jaguar by duplicating a folder with hundreds of items in it.
Don't forget to report your Panther build number along with your results! It's critical you include this with your Jaguar/Puma scores so we have a basis for comparison.
You may choose your own favorite apps to benchmark, but here are some suggestions:
1. Let1kWindowsBloom -- Probably the defacto standard today in measuring the "snappyness" of the window system. Simply and compatibile from OS 8.6 through Panther.
2. XBench
3. Skidmarks suit aka. Basher benchmarks (found on Apple developer tools CD)
4. CineBench2003
5. Boot times between similarly configured Jaguar/Puma and Panther installs
6. Journaling performance on/off in Panther versus Jaguar by duplicating a folder with hundreds of items in it.
Don't forget to report your Panther build number along with your results! It's critical you include this with your Jaguar/Puma scores so we have a basis for comparison.
Comments
Let1KWindowsBloom is bogus for Quartz Extreme. Even the developer says so.
CineBench is bogus for the G5. Even the developer says so.
Expect a 'fixed' version of CineBench out soon, but L1KWB isn't likely to ever be a good benchmark of Quartz. Let it pass quietly into that good night, sweet prince.
Thought I'd save everyone some time, both in testing and in the inevitable arguments that would follow.
XBench is a decent mark, but even that team has requested a week or so more to ensure good consistency with the G5... which, I might add, is a (supposedly) shipping product. Panther ain't.
Boot time? Sure. It's simple, although likely to vary between b44 and 1.0.
The benchmarks will be tweaked, updated, and outright modified when Panther hits the streets for real. Until then, they're at best raw guesses in the dark.
*shrug*
But hey, whatever floats your boat.
Post dem scores.
Originally posted by Kickaha
*sigh*
Let1KWindowsBloom is bogus for Quartz Extreme. Even the developer says so.
CineBench is bogus for the G5. Even the developer says so.
Expect a 'fixed' version of CineBench out soon, but L1KWB isn't likely to ever be a good benchmark of Quartz. Let it pass quietly into that good night, sweet prince.
can you post sources for your claims that Let1KWindowsBloom and CineBench are bogus for QE and the G5 respectively? I couldn't find anything by the authors on their respective websites. Thanks.
Originally posted by cygsid
can you post sources for your claims that Let1KWindowsBloom and CineBench are bogus for QE and the G5 respectively? I couldn't find anything by the authors on their respective websites. Thanks.
Oh lord. The L1KWB fiasco was when Jaguar came out. Try searching back in the forums a bit. Essentially, L1KWB tests the Finder's implementation of messaging more than it does the actual windowing system, IIRC the final conclusion. It's rather like PPC vs. x86 benchmarks... the benchmark is only really relevant when comparing the same architecture... in this case, OS 9's Finder vs. OS X's Finder. There are too many variables to really state that it still tests what it was intended to test. Now that the new Finder is Yet Another Substantial Update, the situation is yet more muddled.
The CineBench one is a bit more recent: http://forums.appleinsider.com/showt...ight=Cinebench
Check towards the bottom for links to discussions on other boards that are quite a bit more detailed. It boils down to: "The G5 sux!" "No, CineBench doesn't take advantage of any of the new architecture, and..." "The G5 sux!" "Uh, no, CineBench needs to be at *least* recompiled with some decent optimizations before we can..." "The G5 sux!" *WHAP*
But, given that we all understand that such benchmarks really don't mean a whole heck of a lot most of the time, post dem scores. Should be interesting at the least, if not illuminating.
Especially the user interface-test is nice:
User Interface Test\t185.36\t
\tElements\t185.36\t63.07 refresh/sec
Beats the reference dual 800 clearly!
Edit: Re-ran the CPU-test while having a bunch of applications open, plus playing mp3, and the vecLib FFT-test now chimed in at
vecLib FFT\t122.30\t1.88 Gflop/sec, 5 times my initial results.
Originally posted by Zapchud
Xbench 1.1 results for a Dual 1GHz running Panther 7B44 (don't ask me about the disk-test):
Turn journalling off?
Updated results
Download it and run it on your G5s and your G4s and your G3s. But especially on your G5s. Go, go, go, go, go!
Barto
this is Xbench 1.1 on a DP1GHz MDD running the newest build of Panther:
Promising results
Originally posted by Existence
Xbench 1.1 is out and is optimized for the G5. Now only if moderators would stop detrailing threads!
Kwitcherbitchen, nancy boy and post your own scores.
Alright, going to try merging these suckers. Hold on, everybody.