Tiger seems faster than Panther...

D

Durbrow

Registered
This French site claims to have tested 10.4 v. 10.3.8 on various machines and reports Macbenchmarks at:

http://www.mactouch.com/systeme/dossier_tiger_article78.html#sommaire_1

I think the longer bars indicate better performance. It seems that 10.4 is slightly faster than 10.3.8 IF I AM READING THE CHART CORRECTLY. However, it is only slightly faster on G5s. Any ideas why? Should we take this chart seriously? Perhaps the differences are not perceptiable for the average users.
 
CPU, threading, memory and UI elements improved approx 30% on the G4 1.25 GHz and 50% on the G4 1.67 GHz. Wow! This alone feels like it's going to be worth the upgrade for me as I run loads of numerically intensive code.

It is an XBench graph, so larger is better. The thing that would be interesting is to see the breakdown of the CPU and threading tests. That way, I'll get to know which parts have actually improved.
 
It definitely seems that the G4 systems got the biggest boost (Which is noticable). Although the G3 system also got a good boost in the UI department... sweet.
 
Surprising. And how does tiger feel like in regular use? I am not a fan of benchmarks. Tiger might be faster in certain easy tasks, but it's completely useless if the system is "stuffed" with eye candies that keep the idle status of the cpu at 30% or so..
 
isn't/wasn't tiger meant to be bloody revolutionary for the 64-bit g5 users? it's the first fully 64-bit compatible system, and the g4s fare better. even g3's work better, and there was some discomfort about whether g3's would even work at all before....
 
right after the release of the G5 processors there were many many discussions if there will be any performance gain at all when using 64bit instead of 32bit. I wasn't expecting any increase. The benefit in 64bit mainly lies on the support for more ram. Currently a 64bit G5 system can have up to 32gb of ram.
Another thing - I would guess - is that the benchmark tools are still running 32bit.
 
Tiger isn't fully 64 bits. As Zammy says, 64 bits doesn't increase the speed of applications, unless your applications really needed more than 4GB of RAM to begin with.

It doesn't make sense to move everything to 64 bits just yet. On the PC side, the move to AMD64 made code run faster due to one of the deficiencies of the x86 architecture. x86 processors have 8 general purpose registers (GPRs). You can think of this as very very fast local storage that hold data and instructions that the CPU is currently working on. 8 GPRs are far too few, and hence the CPU will have to resort to fetching things it needs from the cache. AMD64 increased this to 16 which is why you get a speed boost when moving to 64 bits, because the CPU doesn't need to access the cache as often.

On PowerPC, there are already 32 GPRs. This is true for all the 32 bit processors and 64 bit processors. As such, the move to 64 bits isn't going to benefit PowerPC as much as it did x86.
 
oh all right. but when did you last buy a computer for yourself with those preinstalled. they are very specialized OS's. the mainstream is xp, MacOS and possibly the bigger versions of Linux, but even then, linux is a tiny market share. of those, Tiger is the first 64bit system. windows 64, as we all know, is apparently shite
 
Linux market share is about as big as the Mac's. Tiger isn't the first 64 bit system. It isn't even a true 64 bit operating system. There's no shame in admitting that.

Those operating systems that Lycander mentioned are not specialized nor are they uncommon. Solaris has been there for as long as anyone can remember, and is currently picking up loads of steam on the desktop after being thoroughly trounced by Linux for a while.

Tiger isn't the first 64 bit OS. Nothing wrong with admitting that, and let's not fall for Job's venerable RDF. This doesn't mean that Tiger isn't a good OS, or that it isn't worth buying.
 
Again... you are incorrect. Look I'm not picking on you for being a Mac faithful. I'm just arguing the facts. There have been quite a few Linux distros that are *fully* 64-bit compiled kernels with 64-bit userland apps. This goes back maybe 3 or 4 (or more) months ago. Viro just said that Tiger isn't fully 64-bits. The core may be 64-bit but the GUI parts aren't so no 64-bit GUI apps unless you do some IPC trickery.

Also, Linux is a lot bigger market than you think. :) Possibly bigger than the Mac actually.
 
more and more off topic... could we please stay on topic of comparing the benchmarks and wondering why the UI of that iMac DV 400 got such a boost? ;)
 
On the other hand, that kinda agrees with those who said that 10.0 was HIGHLY inefficient. Maybe it just _was_ too early back then to move to all that eye-candy with layers and shadows and transparencies? I guess a current graphics card could render an OS 9 kind of interface with as little as 1% of its power. ;)
 
Should we take the chart with a grain of salt? I did not read the pdf file but it could be that the speed gains are due to differences in RAM speed and amount. Also, I assume this is not the official version of Tiger. Sorry to be a kill joy.
 
You mean like this was some sort of fast beta, but they found the bug and the final version will be slowed down as usual? ;) ... Or did you actually mean that the final version of Tiger would be even faster... I personally guess that the tests were run on 8A425, and that 8A428 or newer builds are bug fixes only, since 8A425 was a final candidate.
 
After playing around with XBench I don't think I trust these results. They don't add up. The performance of the 1.25 Al book is way off, if you ask me.

The CPU, Thread, memory, Quartz and perhaps the OpenGL results don't look right to me.

On 10.3.8, the Al book in the article gets {94,80,106,136,93} for {CPU, Thread, Memory, Quartz, OpenGL}.

On 10.3.8, my Powerbook 12" G4 1.33 GHz gets {151,117,135,162,113} when set to "highest" performance and hard drive doesn't spin down, {85,76,117,131,106} when "automatic" speed is set, and {81,60,119,110,62} when "reduced" performance is selected.

It is most likely that their Powerbook isn't set to "highest" performance. Instead, they are most likely set to "automatic" since this matches best with the performance of my Powerbook on the same setting. If that is the case, the benchmarks don't really tell us how much faster Tiger really is.

The reason? It could be that Tiger has better power management that Panther, and hence is able to ramp up the CPU speed faster during performance intensive periods. The fact that the G4 Powermac and G5s and all the non-laptop Macs don't exhibit this 'improvement' lends credence to my theory. After all, desktops don't use the aggressive power management that laptops do. Neither do they throttle the CPU speed since they don't need such tricks to prolong battery life.

If true, this would be rather disappointing, especially since many understood these graphs as really improving the speed of existing hardware.

Shame... :(
 

Attachments

  • Xbench results.zip
    13.1 KB · Views: 0
Viro said:
...The fact that the G4 Powermac and G5s and all the non-laptop Macs don't exhibit this 'improvement' lends credence to my theory...
Click to expand...

Maybe it is me just being stupid... but the one thing that really stood out to me from these benchmarks is the improvement to the speed of the User Interface. Over all platforms: desktop and laptop, the other improvements didn't seem that consequential to me.

Personally I was hoping for a better result in the Quartz benchmarks from everything that has been said about the new vector acceleration in Tiger.
 
You must log in or register to reply here.
Back
Top