Tiger Performance


Logik

macrumors 6502a
Apr 24, 2004
616
0
sounds good for my 1.33ghz G4 15"... weee.. glad i preordered.. only 15 days.. WOOOO
 

Daveway

macrumors 68040
Jul 10, 2004
3,375
0
New Orleans / Lafayette, La
[singing praises] My iMac will be awesome with Tiger! [/singing praises]

Memory and UI management is great. 24% increase! I'm guessing this is from putting the UI on the vid card. But why is memory usage soo much better?
 

PlaceofDis

macrumors Core
Jan 6, 2004
19,232
4
Tiger looks more and more promising everyday, now to just get the money together so i can get it......
 

JFreak

macrumors 68040
Jul 11, 2003
3,145
0
Tampere, Finland
stoid said:
I'm certainly looking forward to a nearly 33% speed increase across the board on my 1.25 Ghz PowerBook!! WooHOOO!!
me too, but i cannot unilize it until digidesign approves tiger for protools systems... i'm hoping it won't take so many months ;)
 

Platform

macrumors 68030
Dec 30, 2004
2,881
0
Nice, and this is just twaks from an OS and with all the thigs that it can do :rolleyes: wow :D
 

RandomDeadHead

macrumors 6502
Feb 8, 2003
454
0
fennario
Great, the crippled HD performance of the G5 iMac gets another hit with 10.4, COOL!

Now a 7200 rpm drive on sata in the iMac G5 is only 3 points faster than a 5400 rpm drive on ata in a Pbook. Bottleneck anyone? :mad:
 

bpd115

macrumors 6502a
Feb 4, 2003
808
42
Pennsylvania
RandomDeadHead said:
Great, the crippled HD performance of the G5 iMac gets another hit with 10.4, COOL!

Now a 7200 rpm drive on sata in the iMac G5 is only 3 points faster than a 5400 rpm drive on ata in a Pbook. Bottleneck anyone? :mad:
Why exactly are the HDs in the iMac slower? I xbenched my fathers 1.42 G4 iMac and the 1.8 G5 and the HD scores were better on the G4...
 

Hemingray

macrumors 68030
Jan 9, 2002
2,913
25
Ha ha haaa!
Sky Blue said:
it only has the iBook 1.33 once?
Yeah, figures they'd have all the major models except the one I'm interested in... :rolleyes:

Oh well, I'd imagine there'd be some kind of benefit on the iBooks too, judging from the other scores (that graph leaves much to be desired!)
 

mkrishnan

Moderator emeritus
Jan 9, 2004
29,641
12
Grand Rapids, MI, USA
You guys realize I'm going to have to pre-order Tiger just so I don't have to miss out on all the fun when you're all playing with it at the end of April, right? I guess I can count it as half an entertainment expense, and half a day-early birthday present. :)

Devman, thanks for the table -- it's really interesting. I wish it had two columns for at least one kind of iBook, though. Preferably *looks around innocently* a G4/12"/800MHz/640MB. :D
 

wowoah

macrumors regular
Jul 16, 2003
187
0
Berkeley, CA
Sorry, but I'm colorblind and I can't really read the chart. I'm guessing we're all whooping and hollering because there's some pretty big speed increases? I have a TiBook 1GHz, can I expect an increase too? :)
 

daveL

macrumors 68020
Jun 18, 2003
2,425
0
Montana
invaLPsion said:
Sorry to play Devil's Advocate, but those results were done with xBench and may not be entirely reliable. Looks good though! :D
I agree. Xbench is junk, from my experience. For the G5 1.8, the Quartz and memory test results between the PM and iMac make no sense at all. BTW, I'm not saying you won't see a good performance boost from Tiger.
 

FoxyKaye

macrumors 68000
Well, the closest thing on the chart to my system is the iMacDV, which posted teeny speed increases across the board.

I guess it makes sense - teeny increases for a teeny processor, but it is heartening to see that Tiger seems to have avoided the WinXP bloat factor.

Maybe I'll buy it after all... :)
 

Soulstorm

macrumors 68000
Feb 1, 2005
1,887
1
Actually, I believe the results, despite the fact that xBench is not so reliable. I have heard that many people have observed the same performance tweaks as the chart says.
 

invaLPsion

macrumors 65816
Jan 2, 2004
1,385
0
The Northlands
Soulstorm said:
Actually, I believe the results, despite the fact that xBench is not so reliable. I have heard that many people have observed the same performance tweaks as the chart says.
Yep, I'm sure we will see speed increases, hopefully in OpenGL! :)
 

mkrishnan

Moderator emeritus
Jan 9, 2004
29,641
12
Grand Rapids, MI, USA
sorryiwasdreami said:
Not bad! My system's IHA user interface will be 56.4% higher! Btw, what does IHA mean?
Here are descriptions of the tests, from XBench...I think *IHM* is the French version of the acronym HMI in English (Human-Machine Interface). The list starts at two because one is just a system profile. :)

2. CPU Test
Single processor only - a test of one application doing single-threaded work
A) GCD Recursion - almost entirely limited by the processor's register speed, L1 cache, and integer math
B) Floating Point Basic - measures single precision floating point operations (+, -, *, /)
C) AltiVec Basic - measures single precision floating point operations implemented with AltiVec operations
D) vecLib FFT - measures system vecLib Fast Fourier Transform performance
E) Floating Point Library - measures double precision math library operations like sin, cos, sqrt, etc.

The "FLOP" rating is not directly comparable to theoretical FLOPs, or a real-world FLOPs produced by another benchmark.

3. Thread Test
tests multiple processors - a test of multiple applications, or a single app doing multithreading
A) Computation - measures 4 worker threads performing integer operations, plus some memory bandwidth
- always faster on MP machines
B) Lock Contention - measures 4 threads quickly acquiring and releasing thread locks
- usually faster on MP machines
- locking performance may be important in certain types of multithreaded code.

The thread tests aren't representative of all types of threaded code, but they do measure some of the factors that affect threaded applications.

4. Memory Test
tests ability to perform memory operations
A) System
a) Allocate - measures the system's ability to allocate many varying-sized blocks of data, using standard system calls
b) Fill - measures the system's ability to fill a large block with data, using standard system calls
c) Copy - measures the system's ability to copy data from one block to another, using standard system calls
B) Stream (derived from the standard STREAM benchmark - http://www.cs.virginia.edu/stream/)
These use 64-bit doubles and AltiVec cache prefetching when appropriate
a) Copy - measures copying speed between 2 large buffers,
b) Scale - measures load-float multiply-store operations between 2 large buffers
c) Add - measures load-load-add-store operations between 3 large buffers
d) Triad - measures load-load-multiply-add-store operations between 3 large buffers (scale and add in one step)

These tests will do better with more memory bandwidth. On a system with very little memory, these may perform slowly.

5. Quartz Graphics Test
A) Line - measures drawing lines of varying widths, colors and rotations at 50% alpha
B) Rectangle - measures drawing rects of varying widths, colors and rotations at 50% alpha
C) Circle - measures drawing circles of varying diameter, colors and rotations at 50% alpha
D) Bezier - measures drawing beziers of varying widths, colors and rotations at 50% alpha
E) Text - measures drawing characters of varying font sizes and rotations at 100% alpha

This test does better with a better graphics card, more memory bandwidth, more CPU, altivec, etc. It's up to the system how to optimize the drawing.

6. OpenGL Test
A) Spinning Squares - measures frame rates of a scene containing numerous semi-transparent tiles

This test does better on a better graphics card, or with a faster processor. As it measures windowed OpenGL, it may not accurately measure the performance of full-screen OpenGL.

7. User Interface Test
A) Elements - measures redraw rate of a window containing typical system controls

This test does better on a better graphics card, or with a faster processor. It measures the operating system's performance drawing standard system controls.

8. Disk Test
A) Sequential
These tests measure typical throughput to the drive.
a) Uncached Write - measures writing in 4K and 256k blocks until a 100MB file is filled
b) Uncached Read - measures reading in a 100MB file in 4K and 256K blocks
B) Random
These tests will be more impacted by the disks's seek time
a) Uncached Write - measures writing in 4K and 256k blocks in random locations into a 100MB file
b) Uncached Read - measures reading 4K and 256k blocks at random locations in a 100MB file

Most disks have slower throughput as they become more full and/or fragmented, so an extremely full disk may perform worse than expected. In some cases, restarting a machine may correct inordinately low test scores.
 

Greencardman

macrumors 6502
Apr 24, 2003
489
2
Madison, WI
wowoah said:
Sorry, but I'm colorblind and I can't really read the chart. I'm guessing we're all whooping and hollering because there's some pretty big speed increases? I have a TiBook 1GHz, can I expect an increase too? :)
I don't think that is listed, but the Ti 800 might be close. The bars measure each system twice, once in 10.3 and once 10.4, so the first two bars (starting at the top) are the iMac Dv 400, then the next two are the Ti 800, then the 5th one is the ibook (its the only one not measured twice), and then the 6th and 7th are the Alu 15" 1.25 and on and on and I'm too lazy to list them all. Panther is always listed on top (ie. the first and third bars) and Tiger on the bottom (ie. the second and fourth bars).

If you can figure out what i just said, you're a genius, cause trying to describe a colored chart without using colors is freaking hard!
 

andrewfee

macrumors 6502
Aug 29, 2004
467
2
While it is XBench, I think if you're comparing the same system on a different os, the results should still be comparable; it's when you compare one machine to another that things start getting inconsistent.

For example: I've now got a maxed out 17" 1.67GHz Powerbook, and compared to my iMac, in everything other than hard drive speed, it feels faster, and more responsive. I don't know what a point on the Disk access actually represents, but the two aren't very far apart; that's just plain wrong. The iMac G5's hard drive is very fast from my experience, and while the Powerbook's drive isn't "slow" it's the only thing I notice as holding the system back in some tasks.

As for video performance; the Radeon 9700 Mobility is somewhere between a desktop 9600 Pro and XT in terms of performance; and should be much faster than the crappy iMac's nVidia 5200 card.

For example, in World of Warcraft, while it still doesn't run as well as I feel it should on this machine, my Powerbook is at least 3x the performance of the iMac G5. I couldn't play the game on my iMac; I was getting 10-20 fps outdoors with everything on its lowest setting, and averaging about 14. (with some options actually lowered in the configuration file below what the ingame options minimum settings are)

On my Powerbook I can run it at the same resolution, but with high texture details and I've yet to see the game drop below 20, and it's averaging about 27fps. If I turn up all the options to max other than the shaders and anisotropic filtering, I get 16-25 fps with an average of about 22 fps. (I prefer to sacrifice looks for a better framerate)

Memory performance doesn't surprise me one bit, although I've not actually seen it affect performance adversely on the Powerbook. (but then it's hard to know when memory is the bottleneck)



Those are some fantastic performance improvements all round though. While I don't think you can trust the figures with it being XBench, the percentage values will probably apply. As I said, I've got a maxed out 17" 1.67GHz Powerbook, so I should hopefully see as much as a 50% cpu speed increase. :eek:

I'm hoping gaming performance comes up a bit though. While WoW runs ok, I'm really wanting to play Doom 3 on the Powerbook. Yes, it may not really advance the shooter genre like Half-Life 2 did, but I just love the atmosphere it has, and id games are just plain fun.