Tiger Performance

[RandomDeadHead]

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...

Has something to do with the sata chip. A design flaw that wasn't noticed until the chip was well into production. Sad thing is I think Apple designed it and had them mass produced, than after realizing the flaw decided to use them any way.

 

[frozenstar]

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'm color deficient, but I was able to recognize that the legend is in reverse order compared to the chart.

I'd say you're looking at about a 25-30% increase in general system performance.

 

[mkrishnan]

These are Panther/Tiger XBench numbers from my computer. Panther was from 10.3.9 after a reboot and system cleaning activities. Tiger was after apps had been installed, also on a reboot with system cleaning.

Can anyone else replicate the huge boost in UI score? That certainly *feels* right. Still the lamentable state of my HD. If colors are hard to see, Panther is on the top and Tiger is on the bottom.

EDIT: oops, I didn't change the title. But you get the idea.

 

[daveL]

These are Panther/Tiger XBench numbers from my computer. Panther was from 10.3.9 after a reboot and system cleaning activities. Tiger was after apps had been installed, also on a reboot with system cleaning.

Can anyone else replicate the huge boost in UI score? That certainly *feels* right. Still the lamentable state of my HD. If colors are hard to see, Panther is on the top and Tiger is on the bottom.

EDIT: oops, I didn't change the title. But you get the idea.

I don't know what system you have, but your Xbench results differ fundamentally from the ones reported here:

https://forums.macrumors.com/threads/123249/

Which only goes to show why I completely disregard Xbench as a valid benchmark.

 

[mkrishnan]

I don't know what system you have, but your Xbench results differ fundamentally from the ones reported here:

Differ fundamentally? I'm not sure I understand. They're almost exactly the same. The PMG5 setup also had similar scores in most areas and a dramtic increase in UI. And as for my system, it's in the title of the plot -- iBook G4/800MHz/640MB RAM. Of course, the *raw* scores are different...if I were you, I'd place even *less* faith in any bench that gave a G4 iBook and a G5 PM the same performance rating.

Test	Panther	Tiger
Overall	146	144
(98%)
CPU	146	136
(93%)
Thread	90	89
(98%)
Memory	235	269
(115%)
Quartz	205	194
(95%)
OpenGL	158	135
(86%)
UI	216	307
(143%)
Disk	103	92
(92%)

 

[daveL]

Differ fundamentally? I'm not sure I understand. They're almost exactly the same. The PMG5 setup also had similar scores in most areas and a dramtic increase in UI. And as for my system, it's in the title of the plot -- iBook G4/800MHz/640MB RAM. Of course, the *raw* scores are different...if I were you, I'd place even *less* faith in any bench that gave a G4 iBook and a G5 PM the same performance rating.

Test	Panther	Tiger
Overall	146	144
(98%)
CPU	146	136
(93%)
Thread	90	89
(98%)
Memory	235	269
(115%)
Quartz	205	194
(95%)
OpenGL	158	135
(86%)
UI	216	307
(143%)
Disk	103	92
(92%)

Your Tiger Disk is higher, theirs is lower.
Your Tiger OpenGL is higher, theirs is lower.
Your Tiger Memory is lower, theirs is higher.
Your Tiger Thread is higher, theirs is lower.

Even though you obviously have a different Mac than the Geeks, you would think that, comparing 2 OS revs, the major categories would line up.

As I said, Xbench results just don't cut it.

 

[mkrishnan]

Your Tiger Disk is higher, theirs is lower.
Your Tiger OpenGL is higher, theirs is lower.
Your Tiger Memory is lower, theirs is higher.
Your Tiger Thread is higher, theirs is lower.

Even though you obviously have a different Mac than the Geeks, you would think that, comparing 2 OS revs, the major categories would line up.

As I said, Xbench results just don't cut it.

True, but most of those scores are so close together that I would guess they are within the margin of error. The only three that have big differences between my Panther and Tiger xbenches are Quartz, OpenGL and UI. You're right about the OpenGL -- I saw your comment in the other thread after I replied. That is a little weird. But really, even of those three, in both cases the huge change was UI, and it was very similar on both systems.

But anyway, you're very right that XBench is limited. I don't mean to dispute that. I'm just wondering if other people see any of the same patterns.

 

[daveL]

True, but most of those scores are so close together that I would guess they are within the margin of error. The only three that have big differences between my Panther and Tiger xbenches are Quartz, OpenGL and UI. You're right about the OpenGL -- I saw your comment in the other thread after I replied. That is a little weird. But really, even of those three, in both cases the huge change was UI, and it was very similar on both systems.

But anyway, you're very right that XBench is limited. I don't mean to dispute that. I'm just wondering if other people see any of the same patterns.

Just to be clear, the mixed results are not completely the fault of Xbench, per se. Benchmarks like this really should be run in single user mode, no network, minimum system daemons, etc. The problem is you don't have access to the GUI in single user mode. You could run the CPU, memory, threads, and disk tests, but none of the others. For example, how can you get reliable Thread benchmarks on a running system that has some unknown number of other active threads that exist outside the benchmark itself? Anyway, it just my take on it.

I'll take Barefeats test results anyday.