2006 Octo Mac Pro - 10.5.7 vs. 1.6 XBench Smackdown!

[JPamplin]

Hi Guys,

Just playing around with the WWDC SL Beta, and just wanted to get any kind of an idea what 10.6 has in it in terms of speed increases or decreases for us Mac Pro owners. What I found was interesting, and in one case, unexpected.

My setup is a modified 2006 Mac Pro with 2 quad-core X5355 Xeons (see sig) with 8GB RAM and an ATI 3870. I run 3 Seagate 750GBs in Bays 1-3 as a software RAID0 boot volume, and have another 1TB spanned to a 750GB in the lower CD bay for Time Machine. I ran XBench 1.3, and bolded anything I thought was interesting or significant.

What surprised me what the Floating Point score - 10.6 is DOUBLE the 10.5.7 score. CPU optimizations I guess, but how will that translate into real world/UI snappiness? Anyone?

The ATI drivers continue to disappoint, but I can't fault them YET about such crappy numbers.

I also attached a spreadsheet with the values if you guys want to play with it. Comments?

JP

 

[Eidorian]

Why so XBench? I'm surprised people use it.

 

[JPamplin]

Well, give me another free benchmarking tool (hopefully one that does 64-bit stuff too for comparison). Everything I've seen so far is commercial software.

JP

 

[Eidorian]

http://www.primatelabs.ca/geekbench/

http://www.realtech-vr.com/glview/

 

[sbb155]

http://www.primatelabs.ca/geekbench/

http://www.realtech-vr.com/glview/

yep, i wanna see geekbench and cinebench

 

[JPamplin]

The problem with GeekBench is a) it's a commercial program to get 64-bit benchmarking, and b) it only does CPU and memory benchmarks - no video, no hard disk, etc. Yes, you get video tests with OpenGL Extensions app, but that's just OpenGL, not Core Image benchmarks.

But, I'll pay the $20 and give you your 64-bit CPU-centric benchmarks, if that's what you want.

Back later,

JP

 

[Tesselator]

The problem with GeekBench is a) it's a commercial program to get 64-bit benchmarking, and b) it only does CPU and memory benchmarks - no video, no hard disk, etc. Yes, you get video tests with OpenGL Extensions app, but that's just OpenGL, not Core Image benchmarks.

But, I'll pay the $20 and give you your 64-bit CPU-centric benchmarks, if that's what you want.

Back later,

JP

And GeekBench is even worse than XBench. Those are two of the most sloppy benchmarkers on the planet. Seriously, I would be hard pressed to come up with anything worse on any platform. XBench is better than geekbench though - by miles. It's at least semi-consistent on the same machine. You've used it appropriately in post one IMO. It's good for showing the difference between different configurations on the same machine. It's not good for much else though. Although huge differences in focused tests between several machines can maybe imply real or actual differences.

I think I have all the benchmarkers made for OS X both commercial and free and the only ones I can agree don't totally suck are:

Speedtools - Commercial,
IO Gauge - Free (drives)
Cinebench - Free.

There's Magnus CPUTest (Free) that's good for testing loads and Haxial Benchmark (Free) that's kinda OK'ish for testing CPU+RAM speeds but it too produces inconsistent results almost as bad as XBench. Unfortunately we don't have any nice benchmarking tools on OS X. The PDS developers on OS X seem to be preoccupied with useless BS like writing files back and forth to their iPhones. Sad.

 

[Eidorian]

And GeekBench is even worse than XBench. Those are two of the most sloppy benchmarkers on the planet. Seriously, I would be hard pressed to come up with anything worse on any platform. XBench is better than geekbench though - by miles. It's at least semi-consistent on the same machine. You've used it appropriately in post one IMO. It's good for showing the difference between different configurations on the same machine. It's not good for much else though. Although huge differences in focused tests between several machines can maybe imply real or actual differences.

Please, please do tell.

I've had XBench vary wildy on the same hardware just with different versions of OS X. Under OS X and Windows I've had GeekBench scale accordingly. It doesn't test the full system hardware but it's more consistent in scaling.

 

[Tesselator]

Please, please do tell.

I've had XBench vary wildy on the same hardware just with different versions of OS X. Under OS X and Windows I've had GeekBench scale accordingly. It doesn't test the full system hardware but it's more consistent in scaling.

Do tell what?

That you are a single exception and lucky beyond belief? OK.

BTW, I'm NOT talking about the windoze version... obviously.


Anyway, I'm finished discussing it. You can show private anecdotals all you like but it won't change anything. I'm not trying to be an ass but I've shown this here not less than three times already. And many people who seem to know what they're talking about here at MR agreed. So, I'm kinda done with the topic myself.



.