Geekbech completely unreliable tr4sh?

[damndoe]

Hi,

I noticed that my Geekbench scores went down over time and did not know why. The more I tested the more they went down. I tested by killing all processes etc having a mostly bare OS X etc.

Today I tested it WHILE running Ableton Live 8.2 Suite, a Pro audio application (DAW).
What I found were ... wow NICE Geekbench score? I even had 3GB samples loaded... I wondered WTF... why? So I went on and quit live and re-ran the benchmark. Lots lower score again... switched on Live again... higher score.

=> Geekbench is utter crap. Stay away from it. Useless. Its useless to measure anything. We have to find something else (like Prime maybe? Quake3 (there are some Open Source ports), NovaBench (seems pretty reliable)... Cinebench works awesomely for testing imho.


Here are 3 takes of Geekbench running while Ableton Live runs.

Here are 3 takes of Geekbench running without Ableton Live running.

p.s.: You could say it is OS X's fault, I will test this later on my MBP. If its the same its not OS X fault. Its a standard OS X on a standard MAC => Geekbench's fault then.

 

[damndoe]

3 Additions

1. I tested it with Logic 9 too. Its the same. If Logic 9 is running (empty project, like with ableton) I get 123xx to 126xx if its not running I get 114xx to 118xx.

2. PLEASE do me a favor and try to CONFIRM this IF YOU have Ableton Live or Logic Pro at least. If you do NOT try some other "intensive"(?) apps. I tried Garage Band.app as well. THE SAME. I launched it and ran Geekbench = 124xx, instantly after that I quit it and ran directly Geekbench = 117xx. Broken if you ask me. Gonna try MBP 2moro (if you can you can try the difference on a real Mac as well and report back).

3. Looking for better benchmark tools. I really want to optimize my system for max MAC PERFORMANCE.

 

[yrrr]

Hi there,

I have Live 8.1.1. Suite running stable on my hackintosh.
The Geekbench Scores are nearly the same with and without live (+/- 100) if I run Geekbench 64 bit.
With Geekbench 32bit it's of course a lower score, but also nearly the same.
I tested Geekbench a lot during overclocking my machine and like you say, I'd also absolutly NOT recommend it, if you want to measure your performance exactly. For a little overview it's ok of course.
Have you tried xbench instead?
http://www.xbench.com/
There you have a nice overview of the singe benchmarks for your drive, gpu, cpu etc.
So maybe this would be a better choice for your performance optimization.

good luck,
yrrr

 

[InActive8618]

Xbench is the worst app ever for benchmarking, not recommended at all. We hackintoshers really need a reliable app for this but I guess real Mac owners don't really care that much as they cannot overclock or mess with the system.

 

[damndoe]

Na... real mac owners care too because they care to measure MacBook Air vs MacBook vs MacBook Pro vs iMac vs MacMini vs MacPro etc.

Did you find Novabench to be a lot more reliable?

 

[InActive8618]

They sure care about that, but they are happy with a round number, they don't care if it's 100 +-, they just need to compare. Instead we mess with the hard, overclock stuff and want to constantly know if the result went up and we do care if it's at least a bit higher. In the end it's because we're a bunch of geeks that we're messing with unsupported OSes in the first place and working hard to get it running as we want.
I will try Novabench later to see if it works better but I've heard that it gives point for the hard disk size and that only would make it the worst ever.

 

[damndoe]

its bad anway that you do not get separate points for
- Processing Thoughtput
- Processing Latency (for Real time things)
- 3D Speed
- Harddisk Thoughtput
- Harddisk Speed

All different for different things.

 

[bemoll4-]

3 Additions

1. I tested it with Logic 9 too. Its the same. If Logic 9 is running (empty project, like with ableton) I get 123xx to 126xx if its not running I get 114xx to 118xx.

3. Looking for better benchmark tools. I really want to optimize my system for max MAC PERFORMANCE.

If we are talking benchmarking for audio apps I find evan's logic test really good for providing real world performance benchmarks... did you try it?

i'm curious to know how to optimize a hackintosh for audio, you're talking hardware, bios tweaks, or... ?

 

[damndoe]

CL vs Mhz
QPI vs CoreSpeed
Turbo vs No Turbo

The Logic 9 benchmark... well its not that good the question is when the barrier gets reached... when you have playback issues once or regular only the first time but then indefinetly its good etc.

 

[justr]

Nova bench is a bad test-- because of the HD/SSD size points.

When I went from a 500 GB 7200 RPM drive to a 60 GB SSD (Sandforce 285/275... FAST!)... my Novabench score actually dropped because all of a sudden it reads my boot drive as being so small.

Not only that, the ironic part is that this is in a machine with the following drives in addition to my boot drive: 2 X 500 GB in striped RAID (Video), 2 x 1 TB storage.

In any case, HDD/SSD size should have ZERO impact on benching scores. Truth is, even HDD/SSD speed should probably be kept separate-- because loading times is one thing, but testing the power of a system (generalization here) that one is tweaking should probably focus on pegging CPU, GPU, and RAM.