Bandwidth Place 2

With pre-SP1, I got somewhere around 45GFLOPS, but definitely not 50+, and it is factual that Intel CPUs have gotten more efficient per clock on each generation. So linpack probably doesn't scale as well per clock.

I'd say Cinebench scales very well by clock speed, proof:
http://www.mylilsite.net/images/cineben ... latest.png
My CPU at 3.5GHz = 6.79, adjusted to 4.5GHz is 8.73. Actual 4.5GHz score is 8.71.

So going by your 2.8GHz = 4.83, you'd get 7.76 at 4.5GHz.

Just thought I'd mention my reasoning for why it would be highly unlikely you'd hit 50GFLOPS like in your projection lol

And yeah, AVX more than doubled my score, so it is amazing but probably a best-case scenario.

I know lol. I did say it probably wouldn't scale 1:1

OK, just figured I would continue on this thread. I was going through my settings and saw a spot that read "Total available graphics memory: 4094 MB".
My card is a "Dedicated video memory: 1023 MB GDDR5".
Why and how is "Shared system memory: 3071 MB"?
Not complaining, just wondering about the odd numbers.

NVIDIA System Information report created on: 06/17/2012 16:23:59
System name: **********

[Display]
Operating System: Windows 7 Home Premium, 64-bit (Service Pack 1)
DirectX version: 11.0
GPU processor: GeForce GTX 560 Ti
Driver version: 301.42
DirectX support: 11.1
CUDA Cores: 384
Core clock: 822 MHz
Shader clock: 1645 MHz
Memory clock: 2004 MHz (4008 MHz data rate)
Memory interface: 256-bit
Total available graphics memory: 4094 MB
Dedicated video memory: 1023 MB GDDR5
System video memory: 0 MB
Shared system memory: 3071 MB
Video BIOS version: 70.24.35.00.02
IRQ: 18
Bus: PCI Express x16 Gen2

Mine says the same thing and I never really thought about it :\ Google gets a lot of laptop GPU results which make more sense as those are much more likely to actually use shared RAM... Maybe its existence here is just a placeholder so their universal driver implementation doesn't need to have as many exceptions.