Sign in with
Sign up | Sign in
Your question

Windows 7

  • Windows 7
  • Windows Vista
Last response: in Windows 7
January 15, 2010 10:38:26 AM

I think I have found the reason Windows 7 has that boost over Vista even when they run the same kernel. I have noticed that my cuda apps run much slower that XP/Vista than in Windows 7 by as much as 30%. I have run this on 2 88000gt's, a GTX260, and a GTX285 with the same results, my cuda apps run much slower on Windows 7 almost like there is a cap, I have fought with this for a couple of months now trying to get Windows 7 to at least run my cuda apps like Vista so I can upgrade. I have also noticed that my wife's computer which run the same peocessor as mine and the same Ram (4GB) runs noticeable slower than my own even with the same specs for the exception of one thing I have 2 cuda enabled GPU's and she runs on a on-board video ( Nvidia 7300 ). I need someone to confirm these results. I run a semi large network (4676 machines ) and my boss wants me to upgrade the whole network based on the recent reviews. I have 2 problems with that being first that ANSOC has not approved Windows 7 (thank GOD ) and I am sure there will be a noticeable performance hit upgrading with our older hardware ( our specs have only on board video and almost none are cuda ) . We do have some cuda hardware on our network but us admin do scoop it up when we can. Can you confirm this because I believe MS pulled a fast one using direct compute to accelerate there O/S to get past the reviewers.

Can anyone confirm this ?

More about : windows

a c 209 $ Windows 7
January 15, 2010 3:55:04 PM

Basing a decision to upgrade almost 5,000 machines on issues with your home machine is a mistake, assuming the hardware configuration is different. What you really need to do is to get some machines that are representative of your user base and evaluate the upgrade using them.
a b $ Windows 7
January 15, 2010 3:59:14 PM

n4exodus said:
I think I have found the reason Windows 7 has that boost over Vista even when they run the same kernel.

They don't run the same kernel. (And most users don't run cuda applications, particularly in businesses.)