View Single Post
Old Dec 9, 2011, 12:07 PM   #5
caveman-jim
Deposed King of Rage3D
 
Join Date: Oct 2003
Posts: 48,947
caveman-jim doesn't need no stinkin' badgescaveman-jim doesn't need no stinkin' badgescaveman-jim doesn't need no stinkin' badgescaveman-jim doesn't need no stinkin' badgescaveman-jim doesn't need no stinkin' badgescaveman-jim doesn't need no stinkin' badgescaveman-jim doesn't need no stinkin' badgescaveman-jim doesn't need no stinkin' badgescaveman-jim doesn't need no stinkin' badges


Default

Quote:
Originally Posted by Napoleonic View Post
I'm still confused with APU hype, what real world advantage it has compared to traditional IGP? (assume both use the same GPU)?
Why would you make that assumption? It's nonsensical. But look at Brazos and the low power E-series APU's vs. a CPU + IGP setup. Way more power to get a cpu + chipset with IGP, and harder to make in small form factors like netbooks and ultra portables.

But, that aside, a couple of differences is the shared memory and the balance of TDP. For compute, the CPU cores and the GPU cores can share the same memory - CPU does work, GPU does work; there's no latency/waiting while the dataset is copied from system ram to GPU ram and back again.

For power, if you have a CPU + IGP then both of those chips have a TDP budget. For an APU, that budget is shared, so when the GPU isn't being used heavily it can be clocked down and power saved and that power allocated to the CPU side. Conversely, when GPU power is needed but not CPU power, the GPU can get Turbo'd in the APU and better performance seen.
caveman-jim is offline