[svn:fx-3.x] 10899: Same fix as 10898, just for halo components.

Question

Revision: 10899
Author:   [email protected]
Date:     2009-10-06 14:11:50 -0700 (Tue, 06 Oct 2009)
Log Message:
Same fix as 10898, just for halo components.
QE notes: -
Doc notes: -
Bugs:  SDK-23263
Reviewer: Darrell
Tests run: checkintests, rob ran mustella tests for Window/WindowedApplication in the trunk version of this checkin
Is noteworthy for integration: no
Ticket Links:
    http://bugs.adobe.com/jira/browse/SDK-23263
Modified Paths:
    flex/sdk/branches/3.x/frameworks/projects/airframework/src/mx/core/Window.as
    flex/sdk/branches/3.x/frameworks/projects/airframework/src/mx/core/WindowedApplication.as

Answer 1

Quote
Originally posted by Assaf
Why use 3D Mark as a measure memory preformance ?
3D Mark 2003 isn't affected by the system anyway, it's a pure GPU test.
It resaults the same on a PII350 and on a P43.0 with the same VGA card.
Use SANDRA memory bandwidth to test memory bandwidth, your RAM should be running at 333MHz dual channel giving a mak bandwidth of 5333MB/S.
It should be pretty close to that on the FX/opteron because of the efficient memory controller, but not exactly 5333. Anything above 2666MB/S (~2700) means your in dual channel - as it should be with 4 sticks.
Actually, that's not true.  People tend to call it "dual channel" because the performance scales to twice as the memory bus data width has been increased to 128 bits wide.  Just like you used to have to put 2 32 bit dimms in a Pentium board, you do the same thing for an Opteron.  But since we're better programmers now (joke), you can opt to leave a dimm out and fall back to 64 bit wide mode for easier user compatibility.
Trust me, I've got datasheets, and it doesn't say dual channel anywhere.  AMD doesn't even us the term "channel."