Valley appears to use slightly more Vram on my config, and -certainly- makes both cards in my SLI setup run hotter...
*at this time the Valley benchmark makes my cards run hotter than -anything- else I run on my config
 
Based on the limited runs I have done compared to the other testers I use, Valley has -way- more "points" at which to monitor performance levels and IQ...
As such I have added it to my standard testing and am using it alot
 
For testing drivers and performance I use the following order...
 
(at Stock videocard clocks)   
3DMark 11 (P) default preset
Unigine Heaven (max settings)
Valley (HD preset)
(I will also sometimes repeat the same tests with videocard overclocks)
 
*I also do IQ and FPS testing in the maps below for each new driver
Battlefield 3 Caspian Border map 64, Gulf of Oman 64 and (just added) Kiasar Railroad 64
I have found that BF3 is a MUCH better real world testing program, as drivers can pass the other tests flawlessly and then have "issues" in BF3 (most often when I overclock my videocards)
 
What makes Valley so good in my opinion, is the large amount of points at which to compare performance and IQ levels...
Mark 11 has about 15 points that really count
Heaven has about 30, but the variances between the points with different driversets are not very good sometimes...
Valley has over 50 points and I am finding more as I am running more drivers, and the differences are really stark. That makes it a ideal testing program as IQ, fps, ftr, draw distance and post processing (among others) are much easier to see and monitor.  
 
 
 

I can tell you that Heaven is picky as hell compared to other benchmarks. I can OC as high in Heaven compared to 3D mark 11 or Valley. So with that said it is stressful to "certain point" because of how picky it is for OCing. But anything else it depends.

Heaven is a bit more stressful IMO,