CONCLUSION
The overwhelming consensus is that the 6600GT (both AGP and PCI-E varieties) is
an excellent value by having the features and similar performance to the last
generation's high-end products. All of this comes in a smaller card with lower
power requirements, quieter cooling, and (most importantly) a $200USD price
point.
I'll not argue with any of that but add that while EVGA's offering
in hardware is really no different than many vendors, they are setting
themselves apart. EVGA is doing this with stellar product support and a new
1+1 Year
Warranty Program. All you have to do is
register your card
online and you'll then receive an additional one year limited warranty for
free.
The performance gains going from a single 6600GT to an SLI setup
are clearly varied. You just have to ask yourself if the somewhat reduced performance from the
initial expectations are worth the investment in two cards. We're still at a
relatively early stage in that
NVIDIA currently only lists 16 games that are officially scalable with SLI.
They have also clarified their statement about the expectations for SLI.
The applications in this list scale by as little as 22% and as much as
99%, with the average scaling of these applications being more than 50%.
That statement appears to be very accurate as the benchmarks in this review
clearly illustrate. So, it looks like things on the SLI front are maturing and
improving. A
50% (on average) improvement is nothing to sneeze at.
6600GTs in an SLI setup do have some very bright points. There
also appears to be room for improvement. The upside is that future drivers
improve performance so things can only get better. One exciting fact is
how well Doom 3 already shines with SLI. Upcoming games based on the Doom
3 Engine could hold a lot of promise when paired with SLI.
A single 6600GT is very wise choice for many with its ability to
maintain ~30 frames per second at 1600x1200 with 4xAA/8xAF in Half-Life 2 and
Doom 3. The decision to go SLI is not as clear; hopefully this review provided
some help though.
January 22, 2005 - Clarified that the timedemos were run
multiple times and an average was then taken to achieve final frames per
second for the final result on page 2. Also clarified that anti-aliasing and
anisotropic filtering settings were set to "Application Controlled" within
the Forceware Driver Control Panel.
January 22, 2005 - Updated the Doom 3 benchmarks on page 2
with 1024x768 and 1280x1024 timedemo results to help illustrate the memory
bandwidth restriction points.
January 23, 2005 - Posted
some more pictures of the SLI setup in our forums.
January 25, 2005 - Possible incorrect results shown for Doom 3 demo1
in SLI mode at 4xAA/8xAF on page 2. Confirmed/corrected results will be
re-posted soon.
January 25, 2005 - Confirmed that the Doom 3 demo1 results on page 2
were flawed and have since been corrected. You can read more on this
here.
