vista
(^) revisited
32 q MAXIMUMPC | oct 08 | http://www.maximumpc.com
The Benchmarks
We take a quantitative look at Vista and XP performance to determine exactly what penalty,
if any, you pay when you upgrade to Windows Vista
T
o test Vista versus XP performance,
we built what we think is a fairly
middle-of-the-road rig—an Intel
Q6600 quad core with 2GB of memory and
a GeForce 8800 GTS videocard. We then
ran a battery of benchmarks in three differ-
ent OS environments: XP with Service Pack
3, Vista sans Service Pack 1 (with modern
Nvidia drivers installed), and Vista with
SP1. Our tests measure everything from
overall system performance to network
speed to gaming prowess.
Overall PerfOrmance
Unsurprisingly, Windows XP remains
faster in almost all of our standard system
benchmarks. More noteworthy is how SP
has improved Vista’s performance, narrow-
ing the gap between that OS and XP in key
tests and even allowing Vista to surpass XP
in our MainConcept encoder test.
Unfortunately for Vista, our desktop
benchmarks do reveal areas where Vista
continues to suffer substantial performance
hits compared to XP, namely in ProShow
and Quake 4. We’ve talked to the ProShow
developers, and they don’t know what
causes the slowdown with their app in
Vista, but they’re investigating. We attribute
the Quake 4 performance hit to poor
OpenGL drivers in Vista.
As we mentioned before, we’re
perfectly willing to sacrifice a few
percentage points of performance from
an operating system upgrade. How-
ever, the difference between Vista SP
and XP SP3 in ProShow and Quake 4
reaches a dismal 10 to 25 percent.
GaminG
We didn’t include any DirectX 10 games
in our tests simply because DirectX 10
wasn’t around when Vista launched,
and DirectX 10 graphics still aren’t
supported on Windows XP. Our basic
system benchmarks already include
a pair of games, FEAR and Quake 4,
but we tossed in an additional round
of 3DMark06 to further assess Vista’s
gaming prowess.
The results were informative.
Aside from the already noted Quake 4/
OpenGL deficiency, Vista performed
admirably both with and without
SP1, turning in scores equivalent to
XP’s. This tells us that the poor gam-
ing performance we saw in the early
days of Vista was more the result of
immature drivers than issues with the
OS. Of course, Microsoft can still be
blamed for shoddy coordination with
the graphics-card makers at the time of
Vista’s launch.
netwOrk transfer sPeed
Our final set of benchmarks test
networking performance. We set up
the fastest NAS box we’ve ever tested,
the QNAP TS-109 Pro, and ran our
standard network storage benchmarks
on it. While we saw the same stunning
performance inadequacies from pre-
SP1 Vista that we observed at the OS’s
launch, SP1 and the subsequent up-
dates seem to have solved most of those
issues. The minor gaps of a few seconds
that do exist between XP and Vista SP
are explained by the fact that XP shuts
the file transfer window before the
transfer is confirmed, while Vista waits
until it has checked the copied file.
the takeawaY
With the exception of a couple outlier
applications, Vista’s performance is
within striking distance of XP’s, for the
most part. Thanks largely to a series of
performance enhancements and SP1,
Vista has closed the gap in many areas
where it was deficient. We’re willing to
overlook the poor OpenGL gaming per-
formance simply because there aren’t
very many OpenGL games coming out,
and it seems the ProShow problem is an
isolated incident.
Overall sYstem PerfOrmance
WindoWs XP (sP3) WindoWs VisTa (launch) WindoWs VisTa (sP1)
Premiere Pro cs3 (sec) 924 960 960
Photoshop cs3 (sec) 133 136 139
Proshow (sec) 963 1214 1275
Mainconcept (sec) 1881 1822 1814
Quake4 (fps) 143.5 126.5 125.
FEaR (fps) 65 65 65
GaminG Perf Ormance
WindoWs XP (sP3) WindoWs VisTa (launch) WindoWs VisTa (sP1)
3dMark06 Game 1 (fps) 29 28 28
3dMark06 Game 2 (fps) 26 26 26
netwOrk transfer sPeeds
WindoWs XP (sP3) WindoWs VisTa (launch) WindoWs VisTa (sP1)
network – small to nas (sec) 38 48 43
network – small from nas (sec) 39 68 42
network – large to nas (sec) 139 181 144
network – large from nas (sec) 140 271 142
Best scores are bolded. These are our standard system benchmarks, with one exception. We ran the games at 1920x1200 resolution, with 4x AA and 16x anisotropic filtering on FEAR, no AA and no anisotropic filtering on Quake 4.
Best scores are bolded. Our NAS test consists of two file transfers, which we refer to as the small and large tests. The small test is 659MB spread over 180 files, while the large test is a single 2.8GB file.
Additional reporting by Benson Hong