Not forgetting the software, the nView application included
with the current NVIDIA ForceWare drivers significantly adds
to the usability of the system. Anyone who has used a machine
with NVIDIA graphics under Windows would know that dual monitor
screens are a snap to set up. The nView display properties
gives an assortment of user interface enhancements including
virtual desktops, special effects like taskbar transparency,
display zooming, hotkeys, colour calibration, and more.
Our benchmark system of choice was BOXX
Technologies' 3DBOXX 7108. This dual-Opteron machine
was recently reviewed here on CGNetworks (link)
We used CINEBENCH 2003 to compare the relative performance
of the Quadro FXs in rendering through two scene sequences, "Pump
Action" and "Citygen", with and without hardware
lighting acceleration. 100 CB-GFX units represents MAXON's
reference 1GHz Pentium 4 machine in software render. All
tests were done with NVIDIA driver defaults and with OpenGL
In the software render all three Quadro FXs rendered at
the same rate (360 CB-GFX). The difference comes as hardware
acceleration is applied. In the graph CINEBENCH 2003
- Shading, the first benchmark is done with hardware
transform and shading, but with software lighting. In the
second test everything (transform, shading, and lighting)
is done in hardware. Here the Quadro FX 3000 won out, with
the Quadro FX 1100 not too far behind. The Quadro FX 500
certainly lagged behind, hardware lighting wasn't much quicker
than software lighting with the Quadro FX 500.
The bmark data set scenes were run in 3ds max 6 to gauge
application interactivity. Here again the Quadro FX 3000
took a lead in most of the tests, and a big lead in a few
such as 8spot, rasterize, and textured1 (as
shown in 3ds max 6 - Graphics Script Benchmark graph).
The Quadro FX 1100 trailed closely behind or matched the
Quadro FX 3000 in most of the tests, but lagged significantly
behind on a few others. The Quadro FX 500 lagged seriously
behind in most of the tests, seemingly bottlenecked in a
lot of the intensive ‘fill-rate’ tests.
The popular workstation graphics benchmarks suite SPECViewPerf
7.1.1 was used to help gauge the relative speed at which
the Quadros could handle workstation applications. The data
sets in SPECViewPerf are based off applications tests from
3ds max, Lightscape, DesignReview, Pro/Engineer, DataExplorer,
and UniGraphics (see SPECViewPerf 7.1.1 graph).
Here the Quadro FX 3000 took the lead, while the Quadro
FX 1100 wasn't too far behind. The Quadro FX 500, on the
other hand, definitely wasn't a hot 3D performer in most
|CPUs||Dual Opteron 248 (2.2GHz) |
Tyan Thunder K8W
2GB total (Four 512MB ECC DDR PC3200 modules).
Four 36GB 10,000 RPM Serial ATA (Western Digital
One drive used non-RAID. Remaining three drives in
RAID 5 array via RAIDCore controller.
|RAID Controller ||RAIDCore RC4852 Serial ATA (eight channel) 64-bit PCI-X|
|Software and Settings |
|OS||Windows XP SP1a |
|Video Driver||NVIDIA ForceWare 56.72|
3ds max 6 bmark
|Screen resolution||1280x0124 pixels, 32 BPP |
2003 – Shading test
and Price Performance
According to NVIDIA, all three boards are aimed for different
applications. From our testing it would certainly appear
true. The Quadro FX 500 (currently selling for approximately
US$210) is the cheapest Quadro FX available, and is more
suited to 2D business work and light-duty 3D applications.
The Quadro FX 500 GPU's feature set is the same as its more
expensive brothers, but isn't designed for heavy 3D work
where performance matters.
The Quadro FX 1100 (currently selling for approximately
US$630) came very close to the performance of it's bigger
heavier brother. This was surprising, considering it had
less than half the memory bandwidth but is evidence of the
advantage of its newer GPU.
The Quadro FX 3000 (currently selling for approximately
US$1300) is the card intended for 3D-heavy applications.
Considering its 256MB of memory in particular, this graphics
processor is best for applications with larger textures and
at higher resolutions. The Quadro FX 3000 does take up an
adjacent slot, but thanks to its memory bandwidth its performance
manages to edge out the newer Quadro FX 1100.
All three cards performed seamlessly. Our pick for most
applications is the new Quadro FX 1100 as it performs comparably
to the Quadro FX 3000 at half the cost. For cost-conscious
buyers, the Quadro FX 500 will suit light-duty 3D users,
but the GeForce FX 5200 or even GeForce 4 range may be adequate
for these users too.
Watch out for the Quadro FX 4000 coming out in a couple
of weeks or so. Even though the Quadro FX 4000's GPU is significantly
different to the existing Quadro FX product-line, it will
be slotted into the same family. Going by what NVIDA has
said, don't expect the current Quadro FX 3000 pricing to
change. So the new model may be very expensive indeed.