: News
: Reviews
: Editorials

: Archives
: About
: Contact
: Advertising
: Privacy
: Links
: Forums

Gigabyte 8600 GTS Silent Pipe 3

Quiet Performance

by Josh Walrath

Test System Setup

            Since this is a midrange card, I thought to test it on a midrange system.  No C2D X6800 for this one!  Most other reviews have tested this board against many current cards, such as the aforementioned X1950 Pro.  I thought I would do something a bit different.  I took an old 7800 GTX clocked at 450 MHz core and 625 MHz (1250 effective) memory.  This lines up with the slightly newer 7900 GT.  I really wanted to see if the 8600 GTS would beat the two year old “high end” card from NVIDIA, much as the 7600 GT beat the two year old 6800 GT/Ultra.  The memory bandwidth of the 7800 GTX is at 40 GB/sec, which is 8 GB/sec faster than the 8600 GTS with its 128 bit memory bus.

The full bundle for all to see.  Note the power adapter in the middle.

            Since NVIDIA has driver revisions all over the place, I really couldn’t use the same ones for each card.  For the 8600 GTS I used the 158.42, which fix some issues previous drivers had.  For the 7800 GTX I used the 158.18s, which are the latest available for the GeForce 7 series.  I also used Windows Vista 32 bit edition, with the latest DX10 redistributable (which is incidentally required for the Lost Planet demo).  Texture mip map quality was set to High Quality for both cards.

            To measure power usage I had a Kill-a-Watt from P3 to record current power usage.  This part was installed between the wall and the computer, speakers and monitor were attached to a different outlet.  For temperature I used a Bonjour digital laser thermometer (accuracy within +/- 1.8 degrees C).

 AMD Athlon 64 X2 5000+

BFG 590 SLI Motherboard

2 GB SuperTalent DDR-2 800 (4:4:3:8 timings)

Seagate 7200.10 320 GB SATA II Hard Drive

Toshiba DVD-ROM

Onboard Realtek HD Audio

600 Watt Thermaltake PurePower Supply

Windows Vista 32 bit

Results

3D Mark 2005

            This benchmark is primarily vertex bound, and as such it makes for an interesting comparison between the 8 vertex shader GeForce 7800 GTX and the unified approach to the 8600 GTS.  In an older benchmark such as this, we would expect memory bandwidth and fillrate to be more important than shader power, but the results are quite surprising.  I used the default settings for the benchmark (1024 x 768, no AA, no AF).

            The difference in scores is amazing using this synthetic benchmark.  As I mentioned above, 3D Mark 2005 is notoriously vertex shader bound.  We can see that the 8600 GTS is obviously evening out the pixel and vertex usage in this application, and is simply stomping all over the older 7800 GTX.  This does not mean that the 7800 GTX or 7900 GT users should rush out and get an 8600 GTS, as most games are not vertex bound like this benchmark is.  Unless of course the user only plays 3D Mark 2005… 

3D Mark 2006

            This latest benchmark is not nearly as vertex bound as 3D 2005, but it does make heavy use of complex pixel and vertex shaders as well as HDR content.  We would expect in this synthetic benchmark to see the 8600 GTS outrun the 7800 GTX due to its unified and powerful shader approach, as well as the reworked ROPS that are more able to adequately handle FP16 HDR formats.

            Here we see a little bit more parity than 3D 2005, mainly because 3D Mark 2006 is a lot more balanced between pixel and vertex shader use.  We also see the 8600 GTS still outperform the 7800 GTX by a wide margin.  The 7800 GTX does feature more fillrate and bandwidth than the 8600 GTS, and it does have 24 of the more traditional pixel shaders and 8 vertex shaders running at 450 MHz.  Still, in a controlled condition such as 3D Mark 2006, the 8600 GTS does take a nice lead.

            Synthetic benchmarks are exactly that, synthetic.  While they try to replicate what could be considered “next generation content” the fact of the matter is that we still live in a time when developers are more interested in traditional rendering combined with shaders.  This means that pixel and texture fillrate still score high in many current applications.  As such, the 7800 GTX should perform faster at higher resolutions than the 8600 GTS across a wide variety of applications.

Next:  More Results

Copyright 1999-2007 PenStar Systems, LLC.