Fortunately, as technology progresses, computer power does get cheaper over time, and that’s certainly true in graphics. Case in point: NVIDIA’s brand-new GeForce 6800 GS. This video card, making its debut today, packs the considerable power of a GeForce 6800 GT at the not-so-painful price of $249. Does that combination of price and performance make the GeForce 6800 GS worthy of your consideration? We aim to help you answer that question by flooding you with benchmark results until your head swims and you cry out for sweet mercy. We’ve tested a number of the latest games, including F.E.A.R, Serious Sam 2, and Half Life 2: Lost Coast. Keep on reading to see how the 6800 GS compares to several of its closest competitorsand whether a pair of these puppies will keep up with one of those six-hundred-dollar monsters.
Gran Turismo versus… Good Stuff? General Smoothness?
If you’re like me, you may be a little bit surprised to see NVIDIA cranking out a brand-new GeForce 6-series video card as we head into the holiday buying season, given that the GeForce 7 series has been around for a while in several incarnations. The truth is that the GeForce 6800 GS is not a major new GPU revision in any sense of the word, though, and that the 6800 GT that it ostensibly replaces was most definitely in need of replacement.
The 6800 GS is based on a GPU that NVIDIA has dubbed NV42, and its specs are rather similar to another, already familiar NVIDIA GPU, dubbed NV41. You’ll find an NV41 chip lurking under the garishly decorated heatsink of any recent GeForce 6800 (no extended moniker edition) card with a PCI Express interface. Both the NV41 and the NV42 have five vertex shader units, 12 pixel shaders, and eight render back end units, or ROPs. The difference between them is that the NV41 was built by IBM using a 130nm fab process, and the NV42 is built using a 110nm process. Because the two chips are manufactured using different fab processes that rely on different libraries, NVIDIA estimates their transistor counts differently. Thus, the NV41 purportedly has 190 million transistors, while the NV42 has 202 million transistors, although the two chips “are fundamentally the same architecture,” according to NVIDIA.
You may be wondering: why would you consider a 12-pipe chip like the NV42 a replacement for the GeForce 6800 GT? The answer is simple: higher clock speeds. Although the 6800 GS has fewer pixel shader pipes, it’s practically a dead ringer for the 6800 GT when it comes to key specifications like texel fill rate, thanks to its 425MHz core clock speed. In fact, because I’m still trying to shed those recurring math-class nightmares that plague any liberal arts major, I’m going to hit you with a fancy table laying out those specifications in some detail. This should illustrate how fewer vertex and pixel shaders running at higher clock speeds can make the 6800 GS a pretty close match for the 6800 GT. Of course, these numbers won’t produce exactly comparable performance, especially across different generations and brands of GPU architectures, but they’re still fairly relevant.
The 6800 GS doesn’t quite match up with the 6800 GT in terms of pixel fill rate, but that number isn’t terribly important in today’s games, which require increasing amounts of shading and texturing power per pixel. The 6800 GS offers slightly more vertex processing power, slightly less pixel shading and texturing power, and exactly the same amount of memory bandwidth as the GeForce 6800 GT. Functionally, they’re practically twins.
The NV42, however, has the benefits of youth. The 6800 GT is based on the original NV40 chip, and it needs the help of a separate bridge chip in order to talk over a PCI Express interface. NV40 silicon is also incapable of accelerating WMV video decoding. The NV42’s smaller transistor count and newer fab process should allow the 6800 GS to draw less power and produce less heat than the 6800 GT. The 6800 GS is quite likely cheaper to manufacture, as well.
As you’d expect, the GeForce 6800 GS needs only a single-slot cooler to keep it happy. The board design looks very similar to other GeForce 6-series cards, especially the GeForce 6800 cards based on the NV41.
My only quibble with the NVIDIA reference design pictured abovewhich is the card we testedis the lack of a second DVI output. I think a $249 graphics card ought to have a pair of digital outputs. Heck, I think a $149 graphics card should, too. Let’s hope some of the board makers producing 6800 GS cards decide to do the right thing and convert that VGA output into a second DVI port.
So… who’s the competition?
In order to compare the GeForce 6800 GS with its closest competitors, we’ve had to stare deep into the soul of the video card market and pick out a couple of names from the crowded space around the $249 price point. If you were going to spend about that much on an ATI graphics card right now, you’d more than likely be looking at a Radeon X800 XL. You can find the X800 XL selling at online retailers for a prevailing price of about $259, although there are a few places selling the X800 XL for less. This 16-pipe card started life as a foil to the GeForce 6800 GT, so I suppose it’s fitting that it should now face off against the 6800 GS.
There’s also ATI’s new generation of graphics cards, the Radeon X1000 family. Due out at the end of this month is the Radeon X1600 XT with 256MB of memory, a card that ATI says will list for $249. That should make it a direct competitor to the GeForce 6800 GS, but I’m not so sure about that. We’ve included it in our comparison for the sake of fairness and completeness, but the X1600 XT just doesn’t look to me like a $249 card. Everything about it, from its teensy GPU die size to its four texture units, four render back ends, and its 128-bit memory interface screams “cheaper!”like $169 or so. The X1600 XL looks like it’s outclassed by the likes of the 6800 GS or even ATI’s own Radeon X800 XL. ATI has blessed the X1600 XT with twelve pixel shaders and a stratospheric 590MHz clock speed, though, so perhaps it’s not completely out of its league. We shall see about that.
We have also included a wide range of higher-priced graphics cards from ATI and NVIDIA in our testing today, in part because they’re competing directly with an intriguing option: a pair of GeForce 6800 GS cards in SLI. Could buying two of these new cards and running them in tandem be more cost-effective than ponying up for ATI’s new Radeon X1800 XT? Maybe so.
We’ve also included a range of cards because it was time for an update since we did our last big graphics comparison in our Radeon X1000 series review. Both major players have released new drivers since then, claiming big performance gains, and several excellent new games have arrived, as well.
Our testing methods
As ever, we did our best to deliver clean benchmark numbers. Tests were run at least three times, and the results were averaged.
Our test systems were configured like so:
|Processor||Athlon 64 X2 4800+ 2.4GHz|
|System bus||1GHz HyperTransport|
|Motherboard||Asus A8N-SLI Deluxe|
|North bridge||nForce4 SLI|
|Chipset drivers||SMBus driver 4.50|
|Memory size||2GB (2 DIMMs)|
|Memory type||Crucial PC3200 DDR SDRAM at 400MHz|
|CAS latency (CL)||2.5|
|RAS to CAS delay (tRCD)||3|
|RAS precharge (tRP)||3|
|Cycle time (tRAS)||8|
|Hard drive||Maxtor DiamondMax 10 250GB SATA 150|
|Audio||Integrated nForce4/ALC850 with Realtek 188.8.131.5200 drivers|
|Graphics||GeForce 6800 GS 256MB PCI-E with ForceWare 81.87 drivers|
|Dual GeForce 6800 GS 256MB PCI-E with ForceWare 81.87 drivers|
|XFX GeForce 7800 GT 256MB PCI-E with ForceWare 81.87 drivers|
|MSI GeForce 7800 GTX 256MB PCI-E with ForceWare 81.87 drivers|
|Radeon X1600 XT 256MB PCI-E with Catalyst 5.10a beta drivers|
|Radeon 800 XL 256MB PCI-E with Catalyst 5.10a beta drivers|
|Radeon X1800 XL 256MB PCI-E with Catalyst 5.10a beta drivers|
|Radeon X1800 XT 512MB PCI-E with Catalyst 5.10a beta drivers|
|OS||Windows XP Professional (32-bit)|
|OS updates||Service Pack 2, DirectX 9.0c SDK update (October 2005)|
Thanks to Crucial for providing us with memory for our testing. 2GB of RAM seems to be the new standard for most folks, and Crucial hooked us up with some of its 1GB DIMMs for testing. Although these particular modules are rated for CAS 3 at 400MHz, they ran perfectly for us at 2.5-3-3-8 with 2.85V.
Unless otherwise specified, the image quality settings for both ATI and NVIDIA graphics cards were left at the control panel defaults.
The test systems’ Windows desktops were set at 1280×1024 in 32-bit color at an 85Hz screen refresh rate. Vertical refresh sync (vsync) was disabled for all tests.
We used the following versions of our test applications:
- trdemo2 demo
- Serious Sam 2 2.064b with TR Demo_0008 demo
- Battlefield 2 1.03
- Guild Wars
- FEAR 1.02
- Half-Life 2: Lost Coast with trcoast1 demo
- FutureMark 3DMark05 Build 120
- FRAPS 2.6.4
The tests and methods we employ are generally publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.
We’ve already looked at the math briefly, but we’ll pause again to consider pixel-pushing power, because it still determines performance to a large degree, especially at higher resolutions when video cards tend to hit their limits.
| Core clock
|Peak fill rate
| Peak fill rate
| Memory bus
| Peak memory
|Radeon X1600 XT||590||4||2360||4||2360||1380||128||22.1|
|GeForce 6600 GT||500||4||2000||8||4000||1000||128||16.0|
|GeForce 6800 GS||425||8||3400||12||5100||1000||256||32.0|
|GeForce 6800 GT||350||16||5600||16||5600||1000||256||32.0|
|Radeon X800 XL||400||16||6400||16||6400||980||256||31.4|
|GeForce 6800 Ultra||425||16||6800||16||6800||1100||256||35.2|
|GeForce 7800 GT||400||16||6400||20||8000||1000||256||32.0|
|Radeon X1800 XL||500||16||8000||16||8000||1000||256||32.0|
|Radeon X850 XT||520||16||8320||16||8320||1120||256||35.8|
|Radeon X850 XT Platinum Edition||540||16||8640||16||8640||1180||256||37.8|
|XFX GeForce 7800 GT||450||16||7200||20||9000||1050||256||33.6|
|Radeon X1800 XT||625||16||10000||16||10000||1500||256||48.0|
|GeForce 7800 GTX||430||16||6880||24||10320||1200||256||38.4|
The 6800 GS slots in between the two competitors from ATI. The Radeon X800 XL has more theoretical texel fill rate and the same memory interface type and speed. The X1600 XT, meanwhile, is long on pixel and vertex shading power as we’ve noted, but lacking in the fill rate and memory bandwidth departments.
The cards stack up as expected in 3DMark05’s synthetic test of multi-textured (or texel) fill rate. The 6800 GS has a little bit less peak pixel-pushing power than the Radeon X800 XL, but it has more than twice the peak texel throughput of the Radeon X1600 XT.
We generally used in-game controls when possible in order to invoke antialiasing and anisotropic filtering. In the case of Doom 3, we used the game’s “High Quality” mode in combination with 4X AA.
The 6800 GS takes it to both of its rivals from ATI in Doom 3, pretty much as we’ve come to expect NVIDIA cards to do in OpenGL games. Note, though, that the top single-card solution in the group is ATI’s Radeon X1800 XT. ATI’s newest Catalyst drivers seem to have paid off here.
Also notice that two GeForce 6800 GS cards can outrun any single graphics card available in Doom 3, even a card that lists for twice as much as a single 6800 GS.
Half-Life 2: Lost Coast
This new expansion level for Half-Life 2 takes advantage of high-dynamic-range lighting, and it looks spectacular. HDR lighting requires the use of higher-color texture formats, so it really stresses a card’s throughput.
The Radeon X800 XL ever-so slightly edges out the 6800 GS here, but the difference is negligible. The Radeon X1600 XT looks outclassed.
Serious Sam II
Here’s a new game with a new game engine that takes advantage of all of the hottest lighting and shading techniques. Unfortunately, we decided not to test with high-dynamic-range lighting on this game because it didn’t appear to work correctly on Radeon X1000-series cards. Perhaps a future patch or driver update will resolve the problem.
We did, however, decide to test both with and without 4X antialiasing and 16X anisotropic filtering enabled, because we wanted to see whether our extensive use of AA and aniso was putting the Radeon X1600 XT’s unique, asymmetrical GPU architecture at a particular disadvantage.
The Radeon X1600 XL’s performance doesn’t seem to suffer inordinately when 4X antialiasing and 16X anisotropic filtering are in use, after all; it’s just slow across the board. The 6800 GS, though, can’t quite keep step with the Radeon X800 XL here.
We tested the next few games using FRAPS and playing through a portion of the game manually. For these games, we played through five 60-second gaming sessions per config and captured average and low frame rates for each. The average frames per second number is the mean of the average frame rates from all five sessions. We also chose to report the median of the low frame rates from all five sessions, in order to rule out outliers. We found that these methods gave us reasonably consistent results.
F.E.A.R. is my current game of choice, and incredibly, I enjoyed reenacting the firefight that we used for this test all 80 to 100 times that I played through it for this review. I used lots of slow-mo to take down the bad guys, and it’s freakishly addictive.
Light detail, shadow detail, texture resolution, shaders, effects detail, model decals, and relections were all set to maximum for our testing. Computer performance, water resolution, and volumetric light density were set to medium.
The 6800 GS again finds itself bracketed by the competition from ATI, although the X1600 XT looks a little stronger than usual in this shader-filled game. In SLI mode, the 6800 GS produces average frame rates comparable to a Radeon X1800 XT, but the 6800 GS SLI rig’s lowest frame rates are superior to the X1800 XT’s.
The 6800 GS essentially ties the Radeon X800 XL once more in Battlefield 2.
The 6800 GS puts in a nice showing in Guild Wars, with a single card outperforming a Radeon X800 XL and two cards in SLI beating out the Radeon X1800 XT.
3DMark05 is a rare case where the Radeon X1600 XT looks like it belongs in the same league as the X800 XL and 6800 GS. Here’s how the three individual game tests break down.
Our three contenders around the $249 price point are very evenly matched here, with no card taking a clear lead. The 6800 GS in SLI takes two of three game tests from the single-card champ, the Radeon X1800 XT.
3DMark’s remaining synthetic benchmarks give us a split result. The 6800 GS wins out in the pixel shader test, but the ATI cards are both faster in the vertex shader test. In fact, the X1600 XT has finally found a place where it excels; it beats out the Radeon X1800 XL and the GeForce 7800 GTX in the vertex shader tests. This robust vertex throughput may help explain the X1600 XT’s ability to keep up with the Radeon X800 XL and GeForce 6800 GS in 3DMark05’s three game tests.
We measured total system power consumption at the wall socket using a watt meter. The monitor was plugged into a separate outlet, so its power draw was not part of our measurement. The idle measurements were taken at the Windows desktop, and cards were tested under load running Half-Life 2: Lost coast at 1600×1200 resolution with 16X anisotropic filtering and HDR lighting enabled.
Out of curiosity, I threw in a GeForce 6800 GT for these tests, because I wanted to see how the 6800 GS compared to its predecessor. Will the GS’s smaller transistor count and newer fab process give it lower power consumption than the GT, or will the GS’s higher clock speed offset its advantages?
The system with the 6800 GS card pulls about 16 fewer watts under load than the system equipped with a 6800 GTnot bad. The systems based on the two ATI cards require even less power under load, however. Also, notice that our idea of pairing up two 6800 GS cards to defeat a single high-end card has taken a hit. The SLI rig eats roughly 25 to 30W more power than a system with a single high-end GPU.
We used an Extech model 407727 digital sound level meter to measure the noise created (primarily) by the cooling fans on our two test systems. The meter’s weightings were set to comply with OSHA standards, and the meter itself was mounted on a tripod approximately two feet from our test system.
I would be more confident in these numbers if the chipset fan on our Asus A8N-SLI mobo didn’t rattle and clatter like it does sometimes. Still, the numbers here show the 6800 GS to be a reasonably quiet card by current standards, and that jibes with what my ears tell me. Two of ’em together in SLI make a lot of racket at idle, although they’re comparable to a setup with one Radeon X1800 XL or XT under load.
The always-on, uber-loud fan on this Radeon X1600 XT has got to be an artifact caused by our review sample card’s pre-production status. I hope. Please?
In terms of fan noise, power draw, and overall performance, the GeForce 6800 GS runs neck and neck with the Radeon X800 XL. That’s not bad at all for a card that starts life at the $249 mark. If history is any guide, board makers will offer discounts and “overclocked in the box” clock speeds to make the 6800 GS even more appealing than NVIDIA’s stock design. In fact, I believe those cards should be available for purchase starting today.
The Radeon X800 XL, meanwhile, started life in a higher tax bracket and only recently has dropped into the $259-ish neighborhood. We’ll have to see whether it will stick around and continue to drop to $249 or less over time. I wouldn’t hesitate to recommend the X800 XL as an alternative to the GS, except for the fact that ATI has now embraced Shader Model 3.0 in the Radeon X1000 series. The 6800 GS has Shader Model 3.0 support, and the Radeon X800 XL doesn’t. That fact may not matter now, but it could well become a source of regret down the road for someone choosing an X800 XL over a 6800 GS today. Game developers may shift their attention to SM3.0 at the expense of older cards, especially once ATI joins NVIDIA in pushing them to incorporate more SM3.0-specific features and performance optimizations.
As for the Radeon X1600 XT, let’s hope ATI finds a better way to plug the gap between the $199 and $299 price points than penciling this minor leaguer into its lineup. The Radeon X1600 XT is in many ways a very forward-looking design, but it’s not a good value at its current list price.
The biggest knock on the GeForce 6800 GS, in my book, comes from another source. If you’re going to fork over $249 for a graphics card, you would probably do well to consider shelling out the additional $70-100 for a GeForce 7800 GT instead, if you can. The 7800 GT delivers significantly more performance, thanks in part to the tweaks NVIDIA made between GeForce generations six and seven. Similarly, if Radeon X1800 XL prices settle into the same range as the 7800 GT as supply becomes more plentiful, that card could become a very attractive option. I don’t advocate reaching into a higher price bracket often, but this is one case where it may make some sense.