An army of cooling fans has been whirring nonstop here in Damage Labs for the past week or so to bring you an extensive set of benchmarks of the GeForce 7800 GT. The short version? The 7800 GT offers near-GTX performance for at least a hundred bucks less than NVIDIA’s top-end card, and it pretty much creams anything else in its price range. Keep reading for the full story.
What makes a GT
For those of you familiar with the GeForce 7800 GTX, getting a handle on the GT will be easy. If you’re unfamiliar with the 7800 GTX, well, then you need to go read our mind-enhancing review of NVIDIA’s seventh-gen GPU. The GeForce 7800 GT is based on the same G70 graphics processor that powers the GTX, but NVIDIA has disabled a few of the chip’s functional units for this less expensive card. I have done violence to a block diagram of a G70 GPU in order to illustrate what’s been disabled. Have a look:
Up top there, you can see that one of the G70’s eight vertex shader units, responsible for geometry processing, is crossed out with a big, red “X.” This vertex unit is not functional in the 7800 GT, so the other seven will have to do. Across the middle, one of the six pixel-pipeline “quads” is crossed out, leaving the GeForce 7800 GT with a total of 20 pixel shader units. Without the services of these deactivated vertex and pixel shader units, the 7800 GT will perform a little bit slower than the 7800 GTX.
NVIDIA has also lowered the recommended clock speeds for the 7800 GT. A bone-stock GeForce 7800 GTX runs its GPU at 430MHz and its GDDR3 memory at 600MHz, while stock frequencies on the GT are 400MHz for the GPU and 500MHz for memory. Beyond that, little else has changed. The 7800 GT still has sixteen ROPs capable of writing one pixel per clock to memory, and it still has a 256-bit memory interface.
None of the GT’s spec changes should hamper its performance too much. The thing has more and better pixel pipes than a GeForce 6800 Ultra, for goshsakes. Still, NVIDIA’s board partners are already starting to push past the recommended specs, as they’ve done with the 7800 GTX. Two of the three brands represented in our recent 7800 GTX roundup sell “overclocked in the box” cards, and some of the same suspects are tuning up the GT. BFG Tech’s GeForce 7800 GT OC, for instance, will sell with a 425MHz GPU and 525MHz memory. Meanwhile, the XFX cards we received for review are clocked at 450MHz with 525MHz memory.
In terms of basic layout and cooler design, I’d expect at least the first wave of 7800 GT cards to be entirely similar to the XFX card pictured above. At first glance, this looks very much like a 7800 GTXsame PCI Express connector, same dual DVI ports, same video in/out connector, same six-pin auxiliary power input. However, the 7800 GT board is actually about a half-inch shorter than a 7800 GTX, and the cooler is downsized, as well. That doesn’t mean, of course, that the 7800 GT is a smallish card; it’s still about an inch longer than a Radeon X850 XT Platinum Edition. It is smaller than a GTX, though.
Board makers aren’t just overclocking NVIDIA’s GeForce 7-series cards; they’re also underclocking prices. NVIDIA’s initial estimated price on the 7800 GTX was $599, but cards are already selling for as low as $499 at online vendors. That’s no bargain, but it ain’t $599, either. In the same vein, I’d expect board makers to undercut NVIDIA’s suggested $449 list price somewhat on the 7800 GT. In fact, XFX says it will have its “overclocked in the box” 7800 GT selling at multiple online vendors, today, at $399.
This was kind of a tough call, but because XFX shipped its 7800 GT to usa real retail product as of todaywith a 450MHz GPU clock and 525MHz memory clock, I decided to test the 7800 GT at that speed. Boards running at NVIDIA’s stock reference frequencies will be slightly slower than the XFX card we’re testing, but our results should be directly representative of a 7800 GT card you can buy right now. Had we received these cards and their drivers sooner than this past Friday, I would have been able to test the 7800 GT at stock reference speeds, as well. Unfortunately, we just received them too late.
This decision will no doubt incite controversy because I did not include scores from an “overclocked in the box” GeForce 7800 GTX. If anyone has a coronary as a result, my apologies. You can see how an overclocked 7800 GTX performs in our multi-card roundup. The scores are more or less directly comparable to the ones in the following pages, save for the Splinter Cell: Chaos Theory numbers and a minor driver revision difference.
Speaking of which, you will notice in the test config tables below that we used NVIDIA’s 77.77 drivers for the GeForce 7800 GT and GTX cards, in both single-card and SLI setups. This represents a total refresh of the GTX test results from our GeForce 7800 GTX review and other recent graphics articles. However, the results for the GeForce 6800 Ultra were obtained using NVIDIA’s older 77.62 drivers. Between some serious time constraints and the fact that performance barely changed at all on the 7800 GTX when moving from 77.62 to 77.77, I elected not to retest the GeForce 6800 Ultra with the 77.77 drivers. If the driver revision difference of 0.015 offends your sensibilities, just pretend the 6800 Ultra results aren’t there.
Similarly, I’ve included results from the Radeon X850 XT Platinum Edition using older Catalyst 5.6 drivers. Make of them what you will.
Also, note that I’ve included scores for a Radeon X800 XL 512MB card from Abit. Radeon X800 XL cards from other vendors are currently selling for between $359 and $400, or just below the likely price range of the 7800 GT. That makes the X800 XL competition, though not 100% direct competition. The X850 XT Platinum Edition might be the most direct competition for the 7800 GT in terms of price. I decided to include the X800 XL in part because I wanted to see how it performs at very high resolutions like 2048×1536. Perhaps the 512MB of video memory will be a help.
That said, remember that resolutions above three megapixels are something of a special case where the GeForce 7800s have architectural advantages over older GPUs.
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 system was configured like so:
|Processor||Athlon 64 4000+ 2.4GHz|
|System bus||1GHz HyperTransport|
|Motherboard||Asus A8N-SLI Deluxe|
|North bridge||nForce4 SLI|
|Chipset drivers||SMBus driver 4.45
IDE driver 5.18
|Memory size||1GB (2 DIMMs)|
|Memory type||OCZ EL PC3200 DDR SDRAM at 400MHz|
|CAS latency (CL)||2|
|RAS to CAS delay (tRCD)||2|
|RAS precharge (tRP)||2|
|Cycle time (tRAS)||5|
|Hard drive||Maxtor DiamondMax 10 250GB SATA 150|
with NVIDIA 4.60 drivers
|Graphics||GeForce 6800 Ultra 256MB PCI-E with ForceWare 77.62 drivers||Dual GeForce 6800 Ultra 256MB PCI-E with ForceWare 77.62 drivers||XFX GeForce 7800 GT 256MB with ForceWare 77.77 drivers||Dual XFX GeForce 7800 GT 256MB with ForceWare 77.77 drivers||MSI GeForce 7800 GTX 256MB PCI-E with ForceWare 77.77 drivers||Dual MSI GeForce 7800 GTX 256MB PCI-E with ForceWare 77.77 drivers||Abit Fatal1ty X800XL 512MB PCI-E with Catalyst 5.7 drivers||Radeon X850 XT Platinum Edition PCI-E with Catalyst 5.6 drivers|
|OS||Windows XP Professional (32-bit)|
|OS updates||Service Pack 2|
Thanks to OCZ for providing us with memory for our testing. If you’re looking to tweak out your system to the max and maybe overclock it a little, OCZ’s RAM is definitely worth considering.
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 demos
- Far Cry 1.3 with tr1-volcano and tr3-pier demos
- Splinter Cell: Chaos Theory 1.04 with trpenthouse demo
- The Chronicles of Riddick: Escape from Butcher Bay 1.1 with trdemo4
- FutureMark 3DMark05 Build 120
All the tests and methods we employed are publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.
Pixel filling power
We can handicap the 7800 GT’s performance prospects to some extent by looking at some key theoretical performance numbers. The raw ability to paint pixels onscreen at a certain rate isn’t exactly destiny for today’s graphics cards, but it’s still an important metric. Here’s how the 7800 GT stacks up.
| Core clock
|Peak fill rate
| Peak fill rate
| Memory bus
| Peak memory
|GeForce 6600 GT||500||4||2000||8||4000||1000||128||16.0|
|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 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|
|GeForce 7800 GTX||430||16||6880||24||10320||1200||256||38.4|
A stock 7800 GT has a higher theoretical peak multitextured fill rate (the more important kind, usually) than a GeForce 6800 Ultra, but a little less than a Radeon X850 XT. The XFX card we’re testing today, though, can push more multitextured pixels than even a Radeon X850 XT Platinum Edition. In terms of memory bandwidth, the 7800 GT is closer to the Radeon X800 XL. This mix of pixel fill rate and memory bandwidth puts the 7800 GT in an interesting place.
Let’s see how the theory translates into performance with some basic synthetic fill rate benchmarks.
These results track pretty well with the theoretical capabilities of the cards. As one would expect, the XFX 7800 GT comes out just ahead of the Radeon X850 XT Platinum Edition in multitextured fill rate. The 7800 GT doesn’t trail the 7800 GTX by too terribly much, either. Let’s see how these numbers translate into performance in games.
We’ve conducted our testing almost exclusively with 4X antialiasing and a high degree of anisotropic filtering. We generally used in-game controls when possible in order to invoke AA and aniso. In the case of Doom 3, we used the game’s “High Quality” mode in combination with 4X AA.
Our Delta Labs demo is typical of most of this game: running around in the Mars base, shooting baddies. The new and imaginatively named “trdemo2” takes place in the game’s Hell level, where the environment is a little more varied and shader effects seem to be more abundant.
Notice how that orange line snuggles up just underneath the green one through all four resolutions in both graphs. The 7800 GT just barely trails its big brother, the 7800 GTX, in Doom 3. With SLI, the story is the same, but at much higher speeds. The ATI cards just can’t keep up.
Next up is Far Cry, which takes advantage of Shader Model 3.0 to improve performance. The game also has a path for ATI’s Shader Model 2.0b. Our first demo takes place in the jungle with lots of dense vegetation and even denser mercenaries. All of the quality settings in the game’s setup menu were cranked to the max.
Our next demo takes place on the Volcano level, where there’s a whole lot of heat shimmer going on. Pixel shading power should be at a premium here.
Far Cry is much more of a contest, as the performance of the Radeon X850 XT Platinum Edition tracks very closely with the 7800 GT’s at the three lower resolutions. Also, the Radeon X800 XL 512MB seems to benefit from having more memory onboard here. Note that it surpasses the GeForce 6800 Ultra and the Radeon X850 XT PE at 2048×1536.
The Chronicles of Riddick: Escape from Butcher Bay
This game has a Shader Model 3.0-type mode, but to keep things even for comparison to the Radeon, I ran all cards with the SM2.0 path.
Like Doom 3, Riddick is an OpenGL game. That may explain in part why the ATI cards struggle so much here; NVIDIA’s OpenGL drivers may be better. Whatever the case, the 7800 GT puts in a heckuva showing, again trailing the GTX by only a hair.
Splinter Cell: Chaos Theory
We’re using the brand-new 1.04 version of Splinter Cell: Chaos Theory for testing, and that gives us some useful tools for comparison. This new revision of the game includes support for Shader Model 2.0, the DirectX feature set used by Radeon cards. The game also includes a Shader Model 3.0 code path that works optimally with GeForce 6 and 7-series GPUs. Because SM2.0 and SM3.0 can produce essentially the same output, we’ve tested the ATI cards with SM2.0 and the NVIDIA cards with SM3.0. (The game won’t let NVIDIA cards run in SM2.0 mode, although they are capable of doing so.)
In our first test, we enabled the game’s parallax mapping and soft shadowing effects. In the second, we’ve also turned on high-dynamic-range lighting and tone mapping, for some additional eye candy. Due to limitations in the game engine (and in NVIDIA’s hardware), we can’t use HDR lighting in combination with antialiasing, so the second test was run without edge AA.
The Radeon X850 XT PE is more than competitive with parallax mapping and soft shadows enabled, but the 7800 GT separates itself from the ATI card once HDR lighting comes into play. The ATI cards don’t have dedicated hardware with floating-point texture filtering and blending capabilities, so they have to rely on their pixel shaders to do the job. The G70’s native support for 16-bit floating point filtering and blending probably gives it an edge.
The 7800 GT’s performance in 3DMark’s three main game tests more or less confirms what we’ve seen elsewhere. The GT is faster than the Radeon X850 XT Platinum Edition, and it can nearly keep up with the 7800 GTX. SLI, of course, opens up a whole new realm of speed.
The G70’s revamped pixel shaders allow the 7800 GT to post some astounding numbers in 3DMark’s pixel shader test, well above the fastest card that ATI has to offer and just below the 7800 GTX once again. As for vertex shaders, it seems the 7800 GT’s one disabled vertex unit isn’t much of a handicap. The GT’s scores track very closely with the GTX’s, although the Radeon X850 XT PE is the fastest single card in the vertex shader benchmarks.
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 a Doom 3 timedemo of our “trdemo2” in High Quality mode at 1280×1024 with 4X antialiasing.
The system based on the 7800 GT draws less power, both under load and when idle, than the systems based on cards it outscores in the benchmarks, including the GeForce 6800 Ultra and the Radeon X850 XE Platinum Edition. In fact, at idle, the 7800 GT system pulls fewer watts than anything in the group.
These results aren’t surprising given how quiet the 7800 GT’s cooler is. At idle, the thing is simply inaudible over the sound of the CPU and chipset coolers in our test system. Under load, the cooler is audible on an open test bench, but probably wouldn’t be when placed inside of a computer case.
I was able to overclock a single 7800 GT card to 491MHz core/565MHz memory simply by using the built-in speed detection button in the NVIDIA driversthat’s 41MHz over the XFX’s stock GPU speed and 91MHz over the NVIDIA-recommended speed. Things were a little different with two 7800 GTs in SLI, where NVIDIA’s utility picked a 475MHz max core clock speed with 540MHz memory. In both cases, my attempts to push for higher core speeds manually were unsuccessful. Nevertheless, the cards were stable at the speeds that the driver chose for them, with no visible artifacts and no crashes.
The 7800 GTX’s slight edge evaporates when the 7800 GT is overclocked to its limits. Of course, there’s some overclocking headroom built into the GTX, as well, so that’s no surprise. Still, the 7800 GT’s overclocked performance is impressive.
So, like I said at the outset, the card is fast, and it’s less expensive than the GeForce 7800 GTX. The performance differences between the two were narrowed a little bit in our testing because we used an “overclocked in the box” GeForce 7800 GT and a stock-clocked GTX. Even so, I’d have a hard time finding the justification for spending an extra hundred-plus dollars on a 7800 GTX now that the 7800 GT has arrived. None of these cards are cheap, but the 7800 GT is easily a better value than its pricier sibling. Either card will get you the full feature set of NVIDIA’s G70 GPU, and the performance delta between them isn’t huge.
If it’s absolute performance you crave, then grab yourself a pair of 7800 GTs and run them together in SLI. A 7800 GT SLI rig ought to satisfy even the most insanely graphics-crazed person. Personally, though, I wouldn’t spring for a full SLI setup right now, because few current games really need that kind of power. I’d stick to a one-card setup, but I’d make darn sure my motherboard would allow me to add a second card down the road.
As for the competition from ATI, well, realistically, it hasn’t been introduced yet. The current Radeon lineup remains quite good, but it can’t keep pace with the 7800 GT, let alone the GTX. We’re still waiting for ATI’s CrossFire dual-graphics platform to arrive, and perhaps some next-generation cards not long after that. Until that happens, the GeForce 7800 GT is clearly the card to have at the $400-ish price point. Given the fact that 7800 GT cards should be available immediately at online vendors, ATI definitely has some catching up to do.