Rage3D nVIDIA GeForce 7800 GT Review Rage3D nVIDIA GeForce 7800 GT Review
A few months ago they dropped a bomb on the graphics industry when they released the GeForce 7800 GTX. Everyone, from industry analysts and hardware reviewers to customers and everyone else in between found a lot to like with the new king of the hill. NVIDIA could barely have pulled off a better release; the card is extremely fast, it has a lot of really cool features, and it was immediately available for purchase. The only thing that people took issue with was the price tag; with a MSRP of US$599 it is still the most expensive consumer 3D graphics card on the market.
Thankfully, two weeks ago at Quakecon NVIDIA sought to rectify that. To round out their high-end offering lineup and make the 7-series approachable to more people they announced the GeForce 7800 GT, a slightly slower 7800 with a significantly lower MSRP of $449.
In this review I'll be looking at a reference 7800 GT from NVIDIA and comparing it to a reference 7800 GTX, also from NVIDIA, and a retail boxed Radeon X850XT Platinum Edition from HIS. All three cards represent the current pinnacle of performance offered from both ATI and NVIDIA, so direct comparisons between them should make for an interesting article.
The 7800 GT is built on NVIDIA's G70 core, the same one the 7800 GTX is built on. It's manufactured at TSMC using their 110nm manufacturing process and weighs in at a whopping 302 million transistors, making it the largest graphics chip ever manufactured.
The card itself has been given some tweaks so that the 7800 GT is about a half-inch shorter than the 7800 GTX, but the GT features the same features as capabilities as it's big brother. The changes include slower core and memory speeds, which were tweaked from 430MHz/1.2GHz on the GTX to 400MHz/1.0GHz on the GT, and a different cooler which lacks the heatpipe technology found on the GTX. The most significant difference however is that where the GTX has 6 quads enabled, giving it 24 pixel pipelines and 8 vertex units, the GT has only 5 quads enabled which reduces it to 20 pixel-pipelines and 7 vertex units.
| Specifications | |||||
| GeForce 7800 GT | GeForce 7800 GTX | Radeon X850XT-PE | |||
| Core | G70 | G70 | R480 | ||
| Silicon Process | 110nm | 110nm | 130nm low-k | ||
| Transistor Count (millions) |
302 | 302 | 160 | ||
| Core Speed MHz | 400 | 430 | 540 | ||
| Memory Speed MHz | 500 GDDR3 (1.0GHz) |
600 GDDR3 (1.2GHz) |
590 GDDR3 (1.18GHz) |
||
| Bus Standard | PEG x16 | PEG x16 | PEG x16 | ||
| Bus Width | 256bit | 256bit | 256bit | ||
| Pipeline Configuration Textures/Pixels/Z Samples (Per Clock) |
20/16/32 | 24/16/32 | 16/16/16 | ||
| Vertex Units | 7 | 8 | 6 | ||
| Peak Memory Bandwidth (GB/s) |
32.0 | 38.4 | 37.8 | ||
| Pixel Fillrate (million pixels/sec) |
6,400 | 6,880 | 8,640 | ||
| Texel Fillrate (million texels/sec) |
8,000 | 10,320 | 8,640 | ||
| API Compliancy | DX 9.0c Shader Model 3.0 |
DX 9.0c Shader Model 3.0 |
DX 9.0b Shader Model 2.0 |
||
| MSRP ($US) | $449 | $599 | $549 | ||
NVIDIA® CineFX™ 4.0 Shading Architecture
64-Bit Texture Filtering and Blending
NVIDIA® Intellisample™ 4.0 Technology
NVIDIA® UltraShadow™ II Technology
Composited Desktop Hardware Engine
Advanced Display Functionality
Advanced Engineering
NVIDIA® Digital Vibrance Control™ (DVC) 3.0 Technology
Operating Systems
API Support
The resolutions I decided to test are 1024x768, 1280x1024 (standard native 4:3 LCD panel resolution), 1600x1200 (standard highest resolution most average CRTs can handle), 1920x1080 (widescreen resolution for many widescreen LCD displays and widescreen HD TVs), and 2048x1536 (the highest resolution supported by many high-end displays). Some games do not support custom resolutions, in those cases I selected what was available from the in-game options (typically this meant replacing the 1280x1024 and 1920x1080 resolutions with 1280x960 and 1920x1440, respectively). The actual resolution tested is listed in each chart.
The Windows XP desktop was set to 1280x960 with a 32bit color depth and 85Hz refresh rate for all tests. Refresh rate locks for 3D graphics modes, as supported by both NVIDIA and ATI graphics control panels, was not enabled. V-Sync was forced off via the graphics card control panel as well. All other graphics card control panel settings were left to their default settings unless otherwise noted.
Anti-Aliasing and Anisotropy were applied in the game engine where the options existed. For games that did not support those options natively, the graphics card control panel was used when appropriate.
Custom batch files were used when possible for automated benchmarking (the details of the commands used are outlined for each test). When manual benchmarking was necessary, Fraps version 2.6.2 was used.
Benchmarking was done with Windows set to the "Adjust for best performance" profile, and all unnecessary Windows services and hardware devices were disabled. The latest drivers for each necessary hardware component were installed prior to testing and kept consistent throughout.
The Onboard sound was disabled for all tests unless otherwise noted.
To setup the test machine I installed Windows XP, patched and tweaked it, and installed all the required games, utilities, and hardware drivers needed for the testing procedure except for the graphics drivers. I then made an image of the drive using Norton Ghost and restored that image to two identical hard-drives. The result was two identical installs of Windows XP with the same tweaks and patches on two identical hard-drives. After that I installed the NVIDIA drivers on one hard-drive and the ATI drivers on the other. Testing the videocards was then a simple matter of swapping videocards and hard-drives when required.
For this review I also overclocked my aging A64 3500+ to 2.4GHz from 2.2GHz to try and reduce the CPU bottleneck on these graphics cards.
System specifications are below
Test System Specs
Benchmark Software
Anti-Aliasing Quality
I took a pretty close look at edge AA quality in my 7800 GTX review and was quite impressed with the improved quality and number of AA features offered by NVIDIA. The 7800 GT doesn't offer any less of an image when compared to the 7800 GTX in that regard, but I decided to go ahead and make sure.
I took advantage of a new single player demo of F.E.A.R., an upcoming first person shooter from the Monolith guys (the same guys responsible for the No One Lives Forever series). It's a pretty easy game to work with for these types of tests (even easier than Half-Life 2 which I used for the GTX review) and the graphics are pretty amazing, making for some nice eye-candy. I created a javascript app that will allow you to easily switch between the screenshots and compare image quality for yourself. I modified the screenshots in Photoshop to zoom in on a portion of the game architecture to makes it easy to see the AA quality. Click the link below to launch the app, it will open in a new window.
Transparency Anti-Aliasing
You might have noticed that in many games, even modern ones with cutting-edge graphics engines, there are sometimes partially transparent textures that the developer has used in certain areas. These textures provide an easy way for game developers to add a lot of detail to a scene without worrying too much about the impact on performance. The disadvantage to using them is that they are normally unaffected by the type of anti-aliasing technology used on modern graphics cards which causes them to stand out like a sore thumb when edge AA is applied over rest of the scene.
Transparency AA is a new feature which NVIDIA introduced with the GeForce 7800 GTX that affectively solves this problem. When it is used, it anti-aliases the edges of textures much the same way traditional multi-sample AA would if it were normal game geometry.
TAA doesn't work without AA enabled (either in-game or forced via the control panel), so it's not possible to anti-alias transparent textures without anti-aliasing the regular game geometry as well. The reason for this is probably because the number of samples it uses is tied the AA level that the game is currently running at. So, for example, if you have 4x AA enabled, TAA is using the same 4 sample pattern. If you are using 2x AA, TAA also uses 2x AA. If you have AA disabled, TAA is disabled as well (regardless of whether you have it on in the Forceware control panel or not).
I tested TAA performance in my 7800 GTX review, but I wanted to test it again to see how much of an impact it would have on the slower 7800 GT.
| 1024x768 | 1280x1024 | 1600x1200 | 1920x1080 | 2048x1536 | |
| TMAA % Faster vs No AA |
-0.1% | -0.0% | -0.0% | -0.1% | -0.0% |
| TSAA % Faster vs No AA |
-0.9% | -3.6% | -5.9% | -7.1% | -6.7% |
To really give the TAA engine a workout I recorded a custom Half-Life 2 demo on the d2_coast_07 level, the one where you have to crawl around underneath the train bridge. This level has a huge amount of transparent textures that are used to give the skeletal structure of the bridge a lot of detail.
The demo is about 2 minutes in length (played back normally). In it I simply move back and forth through the structure of the bridge several times while trying to keep as much of it as possible in view. Because of the sheer number of transparent textures and geometry in this demo, I would consider it a worst-case-scenario.
The performance results from this test are shown in the chart below
| 1024x768 | 1280x1024 | 1600x1200 | 1920x1080 | 2048x1536 | |
| TMAA % Faster vs No AA |
-0.7% | -0.8% | -0.5% | -0.6% | -0.4% |
| TSAA % Faster vs No AA |
-34.9% | -49.4% | -51.7% | -47.4% | -47.9% |
The same level I used to test TAA performance also makes for an excellent level to test TAA image quality. To do that I simply saved a game in a position that I thought would really show off the improvements TAA can make. I then loaded this save position, being careful not to move the mouse, and took comparison screenshots with Fraps. To see the results, I created a simple javascript app that will allow you to easily and quickly compare the screenshots. Click the link below to launch the app, it will open in a new window.
I looked at Anisotropic Filtering quality in my 7800 GTX review last month. Like the AA quality however I decided to retest it on the 7800 GT to make sure that it's at least on par with the GTX.
To test AF quality I took a bunch of screenshots in the F.E.A.R. demo and put them all in a simply javascript app that will let you compare the screenshots quickly and easily.
Unsurprisingly the 7800 GT exhibits the same quality of texture filtering as the 7800 GTX. It looks good in screenshots, but as I said in the 7800 GTX review, in-game there is a noticeable line which moves along ahead of you where the mipmap transitions from one texture resolution to the next. Unfortunately for NVIDIA it is this texture filtering quality that has recently become a pretty controversial topic lately.
Most people call this "texture shimmering", and have been talking about it since the GeForce 6-series. The main gripe is that changing the Image Quality slider from the default "Quality" to "High Quality" in NVIDIA's control panel does not fix the shimmering problem on 7-series cards like it did on the previous generation 6-series cards. The reason for it isn't clear yet, but many attribute it to over-zealous optimizations on NVIDIA's part.
German hardware website 3DCenter.org investigated this problem pretty thoroughly and has a lot of examples and further information already online. I highly suggest you read the article and download the sample videos they have provided; it should give you a pretty good handle of what this issue is all about.
This is pretty bad news for current and prospective G70 owners obviously. The good news however is that NVIDIA has already made a statement regarding this issue and has said that a fix is in the works that will appear in a future driver update. Hopefully they can get it worked out.
Update: September 5th, 2005 Texture Filtering IQ and Performance Analysis with ForceWare Beta 78.03
NVIDIA recently released ForceWare 78.03 as a public beta which purportedly address the texture shimmering issues mentioned in the original article above. I took the liberty to test these drivers against the WHQL release version used in this review, checking in several games to see how big of an impact the fix has on performance and, more important, on image quality.
Objective observation of game play by comparing problem areas using both the Quality and High Quality modes of the 77.77 drivers with the same areas using the High Quality mode of the 78.03 drivers showed very good improvements in texture filtering image quality. The shimmering line that moved along ahead of you with the 77.77 drivers is, for all intents and purposes, gone with the 78.03 drivers. When comparing Quality vs Quality modes the image quality exhibited the same texture shimmer as existed before, however this was not unexpected from this release. Note that, as stated above, the main gripe people had with texture shimmering on the 7-series cards was that turning on High Quality mode would not reduce it to acceptable levels such as it did on the 6-series cards. With the 78.03 release NVIDIA seems to have fixed this issue, at least visually.
The next area of concern is performance. After-all, if the fix causes performance issues then there is no point to it in the first place. To test performance impacts I ran the 77.77 and 78.03 drivers head-to-head in Quality and High quality modes.
Note that these tests were conducted on a machine of slightly different specs from the one used in the rest of the review. The results are below:
| [ Quality ] [ High Quality ] |
| [ Quality ] [ High Quality ] |
| [ Quality ] [ High Quality ] |
| [ Quality ] [ High Quality ] |
As you can see from the series of performance charts above, there is virtually no performance drop between the 78.03 drivers and the 77.77 set in both Quality and High Quality modes.
Overall I am very impressed with NVIDIA's ability to get an effective fix out for this problem as quickly as they did and without it impacting performance.
One last thing I would like to see them change is the default mode from Quality to High Quality. As you can see the impact on performance between the two modes is negligible, but the improvement to Image Quality is fairly obvious.
F.E.A.R. Single Player Demo
The results for the FEAR test are below. Click the text links at the top of the chart to change settings.
Splinter Cell Chaos Theory
In the second chart set I enabled Shader Model 2.0 for the X850XTPE, which is a new feature of the version 1.04 patch, and Shader Model 3.0 for the 7800 GT and 7800 GTX.
The third chart set shows the performance results of the cards when running with the full feature set available when you select the SM2.0 or SM3.0 paths. Anti-aliasing is automatically disabled when you enable HDR, so there are no AA results in the third chart.
Click the text links at the top of each chart to change settings.
Half-Life 2
A batch file was used to automate testing; the command line is below for reference. This batch file was used for each card that was tested. The settings surrounded by < > change for each pass:
The Half-Life 2 results are below. Click the text links at the top of the chart to change settings.
Doom 3
Another batch file was used to automate Doom3 testing as well. The command line is below for reference. This batch file was used for each card that was tested. The settings surrounded by < > change for each pass:
The Doom3 results are below. Click the text links at the top of the chart to change settings.
Chronicles of Riddick: Escape from Butchers Bay
Chronicles of Riddick was benchmarked using another custom batch file that ran each of the 5 built in timedemos that are included with the 1.01 patch. The Shader Model 2.0 path was used for all cards, including the 7800 GT and 7800 GTX, and Anti-aliasing and Anisotropy were set in the graphics card control panel.
The Chronicles of Riddick results are below. Click the text links at the top of the chart to change settings.
Far Cry
Again a custom batch file was used to benchmark Far Cry. I used 3 demos included with the newer patches to test performance (from the Cooler, Training, and Volcano levels). Because Far Cry benchmark framerates can vary between each subsequent pass (sometimes fairly significantly), I ran each of the 3 demos 3 times, then averaged the 9 results to get the final score for the detail level.
The command line for the batch file I used to automate Far Cry benchmarking is below. This batch file was used for each card that was tested. The settings surrounded by < & > change for each pass:
The Far Cry results are below. Click the text links at the top of the chart to change settings.
Pacific Fighters
I benchmarked Pacific Fighters by loading the included "N1K1 vs BeauFighter.ntrk" track and logged framerates using Fraps from the beginning of the track for 90 seconds. All the in game details were set to their maximum levels, including "Landscape Detail" which was set to "Perfect", enabling Pixel Shaded water. Video was set to the "NVIDIA GeForce 6800/6600/FX/4/3" profile for the 7800 GT and 7800 GTX, and set to the "ATI Radeon X800/9800/9700/9600/9500" profile for the X850XT-PE. Anti-aliasing and Anisotropy were set via the graphics card control panel.
The Pacific Fighters results are below. Click the text links at the top of the chart to change settings.
Richard Burns Rally
I used Fraps to benchmark Richard Burns Rally using two custom demos, one recorded on the Great Britain Chirdonhead II stage with rain and the other on the USA Frazier Wells track. I started Fraps framerate logging as soon as the marshal launched the car and had it automatically stop logging after 90 seconds. I logged each pass 3 times then averaged the results to get the final score for that setting and resolution (the game setup utility only goes to 1600x1200, but custom resolutions can be set via the RichardBurnsRally.ini file in the game directory). The default replay camera was used and sound was enabled. Anti-aliasing and Anisotropy were set via the graphics card control panel.
The Richard Burns Rally results are below. Click the text links at the top of the chart to change settings.
3DMark03
I benchmarked 3DMark03 using the default settings. I set it up to loop each test three times, it then automatically averaged the results and gave a final score. The results for each test are below.
| 3DMark03 Game Test 1 - Wings of Fury |
||||||||
|
||||||||
3DMark05
Like 3DMark03, I set 3DMark05 up to loop three times to get the final score. I tested 3DMark05 using the default detected settings.
| 3DMark05 Game Test 1 - Return to Proxycon |
||||||||
|
||||||||
Overclocking the card was pretty easy using Coolbits, NVIDIA's built-in overclocking tool (once enabled via the registry). I let it automatically detect the best overclock, and then manually adjusted it as high as possible. The end result was 485MHz core and 1.25GHz memory, which is 85MHz over the default core speed and 125MHz (250MHz DDR) over the default memory speed, a very good overclock indeed. This pushed the 7800 GT past the default 7800 GTX core and memory speed (430c/1.2m) and, as you can see from the performance chart below, the boost in clock speeds was enough to overcome the missing quad and push the 7800 GT past the 7800 GTX in Doom 3.
I said at the end of my 7800 GTX review that your only choice when looking for a high-end graphics card was the GTX. Not everyone can afford to drop that kind of money on a graphics card though, so thankfully we've been given a new choice in the form of the 7800 GT. As you've seen, a slightly slower 7800 is still a smoking fast graphics card, and with a MSRP of $449 (the card is already well below that on the street), the 7800 GT is within the reach of a lot more people.
And there's not much to complain about with the 7800 GT. The overclocking capabilities are good, at least from what I've seen, and the single-slot cooler is quiet and does a decent job of keeping the temperatures in check. It's even a bit shorter than the GTX, so if you were holding off on an upgrade because you were worried that the card wouldn't fit in your sytem, you might take another look now. It's smoking fast too, and has the same features list as the GTX, but unlike the GTX it's not overly expensive and out of the price range of most users.
The only real concern right now, and it might be a temporary concern, is the shimmering textures problem that has become a hot topic lately. NVIDIA has stated that a fix will be available to solve the problem, but it's likely that any such fix will have an unavoidable negative impact on performance. See this section of the review for an important update.
I think the GT hits ATI especially hard, even more so than the GTX did. With a MSRP $100 less than ATI's flagship Radeon X850XT Platinum and with the 7800 GT winning most of the benchmarks, and also leading the way in the features checklists, I think the last nail in ATI's X8-series coffin has been hammered home.
