Rage3D R580 Technology Review Rage3D R580 Technology Review
The X1800 is a solid product that unfortunately ran into one too many delays which shortened its life considerably. It offered good performance combined with a slew of new technologies that finally brought ATI level with, or better than Nvidia. Most significantly, ATI products were finally capable of handling Shader Model 3.0. ATI preached that Shader Model 2.0 was good enough and they were more or less correct as games didn’t really take considerable advantage of SM3.0 features. Even so it made fully recommending their products difficult as it wasn’t as future proof as some of its competitors. The X1800 line also introduced ATI’s all new ring-bus memory architecture that gave considerable potential for performance improvements. Another significant technology introduced was AVIVO technology that has only recently shown its muscle in the video reproduction area. All these really made X1800 a great chip that was available in decent quantities during its short life.
Why replace the X1800 so quickly when it’s a pretty good product? Product planning occurs considerably earlier than product launch, and with certain assumptions made. The X1800 was supposed to launch around May and this was the timeframe used in planning for the refresher part. If all went according to plan the launch of the refresher part would then closely follow the traditional 6-month product refresh cycle. As we know the X1800 didn’t go according to plan and was delayed significantly. Still, competition and product cycles don’t wait for anyone and it’s better to move forward and cut your losses than stand still in a fast paced environment.
With the plan in hand and no hiccups in sight, ATI is launching the X1900 line consisting of four different cards: X1900XTX, X1900XT, X1900 Crossfire, and the AIW X1900. It’s the first time ATI has introduced an AIW edition at the same time as a new GPU launch. The X1900 isn’t just a speed bump over the X1800, it has a few considerable improvements that we’ll get into soon. It’s meant to engage, entertain, and breakthrough. Engage your reflexes; entertain your senses and breakthrough performance barriers. Let’s find out how well it accomplishes this.
The X1900 series doesn’t introduce any significantly new technologies to the table but makes up for it by adding a whole lot of punch. Pixel shading punch. As speculated, ATI has boosted the number of pixel shader processors to 48, which is three times the amount found in the X1800 line. This accounts for the largest change from the X1800 and also for the increase in the amount of transistors. The X1900 has over 380 million transistors, an increase of 20% from the X1800, and draws approximately 150W of power! The maturation of the 90nm process technology allowed ATI to pack more transistors into the same area.
The significance of pixel shader power with respect to performance is still relatively vague in comparison to using the number of pipelines (ROPs) as a traditional indicator. A jump of three times in magnitude of ROPs would’ve meant substantial performance improvement. What does such an increase in pixel shader power mean?
The use of pixel shaders in games has steadily increased over the past few years. ATI research shows that all new titles in 2006 will use pixel shaders in some way. Not only are they used more but they are becoming increasingly more complex, so that pixel shading is fast becoming the main performance bottleneck. This observation easily explains ATI’s design consideration to up the number of pixel shader processors. Old games already run plenty fast on today’s high end cards. Games released recently that use fairly simple pixel shader programs will see a noticeable improvement and future games that use even more PS will see substantial gains.
The X1900 uses the same pixel shader core from the X1800. Each one is capable of performing up to two 3-component vector operations and two scalar operations each clock cycle at full FP32 precision. Each core is equipped with a dedicated Branch Execution Unit that allows flow control to occur seamlessly. Of course ATI’s Ultra-Threaded Pixel Shader Engine has carried over to ensure that each PS Core is always working. (To read more about the Ultra-Thread Pixel Shader Engine check out our X1800 review)
The next improvement again deals with shading but this time on the texturing side. With the X1900 ATI has added Fetch4 capability to the texture units. Fetch4 actually isn’t something new as the X1300 and X1600 have had the capability ever since launch but it hasn’t been widely publicized due to the X1800 lacking the feature. Fetch4 is a new texture sampling method that is useful for shadow map acceleration. Normally texture units sample a colour texture where one colour value consists of four components (red, green, blue and alpha) and each texture unit is only able to sample one value at a time. A shadow map (depth texture) on the other hand is made up of only one component value. Fetch4 takes advantage of this situation and allows a texture unit to sample four values from adjacent address at a time, increasing the texture sampling rate by four. Taking advantage of this feature requires additional programming on the applications part but ATI noted that it is fairly easy to implement.
High resolution users will be happy with the final improvement which benefits them considerably. ATI increased the Hierarchical Z on-chip memory by 50% to handle resolutions such as 1920x1200 all the way up to 2560x1600. In prior cases there would be a large performance impact if the memory wasn’t enough to handle it but this has been eliminated.
Everything else about the X1900 is carried over from the X1800 including its advanced 512-bit ring-bus memory controller and industry-leading AVIVO technology. From a paper stand point the X1900 filled in the one weak point the X1800 had which was less pixel shader processors compared to the competition. The X1900 reverses that weakness and makes it one of its strengths.
As previously mentioned, ATI is releasing four cards at once: X1900XTX, X1900XT, X1900 Crossfire, and the AIW X1900. Each board has the same base configuration of 16 pipelines (ROPs), 16 texture units, 48 pixel shaders and 8 vertex shaders. The R5xx series was designed to be flexible in configuration compared to previous architectures. ATI is able to scale not only in number of pipelines, but also in PS units, texture units and VS units. The X1900 uses a ratio of 3 PS (ALU) for every 1 TU which is the same ratio used in the X1600.
Why not also increase the number of texture units? ATI has analyzed shaders used in 3D games and observed that shaders are increasingly using more arithmetic (mathematical) operations in comparison to texture operations. Early shaders had roughly equal the number instructions between these two types but new shaders are leveraging more and more mathematical operations, approaching the ratio of 5:1. ATI felt the 3:1 ratio of ALU:TU was the ideal balance for current and future games.
Here's a nice chart outlining the new lineup:
| Model | Memory Amount | Core Speed | Memory Speed | Interface | ROP | TU | PS | VS | SM | MSRP ($US) |
| X1900 XTX | 512 MB | 650 MHz | 1.55 GHz | 256-bit | 16 | 16 | 48 | 8 | 3.0 | $649 |
| X1900 XT | 512 MB | 625 MHz | 1.45 GHz | 256-bit | 16 | 16 | 48 | 8 | 3.0 | $549 |
| X1900 CrossFire | 512 MB | 625 MHz | 1.45 GHz | 256-bit | 16 | 16 | 48 | 8 | 3.0 | $599 |
| AIW X1900 | 256 MB | 500 MHz | 960 MHz | 256-bit | 16 | 16 | 48 | 8 | 3.0 | $499 |
| X1800XT | 512 MB | 625 MHz | 1.5 GHz | 256-bit | 16 | 16 | 16 | 8 | 3.0 | $599 |
| X1800XL | 256 MB | 500 MHz | 1.0 GHz | 256-bit | 16 | 16 | 16 | 8 | 3.0 | $399 |
The X1900 XTX is ATI’s extreme high-end card, which sports an equally extreme price of $649. It’s clocked at 650MHz core speed and 775MHz memory speed, an increase of 25MHz from the X1800 XT. It’s the first time ATI has used the XTX nomenclature. Frankly it’s ugly and contains too many Xs, XT Extreme. They could’ve used Platinum Edition but choose not to as it carried a poor image of unavailability. The XTX will be highly available for purchase through all channels. The X1900 XT is clocked similarly to the X1800 XT with the exception in memory which dropped 25MHz. It’s at a more “reasonable” price of $549. The CrossFire edition for the X1900 is clocked at the lower speed of the XT which is opposite to the X1800 CrossFire that was clocked at the higher model. Finally there is the AIW X1900 which is clocked similar to the AIW X1800XL at 500MHz/480MHz. All four cards will be available at launch.
The cooler is still the big heavy dual slot design that ATI originally developed for the X1800XT and, unfortunately, there doesn’t appear to be any improvements made to it for the X1900. It’s relatively loud, certainly louder than the 7800 GTX fan without question, but when the card is in 2D mode it’s fairly quiet and doesn’t annoy all that much. In 3D mode it spins up fairly quickly under the load however, and given the right circumstances can get extremely loud and annoying.
The cooler certainly does seem to do the job it was designed for though. It pumps a ton of heat out the back of your case, which is good considering how hot the exhaust air feels on the palm of my hand. You’d really hate to have all that dumped in your case.
The back of the card is bereft of any notable features save a small springy retention bracket used to keep the ginormous cooler mounted firmly to the surface of the card. On a sidenote, if you want to burn your finger, touch this when the card has been running benchmarks for a couple hours.
 Samsung 1.1ns GDDR3 |
 Rage Theater |
 X1900XTX Backside |
Resolutions
Image Quality Settings
Test System Specs
| Radeon X1900 XTX | Radeon X1900 XT Crossfire | Radeon X1800 XT | GeForce 7800 GTX 512 | GeForce 7800 GTX | |
| Core | R580 | R580 | R520 | G70 | G70 |
| Silicon Process | 90nm low-k | 90nm low-k | 90nm low-k | 110nm | 110nm |
| Transistor Count (millions) |
384 | 384 | 321 | 302 | 302 |
| Core Speed MHz | 650 | 625 | 625 | 550 | 430 |
| Memory Speed MHz (Effective) | 775
(1.55 GHz) |
725
(1.45 GHz) |
750
(1.50 GHz) |
850
(1.70 GHz) |
600
(1.20 GHz) |
| Memory Size | 512 MB | 512 MB | 512 MB | 512 MB | 256 MB |
| Bus Standard | PEG 16x | PEG 16x | PEG x16 | PEG x16 | PEG x16 |
| Bus Width | 256bit | 256bit | 256bit | 256bit | 256bit |
| Pixel Pipelines | 16 | 16 | 16 | 24 | 24 |
| Pixel Shaders | 48 | 48 | 16 | 16 | 16 |
| Vertex Shaders | 8 | 8 | 8 | 8 | 8 |
| Peak Memory Bandwidth (GB/s) |
49.6 | 46.4 | 48.0 | 54.4 | 38.4 |
| Pixel Fillrate (million pixels/sec) |
10,400 | 10,000 | 10,000 | 13,200 | 10,320 |
| Texel Fillrate (million texels/sec) |
10,400 | 10,000 | 10,000 | 8,800 | 6,880 |
| API Compliancy | DX 9.0c Shader Model 3.0 |
DX 9.0c Shader Model 3.0 |
DX 9.0c Shader Model 3.0 |
DX 9.0c Shader Model 3.0 |
DX 9.0c Shader Model 3.0 |
Note that, in lieu of an actual X1900XT, we are using an X1900 Crossfire master card which has the exact same clock speed and specifications of the X1900XT.
Games Benchmarks (click for settings)
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.
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 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.
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, apps, utilities, and hardware drivers needed for the testing procedure except for the graphics drivers. Using Norton Ghost, I then cloned the drive onto a second identical hard-drive. After that I installed the ATI drivers on one hard-drive and the NVIDIA drivers on the other. Testing the videocards was then a simple matter of swapping videocards and hard-drives when required.
F.E.A.R.
Benchmarking FEAR was simply a matter of running the in-game Performance Test. Some of the action in the sequence is random, but for the most part it produces reliable, repeatable results. It doesn't reflect actual game play but it should give us an idea of how these cards will perform relative to each other.
Quake 4
I benchmarked Quake 4 using 3 custom timedemos then averaged the score from each demo to get the final score for that resolution and setting. The game was set to the "High" setting for all cards, with Anti-aliasing and Anisotropy set in the game.
I used a custom batch file which automatically runs each demo, resolution, and setting. The commandline used in the batch file is detailed below.
The Quake 4 results are below. Click the text links at the top of the chart to change settings.
Serious Sam 2
Serious Sam 2 was benchmarked using the three timedemos that come with the game. The results of each demo were averaged to get the final result for the resolution and setting. A custom batch file was used to automate the process.
The first set of charts were run with HDR disabled and the seond set with HDR enabled. All other settings stayed the same. Since only ATI's X1000 family of cards support HDR with Antialiasing, only those cards were tested with AA in the second set of charts.
The commandline used to launch Serious Sam 2 is below:
Battlefield 2
Battlefield 2 benchmarking is a little tricky. It has a built in time demo feature, but the results it produces can be very unreliable because it starts logging frame rate on the menu screen, before the actual demo starts. The most reliable method to get results is to take the frame rate log the time demo produces (.csv file) and sample the last few thousand frames (in our case we sampled the last 7000 frames). You need to sample the last of the frames because the game starts logging timedemo results duing the loading sequence, which greatly skews the results.. I used a custom timedemo to get the results below.
Splinter Cell Chaos Theory
Splinter Cell Chaos Theory was also benchmarked using a custom batch file. Anti-aliasing and anisotropy were set within the game. The first chart shows all three cards using the same Shader Model 1.1 path, which is the setting the game defaults to.
In the second chart the Shader Model 3.0 path was chosen and all the SM3.0 features were enabled. Even though the X1000 family supports HDR with AA, Splinter Cell Chaos Theory needs to be patched for it to work. As yet there is no such patch, so we only have 0/0 and 0/16 results there.
Half-Life 2: Lost Coast
To get results from Lost Coast we recorded a custom timedemo, which actually ran for the whole length of the level, and manually ran each resolution and setting.
Half-Life 2
Testing of Half-Life 2 was done using 4 custom Source Engine 7 time demos from various sections of the game. Because there is some frame rate variance during the Half-Life 2 benchmark process, we ran each timedemo for each resolution and AA/AF setting 3 times then averaged the results to get the final score. Anti-aliasing and Anisotropy were set on the command line.
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:
hover your mouse over the yellow text to get a brief description of what each switch does
Doom 3
Doom3 was tested with 3 custom timedemo and the resutls of each timedemo were averaged to get the final score for that resolution and setting. We benchmarked combinations of Anti-Aliasing and Anisotropy over the resolutions shown in the chart below. Anti-aliasing and Anisotropy were set on the command line.
Another batch file was used to automate Doom 3 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:
Chronicles of Riddick: Escape from Butchers Bay
Chronicles of Riddick was benchmarked using another custom batch file that ran each of the 3 built in timedemos included with the 1.01 patch. The Shader Model 2.0 path was used for all cards and Anti-aliasing and Anisotropy were set in the graphics card control panel.
Far Cry
Again a custom batch file was used to benchmark Far Cry. We used 3 demos included with the newer patches to test performance (from the Cooler, Training, and Volcano levels). Because Far Cry benchmark frame rates 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:
Pacific Fighters
Pacific Fighters was benchmarked by loading the included "N1K1 vs BeauFighter.ntrk" track and logging 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 "ATI Radeon X800/9800/9700/9600/9500" profile for the ATI cards and "NVIDIA GeForce 6800/6600/FX/4/3" profile for the NVIDIA cards. Anti-aliasing and Anisotropy were set via the graphics card control panel.
 |
 Battlefield 2 X1900XTX  Battlefield 2 7800GTX |
 |
 Splinter Cell Chaos Theory X1900XTX  Splinter Cell Chaos Theory 7800GTX |
 |
 Doom 3 X1900XTX  Doom 3 7800GTX |
 |
 Far Cry X1900XTX  Far Cry 7800GTX |
To test overclocking capability we used ATI's Overdrive.
OverDrive, at least in it's current state, has pretty shallow limits on clocks when it comes to the X1900XTX, just 690MHz maximum for the core speed and 800MHz max for the memory. We hit these limits without trouble almost immediately, which tells us that the card can go a lot higher. Unfortunately the only third party tools currently available which can overclock the X1900XTX past these speeds are not as stable as we'd like to see (I did manage to get the memory speed up to 855MHz with ATI Tool, but it would crash when trying to modify the core speed), so for now we're going to test the performance increase with the maximum OverDrive overclock and hopefully revisit X1900XTX overclocking when better tools become available.
So, in summary, we are testing the card at an overclocked core speed of 690MHz (which represents a 6.2% increase over the default speed of 650MHz), and with the memory set to 800MHz (which is a tiny 3.2% increase over the default 775MHz). These clocks we are very likely far under the capabilities of the X1900XTX.
The following charts represent the performance increase in Doom 3 and Splinter Cell: Chaos Theory with the card running at maximum OverDrive.
|
[ No AA / No AF ]
[ No AA / 16x AF ]
[ 4x AA / No AF ]
[ 4x AA / 16x AF ]
|
| 1280x1024 | 1600x1200 | 1920x1200 | 2048x1536 | |
| % Increase | 1.0% | 3.0% | 3.8% | 4.3% |
|
[ No AA / No AF ]
[ No AA / 16x AF ]
|
| 1280x1024 | 1600x1200 | 1920x1200 | 2048x1536 | |
| % Increase | 1.4% | 3.1% | 3.9% | 4.3% |
The X1900 series represents a nice boost over the short lived X1800 series. The X1900 builds on the strong foundation set by the X1800 series but provides an overall more balanced design (increase in pixel shaders) in accelerating today’s games yet also leaving room for greater potential performance in running future games. Best part about the launch is availability. ATI is so confident in availability that they are making available the whole line of X1900s at once, XTX, XT, Crossfire and AIW!
So how does the new line stack up? The XTX on average performs 15.5% faster than the X1800XT and the 1900XT at 11.8%. We can see the impact of increasing the pixel shaders as the X1900XT actually has a lower speed on the memory side compared to the X1800XT. I wouldn’t get a X1800XT unless you can find it cheap on the market. The X1900XTX is definitely not worth the extra $100 premium as it’s only 3.2% faster than the X1900XT.
Compared to the competition the GTX 512 holds up fairly well and if you only play games based on Doom3 engine then that should be your choice. Really though the GTX 512 is basically not available for purchase which makes the X1900 that much better. The X1900 really flexes it muscles with all the bells and whistles on and leaves the GTX 512 behind. The X1900 clearly surpasses the GTX 256.
The X1900 is an all around solid product from ATI that really is bringing them back on track. It can only get better as driver enhancements leave room for improvement because the architecture is still relatively new. ATI continues to put tremendous amount of resources into drivers which has given us solid Catalyst drivers each and every month. The X1900 provides performance, stability, CrossFire capability, high resolution gaming, HDR with AA, high quality video processing technology and crisp display output. Couldn’t ask for more well except for a lower price ;) !
