| Sapphire Radeon HD 4870 X2 Video Card 100251SR |
| Reviews - Featured Reviews: Video Cards | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Written by Olin Coles | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Monday, 18 August 2008 | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Sapphire Radeon 4870 X2Technology doesn't play favorites, and when it reaches an apex there is an indifference towards who develops it. ATI has learned this lesson the hard way, having watched from the sidelines for so long that most have forgotten that AMD could be a leader in technology. The Radeon HD 4870 X2 is the latest evolution in graphics cards, and ATI has delivered something bigger than we've ever seen before. Benchmark Reviews is fortunate to test the Sapphire Radeon HD 4870 X2 Dual-RV770 GPU video card 100251SR against a collection of today's hottest video cards available. Tao Le Ching had it right: the more you know, the less you understand. This notion surrounds the computer hardware industry as much as anything, because I have discovered that the moment my experiences lead me towards one opinion, the industry changes and goes into a new direction. Case in point: AMD / ATI. Phenom processors built for the Spider platform have had a very difficult time building momentum against Intel, relegating AMD to being second-best. At the other end of the corporate conglomerate is ATI, which has taken such a beating from NVIDIA that most enthusiasts would have to agree that their future looked bleak prior to Q2 2008. But that's exactly when things changed. At some undetermined point in late June of 2008, ATI and AMD each gained ground on the competition in small steps. AMD launched several enthusiast-level processors, lifting them up out of the tailspin. Around the same time ATI launched their Radeon HD 4850 video card, which directly competes with the GeForce 9800 GTX and GTX+. Then, after a few on-again off-again launch dates, NVIDIA and ATI did an excellent job of confusing the community with a barrage of product launches. NVIDIA came out swinging with their GeForce GTX 280 and GTX 260 video cards, and ATI retaliated with the Radeon HD 4870 featuring the industries first implementation of GDDR5 video memory. When it comes to ATI products, Sapphire has always offered the most influential graphics cards available. This time around, the new Radeon HD 4870 X2 is no different. For the first time in many years, ATI secures both the top position in the graphics card race and introduces a dual-GPU video card with GDDR5. The Sapphire 100251SR model offers 24x custom filter anti-aliasing (CFAA) on its dual-750 MHz 800-core RV770 GPU's.
We live in a world where a company that produces the most powerful video card available is usually regarded as the leader in graphics technology. There is a certain amount of truth to this to be fair, but it's not always good to be the king. I imagine NVIDIA has to be getting a little tired of constantly polishing the throne in a lonely palace, especially after their recent launch of the NVIDIA GeForce GTX 280 offered just as much to gamers as it did to multimedia editors. ATI has offered excellent value with the Radeon HD 4850, and improved upon it with the Radeon HD 4870 graphics card. For this article, Benchmark Reviews takes a hard look at the newest king of the castle - the Radeon HD 4870 X2. About the company: Sapphire Technology
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
Series |
ATI Radeon HD 4870 | ATI Radeon HD 4850 |
ATI Radeon HD 4850 X2 |
ATI Radeon HD 4870 X2 |
| Stream Processors | 800 | 800 |
800x2 |
800x2 |
| Texture Units | 40 | 40 |
40x2 |
40x2 |
| ROPs | 16 | 16 |
16x2 |
16x2 |
| Core Clock | 750MHz | 625MHz |
625MHz |
750MHz |
| Memory Clock | 900MHz (3600MHz data rate) GDDR5 | 993MHz (1986MHz data rate) GDDR3 |
993MHz (1986MHz data rate) GDDR3 |
900MHz (3600MHz data rate) GDDR5 |
| Memory Bus Width | 256-bit | 256-bit |
2x256-bit |
2x256-bit |
| Frame Buffer | 512MB/1GB | 512MB |
2x1GB |
2x1GB |
| Transistor Count | 956M | 956M |
2x514M |
2x956M |
| Manufacturing Process | TSMC 55nm | TSMC 55nm |
TSMC 55nm |
TSMC 55nm |
Unified Superscalar Shader Architecture
-
1600 stream processing units
-
Dynamic load balancing and resource allocation for vertex, geometry, and pixel shaders
-
Common instruction set and texture unit access supported for all types of shaders
-
Dedicated branch execution units and texture address processors
-
-
128-bit floating point precision for all operations
-
Command processor for reduced CPU overhead
-
Shader instruction and constant caches
-
Up to 320 texture fetches per clock cycle
-
Up to 128 textures per pixel
-
Fully associative multi-level texture cache design
-
DXTC and 3Dc+ texture compression
-
High resolution texture support (up to 8192 x 8192)
-
Fully associative texture Z/stencil cache designs
-
Double-sided hierarchical Z/stencil buffer
-
Early Z test and Fast Z Clear
-
Lossless Z & stencil compression (up to 128:1)
-
Lossless color compression (up to 8:1)
-
8 render targets (MRTs) with anti-aliasing support
-
Physics processing support
Anti-Aliasing Features
-
Multi-sample anti-aliasing (2, 4 or 8 samples per pixel)
-
Up to 24x Custom Filter Anti-Aliasing (CFAA) for superior quality
-
Adaptive super-sampling and multi-sampling
-
Gamma correct
-
Super AA (ATI CrossFireXTM configurations only)
-
All anti-aliasing features compatible with HDR rendering
Texture Filtering Features
-
2x/4x/8x/16x high quality adaptive anisotropic filtering modes (up to 128 taps per pixel)
-
128-bit floating point HDR texture filtering
-
sRGB filtering (gamma/degamma)
-
Percentage Closer Filtering (PCF)
-
Depth & stencil texture (DST) format support
-
Shared exponent HDR (RGBE 9:9:9:5) texture format support
Closer Look: Sapphire 4870
The Sapphire HD 4870 X2 delivers a new level of performance to the mainstream user with its 2GB GDDR5 memory, and nominal clock speeds of 750MHz (core) and 900MHz (memory). The 4870 X2 uses the PCI-Express Gen2 interface, and features dual connectors for CrossFireX cables, allowing two or more cards to be used together on a CrossFireX compatible mainboard for even higher graphics performance.
All SAPPHIRE graphics cards in the HD 4800 series incorporate the latest ATI Avivo HD Technology for enhanced Video display and feature a new generation built in hardware UVD (Unified Video decoder) considerably reducing CPU load and delivering smooth decoding of Blu-ray and HD DVD content for both VC-1 and H.264 codecs, as well as Mpeg files. In addition to two, dual-link DVI outputs, and TV-Out, a dedicated HDMI adaptor delivers both audio and video output on a single cable for direct connection to an HDMI ready display.
The newest edition to the Radeon family takes a whole new approach to component layout and appearance. There's a lot of questions I have when it comes to building a twin-GPU graphics card: how much cooling does it need, and how much power will it consume? These are all things I plan to answer with this article.
The Sapphire Radeon HD 4870 X2 graphics card (SKU 100251SR) comes in classic ATI red and offers a double-slot sized product for the PCI-Express 2.0 bus. There are a few small details that seem to stand out as either interesting or unique, and I'll make sure to discuss each at length. Let's begin with looks.
Sapphire has been (and will likely continue to be) an icon for sex-sells marketing within the VGA industry. Palit has their cyborg-frog, XFX has an armor-clad wolf, and ZOTAC has a fire-breathing dragon; but none match the collection of attractive Ruby characters adorning Sapphire product. Although they have tried to simmer it down in the past few years, I still associate racy characters with their marketing graphics. I suppose that there's nothing wrong with an animated cute she-warrior looking out at you from the decals and retail packages Sapphire exposes them on, because after all this is a geeks industry. On the Radeon HD 4870 X2, Sapphire offers up a Laura Croft-like character with goofy goggles on.
There's no question that the AMD-ATI merger has trickled down into a serious economy-of-design state of mind. As I finish editing this article, the nearest competition to ATI has recently suffered a 31% drop in their traded stock value, something that may put the two chipmakers on a more even playing field. Nevertheless, Sapphire continues on with their mission, and produces some of the best video cards money can buy. Value translates into several different things, however, and he design of the Radeon HD 4870 X2 is not without its flaws.
Sapphire utilizes a full-length cooling unit to keep the dual-RV770 GPUs within operating specifications. While it obviously seems to be working, I am curious how ATI avoided a total meltdown considering how hot the Radeon HD 4870 could become. The blower fan is located at the rear-most position of the cooling unit, and draws air from the opening on the face of the cooler. While the majority of heat produced by the twin GPUs is exhausted out the mounting bracket grill, some residual heat is passively cooled by a very unique pin-style heatsink.
The cooling unit on the Sapphire Radeon HD 4870 X2 video card is held tight to the printed circuit board and two RV770 GPUs with the use of reinforcing brackets and over a dozen screws. The double-height cooler does an adequate job of cooling the 4870 X2, but there is still a tremendous amount of heat that builds up on the PCB. If you're an overclocker, there isn't very much that can be done to help cool the unit from the reverse side of the circuit board, especially since there are no surface-mounted GDDR5 modules exposed to this side of the Sapphire 100251SR Radeon HD 4870 X2.
In truth, the double-size active cooling solution is barely enough to contain the fire-breathing 4870 X2. The RV770 GPU is "merely" overclocked from the Radeon HD 4850 to the 4870, and the increase in heat output is directly obvious. I will go into deeper detail about heat output and power consumption later on, but for now we've got a good look at what you get when you buy the Sapphire Radeon HD 4870 X2.
In our next section we detail our methodology for testing video cards. Following this we offer a cadre of benchmarks to show where the Sapphire Radeon HD 4870 X2 stands against the Radeon 4870, GeForce 9800 GX2, a pair of Radeon HD 4850's in CrossFire, along with a GTX 260 and GTX 280 put in for good measure... so please read on!
Video Card Testing Methodology
At the start of all tests, the previous display adapter driver is uninstalled and trace components are removed using Driver Cleaner Pro. We then restart the computer system to establish our display settings and define the monitor. Once the hardware is prepared, we begin our testing. The synthetic benchmark tests in 3DMark06 will utilize shader models 2.0 and 3.0. In our higher-end VGA products we conduct tests at the following resolutions: 1280x1024 (Standard 17-19" LCD), 1680x1050 (22-24" Widescreen LCD), and 1920x1200 (24-28" Widescreen LCD). In some tests we utilized widescreen monitor resolutions, since more users are beginning to feature these products for their own computing.
Each benchmark test program begins after a system restart, and the very first result for every test will be ignored since it often only caches the test. This process proved extremely important in the World in Conflict and Supreme Commander benchmarks, as the first run served to cache maps allowing subsequent tests to perform much better than the first. Each test is completed five times, with the average results displayed in our article.
Our site polls and statistics indicate that the over 90% of our visitors use their PC for playing video games, and practically every one of you are using a screen resolutions mentioned above. Since all of the benchmarks we use for testing represent different game engine technology and graphic rendering processes, I feel that this battery of tests will provide a diverse range of results for you to gauge performance on your own computer system. Since most gamers and enthusiasts are still using Windows XP, it was decided that DirectX 9 would be used for all tests until demand and software support improve for Windows Vista.
Test System
-
Motherboard: Gigabyte GA-X48T-DQ6 (Intel X48 Chipset) with version F5 BIOS
-
Processor: Intel E8400 Core 2 Duo 3.0 GHz (Overclocked to 3.6 GHz)
-
System Memory: Corsair PC3-14400 DDR3 1800MHz
-
Disk Drive: Mtron Pro 7500 32GB SATA-II SSD MSP-SATA7525
-
Optical Drive: ASUS BC-1205PT SATA Blu-ray Disc Optical Drive
-
Operating System: Windows XP Professional SP-3 (optimized to 16 processes at idle)
Benchmark Applications
-
3DMark06 v1.1.0 (8x Anti Aliasing & 16x Anisotropic Filtering)
-
Call of Duty 4: Modern Warefare v1.7.568 (4x AA/16x Trilinear AF using FRAPS)
-
Crysis v1.21 Benchmark (High Settings, 0x and 4x Anti-Aliasing)
-
Unreal Tournament 3 v1.3 (High Quality, 16x Anisotropic Filtering using HOC Benchmark Tool v1.3)
-
World in Conflict v1.0.0.9 Performance Test (Very High Setting: 4x AA/4x AF)
Video Card Test Products
| Product Series | Sapphire Radeon HD 4870 102-B50701-10-AT | XFX GeForce GeForce GTX 260 GX-260N-ADDU | Sapphire Radeon HD 4850 102-B50102-00-AT x2 (CrossFireX) | ZOTAC GeForce GTX 280 AMP! Edition ZT-X28E3LA-FCP | Sapphire Radeon HD 4870 X2 100251SR | |
| Stream Processors | 800 | 192 | 128 (x2) | 800 | 240 | 1600 |
| Core Clock (MHz) | 750 | 640 | 600 | 625 | 700 | 750 |
| Shader Clock (MHz) | N/A | 1242 | 1500 | N/A | 1400 | N/A |
| Memory Clock (MHz) | 900 | 1150 | 1000 | 900 | 1150 | 900 |
| Memory Amount |
512 MB GDDR5 |
896 MB GDDR3 | 512MB (x2) GDDR3 | 512 MB GDDR3 | 1024 MB GDDR3 | 1024MB (x2) GDDR5 |
| Memory Interface | 256-bit | 448-bit | 256-bit | 256-bit | 512-bit | 256-bit |
-
XFX GeForce GeForce GTX 260 GX-260N-ADDU (640 MHz GPU/1242 Shader/1150 RAM - Forceware v177.79)
-
Gigabyte GeForce 9800 GX2 GV-NX98X1GHI-B (600 MHz GPU x2/1500 Shader/1000 RAM - Forceware v177.79)
-
Sapphire Radeon HD 4850 102-B50102-00-AT CrossFireX Set (625 MHz GPU/993 MHz RAM - Catalyst 8.7)
-
ZOTAC GeForce GTX 280 AMP! Edition ZT-X28E3LA-FCP (700 MHz GPU/1400 MHz Shader/1150 MHz RAM - Forceware v177.79)
-
Sapphire Radeon HD 4870 X2 100251SR (750 MHz GPUs/900 MHz RAM - Catalyst 8.8 Step 2)
Now we're ready to begin testing video game performance on the Sapphire Radeon HD 4870 X2, so please continue to the next page as we start with the 3DMark06 results.
3DMark06 Test Results
3DMark is a computer benchmark by Futuremark (formerly named Mad Onion) to determine the DirectX 9 performance of 3D game performance with graphics cards. 3DMark06 uses advanced real-time 3D game workloads to measure PC performance using a suite of DirectX 9 3D graphics tests, CPU tests, and 3D feature tests.
3DMark06 tests include all new HDR/SM3.0 graphics tests, SM2.0 graphics tests, AI and physics driven single and multiple cores or processor CPU tests and a collection of comprehensive feature tests to reliably measure next generation gaming performance today. Some enthusiasts may note that Benchmark Reviews does not include CPU-bound tests in our benchmark battery, and that only graphic-bound tests are included.
Here at Benchmark Reviews, we believe that synthetic benchmark tools are just as valuable as video games, but only so long as you're comparing apples to apples. Since the same test is applied in the same controlled method with each test run, I believe 3DMark is a very reliable tool for comparing graphic cards against one-another.
More visitors to Benchmark Reviews operate at 1280x1024 resolution than any other, as it represents the native resolution of 19" LCD monitors. Using this resolution as a starting point, the maximum settings were applied to 3dMark06 which for these tests include 8x Anti-Aliasing and 16x Anisotropic Filtering. Low-resolution testing allows the graphics processor to plateau maximum output performance, which thereby shifts demand onto the system components to keep up. At the lower resolutions 3DMark will reflect the GPU's top-end speed in the composite score, indicating full-throttle performance with little load. This makes for a less GPU-dependant test environment, and is helpful in measuring the maximum output performance in the test results.
Right away our test results indicate that 3dMark06 benchmarks really prefer the CrossFireX pair of Sapphire Radeon HD 4850 video cards. But because of the added overhead of combined graphics processors and video frame buffer memory, the light load created by the shader model 2.0 tests have a relatively negative impact on the CrossFireX score. Another way of describing this phenomenon is comparing the combined Radeon HD 4850 video cards to a race car that has only high gears available in the transmission: it will have a faster top-end speed, but it will take longer to get there. The ZOTAC GeForce GTX 280 AMP! Edition video card is appears to also be another high-gear example, as gets nudged-out by the GeForce 9800 GX2 for SM 2.0 tests but outperforms the GX2 in the more complex HDR/SMR 2.0 tests.
In the SM 2.0 tests the Sapphire Radeon HD 4870 X2 keeps just ahead of the Radeon HD 4850 CrossFireX set by a mere 14%, and ahead of the dual-GPU Gigabyte GeForce 9800 GX2 by an astounding 27%. The overclocked GeForce GTX 280 AMP! Edition video card trails behind 28%. However, moving into the more advanced HDR tests the Radeon HD 4870 X2 produces a much more definitive lead over the entire test group, as it can only be matched with a set of Radeon HD 4850's.
At the widescreen resolution of 1680x1050, the scores are practically identical in terms of ratio to all of our previous tests. Once again the shader model 2.0 tests put the Sapphire Radeon 4870 X2 slightly ahead of the CrossFire set of HD 4850's, while the rest of the graphic cards fall in at different level of lesser performance. This is the case until they reach the shader model 3.0 tests, where everything seems to go flat across the board. Twin Sapphire Radeon HD 4850 video cards in a CrossFireX set are still running circles around the competition in every 3dMark06 benchmark we run, and makes the argument for inexpensive multi-card performance. The Sapphire Radeon HD 4870 X2 surpasses CrossFireX set of Radeon 4850's by only 5% while beating the overclocked ZOTAC GeForce GTX 280 AMP! Edition by a hefty 54%!
Finishing up the series of synthetic benchmark tests under heavy load, the Gigabyte GeForce 9800 GX2 video card roughly matches the AMP!'ed GTX 280 SM 2 tests, yet the CrossFireX set of Radeon HD 4850's and Sapphire Radeon HD 4870 X2 dominate over the entire field in the more demanding shader model 3.0 tests.
Two things were made very clear with the 3dMark06 benchmark tests: NVIDIA isn't winning the fight, and the Sapphire Radeon HD 4870 X2 is so far ahead of any other video card that it redefines the high-end of graphics performance.
| Product Series | Sapphire Radeon HD 4870 102-B50701-10-AT | XFX GeForce GTX 260 GX-260N-ADDU | Sapphire Radeon HD 4850 102-B50102-00-AT x2 (CrossFireX) | ZOTAC GeForce GTX 280 AMP! Edition ZT-X28E3LA-FCP | Sapphire Radeon HD 4870 X2 100251SR | |
| Stream Processors | 800 | 192 | 128 (x2) | 800 | 240 | 1600 |
| Core Clock (MHz) | 750 | 640 | 600 | 625 | 700 | 750 |
| Shader Clock (MHz) | N/A | 1242 | 1500 | N/A | 1400 | N/A |
| Memory Clock (MHz) | 900 | 1150 | 1000 | 900 | 1150 | 900 |
| Memory Amount |
512 MB GDDR5 |
896 MB GDDR3 | 512MB (x2) GDDR3 | 512 MB GDDR3 | 1024 MB GDDR3 | 1024MB (x2) GDDR5 |
| Memory Interface | 256-bit | 448-bit | 256-bit | 256-bit | 512-bit | 256-bit |
Take the 3DMark06 tests at face value, because in our next section we begin real-world testing on a cadre of popular video games known for taxing the graphics processor, and the performance curve is expected change. Our first up is Call of Duty 4, so please continue on...
Call of Duty 4 Benchmarks
Call of Duty 4: Modern Warfare runs on a proprietary game engine that Infinity Ward based off of the tried-and-true Q3 structure. This engine offers features such as true world-dynamic lighting, HDR lighting effects, dynamic shadows and depth of field. "Bullet Penetration" is calculated by the Infinity Ward COD4 game engine, taking into account things such as surface type and entity thickness. Certain objects, such as cars, and some buildings are destructible. This makes distinguishing cover from concealment important, as the meager protection provided by things such as wooden fences and thin walls does not fully shield players from harm as it does in many other games released during the same time period. Bullet speed and stopping power are decreased after penetrating an object, and this decrease is calculated realistically depending on the thickness and surface of the object penetrated.
This version of the game also makes use of a dynamic physics engine, a feature which was not implemented in previous Call of Duty titles for Windows PC's. The new in-game death animations are a combination of pre-set static animations combined with ragdoll physics. Infinity Ward's use of the well-debugged Quake 3 engine along with new dynamic physics implementation allows Call of Duty 4 to be playable by a wide range of computer hardware systems. The performance may be scaled for low-end graphic cards up to 4x Anti-Aliasing and 16x Tri-linear anisotropic texture filtering.
Before I discuss the results, I would like to take a moment to mention my general opinion on Fraps software when it comes to game performance benchmarking. If you're not familiar with the software, Fraps (derived from Frames per second) is a benchmarking, screen capture, and real-time video capture utility for DirectX and OpenGL applications. Some reviewers use this software to measure video game performance on their Windows system, as well as record gaming footage. My opinion is that it offers a valid third-party non-bias alternative to in-game benchmarking tools; but there is one caveat: it's not perfect. Because the user must manually begin the test, the starting point may vary from position to position and therefore skew the results.
In my testing with Fraps v2.9.4 build 7039, I used the cut-scene intro to the coup d'etat scene when Al Asad takes over control. First I allowed the level to load and let the scene begin for a few moments, then I would use the escape key to bring up the menu and choose the restart level option, I would immediately press F11 to begin recording the benchmark data. This scene is nearly four minutes long, but I configured Fraps to record the first 180 seconds of it to remain consistent. Once the scene would end, I would repeat the restart process for a total of five tests. So within a 2 millisecond starting point margin, all benchmark results are comparable which is probably as good as it can possibly get with this tool.
In our frame rate results, all five of the collected test scores were within 0.5 FPS of one-another and then averaged for the chart you see above. Call of Duty 4 showed some small degree of difference in graphics performance at the lower resolution of 1280x1024, but it tapered out thereafter for both the 1680x1050 and 1920x1200 resolutions.
There were three distinct groups of results: the lower high-end consisting of the Sapphire Radeon HD 4870 and XFX GeForce GTX 260, the mid high-end consisting of the GeForce 9800 GX2 and ZOTAC GeForce GTX 280 AMP! Edition, and then the Sapphire Radeon HD 4870 X2 occupying the upper high-end. The CrossFireX set of Radeon 4850's were not included in this benchmark because of driver issues.
Call of Duty 4 put a reasonable amount of strain on the Radeon HD 4870 and GeForce GTX 260 video cards, but they both did very well for themselves during our tests. Since the maximum anti-aliasing available in COD4 is 4x, there won't be any problem with the Radeon 4800-series limit of 8x AA.
Previously the ZOTAC GeForce GTX 280 AMP! Edition was the fastest graphics card on the planet at 1920x1200 resolution, however the Radeon HD 4870 X2 outperforms it by 25%. Seems like we've got a new king, and he's way ahead of the other heirs to the throne.
| Product Series | Sapphire Radeon HD 4870 102-B50701-10-AT | XFX GeForce GTX 260 GX-260N-ADDU | ZOTAC GeForce GTX 280 AMP! Edition ZT-X28E3LA-FCP | Sapphire Radeon HD 4870 X2 100251SR | |
| Stream Processors | 800 | 192 | 128 (x2) | 240 | 1600 |
| Core Clock (MHz) | 750 | 640 | 600 | 700 | 750 |
| Shader Clock (MHz) | N/A | 1242 | 1500 | 1400 | N/A |
| Memory Clock (MHz) | 900 | 1150 | 1000 | 1150 | 900 |
| Memory Amount |
512 MB GDDR5 |
896 MB GDDR3 | 512MB (x2) GDDR3 | 1024 MB GDDR3 | 1024MB (x2) GDDR5 |
| Memory Interface | 256-bit | 448-bit | 256-bit | 512-bit | 256-bit |
In our next section, we shall see if the performance-demanding video game Crysis will help strengthen this position.
Crysis Benchmark Results
Crysis uses a new graphics engine: the CryENGINE2, which is the successor to Far Cry's CryENGINE. CryENGINE2 is among the first engines to use the Direct3D 10 (DirectX10) framework of Windows Vista, but can also run using DirectX9, both on Vista and Windows XP.
Roy Taylor, Vice President of Content Relations at NVIDIA, has spoken on the subject of the engine's complexity, stating that Crysis has over a million lines of code, 1GB of texture data, and 85,000 shaders. To get the most out of modern multicore processor architectures, CPU intensive subsystems of CryENGINE 2 such as physics, networking and sound, have been re-written to support multi-threading.
Crysis offers an in-game benchmark tool, which is similar to World in Conflict. This short test does place some high amounts of stress on a graphics card, since there are so many landscape features rendered. For benchmarking purposes, Crysis can mean trouble as it places a high demand on both GPU and CPU resources. Benchmark Reviews uses the Crysis Benchmark Tool by Mad Boris to test frame rates in batches, which allows the results of many tests to be averaged.
The very first thing we discovered in the low-resolution tests was how seemingly poor both of our multi-GPU products performed. The Gigabyte GeForce 9800 GX2 was the lowest of the group at 1280x1024, matched in performance to a single Sapphire Radeon HD 4850 or GeForce 9800 GTX. The CrossFireX set of 4850's suffered the same rapid-response overhead bottleneck and performed almost the same as a single Radeon 4870 or GTX 260. To be fair, none of these video cards will probably ever realistically see game-play at a resolution this low, so this performance illustrates how high-end GPU power can be cut short if the monitor (resolution) doesn't match it.
Low-resolution testing allows the graphics processor to plateau maximum output performance, which thereby shifts demand onto the system components. At the lower resolutions Crysis will reflect the GPU's top-end speed in the composite score, indicating full-throttle performance with little load. This makes for a less GPU-dependant test environment, and is helpful in creating a baseline for measuring maximum output performance in the next few test results. At the 1280x1024 resolution used by 19" monitors, our results show that performance is beginning to really drop despite the small difference is pixels drawn. In terms of general performance, all of these products maintain the same performance ratio as before, except for the 9800 GX2 which seems to beneficially hold its ground.
The CrossFireX set of HD 4850's is going to soon reach it's limit, as they are in last gear and the 9800 GX2 is still shifting up the tree. Sapphire's Radeon HD 4870 flexes GDDR5 muscle at 1680x1080, running alongside the overclocked ZOTAC GeForce GTX 280 AMP! Edition ZT-X28E3LA-FCP with only a 5% difference. For widescreen users, our benchmarks below indicate that the ATI Radeon HD 4870 matches the performance of NVIDIA's GeForce 9800 GX2 video card, although the 4870 stops delivering post-processing effects at 8x AA and the 9800 GX2 can reach 32x AA if the application supports it. The CrossFireX set of HD 4850's appear to be matching performance of a single Radeon HD 4870 video card; at least as far as Crysis is concerned. Testing in high-pressure Crysis also seems to have effected the Sapphire Radeon HD 4870 X2 we tested, which finally succombs to the AMP!'ed GeForce GTX 280.
Reading the results of our 1920x1200 resolution tests using SOYO's DYLM26E6 monitor for testing, Crysis forced 2.3 million pixels to be processed by our graphical test products. At our highest widescreen resolution, the Sapphire Radeon HD 4870 performs on par with the GTX 260. Despite what 3dMark06 previously reported, the CrossFireX set of Radeon HD 4850's is not king; as a close struggle between the GeForce GTX 280 and Sapphire Radeon HD 4870 X2 comes to a head. If only by a small difference, the overclocked ZOTAC GTX 280 enjoys a 5% (2 FPS) lead over the GeForce 9800 GX2, which also seem keep its own 8% (3 FPS) lead over the CrossFireX set.
At the end of our Crysis testing, neither the GeForce 9800 GX2, GTX 280, or CrossFireX HD 4850 set could touch the multi-GPU Radeon HD 4870 X2. The new king is earning his crown, but as the pressure gets higher the challengers rise to the occasion. Before we leave Crysis though, I decided to include a look at post-processing performance with 4x AA enabled at the 1680x1050 and 1920x1200 widescreen resolutions. The chart below shows the average frame rate performance with 4x Anti-Aliasing enabled.
At 1680x1050, the Radeon HD 4870 is no match for the GeForce GTX 260 it's trailed thus far. Additionally, the CrossFireX set of 4850's falls 28% short of reaching the Gigabyte GeForce 9800 GX2. As if it was really unexpected, ZOTAC's GTX 280 AMP! Edition comes out (way) on top with an impressive 18% lead over the Sapphire Radeon HD 4870 X2 video card. But once the Honeywell 22-Inch LCD was swapped out, I began testing at 1920x1200 and the differences were made very clear.
At 1920x1200 the multi-GPU CrossFireX array of 4850's performed almost identically to a single Radeon HD 4870, while the GeForce 9800 GX2 seemed to fall flat. A heavily overclocked GTX 280 stills shows a prevailing strength, as it produces a 12% lead over the dual-GPU 4870 X2.
With only a small dose of anti-aliasing added to Crysis, there are very few products that would make for playable frame rates. Our Island timedemo mixes a some beach and water views so it's going to be on the high side of frame rates when compared to actual game play, but as you can see the Radeon products do very well when post-processing effects are added. Sapphire's hot-potato HD 4870 matched performance of our 4850 CrossFireX set, but trailed behind the GTX 280 by 52%. It appears that the G92 GPU really stands out like a sore thumb against the newer graphics processors, which yield far better frame rates in our Crysis testing.
| Product Series | Sapphire Radeon HD 4870 102-B50701-10-AT | XFX GeForce GTX 260 GX-260N-ADDU | Sapphire Radeon HD 4850 102-B50102-00-AT x2 (CrossFireX) | ZOTAC GeForce GTX 280 AMP! Edition ZT-X28E3LA-FCP | Sapphire Radeon HD 4870 X2 100251SR | |
| Stream Processors | 800 | 192 | 128 (x2) | 800 | 240 | 1600 |
| Core Clock (MHz) | 750 | 640 | 600 | 625 | 700 | 750 |
| Shader Clock (MHz) | N/A | 1242 | 1500 | N/A | 1400 | N/A |
| Memory Clock (MHz) | 900 | 1150 | 1000 | 900 | 1150 | 900 |
| Memory Amount |
512 MB GDDR5 |
896 MB GDDR3 | 512MB (x2) GDDR3 | 512 MB GDDR3 | 1024 MB GDDR3 | 1024MB (x2) GDDR5 |
| Memory Interface | 256-bit | 448-bit | 256-bit | 256-bit | 512-bit | 256-bit |
In our next section, Benchmark Reviews tests with Unreal Tournament 3. Read on to see how a blended high-demand GPU test with low video frame buffer demand will impact our test products.
Unreal Tournament 3
Unreal Tournament 3 (UT3) is a first-person shooter and online multiplayer video game by Epic Games and is the next installment of the Unreal series after Unreal Tournament 2004. It is published by Midway Games and was released in North America for Windows on November 19, 2007.
Unreal Tournament 3 is actually the fourth game in the Unreal Tournament series and the eighth Unreal game, but it has been numbered in terms of the engine it runs on. UT3 is subsequently part of the third generation, because it runs on the Unreal Engine 3, and does not reuse any content from previous versions.
Since Unreal Tournament 3 was designed as a DirectX 9 video game with no current support expected for DirectX 10, we use Windows XP Pro (Service Pack 3) for our benchmark testing. After completing tests on a wide range of products with settings at their highest, it appeared that Unreal Tournament 3 really didn't stress the video cards nearly as much as I would have liked.
Beginning at the low resolution of 1280x1024, the benchmark scores are so close (and high) for some products that it might be time to eliminate this game from our testing process. Nevertheless, it looks like the Unreal Engine 3 game engine doesn't care too much for the ATI Radeon HD 4850 or 4870 video cards. The CrossFireX set was just barely able to keep pace with the others that it surpassed in previous tests. With all High Quality settings and tweaks enabled along with 16x anisotropic filtering, Unreal Tournament 3 doesn't add strain to any of the graphics cards tested like Crysis did.
As the resolution was raised, the once level performance between the GTX 260 and GeForce 9800 GX2 is beginning to split apart. The ZOTAC GeForce GTX 280 AMP! Edition loses just a single frame per second (0.6%) between 1280x1024 and 1680x1050, and the Sapphire Radeon 4870 X2 loses only 2 FPS. For now it appears that just about any graphics card can play Unreal Tournament 3 without issue, but quite frankly I don't know anyone who actually plays this game.
When I tested the Honeywell HWLM2216 recently, I noticed how the 1680x1050 widescreen display resolution of this 22" LCD monitor offered very little strain over a 19" standard display LCD monitor. Comparatively, 1680x1050 produces 1.76 MP and 1280x1024 produces 1.31, so there's only a very small difference expected between performance levels. The biggest difference is in the user experience, because the widescreen monitor comes in very handy for watching multimedia video or playing large world-scape video games.
At 1680x1050 resolution, the differences were beginning to show, but only 1920x1200 will be useful for illustrating how each product performs. Producing 2.3 MP with a 1920x1280 resolution on our 26" SOYO DYLM26E6 test monitor, each product is now separated far enough apart to sort out the winners and losers. The Sapphire Radeon HD 4870 still trails way behind the GeForce GTX 260 and 9800 GX2, but put two 4850 together for a CrossFireX set and the Radeon's look more like contenders. The GeForce 9800 GX2 puts both G92 graphics processors to good use and beats the 4870 and both CrossFireX 4850's by 16%, but the overclocked ZOTAC GTX 280 AMP! Edition video card still pulls off a very narrow victory with a 2 FPS advantage. Uncharacteristically, the Sapphire Radeon HD 4870 X2 falls more in-line with the GeForce GTX 260 in UT3, which could raise some interesting questions.
Similar to our low-resolution tests, Unreal Tournament 3 appears to provide a very minimal load on the high-end video cards we're testing. Thankfully there are several new games arriving to market late into 2008, so with some luck this benchmark will only be used for low-end graphics comparison into the future.
| Product Series | Sapphire Radeon HD 4870 102-B50701-10-AT | XFX GeForce GTX 260 GX-260N-ADDU | Sapphire Radeon HD 4850 102-B50102-00-AT x2 (CrossFireX) | ZOTAC GeForce GTX 280 AMP! Edition ZT-X28E3LA-FCP | Sapphire Radeon HD 4870 X2 100251SR | |
| Stream Processors | 800 | 192 | 128 (x2) | 800 | 240 | 1600 |
| Core Clock (MHz) | 750 | 640 | 600 | 625 | 700 | 750 |
| Shader Clock (MHz) | N/A | 1242 | 1500 | N/A | 1400 | N/A |
| Memory Clock (MHz) | 900 | 1150 | 1000 | 900 | 1150 | 900 |
| Memory Amount |
512 MB GDDR5 |
896 MB GDDR3 | 512MB (x2) GDDR3 | 512 MB GDDR3 | 1024 MB GDDR3 | 1024MB (x2) GDDR5 |
| Memory Interface | 256-bit | 448-bit | 256-bit | 256-bit | 512-bit | 256-bit |
Our last benchmark of the series is coming next, which puts our collection of video cards against some very demanding graphics with World in Conflict.
World in Conflict Results
The latest version of Massive's proprietary Masstech engine utilizes DX10 technology and features advanced lighting and physics effects, and allows for a full 360 degree range of camera control. Massive's MassTech engine scales down to accommodate a wide range of PC specifications, if you've played a modern PC game within the last two years, you'll be able to play World in Conflict.
World in Conflict's FPS-like control scheme and 360-degree camera make its action-strategy game play accessible to strategy fans and fans of other genres... if you love strategy, you'll love World in Conflict. If you've never played strategy, World in Conflict is the strategy game to try.
World in Conflict offers an in-game benchmark; which records the minimum, average, and maximum frame rates during the test. Very recently another hardware review website made the assertion that these tests are worthless, but we couldn't disagree more. When used to compare video cards which are dependant on the same driver and use the same GPU architecture, the in-game benchmark works very well and comparisons are apples-to-apples.
Based on the test results charted below it's clear that WiC doesn't place a limit on the maximum frame rate (to conserve wasted power) which is good for full-spectrum benchmarks like ours, but bad for electricity bills. The average frame rate is shown for each resolution in the chart below. At the lower 1280x1024 resolution the playing field all keeps around 60 FPS, with exception to the GeForce GTX 280 with top performance frame rates of 67. The Radeon 4850 CrossFireX pair and Sapphire 4870 X2 both suffer from overhead bottlenecks, and perform the worst of the bunch at this resolution.
At 1680x1050 the Sapphire Radeon HD 4870 X2 sustains the same frame rate of 59 FPS, but all other products degrade if only slightly. The GeForce 9800 GTX surges ahead to within range of the ZOTAC GeForce GTX 280 AMP! Edition, but it's going to take a much higher resolution to pull these products away from a common position on the charts.
With a balanced demand for CPU and GPU power, World in Conflict just begins to place demands on the graphics processor at the 1920x1280 resolution. I was expecting more results along the same line I've seen so far, and that is pretty much exactly what I got. The performance decay had its hardest impact on the lower high-level video cards: Radeon HD 4870 and GeForce GTX 260, which for all intents an purposes performed extremely well up to this point in our WiC testing. Two HD 4850's in CrossFireX configuration will yield a 46% improvement over using only one and exceed performance of a single Sapphire Radeon HD 4870 by 19%, but still falls just 2 FPS short of matching the WiC-shy 59 FPS of the 4870 X2. The GeForce 9800 GX2 didn't budge one single frame as it scaled from 1.31 MP up to 2.3 MP, showing how much raw power this product delivers in the right application and finally pushes past the overclocked GeForce GTX 280.
| Product Series | Sapphire Radeon HD 4870 102-B50701-10-AT | XFX GeForce GTX 260 GX-260N-ADDU | Sapphire Radeon HD 4850 102-B50102-00-AT x2 (CrossFireX) | ZOTAC GeForce GTX 280 AMP! Edition ZT-X28E3LA-FCP | Sapphire Radeon HD 4870 X2 100251SR | |
| Stream Processors | 800 | 192 | 128 (x2) | 800 | 240 | 1600 |
| Core Clock (MHz) | 750 | 640 | 600 | 625 | 700 | 750 |
| Shader Clock (MHz) | N/A | 1242 | 1500 | N/A | 1400 | N/A |
| Memory Clock (MHz) | 900 | 1150 | 1000 | 900 | 1150 | 900 |
| Memory Amount |
512 MB GDDR5 |
896 MB GDDR3 | 512MB (x2) GDDR3 | 512 MB GDDR3 | 1024 MB GDDR3 | 1024MB (x2) GDDR5 |
| Memory Interface | 256-bit | 448-bit | 256-bit | 256-bit | 512-bit | 256-bit |
In our next section, we discuss electrical power consumption and learn how well (or poorly) each video card will impact your utility bill...
Power Consumption
Life is not as affordable as it used to be, and items such as gasoline, natural gas, and electricity all top the list of resources which have exploded in price over the past few years. Add to this the limit of non-renewable resources compared to current demands, and you can see that the prices are only going to get worse. Planet Earth is needs our help, and needs it badly. With forests becoming barren of vegetation and snow capped poles quickly turning brown, the technology industry has a new attitude towards suddenly becoming "green". I'll spare you the powerful marketing hype that I get from various manufacturers every day, and get right to the point: your computer hasn't been doing much to help save energy... at least up until now.
To measure isolated video card power consumption, Benchmark Reviews uses the Kill-A-Watt EZ (model P4460) power meter made by P3 International. A baseline test is taken without a video card installed inside our computer system, which is allowed to boot into Windows and rest idle at the login screen before power consumption is recorded. Once the baseline reading has been taken, the graphics card is installed and the system is again booted into Windows and left idle at the login screen. Our final loaded power consumption reading is taken with the video card running a stress test using FurMark. Below is a chart with the isolated video card power consumption (not system total) displayed in Watts for each specified test product:
VGA Product Description(sorted by combined total power) |
Idle Power |
Loaded Power |
|---|---|---|
NVIDIA GeForce GTX 480 SLI Set |
82 W |
655 W |
NVIDIA GeForce GTX 590 Reference Design |
53 W |
396 W |
ATI Radeon HD 4870 X2 Reference Design |
100 W |
320 W |
AMD Radeon HD 6990 Reference Design |
46 W |
350 W |
NVIDIA GeForce GTX 295 Reference Design |
74 W |
302 W |
ASUS GeForce GTX 480 Reference Design |
39 W |
315 W |
ATI Radeon HD 5970 Reference Design |
48 W |
299 W |
NVIDIA GeForce GTX 690 Reference Design |
25 W |
321 W |
ATI Radeon HD 4850 CrossFireX Set |
123 W |
210 W |
ATI Radeon HD 4890 Reference Design |
65 W |
268 W |
AMD Radeon HD 7970 Reference Design |
21 W |
311 W |
NVIDIA GeForce GTX 470 Reference Design |
42 W |
278 W |
NVIDIA GeForce GTX 580 Reference Design |
31 W |
246 W |
NVIDIA GeForce GTX 570 Reference Design |
31 W |
241 W |
ATI Radeon HD 5870 Reference Design |
25 W |
240 W |
ATI Radeon HD 6970 Reference Design |
24 W |
233 W |
NVIDIA GeForce GTX 465 Reference Design |
36 W |
219 W |
NVIDIA GeForce GTX 680 Reference Design |
14 W |
243 W |
Sapphire Radeon HD 4850 X2 11139-00-40R |
73 W |
180 W |
NVIDIA GeForce 9800 GX2 Reference Design |
85 W |
186 W |
NVIDIA GeForce GTX 780 Reference Design |
10 W |
275 W |
NVIDIA GeForce GTX 770 Reference Design |
9 W |
256 W |
NVIDIA GeForce GTX 280 Reference Design |
35 W |
225 W |
NVIDIA GeForce GTX 260 (216) Reference Design |
42 W |
203 W |
ATI Radeon HD 4870 Reference Design |
58 W |
166 W |
NVIDIA GeForce GTX 560 Ti Reference Design |
17 W |
199 W |
NVIDIA GeForce GTX 460 Reference Design |
18 W |
167 W |
AMD Radeon HD 6870 Reference Design |
20 W |
162 W |
NVIDIA GeForce GTX 670 Reference Design |
14 W |
167 W |
ATI Radeon HD 5850 Reference Design |
24 W |
157 W |
NVIDIA GeForce GTX 650 Ti BOOST Reference Design |
8 W |
164 W |
AMD Radeon HD 6850 Reference Design |
20 W |
139 W |
NVIDIA GeForce 8800 GT Reference Design |
31 W |
133 W |
ATI Radeon HD 4770 RV740 GDDR5 Reference Design |
37 W |
120 W |
ATI Radeon HD 5770 Reference Design |
16 W |
122 W |
NVIDIA GeForce GTS 450 Reference Design |
22 W |
115 W |
NVIDIA GeForce GTX 650 Ti Reference Design |
12 W |
112 W |
ATI Radeon HD 4670 Reference Design |
9 W |
70 W |
The Sapphire Radeon HD 4870 X2 is clearly no stranger to high power bills based on the fact that it tops our chart for power consumption under load. Even at idle, the Sapphire Radeon HD 4870 X2 video card 100251SR gulps the watts down at a faster pace than the GeForce 9800 GX2 and single 4870 under maximum power conditions. Regardless of consumption, the power requirements for the Radeon HD 4870 X2 consist of only two six-pin PCI-Express power connections to ensure that the twin RV770 receives enough juice to push out the frames in 3D mode. This may leave some middle-market enthusiasts and lower-end gamers in search of a new power supply force feed the Radeon HD 4870 X2 the power it needs.
Most enthusiasts make the mistake of associating a smaller die process with an improved power efficiency. Clearly, the downside to the 55 nm RV770 GPU is it's lack of energy efficient operation. Putting two together on the same PCB doesn't double the consumption of a single Radeon HD 4870, but it sure does try. The power consumption measured under full load doesn't match the performance, but it certainly matched heat output. The idle power draw is extremely high, which is uncommon since emphasis is usually placed on idle/standby mode efficiency and conservation.
Taken as a whole the idle stand-by power consumption is pretty unforgivable, especially since this the condition your equipment will be in the majority of the time. While loaded power consumption is the highest we've ever seen, the price paid to your utility company for gaming would be about the same with just about any other video card. Once upon a time, the computer and gaming consoles seemed like an inexpensive alternative to arcade gaming... but that was before energy costs soared through the roof.
Radeon 4870 X2 Heat Output
Normally I would go into great detail and illustrate where a video card heats up the integrated components with use of a non-contact IR thermometer. However Sapphire kept a short leash on the Radeon 4870 X2 they loaned me, and it was back on sent on its way to the next editor only two days after first receiving it. No worries... I did what I could.
The ambient room temperature was steady at exactly 19.0°C and the inner-case temperature hovered around 32°C. To begin my testing, I used ATITool v0.26 to record GPU temperatures at idle and again at high-power 3D mode.
At idle, the Sapphire Radeon HD 4870 X2 recorded an extremely warm 76°C. This temperature in and of itself was enough to raise concern, but then again I hadn't even begun to use 3D mode which asks for another 220W. Using a full 3D load for roughly twenty minutes, the Radeon HD 4870 X2 didn't waste any time heating up the room. At full load, the twin RV770's produced an operating temperature of 92°C; by far the hottest I have seen any graphics product get.
The GPU nearest the blower fan recorded a low 65°C under load as measured from the backside of the PCB, while the GPU nearest the exhaust rated 78°C. The exhaust vent registered 63°C, while the pin-shaped heatsink bar along the side reached up to 86°C. Hopefully the aftermarket cooling industry won't waste any time designing a better cooling solution for the Radeon HD 4870 X2, because ATI could use some help.
Please continue to the review conclusion in the next section, where I share my final thoughts on the RV770 graphics processor and give my opinion of the new Radeon HD 4800-series product offerings.
Radeon 4800-Series Final Thoughts
There's a lot to like about the new Radeon HD 4800 series of products. To begin with, 800 cores is nothing to scoff at; even if they aren't nearly as efficient as they sound. CrossFireX scales performance very well, and for the first time actually makes multi-card setups worth the money. The most likable part is pricing: As of April 2009 the Radeon HD 4850 currently sells for $99 after rebate, which forces NVIDIA to drop the price of their GeForce GTS 250 (re-labeled 9800 GTX+) to meet with the competition. When ATI launched the Radeon HD 4870 at $300, NVIDIA had to answer back by dropping the GTX 260 to a more affordable price. The Radeon HD 4870 now sells for as low as $134.99 after rebate, making it difficult for the GTX 285. But that's where everything becomes unclear, and the value of ATI's latest product comes into question.
Here's why clarifying how the value has become so tricky is difficult to define. The initial interest in a product like the Radeon HD 4850 lies in the fact that it competes head-on with the GeForce GTS 250 (9800 GTX). But now that they are both priced roughly the same, value takes on a new dimension. In each and every test I conducted, the Radeon HD 4850 kept up with the GeForce 9800 GTX but never outperformed it (until 4x AA was added to Crysis). This would be the main reason why I see value becoming more of an issue outside of video game performance. The GeForce 9800 GTX+ offers HDMI, and so does the Radeon HD 4850. Both offer essentially the same exact sub-features down the line, except for when it comes to multi-card configurations; which is where the CrossFireX configuration really comes to shine.
I am very much aware that NVIDIA offers SLI just like ATI offers CrossFireX, but what I'm talking about is multi-card compatibility with motherboards. AMD Didn't exactly impress the world with Phenom, and thus the world hasn't jumped onboard to use their processors. Instead, Intel scooped up a large share of the consumer base with their P35/X38/P45/X48 chipsets (all launched within about ten minutes from each other). But here's my point: ATI still wins. All of these Intel motherboards, along with all of the AMD motherboards, offer CrossFire support exclusively. NVIDIA is left holding their own hand, because only select few Intel X58-based motherboards are expected to combine AMD's CrossFire technology with NVIDIA's SLI.
I'm not entirely sold on everything that the chipmakers would like for us to believe. I think it's sometimes worth questioning the wisdom, and in this regard I find that AMD is trying to pull one over on consumers by describing their RV770 to have 800 scalar processors. The reality is that ATI's 800 stream processor cores do not compare 1:1 against the competition, especially since the GeForce 9800 GTX can outperform the Radeon HF 4850 with only 128 shader cores. So despite what ATI would like to market, scalar processors they are not as they function exactly like vector processors would. With 800 processor cores residing in five bank location, each series of 5 processor cores process only one vector unit at a time - even if that vector doesn't need to use all five processor cores. 800 Cores are there, but they are far from the efficiency level seen by the competition.
My final thoughts on the 4800-series is where the Radeon 4850 and 4870 come up short. They're both great products just so that we're clear, but for a 55 nm process there's a lot missing from the RV770's arsenal that really should be there. I consider efficiency at the very start of this list, and even though my first lesson on the relationship between die process size and energy efficiency came from Mr. Jen-Hsun Huang, President of NVIDIA, I later researched this through my own testing and discoved that he was correct: reduced die process does not equal increased energy efficiency. Obviously this phenomenon holds up very well against the power consumption results I've tested for this review.
ATI's memory bus architecture also has me questioning their efforts. NVIDIA can produce a 512-bit memory bus making a 1:1 ratio of memory to interface (512MB @ 512-bit), so it's just a little disappointing that ATI did not do the same for their Radeon HD 4870/4890 (which both use GDDR5). Perhaps if I lower my expectations on improvements towards technology, I could accept a 256-bit Radeon HD 4850, but if you're going to make GDDR5 your marketing headline then perhaps you should also do something to match the technical achievements found in competing products. Notwithstanding, memory bandwidth is far from being saturated by today's gaming software (and hardware interface limits), and so my complaint is really more of a moot point, but it still stands to reason that AMD missed an opportunity here.
Radeon 4870 X2 Conclusion
Presentation is always the first summary I give in each product review conclusion, but ironically this mundane topic seldom provides the same outcome. Sapphire, who I have pointed out once already is at the top of the sex-sells marketing game for the computer industry, offers us a full spectrum of tastes... er, personalities... for their product characters. Their latest 4870 X2 series is not really so different, but I don't think they'll be showing off full-fledged lingerie models on their product packaging anytime soon. (I double-dare them.)
Judging the product appearance is a little more difficult than I'm used to. Lately, almost everything has been encased in plastic housings with a contoured finish. This doesn't mean that you can't look sharp without the extra plastic body work, but there's a level of protection all of that stuff provides which goes missing on ATI products. Still, Sapphire has done a very good job mixing red hues with carbon fiber decals, making the Radeon HD 4870 X2 look as fast as it actually performs (unlike most Honda's with the same decals). It's still not entirely clear to me why they added the pin-shaped sidebar heatsink, but combined the entire 100251SR SKU looks sleek.
Construction is solid, but not without some concerns. I absolutely appreciate ATI for protecting components on the back side of the PCB, but at the same time I have to hold both ATI and Sapphire accountable for the cooling unit. If you've read any or my video card reviews, you'll know how much I dislike VGA coolers that exhaust back into the computer case. So it's good news for ATI that they pleased me in this regard on their 4870 X2. Yet even despite their best efforts, the cooling performance is dangerously low. Granted the RV770 can reach 105°C+ before failure (and our sample climbed to "only" 92°C in a cool room), it's too bad that most nearby components will melt by this point.
In terms of performance, the Sapphire Radeon HD 4870 X2 100251SR is the new king of the castle. Throughout our benchmark testing, the Sapphire Radeon HD 4870 X2 performed so far ahead of any other video card that it redefined the high-end of graphics performance. Based on our benchmark tests using Crysis (without AA), The 4870 X2 usally performed around 32% better than the single-GPU Sapphire Radeon HD 4870 and kept only a couple FPS away from the GeForce GTX 280. but when 4x AA was added into the post-processing effects performance favored the competitors new product line. As a graphics processor, the combined RV770's performed incredibly well; but as a hardware component I felt that there was a margin of performance lacking from the cooling solution incorporated on the 4870 X2. With the temperatures under load reaching 92°C, the heat is generated during work as well as rest.
Value is always relative to supply and demand, along with whatever retailers think they can get from the consumer audience at any particular time. As of December 2008, the Sapphire Radeon HD 4870 X2 100251SR is now being sold at NewEgg for $489.99 with an additional $20 mail-in rebate for a short time longer. I've found a few other sources selling around the same price, so it's a good idea to do some price shopping before you buy. Even though the GeForce GTX 280 is priced $100 less, I think the Radeon HD 4870 X2 should see a relative price reduction after the initial glow of a fresh product launch fades away.
In summary, it's my final recommendation that hardcore gamers give the Sapphire Radeon HD 4870 X2 100251SR very serious consideration. If your budget can't afford this top-shelf graphic solution, keep in mind that two Radeon HD 4850's placed into a CrossFireX set often times performed very close to this standalone product. This situation makes a very good argument for multi-card graphics solutions over multi-GPU, and with this taken into consideration you have to look at price and sub-feature offerings. It's no surprise that CrossFire and CrossFireX are available on just about every single motherboard that fits an Intel or AMD processor; the same is hardly true for the SLI. So then it's down to price. Both the Radeon 4850 and 4870 are now offered at very affordable prices, and if shopped for can yield more performance for less cash than the purchase of a single multi-GPU Radeon 4870 X2. Although the drivers I tested with were beta Catalyst 8.8 and couldn't remain completely stable for every test I attempted, they should be user-friendly by the time you're reading this review. The Sapphire Radeon HD 4870 X2 is an outstanding product, and the combined RV770 GPU's raise the bar for graphics performance yet again.
Pros:
+ Very good AA/AF performance for higher-end games
+ Supports DirectX 10 and Shader Model 4.1
+ Twin 750 MHz RV770 GPUs / 2 GB of 900 MHz GDDR5 vRAM
+ Features ATI AVIVO Technology
+ 1080p HDMI Audio and Video supported for HDCP output
+ Very quiet fan under normal operation
+ Introduces new TeraScale Graphics Engine technology
+ Externally-exhausting cooling solution
+ Supports CrossFire functionality
+ 24x Custom filter anti-aliasing (CFAA)
+ 5 GBps PCI Express 2.0 graphics interface
+ Kit includes CrossFireX bridge component
Cons:
- Expensive top-performance product
- Maximum post-processing Anti Aliasing is limited to 8x
- Consumes more power than any other product tested
- Fragile electronic components are exposed on PCB
- Extremely high heat output - very hot under full load
Ratings:
-
Presentation: 9.50
-
Appearance: 9.00
-
Construction: 9.50
-
Functionality: 9.75
-
Value: 7.75
Final Score: 9.1 out of 10.
Excellence Achievement: Benchmark Reviews Golden Tachometer Award.
Questions? Comments? Benchmark Reviews really wants your feedback. We invite you to leave your remarks in our Discussion Forum.
Related Articles:
- HIS HD7950 IceQ Turbo 3GB Video Card
- Diamond Viper ATI Radeon HD 3870 512MB Video Card
- NVIDIA GeForce GTX 670 Video Card Tests
- ASUS ENGTX480/2DI/1536MD5 GeForce GTX 480
- Zotac GeForce GTX-470 Fermi Video Card
- Sapphire ATI Radeon HD 5870 Video Card 21161-00-50R
- XFX Radeon HD5770 Video Card HD-577A-ZN
- MSI N275GTX Twin Frozr OC Video Card
- NVIDIA GeForce GTX 580 Video Card Performance
- ASUS GeForce GTX 660Ti DirectCU-II TOP
