Sapphire Radeon HD 5870 Video Card Review :: Speeds & Feeds

• Engine clock speed: 850 MHz
• Processing power (single precision): 2.72 TeraFLOPS
• Processing power (double precision): 544 GigaFLOPS
• Polygon throughput: 850M polygons/sec
• Data fetch rate (32-bit): 272 billion fetches/sec
• Texel fill rate (bilinear filtered): 68 Gigatexels/sec
• Pixel fill rate: 27.2 Gigapixels/sec
• Anti-aliased pixel fill rate: 108.8 Gigasamples/sec
• Memory clock speed: 1.2 GHz
• Memory data rate: 4.8 Gbps
• Memory bandwidth: 153.6 GB/sec
• Maximum board power: 188 Watts
• Idle board power: 27 Watts

The Sapphire Radeon HD 5870 card is based upon ATI’s RV870 chip codenamed Cypress. This is a high-performance part on TSMC’s 40nanometer process with 2.15 billion transistors. It is in fact bigger in terms of transistor count than any previous graphics chip with the previous record held by NVIDIA’s GT200 chip which had 1.4 Billion transistors. In terms of die size the RV870 has a size of about 334mm2 meaning it is much smaller than the GT200 which had a die size of 576mm2 with 700 million less transistors.

ATI introduced the first 40nm video card in the form of the RV770 or HD4770 earlier this year. That card was a mid-range card but this is the first chip that is considered high end to be built on this process. The lower the process number the higher the number of transistors that can fit into that space and the lower the power consumption for the chip compared to the larger process nodes. In the case of the Cypress chip, the maximum power draw is less than 188W which is a little higher than the HD 4890’s. Idle power consumption has dropped to 27W due to the low-power mode of the GDDR5 memory.

The big push with video cards from NVIDIA and ATI is doing computing tasks on the graphics chip instead of the CPU. NVIDIA is pushing CUDA and ATI calls their version Stream Processing. Stream Processing allows the graphics card to speed up regular tasks such as transcoding and encoding videos

Eyefinity is ATI’s newest multi-monitor technology. Previous ATI cards had the ability to drive two displays simultaneously. ATI’s Evergreen family is their first to natively support up to six monitors on a single GPU assuming there are enough ports on the back panel and enough monitors. ATI showed off Eyefinity at their launch event with 24 monitors being driven by four HD 5870 video cards with six Display Ports on each card. Sapphire’s card supports up to three monitors with its outputs with three monitors up to 7680x4800 resolutions should be supported.

The HD 5870 has doubled the number of Stream Processors when compared to the HD 4890 that previously occupied this space on the ATI lineup to 1600 up from 800. The computing power of this graphics card has risen to 2.72 TeraFLOPs from the 1.2TFLOPS of the HD 4870 card. The TMUs have increased to 80 TMUs; double that of the HD 4890/HD 4870 series. The number of outputted pixels (ROPs) has also doubled to 32 from 16. Here’s a chart summarizing the chips including the HD 5850 which will be released next week, the HD 4850 which it replaces, the HD 4870 X2 and the HD 4890:

One of the big improvements on the HD 5870 over the previous HD 4xxx series is in the image quality department. Video cards since the GeForce 2 have supported multi-sample anti-aliasing. Super-Sampling tends to give better image quality due to using a larger resolution and down-sampling versus using a larger number of samples as in multi-sampling. The performance of this new card allows for real-time frame rates with super sampling enabled making the image quality improved over the HD 4870 series. Anisotropic Filtering has also been improved with angle-independent AF now the norm with the HD 5870.

ATI’s Radeon HD 5870 is the first chip to fully support DirectX 11. This new API includes the following upgrades to the current feature set: Shader Model 5.0 DirectCompute 11, Programmable hardware tessellation unit, accelerated multi-threading, HDR texture compression, Order independent transparency. Tessellation allows more detail to be added to characters in games by using tessellation to add geometry to characters. Multi-threading allows more CPU efficiency by spreading the workload among multiple cores on a processor versus the single core usage of most apps today. Compute Shaders allows for optimized post-processing effects, high quality shadow filtering, depth of field, and high definition ambient occlusion. There are quite a few DirectX 11 games in development including: Dirt 2 which is bundled with the Sapphire and most other HD 5870 cards, Crysis 2, Battlefield: Bad Company 2, Alien Vs Predator, and F1 2010 which are due sometime in 2010 except for Dirt 2 which is scheduled to be released in December.