ATI RADEON 3850/3870 (RV670)
320 Shader Processors and 256-bit Memory Bus
Installation and Drivers
Testbed configuration:
- Intel Core2 Duo (775 Socket) based computer
- CPU: Intel Core2 Duo Extreme X6800 (2930 MHz) (L2=4096K)
- Motherboard: EVGA nForce 680i SLI on NVIDIA nForce 680i
- RAM: 2 GB DDR2 SDRAM Corsair 1142MHz (CAS (tCL)=5; RAS to CAS delay (tRCD)=5; Row Precharge (tRP)=5; tRAS=15)
- HDD: WD Caviar SE WD1600JD 160GB SATA
- Operating system: Windows XP SP2; DirectX 9.0c
- Operating system: Windows Vista Ultimate; DirectX 10.0
- Monitor: Dell 3007WFP (30").
- Drivers: ATI CATALYST 7.10; NVIDIA Drivers 169.04.
VSync is disabled.
Synthetic tests
In our base articles we still use old versions of RightMark3D as well as the new version (2.0) for Direct3D 10 graphics cards in MS Windows Vista. Some previously known tests were rewritten for DX10, new types of synthetic tests are added: modified tests of pixel shaders rewritten for SM 4.0, tests of geometry shaders, vertex texture fetch tests. Previous versions of RightMark will also be used until low level fill rate and other tests appear in the new version.
All our synthetic benchmarks can be downloaded here:
- D3D RightMark Beta 4 (1050) with its description on http://3d.rightmark.org
- D3D RightMark Pixel Shading 2 and D3D RightMark Pixel Shading 3 - tests of Pixel Shaders 2.0 and 3.0, link.
- RightMark3D 2.0 with a brief description is here
RightMark3D 2.0 requires MS Visual Studio 2005 runtime and the latest DirectX runtime update.
Synthetic tests were run with the following graphics cards:
- RADEON HD 3870 with standard parameters (HD3870)
- RADEON HD 3850 with standard parameters (HD3850)
- RADEON HD 2900 XT with standard parameters (HD2900XT)
- RADEON HD 2600 XT with standard parameters (HD2600XT)
- NVIDIA GeForce 8800 GT with standard parameters (GF8800GT)
- NVIDIA GeForce 8800 GTS with standard parameters (GF8800GTS)
We've chosen these graphics cards to analyze performance of the new products from AMD for the following reasons: we compare the new cards with the RADEON HD 2600 XT as the former top Mid-End solution; the old GeForce 8800 GTS will just make the comparison more interesting; the GeForce 8800 GT is the main competitor to the RADEON HD 3870; and the RADEON HD 2900 XT interests us here, because the R600 was used as a basis for the new Mid-End GPU, only its frequencies and memory bandwidth are different.
Direct3D 9: Pixel Filling tests
This test determines peak texel rate in FFP mode for different numbers of textures applied to a pixel:
Most graphics cards from AMD demonstrate results that are close to their theoretical maximum, just a tad lower. As usual, results of synthetic tests do not reach theoretical values in modes with many textures. Texture fetch rates of the HD 3870 and the HD 3850 match those of the HD 2900 XT adjusted for the frequencies. The graphics card from NVIDIA, notable for high texturing speed owing to its improved TMUs, fails to demonstrate maximum results.
In case of few textures per pixel, the HD 3870 and the 3850 compete with the GeForce 8800 GT, because the latter has insufficient memory bandwidth - it's lower than in the RV670 card. But in heavier conditions, AMD graphics cards slow down compared to the main competitor, proving full identity of texture units and their relation to the R6xx architecture. Let's have a look at the fill rate test:
The second synthetic test measures the fill rate. It shows the same situation adjusted for the number of pixels written into the frame buffer. In case of 0 and 1 textures, the new Mid-End solution from NVIDIA is outperformed by one of the AMD solutions. However, it's much stronger in case of many textures per pixel. As we assumed in the previous article, RV670-based graphics cards are outperformed by G92-based cards in texel and fill rates, when performance is not limited by memory bandwidth. Theoretical difference is even more noticeable.
Your link here
