![[H]ardOCP](/images/blank.gif){kind=small}
- Date:
- Tuesday , September 22, 2009
- Author:
- Brent Justice
- Editor:
- Kyle Bennett
- Google +1
AMD's ATI Radeon HD 5870 Video Card Review
AMD’s ATI Radeon HD 5800 series debuts today and the new ATI Radeon HD 5870 is now available. We will give you all the gritty details of the HD 5800 Series, and show you just how well the new flagship ATI Radeon HD 5870 accelerates games. Can you say, “I like twice as much performance in the same power envelope.”
Radeon HD 5800 Series Architecture
The Radeon HD 5800 series is more than just doubling all the good parts; there have been additions, improvements and refining.
The "TeraScale 2" hardware architecture of the Radeon HD 5800 series is an evolutionary upgrade from the Radeon HD 4800 series. There have been no revolutionary changes to the actual architecture involved here. The revolutionary changes came in the previous generation, known as the TeraScale Architecture. TeraScale 2 simply expands upon this architecture, improving performance and adding features. Looking at the slide above you will see the layout is very similar to the Radeon HD 4800 series, with more of everything.
Arguably, DirectX 11 features and support could be considered "revolutionary" if AMD had not already supported some of DX11’s important features like tessellation in their previous generation video cards already. In fact, this is AMD’s 6th generation tessellation engine. It is expected that NVIDIA’s next generation GPU will be its first to support tessellation as is required in DX11.
You see, AMD has been on the DX11 track for a while now, supporting DX10.1 which shares some DX11 features in their previous GPUs, certainly ahead of the curve in comparison to the green team. In fact, one of DX11’s new features was developed by ATI, Gather4 or (Fetch4 previously) is now a DX11 standard.
The Radeon HD 5870 has 2.7 TeraFLOPS of compute power in single precision and 544 GigaFLOPS in double precision. ATI has added the ability to perform a Co-issue MUL, dependant ADD in a single clock cycle plus SAD improvement, DX11 bit-level ops and fused Multiply-Add. Each thread processor contains four streaming processor units and one special function streaming processor unit, a branch unit and registers.
Along with the addition of thread processors ATI has really beefed up the texture units and caches. There are now 80 texture units and the HD 5870 can perform 68 billion bilinear filtered texels per second and up to 272 billion 32-bit fetches per sec. The L1 texture fetch cache bandwidth has been improved and the L2 cache has been doubled to 128KB per memory controller. There are also a few new DX11 texture features such as 32-bit and 64-bit HDR block compression modes.
As we mentioned, the HD 5870 contains full DX11 tessellation which is now ATI’s 6th generation of the technology. Geometry shading has gotten faster and they have even improved OpenGL quality by supporting 12-bit sub-pixel precision. So if you thought just DX11 was receiving improvements, you’d be wrong, OpenGL and OpenCL are fully supported and with improvements.
The ROPs have also been doubled, which has the inherent effect of doubling performance for antialiasing in 32-bit and 64-bit color, which is what is important for you guys. There have been other improvements though beyond this, ATI found a way to also increase CFAA performance. CFAA uses the stream processors for calculation, and does not tax the memory system, but ATI has found through some tweaks of the hardware they were able to increase raw CFAA performance clock-for-clock in this GPU.
Memory Controller
The Radeon HD 5800 series memory controller has also received some optimizations. ATI has now turned on GDDR5’s Error Detection Code (EDC) similar to ECC in DRAM. This allows GDDR5 modules to reach higher frequencies while reducing errors. This could help make video cards as a whole more stable and robust, especially as you clock memory frequencies higher.
One interesting result of enabling EDC however is that overclocking consequences have now changed. In the past, when you overclocked memory on a video card it would ramp up in frequency until the video card would start throwing out artifacts, and then crash. Now when you overclock memory frequency your performance will increase until it reaches its climax and EDC kicks in fixing errors in the memory due to the frequency being too high. This means EDC will fix these problems and your performance will start to actually degrade as you increase frequency and EDC has to work harder at fixing the errors. There won’t be as much flashing polygons and artifacts anymore and no outright crashing from overclocking memory, you will only see a performance decline. Therefore, the goal is to now overclock memory until performance starts declining, then back it off until you’ve reached the point right before the decline, and that is your maximum stable memory overclock.
AA and AF Improvements
AMD did not sit idly by just improving performance, they have made an addition to AA and quality improvements to AF.
An old feature being brought back to life on the Radeon HD 5800 series is Supersample AA! Yes, full Supersample AA is now selectable in Catalyst Control Center. You will be able to select whether Multisample AA or Supersample AA is used in all your games. This of course demands a LOT from the hardware, but it has the performance to at least allow some level of it in most games, at lower resolutions. We will show you how this looks in games on the next page. It is great to have this option now present.
Anyone remember this little program that showed filtering looking down a tunnel? We used to use this quite a bit back in the day because filtering varied greatly between video cards, and there were noticeable differences in games. In recent years however filtering has been so good that all games have been playable at 16X AF and there have been very few discernable differences between video cards. We thought it couldn’t get any better! 16XAF already looked great on the AMD side of things. Well, it just got better.
ATI has improved texture filtering to the point that it looks as good if not better than the reference rasterizer. The new AF algorithm eliminates the angle dependency of the previous generation, with absolutely no performance hit in doing so compared to the previous generation. We will show some screenshots on the following page to see how this compares to the previous generation and to NVIDIA.
DirectX 11
The Radeon HD 5800 series completely support DX11 in every which way. This GPU was built with this feature set in mind, and AMD wanted to make sure this was released hand-in-hand with Windows 7 launch.
As has been beaten into your brain by now, the ATI Radeon HD 5800 series supports DX11 tessellation in full through the Hull and Domain Shaders. One improvement in DX11 that may affect performance is multi-threading improvements which allow the application, DirectX runtime and DirectX driver to each run in separate threads.
Of course the most important aspect of DirectX 11 and DirectCompute 11 within DX11 is when will it be used in games and will it make a difference? Here is a list of games that will support DX11 in some form or another. The very first one is going to be Battleforge which will receive a patch today that enables support for DX11 and uses DirectCompute 11 to accelerate effects. What we saw of this in the demo wasn’t that impressive, but there are some other games that are using DX11 to greater affect.
The list of games includes: S.T.A.L.K.E.R.: Call of Pripyat, DiRT 2, Lord of the Rings Online, Dungeons and Dragons Online, Aliens vs. Predator, and Genghis Khan. There are three gaming engines that we know of in development that will natively support DX11, CryEngine 3 from Crytek, Frostbite 2 Engine from EA DICE and Vision Engine from Trinigy.
We have seen Eyefinity demos of CryEngine 3 running and they are extremely impressive…and running at more than 5 frames per second.
DiRT 2 for PC, due out this December, has been delayed to add in new DX11 features. DiRT2 will be using DirectCompute 11 to provide hardware tessellation of water, cloth, and the crowd and particle systems in the game. There will also be HDAO shadow filtering and DX11 depth of field. In the second screenshot above you can see how the water itself has geometry and deforms as you drive through it creating waves.
Aliens vs. Predator will also use hardware tessellation. In the image above you can see one of the Aliens up close, and the Alien is so Tessellated (so many triangles) that it looks like it has a texture! This demo we saw was pretty amazing to be honest; the level of detail added up close is amazing compared to no tessellation. We can’t wait to see how this game looks and plays. Performance was right up there in the demo we witnessed; no lag was noticeable in doing this hardware tessellation.
Video
Lest we forget, there are also a few video related improvements to the Radeon HD 5800 series.
UVD 2.0 is back, and it has been enhanced to allow to accelerated decode of two 1080p HD video streams at once. HDMI features have also been improved, supporting HDMI 1.3a. There have even be HDMI audio improvements supporting Dolby TruHD and DTS-HD Master Audio with full support for Blu-ray audio formats and up to 8 channels of 192kHz/24-bit audio. Suffice it to say, the Radeon HD 4800 series was already great at decoding video, and the Radeon HD 5800 series is just that little bit better.