The new Tech Champion: HD5870
The new Radeon HD 5000 series should deliver around 2x more performances thsan previous generation Radeon 4000 cards, and brings DirectX 11 support to desktops for the first time.
It does exactly that.
With months ahead of nVidia’s answer to the 5k series, ATI is looking to be the first to market with two high-end DX11 cards. They are effectively first with a whole series of working DX11 cards and the competition doesn’t have anything to respond with.
One first good thing is detectable when looking at both 5870 and 5850: unlike the 4850 in the 4k series, the 5850 had GDDR5 memory and is based on the same exact architecture as the HD 5870.
To see what we deal with, let's compare: while the Radeon HD 4870 featured 800 SPUs along with 40 TAUs (Texture Address Units) and 16 ROPs (Rasterization Operator Units), the new and improved 5870 kind of has doubled everything -- 1600 SPUs, 80 TAUs and 32 ROPs. This has warped compute power from 1.2 TFLOPs to an incredible 2.72 TFLOPs out of a single GPU.
The core clock speed has also been increased from 750MHz to 850MHz compared to the Radeon HD 4870.
The Radeon HD 5870 utilizes GDDR5 memory, which is now clocked at 1200MHz, allowing for a memory bandwidth of 153.6GB/s, or 33% more than the 4870.
From a market point of view, the HD 5870 is made to replace the expensive-to-build and somewhat inefficient dual GPU HD 4870X2. But it's not just that. There are plenty of brand new features that make it a worthy choice.
To start with, we now see the two 'Cypress' series cards (HD 5870 and HD 5850) followed before the new year by the dual GPU Hemlock card which will make use of two 'Cypress' processors.
After that, we'll probably see some mid-range cards and new 32nm tech testing.
The main thing that we see on the cards is of course DX11. DX11 will soon be used in games like Aliens vs. Predator, STALKER: Call of Pripyat and Dirt 2 and a lot more afterwards.
Now we should talk about some Features.
DX11 is focused on reparing where DX10 phucked-up, making DX a worthy API again and removing some of the developer's hear eacs.
So... what we have in DX11:
The possibility to speed up other applications which in the past have not been accelerated using DX runtime. This happens in W7 and a few games devs already started to expand on.
DirectX Compute brings us back the DX. In Vista, there were quite a few problems with the 'indirect' approach of DX. Now things come back to normal, and DX can be called Direct again.
DirectX Compute can increase efficiency of physics processing and can think for the game's A.I. instead of the CPU among other things.
Compute Shader will now be used with Shader Model 5.0 to bring new graphical features such as ray tracing, advanced post-processing, order independent transparency, depth of field and others.
DX11 also brings us 64-bit floating precision (also called double precision), tessellation (the new more efficient version)
OpenCL, the future
So far there have been attempts to harness a GPU's power by engines such as nVidia’s Compute Unified Device Architecture (CUDA) and ATI’s Stream SDK (which in v2.0 supports OpenCL).
They are both limited by their architecture and by the laziness of the programmers.
A language that was universal and available across multiple platforms became imperative.
That’s were the completely open-source OpenCL (Open Computing Language) spawned.
OpenCL suppose to make possible the execution of code across multiple hardware resources such as GPUs and CPUs.
This makes it possible to prioritize workloads and enables developers to write universal cross-hardware code, which it's a good thing because that code will work on anything and will be cheat-proof (no more paying for specific optimizations for a specific architecture).
Now AMD took things even a little further. By combining Havok with Bullet Physics (open-source physics engine) and the PixeLux (with thier Digital Molecular Matter) engine, AMD is truly moving towards an open platform for developers.
Another new thing on the ATi 5k series cards is Eyefinity.
Eyefinity will give users the ability to use up to 6 monitors all running from the same graphics card. This is indeed something for games, for anyone that wants to enjoy HD movies and even for developers (6 monitors should be enough space for all the coding, debugging and testing software one would use for a game).
Over the time, the quality of anisotropic filtering had become pretty good. Now it's the best there is. The 5870 has Angle-Independent Anisotropic Filtering. This is the true way of performing AF.
AMD said that there is no performance hit with the new AF method compared to the old one, so everything should be fine. We just got free uber-AF.
5870 has re-implemented SSAA as a full screen anti-aliasing mode. Now gamers can once again access the higher quality anti-aliasing offered by a pure SSAA mode, instead of being limited to the best of what MSAA + AAA could do. This favors all pre-DX10 games. Good thing they thought about the past too.
Tests and benchmarks
Although there's no true driver for this card yet, I've tested it in quite some recent and/or very known titles.
As test setup, I've used a Q6600 CPU overclocked at 3.6GHz with 4x1 DDR2 A-Data memory running on a quad-RAID 0 out of 4 WD 1TB Black drives.
I took the Radeon 4890 as reference score (100%).
Test were conducted at the highest possible settings, because only there we see the real difference.
The considered FPS for the score was the minimum and the average. All tests used 2560*1600 resolution.
So here we go with the tests:
Call of Duty World at War, 4xAA, 16xAF
Company of Heroes, Opposing Fronts
S.T.A.L.K.E.R Clear Sky, NoAA, NoAF
Call of Juarez: Bound in Blood, Maxed
Enemy Territory: Quake Wors, 8xAA, 16xAF
Farcry 2, DX9, HDR On, 4xAA
Farcry 2, DX10, HDR On, 4xAA
Crysis Warhead DX9
Crysis Warhead DX10, Maxed
Dawn of War 2, Ultra
Fallout 3, Ultra, High-Res Texture Pack
HAWX, DX10, SSAO Very High, 4xAA
Resident Evil 5, DX10, Maxed
Wolfenstein, 8xAA, 8xAF, Maxed
Batman: Arkham Asylum, NoAA, Maxed
Out of 15 tets, the 5870 wins at 9 vs. the dual-GPU card GTX295.
This is very good, considering that the driver for the 5870 in not by far done.
In other then games scenarios, where the true processing power of the card can be seen, in nVidia's Ocean Demo, the 5870 scores almost double then 285 and 295 and in corssfire mode it's even 3 times faster.
As for heat & power, the cards only drained around 27W in idle mode. This is a great achivement. It's also very quiet.
Overall, the Radeon HD 5870 has proven to be a real winner. The only drawback would temporarely be the relatively higher then expected price.
Since nV is having big trouble with it's fab process of the GT300, just as I have forseen, this time it can be said that the Radeon cards have won the round.
Happy shopping !
Respect for AMD/ATi
This card is indeed worthy on all chapters. If any video card cand be called a tech marvel, this is the one.