The BARTS graphics processor
The BARTS graphics processor
Now the Radeon HD 6870 X2 is based on two Barts GPUs. The single GPU cards derived from Barts will in fact be segmented into the mid-range, not high-end as many people expected.
Let's have a quick comparative overview of some of the specifications representing a certain cope of mid-range performance products
Specifications | Radeon HD 5770 | Radeon HD 5850 | Radeon HD 6850 | Radeon HD 6870 | R6870 X2 |
GPU | Juniper XT | Cypress Pro | Barts Pro | Barts XT | Barts XT |
Manufact. tech. | 40nm | 40 nm | 40nm | 40nm | 40nm |
GPU frequency | 850 MHz | 725 MHz | 775 MHz | 900 MHz | 900 MHz |
Stream processors | 800 |
1440 | 960 | 1120 | 1120 |
Memory frequency | 4800 MHz | 4000 MHz | 4000 MHz | 4200 MHz | 4200 MHz |
Memory bus |
128-bit | 256-bit | 256-bit | 256-bit | 256-bit |
Memory buffer |
1 GB GDDR5 | 1 GB GDDR5 | 1 GB GDDR5 | 1 GB GDDR5 | 2 GB GDDR5 |
Power consumption |
108 watts | 151 watts | ~ 127 watts | >150 watts | >280 watts |
Performance |
- | - | Faster than 5830 | Faster than 5850 | Faster than 6970 |
So the two R6800 SKUs each have a separate codename, the R6850 being Barts Pro and the 6870 being the Barts XT.
Both cards are of course updated DX11 class products with a couple of new features. Armed with 1.7 billion transistors the Radeon HD 6850 is pitted against the competition's GeForce GTX 460 and GTX 560 models, the 6850 is clocked at 775 MHz on its core and shader processor domain and comes with a full GB of graphics memory. This memory is 256-bit and running at an effective data-rate of 4000 MHz.
The product has 12 SIMD clusters, 48 texture units, 960 shader processors, 32 ROPs and a TDP of 127W, 19W in idle. Despite the memory bus you can recognize similarities close to the R5770 here. This product however will bring you 1.5 TFLOPS of performance, and combined with that 256-bit gDDR5 memory it will leave the R5770 far behind it.
Pitted against the GeForce GTX 560 1GB version is the Radeon HD 6870, and this is going to be the more interesting product for most of Guru3D's readers we suspect. As you noticed, the overall shader count and clock frequencies are higher then earlier leaked rumors, that applies to the R6870 as well.
The 6870 is clocked a good chunk faster at a nice 900 MHz, it also comes with a full GB of graphics memory. This memory is 256-bit and running at an effective data-rate of 4200 MHz with a peak bandwidth of 134 GB/sec. The product has 14 SIMD clusters, 56 texture units, 1120 shader processors, 32 ROPs and a TDP slightly above 150W with a 19W idle TDP. This product will bring you 2.0 TFLOPS of performance. This means that performance wise the R6870 will sit in-between a R5850 and 5870.
Features wise both graphics cards will be very similar to the last generation products and are merely an advanced update, however some new features like DisplayPort which now follows 1.2 interface specification, HD3D, UVD3 and HDMI 1.4a are introduced. We also spot a new Anti-Aliasing mode (Morphological AA), better Anisotropic Filtering and improved Tessellation performance up-to twice the performance of that of the 5000 series.