While the HD79XX series and HD78XX series of GPU's have different core names, the two cores are very, very similar in design. Similar to the way NVIDIA graphics card compute, the new AMD GCN archetecture has tiny 'cores within a core' - stream processors with each processor working on an instruction set until that work has been completed. This is the make up of AMD's GCN (Graphics Core Next) archetecture which is based on the same 28nm process as the Tahiti cores. There are 20 of these compute units on-board the HD7870 core equalling 1280 stream processors, 160 more than the HD6870 but quite a lot less than the 1536 the HD6970 possessed. Comparing the the HD7870 against the previous gen though is pointless thanks to the new GCN archetecture as even the mighty HD7950 had few more at 1752 and comparing the HD7950 against the HD6970 is like chalk and cheese.



Using the new Pitcairn core, the HD7870 comprises of 20 Graphics Core Next Compute Units (1280 stream processors) which is slightly trimmed down from the higher end 7900 series which has 32 CU's. With 32 ROPs and a 256-bit GDDR5 memory interface (equalling 153.6 GB/s of memory bandwidth), the specifications are twice that of the HD7770 GPU we reviewed earlier. The card also features a 9th generation tessellator which equates to the HD7870 being 425% faster than the HD6970 so programmes that require tessellation should so a big improvement over last generation GPU's.



AMD suggests that the HD7870's performance is expected to sit between NVIDIA's GTX580 and GTX570 cards which makes sense as this where the GPU's price point lies. As you can see from the slide above, the HD7870 has many advantages over its smaller sister card, the HD7850, namely the extra clockspeed and 16 TU's which enhances the remainder of the specifications by a fair margin. Both cards have the same 2GB of GDDR5 and 256-bit memory interface.

Let's take a look at the features of today's review sample...