ASUS Radeon R9-290X DirectCU II OC review - A New Island & Technology

The Radeon R9 290 and 290X

Yes, you guys have been waiting for it for a long time now. AMD is finally releasing the R9-290 series, the aim for these GPUs is to compete with NVIDIA's GeForce GTX 780 and Titan. In the initial phase there will be two 290 cards available, the R9-290 and the 290X. The first being a 2560 Shader processor encounting product, the second one has a whopping 2816 shader processors. Let's have a peek at a quick overview:

The Hawaii processor has a whopping 6.2 Billion transistors all fitted on a 438 mm² Die size. At 28nm AMD applied the GCN architecture again, albeit you will learn that there have been a few tweaks at cache levels as the L2 cache is now 1MB for example. More on that later though. The strongest SKU will be the 290X which packs 2816 shader processors, followed by the 280 with 2560 Shader processors. 

An interesting step from AMD is the move towards a 512-bit memory bus. Nobody really expected it as in the past AMD merely once used it on one of their products. That was the ATI Radeon HD 2900 XT all the way back in 2007, the one with the ring-bus design. They went back to 256 and 384-bit busses because the 512-bit bus simply was complex and using a lot of wires/traces on the PCB. It kind of does make sense right now to make the move towards 512-bit. The GDDR5 being used is cheaper then newer memory formats, and the cheaper 5.0 Gbps in fact saves on overall costs to adapt towards 4 GB graphics memory. AMD makes use of eight 64-bit memory controllers accumulating towards 512-bit. Combined with the graphics memory clocking in at 5.0 Gbps this will bring a 320 GB/sec memory bandwidth towards the graphics card. We'll look at the GCN architecture on a separate page though.

PowerTune 

Starting with Hawaii AMD has now also fully implemented a 'Turbo' feature much like NVIDIA's boost technology on all of their products.

This means that there is no longer a fixed clock on these cards. The GPU clock will get a baseline and maximum frequency, the card will clock up/down in-between these two values based on power draw/limiters, performance and heat. In return the card will adaptive manage fan control, clock frequency and voltages (dynamic). AMD always has been much more reserved with the difference in between baseline and maximum frequency, so the difference in values won't be extremely big.

28nm Technology

The GPU architecture has remained the same. Comparable to the last-generation products, AMD still uses the 28nm process technology, the cards are PCIe gen 3 compatible and there have been significant changes on power consumption. We'll address all the features separately of course. With the launch of the R7 and R9 series you will also see Eyefinity updated towards version 2.0 DDM; audio is now fully supported (you hear audio on the actual monitor it's played off), the 5x1 landscape mode is introduced, and you may now create custom multi-monitor resolutions.
 

Interesting is that AMD now implemented next generation digital multipoint audio with smart channel splitting. So if you setup Eyefinity with three screens the six speakers in your monitors will now be configured as front left/mid/right channels. Very clever.
 

Eyefinity has also received a small update. You can now mix and match any monitor (similar monitors) output and create an Eyefinity setup. So if you have three of the same monitors you could hook up one to HDMI, one to DVI and one to DP. That is the kind of flexibility we like a lot alright.