Tweet
On paper, the task facing AMD's engineers must have looked formidable. Having jostled for supremacy with Nvidia's GTX 780 Ti and Titan, the firm was clearly competitive with its Radeon R9 290X - until the arrival of GTX 970 and GTX 980, both of which outperformed the best that the red team had to offer, and did so with bags of overclocking headroom to spare. In producing the Radeon 300 series, AMD had to match or beat Nvidia's excellent performer - and not only that, it had to do it using existing silicon. We had doubts that it would be possible, but as the benchmarks rolled in, the bottom line became clear: AMD has done it.It's a remarkable achievement bearing in mind that all the evidence points towards the R9 390X being little more than an overclocked version of the outgoing 290X. AMD dubs the 300 series version of the chip 'Grenada' but really and truly, a straight head-to-head strongly suggests it is the same as the existing Hawaii, simply with a 50MHz overclock - upped still further to 100MHz here in the MSI Gaming version of the card we're reviewing here. Now, there've been overclocked 290Xs before, but none of them have troubled the GTX 980 - something else must have changed to see the impressive performance we've looking at here.
That would appear to come down to memory bandwidth. The R9 390X features 6000MHz GDDR5, a 20 per cent improvement in terms of throughput over the 290X (and overclocked still further on this MSI board to 6100MHz) - the result of AMD shifting to a much faster, more capable Hynix memory module. And as a bonus, the firm has decided to double the allocation of VRAM up to a mammoth 8GB. Superfluous? Right now, yes. But in a market where many console ports now require 3GB to service 1080p, who knows what might happen in the months and years to come - especially when cards like the R9 390X are built to service more memory hungry resolutions much higher than the standard full HD?
Read more…
More...