Lets keep things as relevant as possible shall we? We aren't talking about the GTX 1070 or 1080, but 480 against 1060. So, keeping things as relevant as possible using you're logic people can set the power target to its maximum on a 1060 GPU and wow, look at that, it'll consume more power where it has the TDP room to maintain higher clocks, just like the AMD 480 GPU will if you remove its power target. To that end have you ever considered why the 480 uses more power? It simply has more going on under its hood (read: more future proof architecture), but we'll get on to that in a moment.
Again keeping things as relevant as possible you saying the 480 is "weak" because it has more bandwidth is utter nonsense. Sorry but you clearly gone and pulled that out your arse. The 480 not only has more "physical" bandwidth than a 1060 because you know, its just on a wider, more potent bus, it also makes use of delta colour compression just like a 1060 does to further enhance its actual bandwidth. Further, each Polaris geometry engine has a index cache for what AMD calls "small instanced geometry." The cache supposedly reduces the need to move data around, reducing internal bandwidth requirements and improving throughput. Pure and simple, this gives the 480 a advantage. In case you want to dispute this (I somehow get the feeling you will...)
Proof. Check the bandwidth test, the 480 beats out the GTX 980 (sorry, no actual 480 vs 1060 I could find, but supposedly the 1060 is more or less 980 performance so close as possible match) with random textures and is within spitting distance with black textures. As a minor side note some people may also like to look at the polygon throughput test in that link, and the ALU tests, all of which show very potent numbers. Particularly the ALU tests where the 480 beats the 980 in every single one of them by very comfortable margins (you can also extrapolate these results as wins or on par with potential 1060 results). No I'm not conveniently ignoring the texture fillrate test either, just chalking it up to a architectural bottleneck, or hell, even a driver issue as results are very similar to that of a 380X which has less texture units.
Finally, back to that more going on under the 480s hood, just give
THIS a read.
So bottom line, no, the 480 is not a weak GPU and is the more future proof with stronger DX12 and Vulkan showings than the 1060. To be clear, I'm not saying the 1060 is a bad card, its just not the smart choice with its lack of Async Compute performance, very uninspiring DX12 performance, and somewhat poor Vulkan performance. GTX 1060 is for "now", RX 480 is for now and the future, ie; as I already stated, those who don't plan to upgrade their GPU for 3+ years.
Trying to compare apples to apples much? Was the Athlon64 a worse CPU because people couldn't get bananas overclocks out of it like people could do with the Pentium 4? Hell no, you could get a monster OC on a P4 and it still sucked. To a lesser extent the same is true of the 480 it (currently) doesn't clock as high as a 1060 but it doesn't need to it basically gets the same amount done @ 1.3GHz as the GTX 1060 does @ 1.6 - 1.7GHz. Needing more speed to do the same amount of work does not mean a better GPU, it means the absolute opposite, which to spell it out means the 1060 has crap IPC performance. This should not be any surprise to anyone nvidia themselves have said they took away some IPC in a bid to maximise clock frequency.
Bookmarks