RX 480 4gb vs RX 480 8gb vs GTX 1060 3gb vs GTX 1060 6gb

I would get the 480 4gb, the 8gb isn't that much better and really only viable for Crossfire setups where you need the extra VRAM (since it is not shared).

The reference cooler for the RX 480 is not great, I'd probably go for the 480 4GB. The 1060 6GB isn't bad either, but a $40 difference is IMO generally not worth it.

Well, not really. But gaming across 2 screens is not a good idea. It will put the centre of the view right at the bezels of the two screens, and that's pretty bad. Unless the monitors have extremely thin bezels maybe... I still wouldn't want to play that way though.

Also remember that currently the RX 480 has a significant advantage in DX12 and Vulcan APIs. If you are taking the long term view and plan to use this card for many years (when DX12 becomes the norm) the 480 will likely serve you better. However if you are concerned about certain games and applications geared towards Nvidia like Fallout and Overwatch I would go for the 1060 6gb.

Nah, those are old benchmarks. Newer benchmarks show the RX 480 matching the GTX 1060 in DX11, and winning handily in DX12 and Vulkan.

http://www.hardwarecanucks.com/forum/hardware-canucks-reviews/73945-gtx-1060-vs-rx-480-updated-review-23.html

So which one of these would I buy? That will likely boil down to whatever is on sale at a given time but I’ll step right into and say the RX 480 8GB. Not only has AMD proven they can match NVIDIA’s much-vaunted driver rollouts but through a successive pattern of key updates have made their card a parallel contender in DX11 and a runaway hit in DX12. That’s hard to argue against.

Click to expand...

As for DX11 vs DX 12, it is ironic in that in TPUs test, the 1060 did better than the 480. One of the three games could not be tested in DX12 cause they couldn't keep it running.

But while the **current** differences in DX12 and Vulcan are acknowledged, also remember ... or search the forums and you can find hundreds of posts saying the same thing about Mantle ...and the same thing about HBM ... and the same thing about DX11. The first two are historical footnotes after they bombed. As for the last, remember that the games in question did not begin their lives as DX12 games.... with a 3 - 5 year development cycle as a minimum, these games were developed in DX11 and essentially "ported" to DX12. For example, TPU reported that when they tested Hitman's they wrote " The game's DirectX 12 implementation, however, is too riddled with bugs to be integrated into our test bench for now"

However, looking at TPUs tests.... the 1060 tops the 480 in both DX12 games, the 3rd DX 12 bugged title (Hitman) not included.

Just Cause (480 8GB) = 81.1 / Just Cause (480 OC'd) = 87.3
Just Cause (1060 6 GB) = 83.8 / Just Cause (1060 OC'd) = 98.6

Rise of the Tomb Raider (480 8GB) = 62.7 / Rise of the Tomb Raider (480 OC'd) = 67.5
Rise of the Tomb Raider (1060 6 GB) = 70.8 / Rise of the Tomb Raider (1060 OC'd) = 83.3

Overall the test showed the following:

Over the 16 games, the 1060 had a 5.78% advantage, 15.60% when both cards overclocked
Over the 14 DX11 games, the 1060 had a 5.55% advantage, 15.35% when both cards overclocked
Over the 2 DX12 games, the 1060 had a 7.51% advantage, 17.49% when both cards overclocked

It certainly is important to look at the games you play and how each card does in each game.

The question for the driver teams at each company is ... do we invest time chasing the moving target, or is our time better spent waiting for the platform to stabilize and developing for the stationary target. Again, for better or for worse, ATI / AMD has always been the 1st to jump on new APIs and taken in early lead in adapting the new API performance. Those who made purchase decisions on "mantle performance" being better, are having a bit of buyer's remorse today. With Mantle, we saw game demos which were "called out" by enthusiasts because they were designed not so much around what would bring the best performance but around what could best take advantage of mantle's capabilities "for demonstration purposes".

I guess I am a bit jaded after so many "[Insert new AMD technology here] is going to change everything" announcements... and then being disappointed when it falls flat. I'm not saying it won't matter long term, I am saying we been here before too many times and in the past and soon after the hoopla died and it was forgotten.

I am saying we just don't know yet how this will fall out once games begin development from the getgo on these APIs.... just too early to make a definitive conclusion. So, for me, I am just sitting tight on this issue and waiting to see how it falls out.

Lets agree, for the moment, that they did catch up for the purposes of the discussion .... now add 10% for the difference between the 7.7% extra ya get from OCing the 480 versus the 17.7% you get from the 1060 ... to my eyes, that still leaves it 10% behind and that isn't "caught up".

As for the "for the moment" part ... if ya been at this a while both nVidia and AMD have been caught previously tweaking their drivers to perform better in game demos. Then another site tests the same game using a different sequence in the game and the sudden "catch up" evaporates. I am not saying this is what happened here, but "once bit so warned" as the saying goes. That test is only 4 days old.... so before accepting anything, ya want to wait a bit and see what popes up. And now we have that.

Since the implication is that the above techpowerup graph is no longer invalid, lets look at what techpowerup has to say about this. In July, the 1060 held an 11% lead at TPU. Now lets look at the TPUs (Dec 8th) latest test which uses the same games in the graph plus a few more for a total of 21 games.... 3 times what Canucks used

Here's the graph, which says they picked up 2.1%

Average Performance Gain: +2.1%

Click to expand...

So, let's compare and see whether the chart pictured above invalidates the relative rankings.

1. The original graph shows an 11% advantage for the 1060 GB over the 480 8GB

2. Techpowerup shows a 2.1% pickup in performance since the last driver ... and 4-6% since the 16.7 drivers used in the above graph

https://www.techpowerup.com/reviews/AMD/Radeon_Crimson_ReLive_Drivers/8.html

The new drivers don't particularly offer significant performance gains over previous drivers; however, AMD's 4-6 % performance gain claims over the 16.7 drivers its Polaris 10 hardware launched with do appear to hold true. Performance isn't really the focus of this release, and they seem to have fixed a few game-specific bugs, so this isn't a major area of concern for us.

Click to expand...

So what do we see:

a) The 1060 was ahead by 11% in July when the 480 used the 16.7 drivers

b) The 480 showed a 4-6% improvement over the 16.7 driver since then .. let's call it 5%

c) The data doesn't say how much nVidia drivers may have improved since then, but for the sake of argument, let's call it 0%

d) The difference on OC headroom is a 10% advantage for the 1060

What can we conclude from that ? Let's use the July 480 performance as the base of 100

480 =100 x 1.05 (driver improvement) x 1.077 (OC) = 113.09
The 480 AIB cards overclocked now are 13.09 % faster than the 480 stock card than it was in July

1060 =111 x 1.0 (we don't know driver improvement, using 0%) x 1.177 (OC) = 130.65
The 1060 AIB cards overclocked now are 30.6 % faster than the 480 stock card than it was in July

Relatively speaking .... today the 1060 OC'd is 15% faster then the 480 OC'd in those 14 games.

As I have said before, you can hand pick whatever games you want to try and make any chosen card look better. My point relates to the claim that the relative rankings, using the games in that test, remain valid. The relative % may have changed slightly but the graph above is still valid with respect to the relative position in the rankings. The gap has been narrowed, but even at stock settings 9and ignoring any relatively recent improvements in nVidia drivers) a 5% improvement does not erase an 11 point advantage in those 14 games.

No. And even accepting your bogus numbers, the 480 is still better value for money.