27 game shoot out: RX 480 vs. 580 vs. GTX 1060 [tpu]

[tviceman]

http://www.techspot.com/review/1393-radeon-rx-580-vs-geforce-gtx-1060/

TL;DR RX 580 comes out 2-3% ahead of the GTX 1060, which comes out 2-3% ahead of the RX 480. 9gbps vram makes very little difference for the GTX 1060, and the RX 580 requires 60% more power draw for it's 2-3% advantage. No overclocking was tested.

As I've personally argued for the past several months, you can't go wrong with either choice between the cards in question. Each has its own strengths and weaknesses, but by and large perform about the same outside of 4-5 outliers.

 

[Tweak155]

Seems like you should just pick the best deal of the 3 at the time of ordering.

 

[guskline]

Steve does a great job in this video. I agree with Tweak155 above about picking the best deal.

 

[raghu78]

yeah a Rx 580 8 GB at USD 230 is probably the card to beat for price/perf.

https://www.newegg.com/Product/Product.aspx?Item=N82E16814131714

The widespread availability of Freesync monitors and at cheaper price points vs Gsync is a key factor in making the Rx 580 a damn attractive GPU at USD 230.

 

[Head1985]

I was expecting more from 1060 9Ghz.I know that on GTX1070 oc memory to 9500mhz increase average performance in every game by 8-9%.

Why Nv do this?Put 9Ghz on 1060 which it seems 1060 dont need at all but left 1070 with slow 8Ghz.9Ghz memory will help alot in 1070 case.

 

[raghu78]

I was expecting more from 1060 9Ghz.I know that on GTX1070 oc memory to 9500mhz increase average performance in every game by 8-9%.

The GTX 1070 with 1920 cuda cores has 50% more cuda cores than GTX 1060 with 1280 cuda cores. But GTX 1070 with 256 GB/s has only 33% more bandwidth vs the GTX 1060 (192 GB/s). So its quite understandable that GTX 1070 gains more from increased bandwidth as it has lesser bandwidth /core in stock config.

 

[Head1985]

The GTX 1070 with 1920 cuda cores has 50% more cuda cores than GTX 1060 with 1280 cuda cores. But GTX 1070 with 256 GB/s has only 33% more bandwidth vs the GTX 1060 (192 GB/s). So its quite understandable that GTX 1070 gains more from increased bandwidth as it has lesser bandwidth /core in stock config.

Yes but performance Gap is only 30% whitch indicate GTX1060 is memory bandwidth bottleneck.It should be close to 50% if GTX1060 is shader bottleneck.

But tell me why Nv didnt put 9Ghz or even 10Ghz GDDR5x memory on 1070?

Btw look at those gaps:
1060 to 1070-30%
1070 to 1080-27%
1080 to 1080TI-27%

Its pretty much equal for every tier.

 

[LTC8K6]

Seems like you should just pick the best deal of the 3 at the time of ordering.

The 60% more power draw of the 580 for such a small performance gain would seem to make the 1060 the best overall.

 

[raghu78]

Yes but performance Gap is only 30% whitch indicate GTX1060 is memory bandwidth bottleneck.It should be close to 50% if GTX1060 is shader bottleneck.

But tell me why Nv didnt put 9Ghz or even 10Ghz GDDR5x memory on 1070?

Btw look at those gaps:
1060 to 1070-30%
1070 to 1080-27%
1080 to 1080TI-27%

Its pretty much equal for every tier.

Performance never scales linearly with added shaders. Even if a GPU doubled every resource you rarely get perfect 100% perf increase. Moreover GTX 1060 has 48 ROPs while GTX 1070 has 64 ROPs. So even though there are 50% more cuda cores in GTX 1070 it has only 33% more ROPs and bandwidth compared to GTX 1060. Different games behave differently as they stress different aspects of the GPU - shaders, ROPs, bandwidth. Moreover the average perf which you mention from tpu is over a suite of 15+ games which use different APIs - DX11,DX12,Vulkan . You are never ever going to get perfect scaling over such a large game suite. If you have doubts compare GTX 1080 Ti vs GTX 1080 and you will see perf increase is not perfectly linear to increased cores, ROPs or bandwidth.

The 60% more power draw of the 580 for such a small performance gain would seem to make the 1060 the best overall.

Overall system power consumption is up 22% (60w). In a desktop environment with quality 500w psu thats not a issue. If you want a mini ITX / HTPC card the GTX 1060 is the only option. So which card you choose depends on the user requirements.

 

[Termie]

While I respect the work Steve puts into his GPU reviews, he really made a big mistake in testing all these cards at their factory overclocked levels and then not listing the specific cards and clocks in the graphs. It basically makes the results meaningless. Putting aside the red vs. green debate, the standard GTX 1060 ran a huge factory overclock (it was an EVGA FTW), while the GTX 1060 9Gbps ran a small factory overclock (it's an MSI Gaming X+), meaning you can't even tell from Steve's results whether the faster memory makes a difference, as the two cards perform the same. Clearly it does, but casual readers will conclude it does not.

And truth be told, we all know the RX 580 is clocked very close to its maximum, while the 1060 9Gbps probably had another 10% headroom, so that too would affect the results, and should have been considered if he was going to bother testing overclocked cards.

 

[TeknoBug]

I'm happy with my RX 580.

 

[nathanddrews]

I can't express how much I love this graph, I wish all reviews ended this way:

Spoiler: Percent Difference

 

[96Firebird]

yeah a Rx 580 8 GB at USD 230 is probably the card to beat for price/perf.

https://www.newegg.com/Product/Product.aspx?Item=N82E16814131714

I dunno, a 1060 6GB at $207 seems like a better price/performance deal, considering the results from the article in the OP...

https://www.walmart.com/ip/EVGA-GeF...26386740870416659407&affillinktype=10&veh=aff

 

[poofyhairguy]

If you want a mini ITX / HTPC card the GTX 1060 is the only option.

I use my blower model 480 in a Mini ITX HTPC just fine. I think the form factor matters more for Mini ITX. I will admit a blower model 1060 would probably be better but they are expensive.

 

[Bacon1]

I can't express how much I love this graph, I wish all reviews ended this way:

Spoiler: Percent Difference

What do most of the bottom ones have in common? GameWorks.

DH2, GoW4, GTAv (this has GW + Amd shadows), WD2, TW3, Mirrors Edge Catalyst, Ghost Recon Wildlands. 7 out of the 9 games where Nv is faster are GW titles.

Interesting to see FCP and Mafia 3 swap places since Far Cry Primal launched much better on AMD and Mafia 3 much better on NV.

 

[LTC8K6]

Performance never scales linearly with added shaders. Even if a GPU doubled every resource you rarely get perfect 100% perf increase. Moreover GTX 1060 has 48 ROPs while GTX 1070 has 64 ROPs. So even though there are 50% more cuda cores in GTX 1070 it has only 33% more ROPs and bandwidth compared to GTX 1060. Different games behave differently as they stress different aspects of the GPU - shaders, ROPs, bandwidth. Moreover the average perf which you mention from tpu is over a suite of 15+ games which use different APIs - DX11,DX12,Vulkan . You are never ever going to get perfect scaling over such a large game suite. If you have doubts compare GTX 1080 Ti vs GTX 1080 and you will see perf increase is not perfectly linear to increased cores, ROPs or bandwidth.



Overall system power consumption is up 22% (60w). In a desktop environment with quality 500w psu thats not a issue. If you want a mini ITX / HTPC card the GTX 1060 is the only option. So which card you choose depends on the user requirements.

Efficiency sells these days. I like to reward it when I can.

 

[Elfear]

Efficiency sells these days. I like to reward it when I can.

Efficiency is nice and, ceteris peribus, I would choose the most efficient card. For a lot of people though price/performance is the #1 priority. 1060's normally tend to run ~$250 where you can get a 480 for ~$200. The 1060 posted above by 96firebird is a great deal but in general the 480 makes more sense from a price/performance perspective.

 

[Guru]

What do most of the bottom ones have in common? GameWorks.

DH2, GoW4, GTAv (this has GW + Amd shadows), WD2, TW3, Mirrors Edge Catalyst, Ghost Recon Wildlands. 7 out of the 9 games where Nv is faster are GW titles.

Interesting to see FCP and Mafia 3 swap places since Far Cry Primal launched much better on AMD and Mafia 3 much better on NV.

That is hardly a logical and reasonable way to look at things. Most of the top games where RX 580 wins are AMD sponsored titles.

I think its right to say that the RX 580 wins in terms of performance over the 1060 6GB overall, but its unreasonable to say the games the 1060 6GB wins is due to them being Nvidia sponsored games. Even if that was the case, you have to post the AMD sponsored games and say the same thing, not be biased like that.

Unfortunately for AMD they have no competition with the 1070, 1080, 1080TI and Titan XPm while the RX 580 just now started winning the overall war. According to computerbase.de and pcgameshardware.de the 1060 6GB was slightly better than the RX 480 8GB, about 2-4%.

Now the RX 580 8GB is likely to come out ahead by 1-2%.

 

[Malogeek]

Looking forward to the entire thing done on the Ryzen platform as well, which was mentioned in the review as in the works.

 

[Bacon1]

That is hardly a logical and reasonable way to look at things. Most of the top games where RX 580 wins are AMD sponsored titles.

Doom, COD, RE 7, Mafia 3, Titanfall 2, The Division, Warhammer DOW 3, aren't sponsored by AMD.

Only TW:WH, DX:MD, BF1 and AOTS are.

So 4/11 are sponsored by AMD vs 7/9 of the bottom ones sponsored by Nvidia.

 

[poofyhairguy]

Interesting to see FCP and Mafia 3 swap places since Far Cry Primal launched much better on AMD and Mafia 3 much better on NV.

I find that fascinating too. Hawaii slayed Maxwell in Far Cry Primal.

 

[Headfoot]

Doom, COD, RE 7, Mafia 3, Titanfall 2, The Division, Warhammer DOW 3, aren't sponsored by AMD.

Only TW:WH, DX:MD, BF1 and AOTS are.

So 4/11 are sponsored by AMD vs 7/9 of the bottom ones sponsored by Nvidia.

#Rekt

 

[Bacon1]

I find that fascinating too. Hawaii slayed Maxwell in Far Cry Primal.

I wonder if it isn't a benchmark vs gameplay thing at least for FCP. I'm not sure if Mafia 3 has a benchmark.

I remember in Digital Foundry's video though the 390 just blew away the 970. Maybe the benchmark is different or maybe the game just performs much better on a 1060 vs the 970.

 

[kawi6rr]

That is hardly a logical and reasonable way to look at things. Most of the top games where RX 580 wins are AMD sponsored titles.

I think its right to say that the RX 580 wins in terms of performance over the 1060 6GB overall, but its unreasonable to say the games the 1060 6GB wins is due to them being Nvidia sponsored games. Even if that was the case, you have to post the AMD sponsored games and say the same thing, not be biased like that.

Unfortunately for AMD they have no competition with the 1070, 1080, 1080TI and Titan XPm while the RX 580 just now started winning the overall war. According to computerbase.de and pcgameshardware.de the 1060 6GB was slightly better than the RX 480 8GB, about 2-4%.

Now the RX 580 8GB is likely to come out ahead by 1-2%.

Wonder whose alias this is?

 

[Guru]

Doom, COD, RE 7, Mafia 3, Titanfall 2, The Division, Warhammer DOW 3, aren't sponsored by AMD.

Only TW:WH, DX:MD, BF1 and AOTS are.

So 4/11 are sponsored by AMD vs 7/9 of the bottom ones sponsored by Nvidia.

Doom, COD, Warhammer DOW 3, BF1 are all AMD sponsored titles, what are you talking about? http://www.amd.com/en-us/markets/game/featured