Digital Foundry: DX12 faceoff between GTX 1060 and AMD 480

[Mondozei]

https://www.youtube.com/watch?v=dTJPp7WdEBs

Interesting, short video.

 

[crisium]

Summary of results for those of us at work?

 

[Hitman928]

Summary of results for those of us at work?

At 1080p the 1060 edges out the 480 in Timespy. They test async on and off, 1060 gains a little, 480 gains a little more, that's it. DX12 faceoff is not accurate, it's a short timespy test video.

 

[crisium]

Oh, synthetic benchmark comparisons. Several sites already have real, live, actual DX12 game comparisons...

 

[AnandThenMan]

Fix your thread title this is not a DX12 faceoff but a single synthetic benchmark.

 

[poofyhairguy]

Real faceoff results:

 

[dogen1]

Fix your thread title this is not a DX12 faceoff but a single synthetic benchmark.

I assumed that was the name of the video, but yeah the thread title isn't very accurate.

 

[Hitman928]

I assume that was the name of the video, but yeah the thread title isn't very accurate.

The title of the video is:
"3DMark DirectX 12 Time Spy: GTX 1060 vs RX 480 Analysis"

 

[boozzer]

3DMark DirectX 12 Time Spy: GTX 1060 vs RX 480 Analysis

video title, much better.

 

[LTC8K6]

Real faceoff results:

That's basically saying the 1060 and 480 are tied overall in DX12?

 

[poofyhairguy]

That's basically saying the 1060 and 480 are tied overall in DX12?

I think the fair statement to make is the 1060 is more ahead in Directx 11 than the 480 is ahead in Directx 12.

 

[dogen1]

The title of the video is:
"3DMark DirectX 12 Time Spy: GTX 1060 vs RX 480 Analysis"

I meant to say assumed.

 

[swilli89]

That's basically saying the 1060 and 480 are tied overall in DX12?

Well since 1060 is 5-7% faster in DX11 i guess they are saying they are tied in dx11 as well.

 

[LTC8K6]

Well since 1060 is 5-7% faster in DX11 i guess they are saying they are tied in dx11 as well.

I was referring to the chart that was posted. It doesn't say a single word about DX11 as far as I can tell.

But, if the cards are tied in gaming overall, then that's a win for NV due to power consumption.

 

[poofyhairguy]

But, if the cards are tied in gaming overall, then that's a win for NV due to power consumption.

You forgot the three Vs:

Less Vram, lower Vulkan performance, 480 performance boost from under-volting

 

[RussianSensation]

Well since 1060 is 5-7% faster in DX11 i guess they are saying they are tied in dx11 as well.

1060's DX11 lead on many review sites is itself skewed because they tested Doom under OpenGL giving 1060 a huge 20% advantage. If Doom was to be tested under Vulkan, 480 would be leading by 20-30% instead. This is a major outlier that can swing the averages greatly!

It's shocking how biased Guru3d was in their testing by first showing off how much RX 480 gains under Vulkan in Doom but then tested RX 480 under OpenGL in the 1060 review. This video covers all the basics about how incompetent/biased a lot of the North American review sites are nowadays:

https://www.youtube.com/watch?v=V54W4p1mCu4

Guru3d
RX 480 Vulkan
1080p = 116 fps
1440p = 73 fps
http://www.guru3d.com/news-story/new-patch-brings-vulkan-support-to-doom.html

1060 review:
1080p
GTX 1060 = 108 fps
RX 480 Open GL = 94 fps

1440p
GTX 1060 = 69 fps
RX 480 Open GL = 61 fps
http://www.guru3d.com/articles_pages/geforce_gtx_1060_review,17.html



It's amazing how the PC gaming community is using averages in performance reviews while ignoring blatant reviewer bias/incompetence used in deriving those averages.

This is unforgivable because both GTX1060 and RX 480 run Doom faster under Vulkan. So what's the reason to test RX 480 under Open GL in a 1060 review? Had Guru3D and other sites, tested RX 480 under Vulkan in Doom it would have beaten the 1060. :whiste:

I was referring to the chart that was posted. It doesn't say a single word about DX11 as far as I can tell.

But, if the cards are tied in gaming overall, then that's a win for NV due to power consumption.

RX 480 = $200
vs.
1060 = $250

If you want to pay $50 more for lower power usage, sure it's a win then.

It's funny how the entire PC gaming community is blatantly ignoring the RX 480 4GB version at $199.99 because what, reasons? Same people who recommended GTX950 2GB, 960 2GB over R9 280X 3GB / R9 380X 3GB, R9 380 4GB and GTX970 3.5GB over R9 290 4GB/R9 390 8GB and had no issues recommending 980 4GB or 980 SLI even are now ignoring a $200 RX 480 that's 90% as fast as a $550 980 that came out less than 2 years ago.

NV PR/Marketing at its finest.

 

[96Firebird]

Well since 1060 is 5-7% faster in DX11 i guess they are saying they are tied in dx11 as well.

But that same website says 8-12%...

 

[LTC8K6]

You forgot the three Vs:

Less Vram, lower Vulkan performance, 480 performance boost from under-volting

I think you forgot that I don't actually think the cards are tied...

 

[LTC8K6]

1060's DX11 lead on many review sites is itself skewed because they tested Doom under OpenGL giving 1060 a huge 20% advantage. If Doom was to be tested under Vulkan, 480 would be leading by 20-30% instead. This is a major outlier that can swing the averages greatly!

It's shocking how biased Guru3d was in their testing by first showing off how much RX 480 gains under Vulkan in Doom but then tested RX 480 under OpenGL in the 1060 review. This video covers all the basics about how incompetent/biased a lot of the North American review sites are nowadays:

https://www.youtube.com/watch?v=V54W4p1mCu4

Guru3d
RX 480 Vulkan
1080p = 116 fps
1440p = 73 fps
http://www.guru3d.com/news-story/new-patch-brings-vulkan-support-to-doom.html

1060 review:
1080p
GTX 1060 = 108 fps
RX 480 Open GL = 94 fps

1440p
GTX 1060 = 69 fps
RX 480 Open GL = 61 fps
http://www.guru3d.com/articles_pages/geforce_gtx_1060_review,17.html



It's amazing how the PC gaming community is using averages in performance reviews while ignoring blatant reviewer bias/incompetence used in deriving those averages.

This is unforgivable because both GTX1060 and RX 480 run Doom faster under Vulkan. So what's the reason to test RX 480 under Open GL in a 1060 review? Had Guru3D and other sites, tested RX 480 under Vulkan in Doom it would have beaten the 1060. :whiste:

There's still the question of the Vulkan library version differences, though. NV and AMD use different ones for Doom, apparently. NV using an older revision.

Whether it makes any difference or not, I haven't heard and I don't have a clue.

I also haven't heard if it was resolved, or what.

 

[RussianSensation]

I think you forgot that I don't actually think the cards are tied...

They are not. The cheapest RX 480 is $200 and the cheapest GTX1060 is $250. GTX 1060 is 9-10% faster for 25% more $$. The RX 480 is actually faster or ties the more expensive 1060 under current DX12/Vulkan games if we ignore the broken DX12 implementation of Rise of the Tomb Raider.

RX 480 also earns $$$ under Windows allowing a PC gamer to pay for itself in less than 6 months. Of course an AMD's card ability to make $ will be 100% ignored by the Pro-NV crowd because they don't want to do the leg work to set this pro-AMD feature up.

So for the 8th consecutive year since HD4870 came out the real story for North American consumers paying $0.20 per kWh or less, who are open minded to learning something new in life, over the next 6 months it will be like this:

$0-200 RX 480 vs. $250 GTX1060

:biggrin:

PhysX, TXAA, ShadowPlay = NV features
Videocards that make $ under Windows 10 = AMD feature

I love it how NV users continue to ignore this amazing AMD feature over the last 8 years.

Now it's even possible to load the money earned onto a prepaid debit card and use it anywhere in the world:

https://www.cryptocompare.com/wallets/#/cards

 

[swilli89]

I was referring to the chart that was posted. It doesn't say a single word about DX11 as far as I can tell.

But, if the cards are tied in gaming overall, then that's a win for NV due to power consumption.

So future Vulkan/DX12 titles, Price, and VRAM concerns aren't part of this win formula huh? That's a little crazy, but I know you already know that.

 

[LTC8K6]

So future Vulkan/DX12 titles, Price, and VRAM concerns aren't part of this win formula huh? That's a little crazy, but I know you already know that.

Well, I have no concerns about that, but you are welcome to yours.

 

[poofyhairguy]

RX 480 also earns $$$ under Windows allowing a PC gamer to pay for itself in less than 6 months.

Amen. Hell after just two months my 480 will mine $100+ of profit. For a $199 card that is huge!

 

[Kenmitch]

There's still the question of the Vulkan library version differences, though. NV and AMD use different ones for Doom, apparently. NV using an older revision.

Whether it makes any difference or not, I haven't heard and I don't have a clue.

I also haven't heard if it was resolved, or what.

But weren't the Radeon cards using a older OpenGL version? At least I remember reading so. Guess it may have changed from the initial launching of the game.

 

[RussianSensation]

Well, I have no concerns about that, but you are welcome to yours.

Computerbase.de shows that the reference 480 averages only 1200mhz or so under GPU boost.

AIB 480s will ship with 1.34+Ghz clocks, or a 13-16% performance boost.

https://www.overclockers.co.uk/sapp...ddr5-pci-express-graphics-card-gx-37b-sp.html

At 1400mhz, RX 480 beats both the R9 390 @ 1150mhz and the GTX970 @ 1500mhz.

Sapphire RX 480 4GB Nitro + 1.3Ghz is already up for pre-order in the UK at 200 pounds.

TechSpot shows no performance difference between an RX 480 4GB and the 8GB version.

Computerbase has the 1060 beating 480 by only 6% at 1080p and an AIB 1060 beats the 480 by 13%.
https://www.computerbase.de/2016-07...hnitt_benchmarks_in_1920__1080_und_2560__1440

These differences will be easily made up by after-market RX 480 cards but the RX 480 4GB should still end up cheaper than GTX1060.

Amen. Hell after just two months my 480 will mine $100+ of profit. For a $199 card that is huge!

I am at $750 USD on my Hawaii cards and I added the R9 295X2 only in early April. That means I am just $50 away from $0 GTX 1070 SLI, which is only 11-12 days of mining. By the time next gen flagships cards come out in 2017, I'll have enough to buy Big Pascal SLI for $0 (if I wanted to do that hypothetically).

If a $200 Intel CPU earned $50 a month after electricity, it would be headline news.