Geforce GTX 1060 Thread: faster than RX 480, 120W, $249

[BSim500]

Wow, that is quite amazing.

half the power and still 80% of the performance.

Indeed. And I'm not even sure if all THG did was just reduce Power Limit (which adjusts the "stock" freq/volt curve). There's actually a neat Turbo Boost trick with Maxwell's where if you overclock whilst simultaneously reducing Power Limit below 85% or so, it will start undervolting faster than it underclocks, ie, performance below 100w or so could even be slightly higher. I played around with this stuff when helping a friend set up her HTPC. Turned a "110w" GTX 960 1,253MHz into an 82w 1,328MHz card (+5% OC at 2/3rds the power). If that extends to Pascal / Turbo Boost 3.0, the green lines on that THG chart could be even flatter. It's going to be one hell of a mobile chip.

 

[sirmo]

Wow, that is quite amazing.

half the power and still 80% of the performance.

Metro Last Light is about the worse case scenario for rx480.. only game worse would have been Project Cars.

Of course it's going to look inefficient when it's significantly slower than 1060 in that game.

rx480 uses 20% more power, yet it's up to 30% faster in Vulkan Doom.. it also powers 2gb more Vram.. which equals 10 watts handicap.. If they used that game as the comparison rx480 would have been ahead.

Using Gameworks titles for power efficiency comparisons is flawed.

 

[Erenhardt]

Wow, that is quite amazing.

half the power and still 80% of the performance.

Going by this trend I would rather have 50% performance at 0 watts of power thank you very much.

 

[Sweepr]

rx480 uses 20% more power, yet it's up to 30% faster in Vulkan Doom..

AMD exclusive low-level optimizations (including GCN shader extensions) + Async Compute only on their cards. Even Rise of the Tomb Raider is more neutral than Doom Vulkan right now.

http://steamcommunity.com/app/379720/discussions/0/359543542243503468
http://gpuopen.com/gcn-shader-extensions-for-direct3d-and-vulkan

it also powers 2gb more Vram.. which equals 10 watts handicap.. If they used that game as the comparison rx480 would have been ahead.

Sure...

MSI Gaming X 1060 vs Sapphire Nitro+ RX 480

4GB model

8GB model

85W more than 1060 even with 4GB. RX 480 is nowhere near 1060 efficiency overall. And let's not forget 1070/1080 are even better than 1060 here.

 

[Sweepr]

Talking about GTX 1060 mobile, first notebook listed in Europe:

ASUS ROG G752VM introduces laptop with GeForce GTX 1060

As I mentioned, laptop ASUS G752VM will be equipped with GeForce GTX 1060. Before dozen days I informed about its specifications. It is known so that the card will receive 6 GB of memory on a 192-bit bus, which gives a bandwidth of 192 GB / s. The base clock speed compared to desktop variants will be reduced from 1506 MHz to 1405 MHz. The GPU Boost 3.0 mode, frequency will be 1671 MHz however, that is 38 MHz less than the version for the PC. There is it because 1709 MHz. But I can tell you that depending on the game and the level of overclocking, the GeForce 1060 GTX laptop can safely cross the threshold of 1900 MHz.

www.purepc.pl/notebooki/asus_wprowadza_laptopa_rog_g752vm_z_geforce_gtx_1060

 

[BSim500]

Metro Last Light is about the worse case scenario for rx480.. only game worse would have been Project Cars.

LOL. Everything where AMD isn't ahead is "biased" to some... Your own chart shows Metro:LL running +15% faster. Given there are differences of +23% in Crysis 3, +38% in BF4, +32% in AC:Syndicate, +22% Overwatch, +27% Tomb Raider DX11, +21% Mirror's Edge Catalyst, +31% ARMA 3, +25% Mad Max, etc, and many other mainstream DX11 titles vs the RX480 at 1080p, that's hardly worst case. Real cherry picking would be something like Anno 2205 (the same +30% outlier as your own inverted "correct" selection of DX12-only titles).

it also powers 2gb more Vram.. which equals 10 watts handicap.

4GB vs 8GB RX 480 has been measured to be a mere 3-5w (and that's with the larger quantity clocked at higher speeds):-
http://www.anandtech.com/show/10446/the-amd-radeon-rx-480-preview/6

^ And that's why when measurements that turn out half to a third as much as "it WOULD perform..." forum guesstimates, the only real answer is "show me" (preferably in a wide spread of +30 games that aren't equally obviously cherry-picked for DX12-only AMD bias, and are measured "like for like", ie, you undervolt both or neither...)

 

[Keysplayr]

LOL. Everything where AMD isn't ahead is "biased" to some... Your own chart shows Metro:LL running +15% faster. Given there are differences of +23% in Crysis 3, +38% in BF4, +32% in AC:Syndicate, +22% Overwatch, +27% Tomb Raider DX11, +21% Mirror's Edge Catalyst, etc, and many other mainstream DX11 titles vs the RX480 at 1080p, that's hardly worst case. Real cherry picking would be something like Anno 2205 (the same +30% outlier as your own inverted "correct" selection of DX12-only titles).


4GB vs 8GB RX 480 has been measured to be a mere 3-5w (and that's with the larger quantity clocked at higher speeds):-
http://www.anandtech.com/show/10446/the-amd-radeon-rx-480-preview/6

^ And that's why when measurements that turn out half to a third as much as "it WOULD perform..." forum guesstimates, the only real answer is "show me" (preferably in a wide spread of +30 games that aren't equally obviously cherry-picked for DX12-only AMD bias, and are measured "like for like", ie, you undervolt both or neither...)

Wow, that certainly paints a more, non cherry picked picture of 1060.

 

[fingerbob69]

...games that aren't equally obviously cherry-picked for DX12-only AMD bias...

Given we're moving into a dx12/Vulcan world ...that's only going to get tougher ...more "biased".

 

[lukart]

Given we're moving into a dx12/Vulcan world ...that's only going to get tougher ...more "biased".

I wouldn't say biased but more adapted to the console / PC gaming overall.
They share a lot more than 5 years ago. Good for all of us I guess

 

[antihelten]

4GB vs 8GB RX 480 has been measured to be a mere 3-5w (and that's with the larger quantity clocked at higher speeds):-
http://www.anandtech.com/show/10446/the-amd-radeon-rx-480-preview/6

Note that this is not really a test of the 4GB version, it is a test of the 8GB version with a special BIOS version that disables 4GB worth of VRAM (this is what all the reviewers received).

If you removed half the memory chips from the RX 480, then you could probably save 15-20W, however it is important to note that this is not actually what happens with the real 4GB version, instead it simply uses the same number of chips but each one at half the size (i.e. 4gbit chips instead of 8gbit chips). I have no idea how big of an influence memory chip density has on power usage.

 

[BSim500]

Note that this is not really a test of the 4GB version, it is a test of the 8GB version with a special BIOS version that disables 4GB worth of VRAM (this is what all the reviewers received).

If you removed half the memory chips from the RX 480, then you could probably save 15-20W

Problem is everyone keeps churning out these "probablies" and literally making up numbers and throwing them around every day as absolute facts with nothing to back it up. 40w for 8GB of modern 1.3v VRAM sounds far too high. Even if it were right though, given the GTX 1060 in Toms review was scoring more fps at "32w" (ie, "62w" total GPU minus "30w" of 6GB VRAM) vs the 8GB RX 480 at "100w" ("140w" total GPU minus "40w" of 8GB VRAM), you end up up with silly "core" figures to try and "talk around" the obvious disparity in architectural efficiencies (by blaming it on the memory), ie, the "thirstier" the VRAM as a percentage of total card power usage, the more it actually makes the whole card disparity worse which I'm pretty sure is the exact opposite of what some are trying to "demonstrate"...

 

[PhonakV30]

AMD exclusive low-level optimizations (including GCN shader extensions) + Async Compute only on their cards. Even Rise of the Tomb Raider is more neutral than Doom Vulkan right now.

http://steamcommunity.com/app/379720/discussions/0/359543542243503468
http://gpuopen.com/gcn-shader-extensions-for-direct3d-and-vulkan

Look what he said! ROTTR more neutral than Doom ? are you kidding me ?

 

[zinfamous]

Look what he said! ROTTR more neutral than Doom ? are you kidding me ?

I don't think dude has ever been called subtle. Though in this context, it is fair to call ROTTR neutral--it generally performs equally poorly on both brands--but it seems to have been patched, right?

Though now claiming that low level optimizations including async compute are only available on AMD? That is odd: for the last year up until that post, we have been hearing that nVidia will allow async compute.

Any day.

Heck, even Ryan talked about it in the 1080 review. Saying it isn't possible on Pascal is, well....not very subtle.

 

[antihelten]

Problem is everyone keeps churning out these "probablies" and literally making up numbers and throwing them around every day as absolute facts with nothing to back it up. 40w for 8GB of modern 1.3v VRAM sounds far too high. Even if it were right though, given the GTX 1060 in Toms review was scoring more fps at "32w" (ie, "62w" total GPU minus "30w" of 6GB VRAM) vs the 8GB RX 480 at "100w" ("140w" total GPU minus "40w" of 8GB VRAM), you end up up with silly "core" figures to try and "talk around" the obvious disparity in architectural efficiencies (by blaming it on the memory), ie, the "thirstier" the VRAM as a percentage of total card power usage, the more it actually makes the whole card disparity worse which I'm pretty sure is the exact opposite of what some are trying to "demonstrate"...

40W for 8 GDDR5 modules running at 8 gbps is not that crazy actually (AMD claimed 35W when they were advertising HBM).

But I agree that no matter how much less power power the "true" 4GB version uses, it's realistically speaking not going to be enough to match the efficiency of the 1060.

If one were to only look at a game like DOOM in isolation then the 4GB 480 might get close to matching a 1060*, but that's just a single game.

*A 1060 is normally 55% more efficient and 10% faster than the RX 480. If the RX 480 is 30% faster in DOOM and both cards use the same amount of power as in Metro: LL (the game TPU uses for power tests), then that would bring the RX 480 within 10% efficiency wise, which is a small enough gap that a different memory config could possibly account for it.

Look what he said! ROTTR more neutral than Doom ? are you kidding me ?

ROTTR is actually fairly neutral now after the latest patch:

http://www.overclock3d.net/reviews/...e_tomb_raider_directx_12_performance_update/3

 

[sirmo]

AMD exclusive low-level optimizations (including GCN shader extensions) + Async Compute only on their cards. Even Rise of the Tomb Raider is more neutral than Doom Vulkan right now.

That's my point. For new LL API games rx480 is pretty efficient. Not so in games which don't take advantage of all that rx480 has to offer.

This should be common sense, no? A single core CPU was more efficient if all you ran was single threaded code.

 

[frozentundra123456]

What is with the new trend to subtract vram usage from power consumption?? Can your PSU somehow magically tell itself not to send power to unused vram? Of course not. If AMD puts out a true 4gb card, then it is valid to test that and see if it is more efficient. Otherwise, it is just pointless speculation. A card "uses what it uses" to paraphrase the old saying.

 

[sirmo]

What is with the new trend to subtract vram usage from power consumption?? Can your PSU somehow magically tell itself not to send power to unused vram? Of course not. If AMD puts out a true 4gb card, then it is valid to test that and see if it is more efficient. Otherwise, it is just pointless speculation. A card "uses what it uses" to paraphrase the old saying.

When it comes to comparing 2 cards yes. When it comes to comparing architectures and looking at the power consumption analysis it's important to understand the underlying details.

 

[OCNewbie]

EVGA GTX 1060 SC is currently in stock at Newegg for $259.99 with a $10 mail-in rebate. Add the $25 off $200 using paypal coupon to make it $224.99:

http://www.newegg.com/Product/Product.aspx?Item=N82E16814487261

 

[OCNewbie]

EVGA GTX 1060 SC is currently in stock at Newegg for $259.99 with a $10 mail-in rebate. Add the $25 off $200 using paypal coupon PP2016BTS to make it $224.99 (must pay with paypal) after the $10 mail-in rebate:

http://www.newegg.com/Product/Product.aspx?Item=N82E16814487261

Non-SC is also in stock for $249.99:

http://www.newegg.com/Product/Product.aspx?Item=N82E16814487260

 

[Thala]

Quick calculation shows, that GTX1060 needs 30% less power on SoC level (120W vs 85W) and 25% less power on board level (160W vs 120W) compared to RX480. Assuming RX480 has 20% more shader performance, which will get visible in DX12 and Vulkan games GTX1060 is still more efficient. For the average DX11 game including gameworks titles the higher power efficiency of GTX060 is just more obvious. For DX12 games the advantage is apparently shrinking but should be still there in average.

 

[rh71]

EVGA GTX 1060 SC is currently in stock at Newegg for $259.99 with a $10 mail-in rebate. Add the $25 off $200 using paypal coupon PP2016BTS to make it $224.99 (must pay with paypal) after the $10 mail-in rebate:

http://www.newegg.com/Product/Product.aspx?Item=N82E16814487261

Non-SC is also in stock for $249.99:

http://www.newegg.com/Product/Product.aspx?Item=N82E16814487260

My price alert via email got me there and I happened to get the very last one of the non-sc. Hope it doesn't get cancelled. $230 shipped is insane for a modern GPU. Coming from a Radeon 6850.

 

[railven]

My price alert via email got me there and I happened to get the very last one of the non-sc. Hope it doesn't get cancelled. $230 shipped is insane for a modern GPU. Coming from a Radeon 6850.

Nice upgrade. My nephews went from a 5870 2GB to GA-1060 WF. Card was longer than I thought but their case took it like a champ. EDIT: And it had a backplate which I wasn't aware of until I unboxed. Decent card for $250.

Good to see cards hitting that MSRP price point.

 

[rh71]

^ don't shoot me, but I'll be putting it in a Q6600 machine until I upgrade the rest of it. I was getting directX errors left and right when gaming with my out-of-support Radeon and this upgrade made sense.

 

[sze5003]

^ don't shoot me, but I'll be putting it in a Q6600 machine until I upgrade the rest of it. I was getting directX errors left and right when gaming with my out-of-support Radeon and this upgrade made sense.

I think I still have a q6600 laying around somewhere. I liked that thing, ran nice with my 8800GT.

 

[hsjj3]

Who in their right mind thinks that 8GB of 2000MHz VRAM consumes 40W of power?

If this was any true, the 4GB version of the RX 480 ought to be consuming more than 20W less than it's big brother, and that's far from the truth.

Even in the worst case scenario, the 6GB GTX 1060 would have a 10W advantage, but it's probably worth 5W or less.