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

[Arachnotronic]

if it was like that, they could have say it was just a AMD Vulkan patch, they did not, and the way they killed kepler, that was no accident, not even the worse Gamework implementation ever could have managed to do that.

If they did that, they would appear biased and it would ruin the whole point of the marketing that AMD is doing. The idea is to make it look as though id optimized equally for both in its Vulkan renderer, but in reality AMD GPUs got the most love because they were probably able to take all of the low level optimizations that they did for the Xbox One/PS4 and implement them in the Vulkan codepath.

id wins because the work to squeeze the most out of AMD's architecture is probably trivial compared to the low level work that would need to be done to squeeze the most out of NVIDIA's architectures, and AMD wins because it gets people talking on forums about how RX 480 is "more future proof."

It's not great and I don't like it as a consumer, but from a biz perspective, this is good marketing/optics for AMD.

 

[Bacon1]

Yep id and Doom is totally anti Nvidia, thats why they intially showed off the Vulkan path while on stage at the 1080 release event

http://www.pcgamer.com/id-software-shows-doom-running-on-gtx-1080-using-vulkan/

But blame id not Nvidia for providing updated drivers to support Async Compute and other low level APIs.

 

[Abwx]

Completely wrong, read the article.

You, you should read the article better :

Originally Posted by JPR
JPR’s AIB Report tracks computer add-in graphics boards, which carry discrete graphics chips (GPUs). AIBs used in desktop PCs, workstations, servers, and other devices such as scientific instruments. They are sold directly to customers as aftermarket products, or are factory installed. In all cases, AIBs represent the higher end of the graphics industry using discrete chips and private high-speed memory, as compared to the integrated GPUs in CPUs that share slower system memory.

So they are either in the PCs sold by the OEMs or delivered by thoses same OEMs after the PCs have been bought, as said this doesnt include items sold by the Neweggs and other Micro Centers unless that s cards provided by PC manufacturers as add on for the PCs sold by these shops.

 

[96Firebird]

Are you reading it wrong? Your summary is contradicted by the paragraph you quoted.

Does it include pre-built PCs? Sure. Is it exclusive to those? No.

The cards you can buy from retailers yourself are called add-in boards.

 

[raghu78]

What appears to have happened is that NVIDIA optimized their OpenGL driver/worked with id mainly on the OpenGL path ahead of the launch so that it could deliver a good day 1 experience for customers, and they achieved this goal.

AMD's launch day performance was terrible and even after they got that initial patch in, their cards badly under-performed in OpenGL relative to what the hardware really should have been able to do.

It's clear that AMD went all-in on the Vulkan code path, sacrificing day 1 OpenGL performance. This hurt them when the RX 480 first launched but it allowed them to make a lot of PR noise when id finally released the Vulkan renderer. The big underperformance of their GPUs in OpenGL made the gains in going to the Vulkan renderer (which has additional "secret sauce" thanks to Async Compute support and use of GCN-specific shader intrinsics) look even bigger than they should have been had their OpenGL driver been properly done.

This big gain allows AMD to market to people that once Vulkan/DX12 is ubiquitous, its hardware is really going to shine.

I suspect we will see id improve the Vulkan codepath to better take advantage of Pascal (adding async compute support should give another 5% boost or so and I wouldn't be surprised if there are other optimizations to be done to take better advantage of NVIDIA's hardware), but who knows when that updated renderer will arrive.

p.s. I was shocked when I saw id's lead rendering engineer retweeting "#BetterRed" posts from AMD marketing personnel on Twitter. This is easily the most obvious "smoking gun" with respect to id giving AMD hardware more love in their Vulkan renderer than they did NVIDIA hardware.

Doom is a classic case of GPUOpen at work. AMD started this initiative to leverage their console wins and to help developers reuse the console low level optimizations on PC. In fact GPUOpen is even lower than DirectX 12 or Vulkan. So here we see the AMD leverage the console wins in the PC space as developers are already well versed with the GCN architecture and have already put in the hard work to optimize on consoles. GPUOpen allows the developer to pretty much reuse their console optimizations on PC.

https://en.wikipedia.org/wiki/GPUOpen

"AMD designed GPUOpen to be a competing open-source middleware stack released under the MIT License. The libraries are intended to increase software portability between video game consoles, PCs and also High-performance computing.

GPUOpen unifies many of AMD's previously separate tools and solutions into one package, also fully open-sourcing them under the MIT License.[3] GPUOpen also makes it easy for developers to get low-level GPU access.

Additionally AMD wants to grant interested developers the kind of low-level "direct access" to their GCN-based GPUs, that surpasses the possibilities of Direct3D 12 or Vulkan. AMD mentioned e.g. a low-level access to the Asynchronous Compute Engines (ACEs). The ACE implement "Asynchronous Compute", but they cannot be freely configured neither under Vulkan nor under Direct3D 12.

Raja Koduri talked about why the consoles provide the best gaming experience / $ . Its all about the low level GPU access and the software toolchains which help developers extract the most out of the GPU hardware. GPUOpen which is open source is the definitive answer to Gameworks and it allows AMD to leverage the console wins in the PC GPU market. We are only seeing the tip of the iceberg with Doom. For developers giving them the same low level access as they have on the consoles and a unified toolchain allows them to bring all of the effort they put into the consoles on to the PC with minimal effort. I would not be surprised if AMD pushed GPUOpen based very low level optimizations heavily with upcoming AMD Gaming Evolved titles like Deus Ex Mankind Divided, Battlefield 1, Watch Dogs 2, Civilization 6.

 

[AtenRa]

What appears to have happened is that NVIDIA optimized their OpenGL driver/worked with id mainly on the OpenGL path ahead of the launch so that it could deliver a good day 1 experience for customers, and they achieved this goal.

Two months ago, DOOM Vulkan on GTX 1080 event

https://www.youtube.com/watch?v=0y3LaLJoo0s

id - We are going to be playing DOOM on the GTX 1080 running Vulkan, which is the first time we have announced that.

NV rep - How cool is that, yea DOOM, 1080 , Vulkan

NV rep - The game is awesome

id - We really wanted to be the first AAA game to really openly embrace Vulkan and show off its capabilities and its capabilities on Pascal are pretty stunning.

Present day,

 

[Sweepr]

Present day,

Selective posting at its best.

Developers posting #BetterRed on Twitter, praising Polaris in AMD events, not allowing Async Compute on any NVIDIA GPU, exclusive low level optimizations, being released coincidentally just a few days before GTX 1060 reviews. They turned a neutral game into a marketing tool overnight, and it doesn't get more obvious than this.

 

[mohit9206]

I think GTX1060 will outperform 980Ti in less than a year.Its only 15% faster now.Basically the same thing as how 970 outperforms 780Ti. So GTX1060 is better buy than used/new 980Ti.

 

[coercitiv]

I think GTX1060 will outperform 980Ti in less than a year.Its only 15% faster now.Basically the same thing as how 970 outperforms 780Ti. So GTX1060 is better buy than used/new 980Ti.

But does it matter how this card will perform in games against 980Ti in a year or two from now?

 

[3DVagabond]

Selective posting at its best.

Developers posting #BetterRed on Twitter, praising Polaris in AMD events, not allowing Async Compute on any NVIDIA GPU, exclusive low level optimizations, being released coincidentally just a few days before GTX 1060 reviews. They turned a neutral game into a marketing tool overnight, and it doesn't get more obvious than this.

"Not allowing async compute". They are working with nVidia to get it working and are waiting on a driver. It's not like we haven't heard that before.

 

[antihelten]

I think GTX1060 will outperform 980Ti in less than a year.Its only 15% faster now.Basically the same thing as how 970 outperforms 780Ti. So GTX1060 is better buy than used/new 980Ti.

The 980 Ti is actually about 20% faster than the 1060, and the 780 Ti was only about 10% faster than the 970 when the latter launched, and today the 970 doesn't outperform the 780 Ti, they are basically tied.

So if the 1060 follows the same pattern as the 970, then it would still be roughly 10% slower than the 980 Ti 1-2 years from now.

 

[3DVagabond]

I think GTX1060 will outperform 980Ti in less than a year.Its only 15% faster now.Basically the same thing as how 970 outperforms 780Ti. So GTX1060 is better buy than used/new 980Ti.

The 780ti is being outperformed by Tahiti. Future performance on nVidia is not something to look forward to.

 

[AtenRa]

Developers posting #BetterRed on Twitter, praising Polaris in AMD events, not allowing Async Compute on any NVIDIA GPU, exclusive low level optimizations, being released coincidentally just a few days before GTX 1060 reviews. They turned a neutral game into a marketing tool overnight, and it doesn't get more obvious than this.

So it is always the Developers and not NVIDIA, always someone else and not NVIDIAs fault. It has nothing to do with NVIDIAs Hardware and crappy DX-12/Vulkan drivers.

I like that, poetic justice

 

[railven]

I swung by my nephews house. Seems the GTX 1060 actually plays their more demanding games @ 1080p/60. Go figure.

Not a single one told me I should have gotten them a RX 480 because someone at school said it would perform better in the future.

Only complaint with the GA-1060-WF is it's so long. It's basically a full size card but seeing other GTX 1060 (like that EVGA SC) makes me wish I had waited to snag a smaller one. I know my nephews, that case has hit the floor once or twice over the years.

Welps the 5870 2GB survived their torture.

 

[Silverforce11]

I swung by my nephews house. Seems the GTX 1060 actually plays their more demanding games @ 1080p/60. Go figure.

Not a single one told me I should have gotten them a RX 480 because someone at school said it would perform better in the future.

Only complaint with the GA-1060-WF is it's so long. It's basically a full size card but seeing other GTX 1060 (like that EVGA SC) makes me wish I had waited to snag a smaller one. I know my nephews, that case has hit the floor once or twice over the years.

Welps the 5870 2GB survived their torture.

And if you had bought your nephew a RX 480, he would still play his demanding games at 1080p/60 and wouldn't know any different. Except in 2 years time, you wouldn't need to give him a new GPU upgrade, he'll still be gaming at 1080p/60 in DX12/Vulkan demanding titles.

 

[railven]

And if you had bought your nephew a RX 480, he would still play his demanding games at 1080p/60 and wouldn't know any different. Except in 2 years time, you wouldn't need to give him a new GPU upgrade, he'll still be gaming at 1080p/60 in DX12/Vulkan demanding titles.

In 2 years they'll probably have a Titan XP if I get my way this upgrade cycle!:twisted: At worse, a GTX 1080.

I'd rather give my oldies to family/friends then deal with the hassle of selling them. But that's just me.

 

[Shivansps]

So it is always the Developers and not NVIDIA, always someone else and not NVIDIAs fault. It has nothing to do with NVIDIAs Hardware and crappy DX-12/Vulkan drivers.

I like that, poetic justice

On DX12/Vulkan its devs who most provide a implementation that works at least at the same speed or better than the higher level one, there is no much nvidia can do about it besides putting money for devs to do it, like AMD.

This is funny, i remember people like you bashing Gameworks for cripping old Nvidia cards performance, then ID Tech comes forward and completely destroy Kepler performance on Vulkan, with a path in theory dedicated to provide better performance to old cards, and it looks fine to you, there is no biased devs here.

Some day you must realise that AMD is just another corporation that puts money to do stuff like Doom devs did, its so hard to belive that AMD is using Nvidia tactics? AMD is your hero or something? im sorry.

 

[sze5003]

In 2 years they'll probably have a Titan XP if I get my way this upgrade cycle!:twisted: At worse, a GTX 1080.

I'd rather give my oldies to family/friends then deal with the hassle of selling them. But that's just me.

Yea I do the same. My brother had a gtx 555 and now he has my 7970.

 

[railven]

Yea I do the same. My brother had a gtx 555 and now he has my 7970.

Original plan was to give em my 780 Lightning, but I ran into picking a card for myself issue and I didn't want them to wait for me to upgrade. Since GF is getting my 980 Ti.

I'll just toss it to my brother-in-law who still has one of my 660 Tis.

 

[sze5003]

Only thing is now I'm stuck without a card. Had a 1070 ordered last night and minutes later newegg emails me they cancelled my order due to stock issues.

 

[railven]

Only thing is now I'm stuck without a card. Had a 1070 ordered last night and minutes later newegg emails me they cancelled my order due to stock issues.

Exactly why I caved and bout them a GTX 1060. If I stuck to the original plan, I'd be out of a card until whatever I decide gets in stock.

I love my family, but screw them if they think I'll go cardless for more than a week. Haha.

 

[Face2Face]

I think GTX1060 will outperform 980Ti in less than a year.Its only 15% faster now.Basically the same thing as how 970 outperforms 780Ti. So GTX1060 is better buy than used/new 980Ti.

Could be the case in DX12 games, but since Pascal is essentialy 14nm Maxwell, I doubt Maxwell will fall off as quickly as Kepler did.

 

[AtenRa]

On DX12/Vulkan its devs who most provide a implementation that works at least at the same speed or better than the higher level one, there is no much nvidia can do about it besides putting money for devs to do it, like AMD.

This is funny, i remember people like you bashing Gameworks for cripping old Nvidia cards performance, then ID Tech comes forward and completely destroy Kepler performance on Vulkan, with a path in theory dedicated to provide better performance to old cards, and it looks fine to you, there is no biased devs here.

Some day you must realise that AMD is just another corporation that puts money to do stuff like Doom devs did, its so hard to belive that AMD is using Nvidia tactics? AMD is your hero or something? im sorry.

http://www.pcgameshardware.de/Doom-2016-Spiel-56369/Specials/Vulkan-Benchmarks-Frametimes-1202711/

is the current driver Geforce 368.81 WHQL an image synchronization active

Since this can be done in various ways on User Modes, the problem with the image output in our opinion is to be sought in the latest Nvidia drivers.

The volcano API anyway also supports a non-synchronized image output. The fact therefore should be relatively easy to resolve by Driver Fix.

Again, its the driver and not the developer. And this has nothing to do with GameWorks and Gaming Evolved, DOOM is neutral, get your facts straight.

edit: This is the same review at 3840x2160. Since the fps doesnt go higher than 60 we dont have the image synchronization NVIDIA problem. GTX 780Ti performance is horrendous with or without Vulkan, its the hardware that is not able to perform and not the developer.

 

[sontin]

It is not the driver. it is the broken Vulkan path on nVidia hardware.

Talos for example runs without "problems". :\

 

[Silverforce11]

It is not the driver. it is the broken Vulkan path on nVidia hardware.

Talos for example runs without "problems". :\

Talos runs Vulkan slower, I say that's a problem.

DOTA 2 is a better example. It can be faster.