Support NeoGAF

[Bolivar687]

I know you are an AMD fan and that's why you should not try to be an apologist, you should kick them in the nuts. Blaming the circumstances just does not work and it goes without saying the same applies to Nvidia. They have noone but themselves to blame for their underwhelming performance in some price brackets, they should have done better.

I don't know exactly what anyone is supposed to be kicking them in the nuts for. They've had amazing priceerformance for a while now, and they've addressed most of the concerns with thermals, efficiency, and frametimes this year. Reading this thread, you see tons of initiatives that already have and will continue to positively benefit consumers, from Mantle evolving into DX12, async compute, LiquidVR, Freesync, Crimson, the open initiatives, it just goes on and on. I have to ask where the "underwhelming performance in some price brackets," is coming from when they're outclassing Nvidia in nearly every one, unless you're talking about the Fury X.

AMD is one of those topics where people go in threads to rattle off the same talking points without critically thinking about the thread topic (obviously not you Kezen), and the Fury X, as the flagship debut of HBM, understandably persists as one of main culprits of that mentality. However, just looking at the two companies, it should be obvious that AMD is not going to come out and completely trash Nvidia with vastly more powerful cards for lower prices. And they shouldn't be measured against an unrealistic yardstick just to get "back in the game". It's incredibly disheartening that great products with pro-consumer initiatives aren't enough. Instead, it sounds like they rather need the right marketing firm to begin shifting the narrative.

 

[Sir Abacus]

Holy shit I'm actually getting dumber by reading this. I couldn't give a flying fuck about inspiration. When I go to buy my next card it'll be the card that pumps out the most frames with the least amount of bullshit. AMD have failed dramatically on this for the last couple of generations.

It's no coincidence that the last time AMD was on top was the R300 series.

 

[AmyS]

Polaris is a cool name, I'll give them that at least - When I was a kid, one of my favorite LEGO sets was this:

lol.

 

[Besthelboy]

Probably will replace my 290.

 

[aeolist]

i'd be surprised if they're just starting to talk about this now and it's on the slate for release in 2016. usually you hear about major changes like this well in advance.

 

[Crayon]

https://www.youtube.com/watch?v=ZcF36_qMd8M

Some factual errors, but overall Nvidia is a jerk.

Wow. I don't really follow gpus or the business. I was just wanting to ask for something like this but didn't know how to ask. I look into it a bit more.

 

[Rushersauce]

Holy shit I'm actually getting dumber by reading this. I couldn't give a flying fuck about inspiration. When I go to buy my next card it'll be the card that pumps out the most frames with the least amount of bullshit. AMD have failed dramatically on this for the last couple of generations.

It's no coincidence that the last time AMD was on top was the R300 series.

What nonsense is this?

 

[AmyS]

https://www.youtube.com/watch?v=ZcF36_qMd8M

Some factual errors, but overall Nvidia is a jerk.

Yep, but the CEO...

 

[cebri.one]

So another rebrand of older cards?

 

[chaosblade]

So another rebrand of older cards?

How can a new architecture be a rebrand? Even if it's "just" GCN2.0 under a different name it's still new cards.

I agree that AMD is getting to the point it doesn't matter though. Their mindshare is so incredibly poor compared to NVidia that they basically have to rely on NVidia to blow it to regain some momentum - the quality of AMD's products doesn't really matter.

 

[Locuza]

Mixed precision is not for games, it's for scientific apps.

FP16 could be great for games.

What was previously stated, from AMD slides was that Arctic Islands was going to be GCN 2.0. Sounds like what they are doing is rebranding that into Polaris.

AMD never stated publicly Arctic Islands, nor GCN "2.0".

Regardless, the perf/watt of most AMD cards is not bad. The R9 380/380X score well there.

Actually every AMD GPU besides Fiji is scoring relatively bad there.

aside from power/performance, I hoping they also focus on their open source "eye candy" features like Physx. with all consoles using their GPU, I think most PC port will have a huge chance using the features if they light and good enough for consoles.

And yeah, I chose nVidia because their Physx. I love eye candies!

The main thing about PhysX is the exclusivity.
AMDs open-source stuff would need to be running much faster on their hardware, than on Nvidia Hardware, to be an argument.

 

[clav]

Linux Kernel starting from 4.4 will change everything as it looks like AMD has been working close with the open source community.

 

[Odrion]

I can't wait to see how they'll fuck it up this time.

 

[Applecot]

If they are inspired by the efficiency of stars in their ability to produce light; it's no wonder their recent cards have been so inefficient.

 

[jfoul]

2016 is a big year for AMD. From what we've seen recently it looks like the flagship AMD GPU will be running on 14nm (Samsung), while Nvidia will be on 16nm (TSMC). I can't wait to get my hands on the next offerings from AMD & NV, with the die shrinks and HBM.

 

[pottuvoi]

Mixed precision is not for games, it's for scientific apps. Nvlink is not for gaming either, I'm not sure Nvidia willfully created fake hype around their GPUs, it's mostly the mainstream outlets not getting it and being desperate for clicks. To my knowledge Nvidia never claimed Pascal would be twice as fast as Maxwell. I guess some read what they want to read.

If one can get ~1.5x throughput simply by using FP16 in cases where full precision of FP32 is not needed, certainly developers are interested.
Really hope that AMD and Intel will have proper support for it as well.

Anyway GCN2 should be interesting, hoping them to open all those fun bits they have in it and in GCN with Vulkan. (Really shame that we have great GPU, but developers can not access all the fun bits with DX or GL.)

 

[davidwhangchoi]

amd, just give me drivers and not a year later when the game is on steam for $10

 

[Unknown Soldier]

2016 is AMD's year of last chances. Polaris and Zen need to be not just competitive but superior to the Nvidia and Intel equivalents if the company is to survive. This really is the end of the line for AMD. The entire fate of the company now depends on Polaris and Zen, they have no money left to limp along with another line of failed products the way they did with Bulldozer and rebrand the same thing forever with GCN all these years and their marketshare and mindshare were nearing zero in 2015.

Do or do not, AMD. There is no try.

 

[Possumowner]

I feel AMD are in the last chance saloon in 2016,let's hope for competitions sake they produce the goods...unless they get bought out

 

[Kezen]

If one can get ~1.5x throughput simply by using FP16 in cases where full precision of FP32 is not needed, certainly developers are interested.
Really hope that AMD and Intel will have proper support for it as well.

Anyway GCN2 should be interesting, hoping them to open all those fun bits they have in it and in GCN with Vulkan. (Really shame that we have great GPU, but developers can not access all the fun bits with DX or GL.)

Ah, I did not know it could be truly practical for games.

 

[Sir Abacus]

If one can get ~1.5x throughput simply by using FP16 in cases where full precision of FP32 is not needed, certainly developers are interested.
Really hope that AMD and Intel will have proper support for it as well.

They already could use FP16 if they really wanted to. They don't do it (anymore) because FP32 is so obnoxiously better at the job. Pre-shader days you only had a couple of layers of multi texturing at most and you could get away with shittier colors and rounding errors. Shaders like water reflection/refraction, bump mapping, specular highlighting all demand extreme accuracy and precision to not look "weird". The only time you'll see FP16 now is on mobile devices and that's because performance efficiency well outweighs the image quality on a smaller screen.

 

[Kezen]

I don't know exactly what anyone is supposed to be kicking them in the nuts for. They've had amazing priceerformance for a while now, and they've addressed most of the concerns with thermals, efficiency, and frametimes this year. Reading this thread, you see tons of initiatives that already have and will continue to positively benefit consumers, from Mantle evolving into DX12, async compute, LiquidVR, Freesync, Crimson, the open initiatives, it just goes on and on. I have to ask where the "underwhelming performance in some price brackets," is coming from when they're outclassing Nvidia in nearly every one, unless you're talking about the Fury X.

Let me clarify : "underwhelming" performance in the sub 200 bracket was referring to Nvidia. In regards to the rest of points it seems to me that many PC gamers simply do not care much about open initiatives. It's cool that AMD are launching their new GPU open library and bet on open standards but I suspect this is going to entice developpers much more than consumers who apparently value the end result more.
I like Gameworks and I don't want it to go away, yet it's closed library but I believe it can perfectly live alongside AMD's GPU open. The former is intended to be easier to implement while open source is typically more challenging, or at least this is how it has been presented to me.
I'm not certain that Mantle evolved into DX12 however. For sure Mantle was a good proof of concept, but DX12 was always going to be a low level API as Microsoft and their partners knew that was the way forward. AMD stole their thunder for a while, that's good PR play.

AMD is one of those topics where people go in threads to rattle off the same talking points without critically thinking about the thread topic (obviously not you Kezen), and the Fury X, as the flagship debut of HBM, understandably persists as one of main culprits of that mentality. However, just looking at the two companies, it should be obvious that AMD is not going to come out and completely trash Nvidia with vastly more powerful cards for lower prices. And they shouldn't be measured against an unrealistic yardstick just to get "back in the game". It's incredibly disheartening that great products with pro-consumer initiatives aren't enough. Instead, it sounds like they rather need the right marketing firm to begin shifting the narrative.

"Get back in the game" obviously refers to them regaining ground (markershare). That excluded they never left the game proper.
And yes better marketing is needed for them, look around you, do you think PC gamers care one bit about "ethics" and what is "fair" ?

 

[Locuza]

Really hope that AMD and Intel will have proper support for it as well.

Starting with Broadwell Intels iGPUs have double FP16 throughput and Skylake improved the implementation.

The only time you'll see FP16 now is on mobile devices and that's because performance efficiency well outweighs the image quality on a smaller screen.

The "old days" are coming back, because efficiency is king nowadays and with FP16 you can get in certain cases double throughput (depending on the implementation) and save some registers and decrease the register bottleneck.

The developers of course don't have to use FP16 precision everywhere, simply only for cases, where higher precision doesn't really matter.

 

[Renekton]

http://www.eurogamer.net/articles/digitalfoundry-2015-amd-radeon-r9-380x-review

Not a bad turnout for Tonga. Beats out the 960 easily.

Nvidia's branding is too strong tho, so this will get overlooked.

 

[Irobot82]

I'm ready to hear when AMD/Nvidia are actually releasing new cards. Q2 2016 sounds so faaaar away. I need a new card guys hurry up!

AMD threads are always hilarious. Best posts are people who believe AMD's new cards will be rebrands. Even if we assumed GCN 1.0 to GCN 2.0 it's no different then Kepler to Maxwell. Maxwell wasn't built from the ground up from scratch, it was derived from Kepler.

 

[Locuza]

I like Gameworks and I don't want it to go away, yet it's closed library but I believe it can perfectly live alongside AMD's GPU open. The former is intended to be easier to implement while open source is typically more challenging, or at least this is how it has been presented to me.

In the end it depends on how the code is written and not necessarily if it's closed or open.
If the code quality is roughly equal, open-source will of course always win.

http://www.eurogamer.net/articles/digitalfoundry-2015-amd-radeon-r9-380x-review

Not a bad turnout for Tonga. Beats out the 960 easily.

Nvidia's branding is too strong tho, so this will get overlooked.

Tonga is one of the, if not the weakest part, AMD ever released in the last decade.
Everything about it is so poorly executed.

The timing, the SKUs, the performance, the energy-consumption, the driver issues with it.

 

[Ville Pukema]

yeah my numbers arent factual, i just cant fathom any logical reason why power draw would matter. the only rationale ive heard people throw out is "i save a few bucks a year on my power bill" which is ridiculous. its absolutely bias. ill never understand how people can be biased to a hardware company. heres a very recent comparison of the 960 to said cards as well as the new 380x.

http://www.eurogamer.net/articles/digitalfoundry-2015-amd-radeon-r9-380x-review

you would be a fool to purchase a 960.

I got 960 over 380X, would have been idiotic to buy 380X. Was over 140€ less, or would someone say 380X is worth 140€ more then 960?

 

[Irobot82]

I got 960 over 380X, would have been idiotic to buy 380X. Was over 140€ less, or would someone say 380X is worth 140€ more then 960?

In America both of those cards cost around $200-$250, so they compete against each other here.

 

[grumble]

Rebrand or not the R9 300 series are very good GPUs that match Nvidia's current offering in performance. The Fury line does not do as well as I thought.

I don't get the rant about rebrands, who gives a fuck as long as the performance is there and it is in the case of the 300 series. The 390 beats the 970 sometimes (Battlefront comes to mind), the 390x edges out the 980 in some of the latest AAA games.
I mean I don't see how this series is a disappointment, Nvidia did not reinvent the wheel with Maxwell. There are some cool new features which are not used anyway.

I don't get the doom and gloom in the GPU space, marketshare is not too hot but is it beyond hope ? I don't think so, PC gamers are not Nvidia drones.

DX12 is also on the horizon, we will see how it will alter the current state of affairs. Async compute (which as far as I understand is not really "doable" on Nvidia) vs Maxwell rendering features, it's safe to say async will be used more regularly as consoles support it as well.

Agreed, a lot of people are under the impression that nvidia cards are better. They aren't really, especially for the price point.

 

[General Lee]

We'll see how much Polaris will change the architecture, but I have no doubt it'll be more of an evolution rather than a revolution. GCN isn't bad, but AMD hasn't changed the basic setup that much in past years, so it's due some bigger changes and the new brand name is certainly an indicator of that. Of course you could be a cynic and think it's just another rebrand with no real changes.

If anything's a rebrand it's Pascal. According to rumors it's just Maxwell @16nm FF+ with Double Precision capabilities put back to make new Teslas. This doesn't paint a very promising picture in terms of improvement for gaming, but the new node will be enough to bring more performance regardless, and they can throw that 2x perf/w number just based on it alone.

I think for AMD this is the time to strike, and one can only hope Polaris will prevail and regain some of that lost trust in them. AMD has dug themselves a deep hole, and it's hard to climb back out of it even with good products. As long as Nvidia is at least close to equal in terms of perf/power/price, there's very little reason for people to switch. RTG really needs their own 970 success story, preferably without any 3.5 GB incident.

 

[Locuza]

Agreed, a lot of people are under the impression that nvidia cards are better. They aren't really, especially for the price point.

I would claim they are overall.
Nvidia executes fast and get it right.

Just to name the things, Nvidia had a robust Dev-Relationship program with the way is meant to be played, many years before ATI/AMD put a bunch of people in a similiar support group.

They knew where GPGPU where going, they established CUDA and enjoy most of the HPC marketshare.
They have the best OGL-Drivers, for Games, for Workstation Applications.
They got their 3D Vision stuff right.
Their Driver got solid frame-pacing for single and multi-GPUs.
They have many settings build in, for AO, pre-rendering-limit.
Shader-Cache, DSR and so on and so on.
And G-Sync with frame-doubling.

This company knows how to execute, how to implement, how to market it.
ATI/AMD are very often in the uncomfortable situation in catching up to Nvidia.

There are many solid reasons why Nvidia is the favorite child.

 

[Chumpion]

Polaris + AMD open-sourcing frenzy + Vulkan + Steam = YEAR OF LINUX ON THE DESKTOP?!!

 

[StardustPain]

I just sell my old 780ti, but I'll wait until the Vive came out to buy a new one.., so I'll make sure I choose the most VR ready, but as many of you say, I would like to support AMD

 

[StardustPain]

Polaris + AMD open-sourcing frenzy + Vulkan + Steam = YEAR OF LINUX ON THE DESKTOP?!!

You're missing the games.., steam alone doesn't do much =(.
(I use mac a lot, and most of the games aren't there)

 

[Kezen]

I would claim they are overall.
Nvidia executes fast and get it right.

Just to name the things, Nvidia had a robust Dev-Relationship program with the way is meant to be played, many years before ATI/AMD put a bunch of people in a similiar support group.

They knew where GPGPU where going, they established CUDA and enjoy most of the HPC marketshare.
They have the best OGL-Drivers, for Games, for Workstation Applications.
They got their 3D Vision stuff right.
Their Driver got solid frame-pacing for single and multi-GPUs.
They have many settings build in, for AO, pre-rendering-limit.
Shader-Cache, DSR and so on and so on.
And G-Sync with frame-doubling.

This company knows how to execute, how to implement, how to market it.
ATI/AMD are very often in the uncomfortable situation in catching up to Nvidia.

There are many solid reasons why Nvidia is the favorite child.

A similar case could be made for AMD which historically were the first to introduce DX9 (9700 pro if memory serves), GDDR5, HBM, unified shaders, or low level APIs for PC.

 

[Dream_Journey]

Good. Hurry up and give me something I won't regret buying. I really want to support AMD.

Yea, same.

 

[Jebusman]

http://www.eurogamer.net/articles/digitalfoundry-2015-amd-radeon-r9-380x-review

Not a bad turnout for Tonga. Beats out the 960 easily.

Nvidia's branding is too strong tho, so this will get overlooked.

This doesn't shock me when the 380x is a 190W part and the GTX 960 is a 120W part.

If it "didn't" blow the GTX 960 out of the water with that much more power available to it, it would've been a travesty.

 

[rambis]

The last time AMD changed shaders it was from VLIW to GCN which was a huge departure. So why are people trying to shrug this off as nothing? I can see if this was just them announcing a new line of cards but I seriously doubt they are rebranding anything here.

 

[Locuza]

A similar case could be made for AMD which historically were the first to introduce DX9 (9700 pro if memory serves), GDDR5, HBM, unified shaders, or low level APIs for PC.

If it really brought an edge against the competition then of course.
The r300 is often named legendary, because ATIs Architecture was very solid and Nvidia screwed up.
GDDR5 was a great help for AMD, they got very competitive with a way smaller chip.

But HBM is little help for them.
Besides the small form factor with nano, they can't harness any advantage against the competition.

Mantle was potentielly a great exklusive point for AMD, but it didn't stick for long.

 

[LiquidMetal14]

The last time AMD changed shaders it was from VLIW to GCN which was a huge departure. So why are people trying to shrug this off as nothing? I can see if this was just them announcing a new line of cards but I seriously doubt they are rebranding anything here.

People like to shrug off AMD. I want them to thrive like nvidia.

These graphics card comments are similar to gsmvoy wars at times.

I've been nvidia for my last 3 builds and also on the 3 PC's I've built on the side. I still am eager to see AMD bring the heat to nvidia.

 

[KillaKalz347]

Everyone wants to "support AMD" but how many of you would actually buy their cards? You'd surely find an excuse like the power draw which is literally meaningless, when you look at the way Nvidia misleads you with their power consumption, or drivers which AMD is making great gains in.

 

[Teletraan1]

Anecdotal indeed, in my office I'm the only one with a NVIDIA gpu.

If AMD manages to produce a GPU that beats whatever NVIDIA offers, I'll be jumping ship back to the red camp.

Do you work in an office in Canada because I have found that a lot of Canadians still have some sort of strange brand loyalty to ATI->AMD.

 

[Locuza]

Everyone wants to "support AMD" but how many of you would actually buy their cards? Your surely find an excuse like the power draw which is literally meaningless, when you look at the way Nvidia misleads you with their power consumption, or drivers which AMD is making great gains in.

I would, if I had confidence in the product I'm buying.
If not, I will never buy the worse product, just because I would like to support a company which is not directly more or less human to me, than the other.

And it's not like energy-consumption, drivers and so on, are an "excuse".
These are valid points for many people, in many different markets.
So AMD has no choice than to improve and get out of the hole.
It's of course incredible difficult for them, since they are in this vicious circle.

 

[The Hermit]

I switched team recently because I was gifted a Freesync monitor. Only later I realized it wasn't the wisest decision ( not because of the card, but because Freesync wasn't as good as I though). Still, I hope AMD get their shit together.

 

[tuxfool]

The last time AMD changed shaders it was from VLIW to GCN which was a huge departure. So why are people trying to shrug this off as nothing? I can see if this was just them announcing a new line of cards but I seriously doubt they are rebranding anything here.

Because they aren't going to abandon GCN. Contrary to popular opinion GCN is a very good architecture (ignoring the fact that they'll want some compatibility with GCN 1.x). It will serve as the basis for any future iteration. This is just like Nvidia, their current architecture can be traced all the way back to Tesla.

 

[rambis]

Because they aren't going to abandon GCN. Contrary to popular opinion GCN is a very good architecture. It will serve as the basis for any future iteration. This is just like Nvidia, their current architecture can be traced all the way back to Tesla.

I have no doubt that whatever this is will be based off GCN. That doesn't make this a rebrand. People should also realize that there is no source that this will be in Artic Islands. The original leaker only posted this picture and then shortly removed it.

 

[Lactose_Intolerant]

Everyone wants to "support AMD" but how many of you would actually buy their cards? Your surely find an excuse like the power draw which is literally meaningless, when you look at the way Nvidia misleads you with their power consumption, or drivers which AMD is making great gains in.

I just got a 390. It seemed like a better buy to me than a 970 going by fps. It is just huge and draws more power.

Until I need to buy a new monitor, it is fine for me.

 

[Thorgal]

this is pretty much AMD 's last chance .

if they cannot sway people away from Nvidia with these cards in 2016 they are done , Nvidia will have their monopoly of the GPU market in all but word just as Intel has an monopoly on the CPU market in all but word .

 

[Akronis]

https://www.youtube.com/watch?v=ZcF36_qMd8M

Some factual errors, but overall Nvidia is a jerk.

Some factual errors? More like dude has a huge fucking bias.

https://www.reddit.com/r/nvidia/comments/3xulmg/nvidia_stop_being_a_dick_by_2kliksphilip/cy7zzdy

This video is basically like saying that NVIDIA sabotaged AMD by using tessellation for everything, when AMD just straight up didn't bother fixing their shitty DX11 drivers.

 

[dr_rus]

I have no doubt that whatever this is will be based off GCN. That doesn't make this a rebrand. People should also realize that there is no source that this will be in Artic Islands. The original leaker only posted this picture and then shortly removed it.

Whether this will or won't be a rebrand will depend on how much changes there will be in Polaris compared to GCN. Considering AMD's financial conditions and their continuing reduction in R&D expenditures I'm not expecting much but I do hope that they will at least improve their perf/watt with Polaris not only because of a new process.