AMD have begun shipping Polaris GPUs - 2 different Polaris chips and 4 different GPUs spotted

[Humbug]

AMD shows of 14nm Polaris desktop GPU at CES

As expected the power efficiency is far better than Nvidia Maxwell and looks set to go head-to-head with Pascal.

Higher end models will use HBM, lower end models will use GDDR5.

 

[Notional]

 It would make sense that they will produce the lower end cards at GlobalFoundries so they can release them earlier then if they made them at TSMC which will be swamped with Nvidia as well.

 

 My guess is that they will release a lower end GPUs to showcase a new architecture and get some advantage in portable devices and at the later date they'll release high end stuff from TSMC who are used to producing GPUs that complex.

 

I personally doubt we will see any 16nm TSMC produced AMD products. AMD has probably secured all of Glofo's 14nm FF production, and we know that Samsung will produce some of the products too (probably the low end/mobile parts).

 

That's what makes everything so interesting this year. Not only do we get 2 node shrinks, HBM and DP 1.3. We get two different manufacturers for AMD and NVidia for the first time in forever.

 

Pretty sure x86 CPU's are a lot more complex to produce than GPU's. So GloFo would have a leg up there. However I think it has more to do with the actual node/way of producing the chip that matters.

[Jahramika]

AMD had the Crimson drivers mess with fan speeds and gpu's were overheating. That's not good either.  That was recent.

That was fixed within a week. 

[Fetzie]

That was fixed within a week.

Within a day.

[Sakkura]

 It would make sense that they will produce the lower end cards at GlobalFoundries so they can release them earlier then if they made them at TSMC which will be swamped with Nvidia as well.

 

 My guess is that they will release a lower end GPUs to showcase a new architecture and get some advantage in portable devices and at the later date they'll release high end stuff from TSMC who are used to producing GPUs that complex.

 

Plus a smaller GPU is easier to pull off with regard to yields.

[TechGod]

Why are you being so defensive and agressive?

I'm saying the truth. The people over here pick on Apple fanboys and pick on them specifically.

[Big-G]

Why would reference and non-reference make a difference? Put either PCB under a water block and compare.

Non-Reference PCB's usually have improved designs in terms of power delivery and components

[Nodyi]

For example they don't try to stop people from overclocking their mobile GPU and then pull a 360 after consumer backlash. Note its a 360 not a 180

Inb4 someone defends Nvidia like the sheep who defend Apple.

 

As a consumer, yeah, that is shitty but in practice it means very little since we are talking about highly volatile computers that shouldn't be overclocked at all. Though they have had other moves like this (Like GSYNC not being open to AMD cards and asking for $$$),things like this is why Nvidia and Apple are good companies for their shareholders/owners and AMD,today, is a clusterfuck being untangled that hopefully might finally catch up. Yeah, I like that AMD has open-source technologies. I don't like Nvidia gameworks clusterfucks. None of these mentioend arguments for either side, don't detract from the fact that AMD GPUs are absolute crap for laptops. They are non-existant on that market. If you have a new AMD APU+GPU (cool technology, sadly it is still lag behind a single uniform Nvidia mobile GPU), you either wasted way too much money or caught a killer deal. Most likely the former one. So you can bash and talk shit all you want, but that isn't something to be taken lightly. 

 

And in the desktop world of GPUs, they are competetive, but with a net loss on their investment. Which to be honest, to us, as customers shouldn't mean much. This was their conscious decision to stay in the market somehow because they are getting hammered in other areas. They are also non-existant in the higher tiers of desktop CPUs. If AMD was competetive in all the mentioned markets, and not just lower-to-medium CPU market and desktop-GPUs, and with a profit for its owners, Nvidia would be forced to up its game. Instead AMD rebrands old CPUs at lower prices and Intel ups every generation by exactly 10% because they control the market. So, you get to pay a premium for every little improvement. Oh, well. Better go bash on the Internet while knowing next to nothing.

 

Oh and the whole Apple bashing things on this forum. It is almost as if my Apple stocks rises everytime somebody bashes them, while all the other companies are still oblivious how exactly they got their asses handed over by them. And all that while having a fuck-the-consumer policy. Brilliant. 10 points to Gryffindor.

[Clyne]

It just wouldn't make sense for NVIDIA to intentially gimp Kepler. If they got caught, the least worst thing is that massive amounts of consumers won't buy their things for a very long time. The worst result is that they get sued and have to pay potentially billions of dollars. Seeing how everyone, their dogs, and their puppies got a GTX 900 series card, it wasn't exactly like NVIDIA was hurting for sales.

There isn't any proof that Nvidia is gimping kepler, we could say that their not as focused on Kepler as they are on Maxwell.

[VagabondWraith]

There isn't any proof that Nvidia is gimping kepler, we could say that their not as focused on Kepler as they are on Maxwell.

Fanboys like to parade around nonsense and FUD. It's best left for the absent minded.

[Mahigan]

It just wouldn't make sense for NVIDIA to intentially gimp Kepler. If they got caught, the least worst thing is that massive amounts of consumers won't buy their things for a very long time. The worst result is that they get sued and have to pay potentially billions of dollars. Seeing how everyone, their dogs, and their puppies got a GTX 900 series card, it wasn't exactly like NVIDIA was hurting for sales.

NVIDIA didn't gimp Kepler, countless reviews have proven that. What has happened is that video games have become, slowly but surely, more compute oriented over time. GCN is a better compute architecture. So GCN has gotten progressively better over time.

This trend will continue throughout 2016 where we should see a FuryX outpace a GTX 980 Ti. Who cares though, as most people will be sporting Polaris or Pascal based GPUs.

[ivan134]

NVIDIA didn't gimp Kepler, countless reviews have proven that. What has happened is that video games have become, slowly but surely, more compute oriented over time. GCN is a better compute architecture. So GCN has gotten progressively better over time.

This trend will continue throughout 2016 where we should see a FuryX outpace a GTX 980 Ti. Who cares though, as most people will be sporting Polaris or Pascal based GPUs.

Kepler is better than Maxwell at compute.

[Clyne]

Fanboys like to parade around nonsense and FUD. It's best left for the absent minded.

Thing is, Kepler is better than Maxwell IMO anyway.

Once support for maxwell ends then it wouldn't perform well as it would have.

Cutting down the GPU and using software optimization...since Fermi is a "hardware" card, it still stands up today.

[VagabondWraith]

Thing is, Kepler is better than Maxwell IMO anyway.

Once support for maxwell ends then it wouldn't perform well as it would have.

Cutting down the GPU and using software optimization...since Fermi is a "hardware" card, it still stands up today.

Maxwell is fine. It does what nVidia focused on it to do: gaming. Its cool, its efficient and overclocks like a champ. And that's all that matters.

[Clyne]

Maxwell is fine. It does what nVidia focused on it to do: gaming. Its cool, its efficient and overclocks like a champ. And that's all that matters.

I'm just going to leave it there before any more "heated" debates come up. 

[VagabondWraith]

I'm just going to leave it there before any more "heated" debates come up.

I'm not debating, nor looking for an argument. Just stating my opinion.

[Kendrick]

R9 390>GTX 970>R9 380X>R9 280X>R9 380>GTX 960

[Blake]

Classic fanboy. Says the guy with a Radeon card avatar. lmao what a hoot.

And I'm still waiting on proof. Not a bunch of conjecture and hypotheticals. Just pure, solid proof.

1. Just be quiet. You asked for proof, there was a youtube video (or two) posted, and you claimed 'cherry picked'. Stock v Stock 980ti v fury X is equal. Overclocked 980ti does win (and normally factory overclocks are higher/actually exist with the 980ti). The burned of proof is now on YOU, as YOU claimed the benchmarks provided where cherry picked. If you are unable to find any others that have recent drivers, you have every right to remain skeptical, just don't go acting like a monkey and start throwing your shit everywhere.

 

2. Claims bias for having Radeon Avatar, yet has "Intel / nVidia Enthusiast" as forum title. "Your a hypocrite, 'arry".

 

3. If you want to be taken seriously, refute the central point of other arguments (https://upload.wikimedia.org/wikipedia/commons/7/7c/Graham%27s_Hierarchy_of_Disagreement.svg).

[Sakkura]

NVIDIA didn't gimp Kepler, countless reviews have proven that. What has happened is that video games have become, slowly but surely, more compute oriented over time. GCN is a better compute architecture. So GCN has gotten progressively better over time.

This trend will continue throughout 2016 where we should see a FuryX outpace a GTX 980 Ti. Who cares though, as most people will be sporting Polaris or Pascal based GPUs.

 

 

Kepler is better than Maxwell at compute.

 

Yeah. If anything, it's that games have put more emphasis on tessellation. Where GCN IIRC beats Kepler, but Maxwell beats GCN. That's also why some people suspect Nvidia is intentionally influencing developers to put more tessellation in their games, even where it doesn't improve the visuals.

 

Just like with Crysis 2, back when Fermi was better at tessellation than VLIW5/VLIW4.

[Tomsen]

Yeah. If anything, it's that games have put more emphasis on tessellation. Where GCN IIRC beats Kepler, but Maxwell beats GCN. That's also why some people suspect Nvidia is intentionally influencing developers to put more tessellation in their games, even where it doesn't improve the visuals.

 

Just like with Crysis 2, back when Fermi was better at tessellation than VLIW5/VLIW4.

Bingo.

[don_svetlio]

Yeah. If anything, it's that games have put more emphasis on tessellation. Where GCN IIRC beats Kepler, but Maxwell beats GCN. That's also why some people suspect Nvidia is intentionally influencing developers to put more tessellation in their games, even where it doesn't improve the visuals.

Just like with Crysis 2, back when Fermi was better at tessellation than VLIW5/VLIW4.

GCN 1.1 > Keller,> GCN 1.0 in terms of Tess performance. Keller is a better design than Maxwell

[ivan134]

Yeah. If anything, it's that games have put more emphasis on tessellation. Where GCN IIRC beats Kepler, but Maxwell beats GCN. That's also why some people suspect Nvidia is intentionally influencing developers to put more tessellation in their games, even where it doesn't improve the visuals.

 

Just like with Crysis 2, back when Fermi was better at tessellation than VLIW5/VLIW4.

This is essentially what I mean when I and some others say Kepler is being gimped, but for some reason people translate it as us saying Nvidia is changing code in the drivers to make Kepler run worse. Just about every GameWorks game since the launched of Maxwell is tessellated up the ass with no actual visual improvement from using such high levels of tessellation. A combination of that and not bothering with performance optimization for anything below the 780. It's insane how badly the 770/680 and 760 perform today, but apparently it doesn't matter since apparently buying a new card every generation is now considered normal.

[Castdeath97]

Great to see AMD closing to the release date, considering the current rate they might be able to release the new GPUs before or just around the same time as Nvidia.

 

Yup the 980ti > Fury X bandwagon ended 3 months ago. Just like how damn near everyone reccodmended a CM 212 Evo a month ago. Maybe 2 people and I would reccodmend a PURE ROCK or cryorigH7

 

Well the evo cheaper isn't it?

[VagabondWraith]

1. Just be quiet. You asked for proof, there was a youtube video (or two) posted, and you claimed 'cherry picked'. Stock v Stock 980ti v fury X is equal. Overclocked 980ti does win (and normally factory overclocks are higher/actually exist with the 980ti). The burned of proof is now on YOU, as YOU claimed the benchmarks provided where cherry picked. If you are unable to find any others that have recent drivers, you have every right to remain skeptical, just don't go acting like a monkey and start throwing your shit everywhere.

2. Claims bias for having Radeon Avatar, yet has "Intel / nVidia Enthusiast" as forum title. "Your a hypocrite, 'arry".

3. If you want to be taken seriously, refute the central point of other arguments (https://upload.wikimedia.org/wikipedia/commons/7/7c/Graham%27s_Hierarchy_of_Disagreement.svg).

If you don't understand the context of the discussion between me and another poster, please excuse yourself and don't butt in others peoples business.

Thank you.

[Humbug]

There isn't any proof that Nvidia is gimping kepler, we could say that their not as focused on Kepler as they are on Maxwell.

Ya a lot of Nvidia users are convinced it is happening but I don't think so (not an Nvidia user though).

It's just that modern AAA games are very reliant on game specific driver optimizations for maximum performance. Nvidia's driver team has been focused on the GPUs which they are currently selling (maxwell).

Whereas In AMD's case of continuous GCN revisions (and some rebrands) even people who purchased first gen GCN cards are still enjoying game specific driver optimizations for the newest games (even if it takes a few more days).

 

So this ends up with people asking questions like if the GTX 680 used to be on par with a HD 7970 in 2011 games then why isn't it the same for 2015 games?

After the witcher 3 fiasco and mass protests on the support forum though Nvidia did do some improvements to Kepler performance, so the situation has improved actually..

 

One of the objectives of DX12 and Vulkan is to become less reliant on game-specific driver tuning. Here's to hoping...