AMD Polaris architecture – GCN 4.0

[flopper]

I'm not interested in upgrading to the same chip as I have but on a smaller process, so Polaris offers me nothing of interest.

Well check what amd did with the 290 and tweaked it into a 390 series which matured and such was a upgrade in spite of a lot of people saying it wasnt before it was out. shame on them.

Polaris offers a lower power draw and likely better OC ability and the recent features of async compute and dx12 with DP 1.3 along the way.
I dont buy high end entusiast cards, and I am a fan of gpu´s but the price points are simply awful much like what Intel recently did with their cpu added a ton of more price to it and the same time laid off 20000 people.

I just wait until amd present their line up.

The paper launch with Nvidia PR, none took notice that they didnt talk, Async compute in hardware??? why? maybe....

 

[Steampunk]

Ellesmere XT is more than just Hawaii/Grenada on a smaller process. Even if it has similar number of CU's, they have added many performance enhancing features that will make it a faster part at the same clocks. And that is before any overclocking headroom added by the new node and finfetts being taken into account.

Ellesmere XT has nothing to worry about against the 1080, it will clock up and match or exceed it at a lower price point.

It is like people saying that Baffin XT will be 950 performance, Baffin XT has Pitcairn numbers of CU's and Pitcairn already has a lead on the 950 by a decent margin. And that is before any of the GCN2-4 enhancements or clock improvements from the node shrink, considering Pitcairn is still GCN1 vs Maxwell2 950.

Probably in an ideal world when using DX12. What about the games I'm playing now and the ones I'll be playing in the next year or two? For that I'm going to need more than a midrange part to make an upgrade from my 290 worthwhile.

It's all very well saying Polaris 10 is smaller, more efficient and fairly cheap, but if it's only giving me 10-15 percent improvement on current games, it's not worth spending a few hundred pounds on it. If I'm waiting for devs to do DX12 properly some time down the line (ie "jam tomorrow") to get real benefit, then I might as well wait to see what Vega brings.

It's just not sensible for me to spend money and move down the AMD product stack just because it's the only thing that AMD have to sell me. At the same time Nvida is doing the song and dance about how their new high end product are coming and raising expectations, where AMD looks like they have little faith in their products because they won't stand up and be proud of them.

Being "more efficient" is simply not impressive unless you use that efficiency to build more graphical power into your products, just like simply dropping a process node is no benefit unless you build a better/faster/cheaper product with it. AMD needs to start pushing the high end too, not just the midrange.

 

[CAT-THE-FIFTH]

Probably in an ideal world when using DX12. What about the games I'm playing now and the ones I'll be playing in the next year or two? For that I'm going to need more than a midrange part to make an upgrade from my 290 worthwhile.

It's all very well saying Polaris 10 is smaller, more efficient and fairly cheap, but if it's only giving me 10-15 percent improvement on current games, it's not worth spending a few hundred pounds on it. If I'm waiting for devs to do DX12 properly some time down the line (ie "jam tomorrow") to get real benefit, then I might as well wait to see what Vega brings.

It's just not sensible for me to spend money and move down the AMD product stack just because it's the only thing that AMD have to sell me. At the same time Nvida is doing the song and dance about how their new high end product are coming and raising expectations, where AMD looks like they have little faith in their products because they won't stand up and be proud of them.

Being "more efficient" is simply not impressive unless you use that efficiency to build more graphical power into your products, just like simply dropping a process node is no benefit unless you build a better/faster/cheaper product with it. AMD needs to start pushing the high end too, not just the midrange.

Well you do realise there are big changes to the way tessellation is done in Polaris and a new command processor too which will improve DX11 performance??

Maybe AMD is worried about Polaris? Maybe they cannot beat a Fury and its GTX1060 level? Or it could be another HD4870. ATI waited weeks after Nvidia launched the GTX280.

Within a month Nvidia had to cut prices very quick on a card with a massive chip. Remember if Polaris even manages to get close to GTX1070,Nvidia is using a 40% larger chip.

After all we have seen multiple public demos of the cards,the last just a few weeks ago in Corsair Air 240 cases.

The performance predictions even I made of it being r9 390 to Fury level also are on the low side since Polaris 10 should have more transistors than Hawaii and also not have as much DP performance.

OFC,GF could have ***** up and clockspeeds are subpar - that is my main worry.

 

[Seanspeed]

Why? NV fanboys have to buy a 10x0 to beat the FuryX at DX12 and stop their tears running over the last few months. Also they have to avoid the natural driver degradation of their one year old 980Ti. Something that doesn't apply to the AMD products (aka see the 290X performance nowadays)

To understand their IQ see them right now on the other threads. They even celebrate that NV supports Vulcan on DOOM and some say "AMD sucks because they are stick to DX12". Is futile trying to explain that Vulcan is basically Mantle in it's core and fully supported by AMD. But after all those decades, I do not care any more. Let them spend their money on the new 10x0 and then in 6 months, another amount of the next NV product, trying to keep up with times.

Closing, explains a lot that NV is something like religion to them, because the new NV cards going to support DP1.4/1.3 which means on hardware level the vesa adaptive sync is there, but not supported by their drivers, just to milk them like cows on Gsync modules.
Any other company in electronics or not, following such tactic, should have been crucified publicly by them.

You sound as bad or worse than anybody you think you are criticizing here man. Total 'us vs them' condescending perspective that only serves to further reinforce polarizing attitudes.

 

[CAT-THE-FIFTH]

You sound as bad or worse than anybody you think you are criticizing here man. Total 'us vs them' condescending perspective that only serves to further reinforce polarizing attitudes.

TBF,if you look a the threads on Vulkan,you could see all the argueing since it was Mantle based. Now a full 180 and it's the best thing in the world.

The irony.

 

[andybird123]

TBF,if you look a the threads on Vulkan,you could see all the argueing since it was Mantle based. Now a full 180 and it's the best thing in the world.

The irony.

I'm still of the same opinion on Vulkan, i dont think we will see masses of games using it.

The argument over vulkan wasnt whether it was good or not, the main argument was that it was reskinned mantle (its not) and that AMD would benefit more from it (because it was reskinned mantle)

 

[Seanspeed]

TBF,if you look a the threads on Vulkan,you could see all the argueing since it was Mantle based. Now a full 180 and it's the best thing in the world.

The irony.

I'm not defending any faulty reasoning on behalf of certain fanboys by any means(though there's more to it than he was saying). It's just annoying to see somebody acting like a total political party campaigner talking about 'those people' and how they're all the silly ones, not me and my people. 'Us vs them' type of crap.

 

[Mauller]

Probably in an ideal world when using DX12. What about the games I'm playing now and the ones I'll be playing in the next year or two? For that I'm going to need more than a midrange part to make an upgrade from my 290 worthwhile.

It's all very well saying Polaris 10 is smaller, more efficient and fairly cheap, but if it's only giving me 10-15 percent improvement on current games, it's not worth spending a few hundred pounds on it. If I'm waiting for devs to do DX12 properly some time down the line (ie "jam tomorrow") to get real benefit, then I might as well wait to see what Vega brings. .

GCN4 has many improvements to single threaded and multi threaded shader utilisation. Then it also has the Primitive discard accelerator on top to help reduce the load on the shaders and geometry processors as well.

So before any clock differences the GCN4 parts will be faster at the same clocks.

 

[TheRealDeal]

I'm still of the same opinion on Vulkan, i dont think we will see masses of games using it.

The argument over vulkan wasnt whether it was good or not, the main argument was that it was reskinned mantle (its not) and that AMD would benefit more from it (because it was reskinned mantle)

The main win for AMD having Mantle in Vulkan should be that they should perform well in games using it. AMD with open GL were behind Nvidia. So now AMD should be strong on all gaming Api's like Nvidia are, assuming people start using Vulkan instead of Open GL.

 

[Simon Belmont]

I'm guessing that's why we haven't seen much of Drunkmaster recently.

Cause he thinks he's a knows it all. Like a few other members here.

 

[Steampunk]

GCN4 has many improvements to single threaded and multi threaded shader utilisation. Then it also has the Primitive discard accelerator on top to help reduce the load on the shaders and geometry processors as well.

So before any clock differences the GCN4 parts will be faster at the same clocks.

How much faster? In what kinds of games? What are those clocks going to be? At what prices?

We just don't know because AMD aren't saying and the only narrative out there is controlled by rumours and Nvidia's latest launch. AMD need to start managing their customers' expectations before they lose yet more market share to "new and shiny" from their competitors.

 

[Seanspeed]

How much faster? In what kinds of games? What are those clocks going to be? At what prices?

We just don't know because AMD aren't saying and the only narrative out there is controlled by rumours and Nvidia's latest launch. AMD need to start managing their customers' expectations before they lose yet more market share to "new and shiny" from their competitors.

If Polaris 10 is actually going to compete with GP104, AMD still has a few weeks to get some decent info out about it to sway customers to not buy Nvidia just yet.

I do expect lots of DX12 benches, though....

 

[Steampunk]

If Polaris 10 is actually going to compete with GP104, AMD still has a few weeks to get some decent info out about it to sway customers to not buy Nvidia just yet.

I do expect lots of DX12 benches, though....

I just have a suspicion that all this concentration on Polaris 10 efficiency means that instead of faster chips in the midrange, it's going to be smaller, cheaper chips, more profit for AMD, but performance capped at this "60FPS VR minimum". ie "more efficiency = less transistors = smaller chips & good enough performance". We will have to wait for Vega to get "more efficiency = more transistors packed in = more graphics power & exceptional performance".

This would not be an issue if Vega launched at the same time, but it's at least half a year away, all in the face of Nvidia coming in at the high end already. To me, it's like AMD is only talking about the mid/low end of their product stack, and the high end is a long way off. With Nvidia's high end announcements, they will steal the halo marketing and again be seen to be ahead and better purely because they got their message (and likely products) out first.

It frustrates me no end that AMD stick with this philosophy of keeping their customers in the dark, letting the rumour mill work overtime, and then leaving their customers disappointed. It's been failing for years, and yet AMD are still stuck in the same situation with Nvidia running rings ariound them. It's all the more annoying because AMD GPUs are actually very good, and their products deserve to be supported and advertised better than this. Their existing and potential customers certainly deserve better too, if AMD want people to put their hands in the pockets and give AMD their money.

 

[Panos]

The paper launch with Nvidia PR, none took notice that they didnt talk, Async compute in hardware??? why? maybe....

I was looking at their posted data and indeed couldn't see that.

It will be funny if there is not hardware level support for async compute.

 

[SiDeards73]

Lol @ people expecting AMD to beat the 1070 on price and performance, it's not going to happen, they won't be releasing a 980ti beater that's £280, just is not going to happen, I expect their card that bears the 980ti to be around £400 and still be worse than the 1080, and not arrive til 2017

I hate the fact I'm going to have to buy an Nvidia card and a gsync monitor as my only alternative will be to give up pc gaming at the high end.

I'm an ideal world AMD deliver a truly earthshattering card that is not only priced well but delivers on all fronts, performance, power draw and low heat.

Unfortunately I look at past launches like the 290, 390 and the Fiji stuff and think that people in charge of the critical decisions at AMD don't know their butts from their elbows

 

[Stanners]

I think maybe AMD are going for the low-mid sector with volume sales(Polaris). While it may not be good for people in the high end market it is where the money is.

If they could shift a lot of units in the low to mid markets it would bring the revenue in and enable them to start fighting in the high end(Vega), plus it would claw market share back. Of course this is all my own opinion and maybe miles off the mark

 

[Steampunk]

I was looking at their posted data and indeed couldn't see that.

It will be funny if there is not hardware level support for async compute.

It wouldn't be the first time Nvidia have quietly moved stuff off the GPU and onto the CPU. More profit for them, and it's not like their customers care, especially if Nvidia pay devs not to use async, or hide all this lack of hardware support inside Gameworks.

 

[Steampunk]

I think maybe AMD are going for the low-mid sector with volume sales(Polaris). While it may not be good for people in the high end market it is where the money is.

If they could shift a lot of units in the low to mid markets it would bring the revenue in and enable them to start fighting in the high end(Vega), plus it would claw market share back. Of course this is all my own opinion and maybe miles off the mark

How did that work out for AMD up against Nvidia in the 290/390/Fury generation of products?

 

[Stanners]

How did that work out for AMD up against Nvidia in the 290/390/Fury generation of products?

290 and fury were both high end cards, nor sure what you are asking?

 

[LoadsaMoney]

I don't wish to go without freesync but I will if I have to.

Im going to do without it, as i have to go Pascal me, only option, i can't wait until sometime next year for an upgrade from AMD.