AMD Confirms RX 480 At $199 USD, Other APU & Polaris Announcements

[Fl0r!an]

AMD gets a large improvement going from DX11 to DX12 because their DX11 driver is pretty crap. AMD gets a large improvement going from OpenGL to Vulkan because their OpenGL driver is pretty crap

While it's true that AMDs OpenGL drivers are performing rather subpar (just look at Doom benchmarks), one has to admit that the performance in modern APIs is really awesome.

I realized that when I swapped in my old R9 280: It outperforms my GTX 780 (which is usually MUCH faster) by at least 20% in Doom (Vulkan)! Quite impressive to see >60FPS@1440p in a modern game from such a old card. A GK104 card (direct competition back in the days) is getting nowhere near that.

 

[xsmi123]

I SPECIFICALLY said Vulkan and DX 12

$450 vs. $240? Ah, NO!!

 

[koyoot]

http://semiaccurate.com/2016/08/31/amd-goes-7nm-new-global-foundries-agreement/

Today is the day when there is also mass production starting of 14 nm HP. This is a process which combines FinFET and SOI technology, and is around 20% better than LPP, however it is isnt mostly 20% one way or another. It delivers 10% lower power consumption for 10% higher frequency.

Secondly. This is most likely the process on which respin of Polaris architecture can happen, and AMD already touted this with their naming scheme. RX 4X5 GPUs.

 

[koyoot]

http://arstechnica.com/gadgets/2015/07/ibm-unveils-industrys-first-7nm-chip-moving-beyond-silicon/
http://www.fudzilla.com/news/processors/38193-ibm-delivers-7nm-to-gloflo

To add to the previous post. This is what we will see in 2018/2019.

 

[ManuelGomes]

Maybe this will help Polaris with power consumption and clear the matter for once.

 

[Stacc]

It looks like AMD is trying to move some of their products away from Global Foundries. It seems they are having a hard time keeping their process competitive with Intel and TSMC. It looks pretty clear at this point that Polaris was manufactured at Global Foundries due to it being cheaper for AMD and not because it results in faster/more efficient products.

 

[koyoot]

It looks like AMD is trying to move some of their products away from Global Foundries. It seems they are having a hard time keeping their process competitive with Intel and TSMC. It looks pretty clear at this point that Polaris was manufactured at Global Foundries due to it being cheaper for AMD and not because it results in faster/more efficient products.

Yeah, cheaper. And 95% of industry is manufacturing on TSMC, because it is more expensive? Sometimes people post things on this forum without thinking it through... GloFo has only 2 clients: AMD and Qualcomm. TSMC: Apple, AMD, Nvidia, Qualcomm, and 3 other clients. There are resons for this, and are not fully related to "technological advantage".

AMD is not moving away from GloFo. Zen will be produced on the same process as Power9 in the same fab nr 8 in Malta, NY. Mass production of final silicon started today.

If the reason why Apple would go for manufacturing their products at specific OEM was technological advantage they would ALWAYS have picked Intel, before anyone. TSMC has the lowest prices in the industry, because they go for volume.

http://semiaccurate.com/2016/09/01/intel-finally-narrows-14nm-process-technology-gap-samsung/

 

[Stacc]

Yeah, cheaper. And 95% of industry is manufacturing on TSMC, because it is more expensive? Sometimes people post things on this forum without thinking it through... GloFo has only 2 clients: AMD and Qualcomm. TSMC: Apple, AMD, Nvidia, Qualcomm, and 3 other clients. There are resons for this, and are not fully related to "technological advantage".

AMD is not moving away from GloFo. Zen will be produced on the same process as Power9 in the same fab nr 8 in Malta, NY. Mass production of final silicon started today.

If the reason why Apple would go for manufacturing their products at specific OEM was technological advantage they would ALWAYS have picked Intel, before anyone. TSMC has the lowest prices in the industry, because they go for volume.

http://semiaccurate.com/2016/09/01/intel-finally-narrows-14nm-process-technology-gap-samsung/

95% of what industry? This article states that TSMC has 12% market share. There is evidence that TSMC's 16nm process is more efficient than Global Foundries 14nm process. Nvidia saw much better gains in efficiency moving from TSMC's 28nm process to their 16nm process than AMD did going from TSMC 28nm to Global Foundries 14nm. When Apple had its A9 (I think thats the right number) produced at both fabs, the TSMC one had better efficiency and performance.

It looks clear to me that if you are trying to produce large, high performance dies you should be on TSMC. If Zen is produced at Global Foundries it will be at a significant process disadvantage compared to Intel. I think its been already confirmed that Vega will be produced at TSMC for similar reasons.

Intel won't be able to produce ARM chips until its 10nm process comes online. Thus, Apple has been limited to choosing between Global Foundries and TSMC.

 

[koyoot]

95% of what industry? This article states that TSMC has 12% market share. There is evidence that TSMC's 16nm process is more efficient than Global Foundries 14nm process. Nvidia saw much better gains in efficiency moving from TSMC's 28nm process to their 16nm process than AMD did going from TSMC 28nm to Global Foundries 14nm. When Apple had its A9 (I think thats the right number) produced at both fabs, the TSMC one had better efficiency and performance.

It looks clear to me that if you are trying to produce large, high performance dies you should be on TSMC. If Zen is produced at Global Foundries it will be at a significant process disadvantage compared to Intel. I think its been already confirmed that Vega will be produced at TSMC for similar reasons.

Intel won't be able to produce ARM chips until its 10nm process comes online. Thus, Apple has been limited to choosing between Global Foundries and TSMC.

What I mean is the amount of clients manufacturing on TSMC. Not how many chips particular Fab company can push out.

TSMC has larger overall capacity than GloFo. 80k Wafers/month vs 60K wafers/month. Which can be reflected in the numbers of chips produced by particular company.

I am not questioning the technological advancement. Read exactly what I have written. TSMC is chosen because it is the cheapest of the lot in the first place. That is what I have written. Apple chose TSMC for A9 production because... production of only A9X would not consume all of the capacity that Apple had to buy from TSMC in order to even start production. Thats why they split the production between the two companies.

We will see the end results to tell which one of the processes is better: Intel 14nm or IBM SOI/GloFo FinFET hybrid.

 

[ManuelGomes]

http://www.anandtech.com/show/10642/hdmi-alt-mode-for-usb-typec-announced

 

[Stacc]

The Steam hardware survey for august is out. This gives us some real data on well these new GPUs are selling to gamers. Lets pull out some of the new GPUs.

GPU - Release - Steam Usage Percentage
Nvidia GTX 1080 - 5/16 - 0.49%
Nvidia GTX 1070 - 6/16 - 0.71%
Nvidia GTX 1060 - 7/16 - 0.24%
AMD RX 480 - 6/16 - 0.11%
AMD RX 470 - 8/16 - 0.01%

Basically, Nvidia is doing very well. Despite the RX 480 being on the shelf for a few weeks longer than the GTX 1060, the 1060 is used more on steam by more than a 2:1 ratio. Its unclear from this data whether this is a supply problem from AMD or if Nvidia's product is just more popular. Nvidia is also dominating the high end sales, as they don't have any competition here.

 

[koyoot]

Only with Steam users. Which is small margin of computer users around the world. Bare that in mind, before you will draw any conclusions. Also the survey does not survey every single computer with Steam installed.

My primary gaming platform is Battle.net for example. I do not use Steam, at all. League of Legends is not on Steam. Only most popular Steam games are DOTA 2 and CS:GO. Which combined(37 mln, 25 mln DOTA 2, 12 mln CS:GO) are around active accounts on LoL(44 mln active accounts), and much smaller than active accounts on B.net(6 mln WoW, 9.5 mln Overwatch, 30 mln Hearthstone, 15 mln Heroes of the Storm, 12 mln Starcraft 2, 18 mln Diablo 3, with total of around 60 mln active accounts and players per day, around the world).

I have made a slight error. Steam has currently around 150 mln accounts. However, bare there in mind that again, not only every computer is surveyed, secondly, not everyone is using Steam as main gaming platform. Polish magazine CDAction offers full versions of games for around 5$ with the magazine, and they mostly require Steam to run. Overall market for gaming is much bigger than Steam.


Edit: http://wccftech.com/alienware-rx-470-laptop/ I have to say. I have not been expecting this. This is huge surprise.

 

[Stacc]

Only with Steam users. Which is small margin of computer users around the world. Bare that in mind, before you will draw any conclusions. Also the survey does not survey every single computer with Steam installed.

My primary gaming platform is Battle.net for example. I do not use Steam, at all. League of Legends is not on Steam. Only most popular Steam games are DOTA 2 and CS:GO. Which combined(37 mln, 25 mln DOTA 2, 12 mln CS:GO) are around active accounts on LoL(44 mln active accounts), and much smaller than active accounts on B.net(6 mln WoW, 9.5 mln Overwatch, 30 mln Hearthstone, 15 mln Heroes of the Storm, 12 mln Starcraft 2, 18 mln Diablo 3, with total of around 60 mln active accounts and players per day, around the world).

I have made a slight error. Steam has currently around 150 mln accounts. However, bare there in mind that again, not only every computer is surveyed, secondly, not everyone is using Steam as main gaming platform. Polish magazine CDAction offers full versions of games for around 5$ with the magazine, and they mostly require Steam to run. Overall market for gaming is much bigger than Steam.

Its not a perfect metric, but its about as good as you can get without getting direct access to sales numbers.

Regardless, its troubling that AMD's strategy was to take the mainstream market first, but is losing dramatically to Nvidia's GTX 1060. This is aided by the fact that they could only beat Nvidia to market by a few weeks and evidently were supply constrained.

 

[koyoot]

Do not draw any conclusions about marketshare based on only Steam survey numbers that does not survey ALL of the computers using Steam.

Your point is valid, however, as has been discussed in any gaming forum out there, it is just only small part of MUCH larger picture.

 

[Asgorath]

The Steam hardware survey gets a lot of data points (i.e. a good representative sample size). While the only way to know for sure would be to get full sales numbers, I don't think you can simply argue that the Steam results aren't accurate. You don't need to survey ALL Steam computers to get meaningful data that correlates with reality.

 

[lowendlinux]

Steam HW survey has never really been accurate

 

[WestonHarvey1]

They fixed it by offering a reduced performance "PCIe compliant" mode. A Band-Aid for a failed design.

You're talking about Compatibility Mode, which is the alternative fix that is not on by default. The standard fix just rebalances the power draw to pull less off the bus and more off the 6 pin connector. While it is still slightly out of spec in this mode there's no performance hit.

No RX 480 owner cares about this non-issue. Mine's doing great, PC hasn't caught fire, don't give a rats ass about a sometimes extra 15 watts of power consumption.

 

[ActionableMango]

Basically, Nvidia is doing very well.

Unless I'm mistaken, in the Post-Fermi era Nvidia/AMD has been about 3:1 ratio (Nvidia getting 75% of market share). So a 2:1 ratio (66% Nvidia) would actually indicate decreasing market share for Nvidia.

I generally pick Nvidia cards for my own use, but for healthy competition I'd actually like to see the two companies neck and neck.

 

[koyoot]

The Steam hardware survey gets a lot of data points (i.e. a good representative sample size). While the only way to know for sure would be to get full sales numbers, I don't think you can simply argue that the Steam results aren't accurate. You don't need to survey ALL Steam computers to get meaningful data that correlates with reality.

Have you ever considered things like in countries that are out of the richest ring of countries like USA, Australia, Canada, GErmany people rarely buy games? Have you ever considered places like India, China, African countries, Ukraine etc where people cannot afford GTX 1080/1070 GPUs?

As usual on gaming forums: Steam surveys are pushed around by fans of particular brands to show their favourite brand is better. No wonder the same thing will happen when AMD will have higher marketshare than Nvidia.

 

[tuxon86]

HIndia, China, African countries, Ukraine etc where people cannot afford GTX 1080/1070 GPUs?.

They can't afford $300 cards either. They mostly go with tghe iGPU solution or older hardware.

 

[Stacc]

As usual on gaming forums: Steam surveys are pushed around by fans of particular brands to show their favourite brand is better. No wonder the same thing will happen when AMD will have higher marketshare than Nvidia.

The reason I posted them is so we can stop making claims with no evidence to back them up. This is the first indication that we have of how successful these new GPUs from Nvidia and AMD are. Is it absolutely representative of the entire industry? No. Does it provide some insight into which cards are more successful? Yes. Is it better than using random message board postings or unsubstantiated claims? Yes.

 

[koyoot]

The reason I posted them is so we can stop making claims with no evidence to back them up. This is the first indication that we have of how successful these new GPUs from Nvidia and AMD are. Is it absolutely representative of the entire industry? No. Does it provide some insight into which cards are more successful? Yes. Is it better than using random message board postings or unsubstantiated claims? Yes.

Yeah, you post something, and then draw conclusion about whole situation by looking at one part of it. What is it? Making claims without eveidence to back them up.

Basically, Nvidia is doing very well. Despite the RX 480 being on the shelf for a few weeks longer than the GTX 1060, the 1060 is used more on steam by more than a 2:1 ratio. Its unclear from this data whether this is a supply problem from AMD or if Nvidia's product is just more popular. Nvidia is also dominating the high end sales, as they don't have any competition here.

This is only about steam. Yet, you claim it is reflection of WHOLE WORLD.

 

[Asgorath]

Have you ever considered things like in countries that are out of the richest ring of countries like USA, Australia, Canada, GErmany people rarely buy games? Have you ever considered places like India, China, African countries, Ukraine etc where people cannot afford GTX 1080/1070 GPUs?

As usual on gaming forums: Steam surveys are pushed around by fans of particular brands to show their favourite brand is better. No wonder the same thing will happen when AMD will have higher marketshare than Nvidia.

I'm not pushing the survey results, I'm simply stating that you can't just ignore them outright. Why are people buying an RX 480 or a GTX 1060 if they don't play games? So, for those people who play games and run Steam, there seems to be a 2:1 ratio of folks with the NVIDIA card versus ones with the AMD card. I'd be surprised if the overall sales aren't in that general ballpark.

I get that you're and AMD fan and you're entitled to your opinions. However, I think it's hard to argue that the Pascal GPUs aren't selling very, very well. You can enjoy your RX 480, because I'm really enjoying my GTX 1080.
[doublepost=1472843490][/doublepost]

Yeah, you post something, and then draw conclusion about whole situation by looking at one part of it. What is it? Making claims without eveidence to back them up.
This is only about steam. Yet, you claim it is reflection of WHOLE WORLD.

Do you know anything about statistics? You don't need to poll every single person in a country or the world to get meaningful data. Once you have enough samples, you start converging on a result and rapidly approach diminishing returns (i.e. once things have leveled off, you can double or triple your sample size and see very little change in your results). There's a reason why the Steam HW survey doesn't use every single computer running Steam: they don't need to in order to get their data.

 

[tuxon86]

Yeah, you post something, and then draw conclusion about whole situation by looking at one part of it. What is it? Making claims without eveidence to back them up.
This is only about steam. Yet, you claim it is reflection of WHOLE WORLD.

Not even close to what he said. He claim it was the only available metric that we have at the moment and it does demonstrate a trend. We get it that it goes about your personal and unsubstanciated belief... Eh at least it isn't a silly anonymous blog post from Poland!

 

[koyoot]

I'm not pushing the survey results, I'm simply stating that you can't just ignore them outright. Why are people buying an RX 480 or a GTX 1060 if they don't play games? So, for those people who play games and run Steam, there seems to be a 2:1 ratio of folks with the NVIDIA card versus ones with the AMD card. I'd be surprised if the overall sales aren't in that general ballpark.

I get that you're and AMD fan and you're entitled to your opinions. However, I think it's hard to argue that the Pascal GPUs aren't selling very, very well. You can enjoy your RX 480, because I'm really enjoying my GTX 1080.

First of All, one of the reasons why I discuss here, and take AMD side, is because I hate fanboyism in any way shape or form, to counter typical "opinions" on forums. That is the reason why I hate SemiAccurate forum, because it is 95% red tinted all over it, and I hate Anandtech forum, because the most vocal users are Nvidia fanboys.

Secondly. I do not have RX 480, currently I use MBP mid 2012 with Geforce GTX 650M, and linux workstation with Xeon E3 and... Nvidia GTX 980.

Get the point what I am writing. Steam is not the whole world. Steam is used in richest countries. Nvidia sells the most cards in richest countries. And over 70% of RX 480 sales have been in other countries than the richest ones. Thats what have consumed the availability and rised up prices for AMD GPUs. I never think as simply as most people on this board do, I want to compare data from other points of view.

Do you know anything about statistics? You don't need to poll every single person in a country or the world to get meaningful data. Once you have enough samples, you start converging on a result and rapidly approach diminishing returns (i.e. once things have leveled off, you can double or triple your sample size and see very little change in your results). There's a reason why the Steam HW survey doesn't use every single computer running Steam: they don't need to in order to get their data.

Fair enough about numbers. But it cannot be used then to draw any conclusions about whole situation about sales. If you are doing this, in any analysis, and take those numbers seriously, you should be fired from your work.