14nm r9 nano question

[tential]

This will be a game changer for PC gaming. 80watt boards that run cool/silent and have the performance of an R9 Nano today, but a few months from now.

AMD has squeezed all they can out of 28nm with the R9 Nano. So 14nm cannot come soon enough. Costs will plummet as well.

Costs will not plummet. Where did you get that piece of misinformation from?

 

[Hugo Drax]

Costs will not plummet. Where did you get that piece of misinformation from?

Its pretty obvious. This 650 dollar R9 nano 2015 performance level will cost 350 dollars.

And the new 650 dollar R9 Nano 175w card with twice the performance will still cost the same but it will be double the performance of R9 nano 1.0 2015

 

[golem]

Its pretty obvious. This 650 dollar R9 nano 2015 performance level will cost 350 dollars.

And the new 650 dollar R9 Nano 175w card with twice the performance will still cost the same but it will be double the performance of R9 nano 1.0 2015

I think it's more likely nano 1 will disappear when nano 2 comes out. It not like there's a lot of stock right now that will have to be sold off at fire sale prices.

 

[tential]

Its pretty obvious. This 650 dollar R9 nano 2015 performance level will cost 350 dollars.

And the new 650 dollar R9 Nano 175w card with twice the performance will still cost the same but it will be double the performance of R9 nano 1.0 2015

I'll bet your current r9 nano that it will not cost 350 in 2016 in a nano for factor.

Realistically, I'll bet any amount that isn't happening. Just name it.

Both comments are just wrong tbh..

If you think we'll see double the performance of anything in 2016 then just lol.....
Not saying I don't wish it. But if double the performance of fury x, nano, 980ti is out in 2016 I'll deliver my money and a lot more....

 

[Techhog]

I think it's more likely nano 1 will disappear when nano 2 comes out. It not like there's a lot of stock right now that will have to be sold off at fire sale prices.

"Performance level," not the card itself.

 

[Silverforce11]

If you think we'll see double the performance of anything in 2016 then just lol.....
Not saying I don't wish it. But if double the performance of fury x, nano, 980ti is out in 2016 I'll deliver my money and a lot more....

We probably won't see 2x performance across the board, certainly not with mostly DX11 titles still being used in benchmark summaries.

But in DX12 titles? I'm expecting 2x.

At the recent SIGGRAPH conference, where devs, NV, AMD, MS were talking about the future APIs such as DX12 & Vulkan, there was a comment that "next-gen GPUs are 200% faster in DX12".

 

[golem]

"Performance level," not the card itself.

If hbm 1 and 2 are supply constrained and expensive as seems to be the case. Why would AMD use what little of it is available on cheaper cards?

 

[Hugo Drax]

I'll bet your current r9 nano that it will not cost 350 in 2016 in a nano for factor.

Realistically, I'll bet any amount that isn't happening. Just name it.

Both comments are just wrong tbh..

If you think we'll see double the performance of anything in 2016 then just lol.....
Not saying I don't wish it. But if double the performance of fury x, nano, 980ti is out in 2016 I'll deliver my money and a lot more....

At 14nm they will be able to make a lot more chips, with HBM eventually no longer being constrained why do you think none of this is possible?

Desktops used to cost 4K for a 486 now for 500 dollars you get a computer that outperforms a Cray Y-MP that cost several millions.

 

[Hugo Drax]

If hbm 1 and 2 are supply constrained and expensive as seems to be the case. Why would AMD use what little of it is available on cheaper cards?

They will not longer be constrained once they get the process down and they can fab these chips en masse.

HBM is cutting edge today, tomorrow it will be on every GPU once they start pumping them out.

 

[tential]

They will not longer be constrained once they get the process down and they can fab these chips en masse.

HBM is cutting edge today, tomorrow it will be on every GPU once they start pumping them out.

Not literally tomorrow.... Did you not hear nvidia might use ddr5x instead of hbm?

I think you're vastly miscalculating how long it takes for hbm and this new node to ramp up to their maximum potential as well as the performance increase you'll get.

I wish it'll be 2x though.... Very good for me since I plan on getting the dual amd chip

 

[bystander36]

Not literally tomorrow.... Did you not hear nvidia might use ddr5x instead of hbm?

I think you're vastly miscalculating how long it takes for hbm and this new node to ramp up to their maximum potential as well as the performance increase you'll get.

I wish it'll be 2x though.... Very good for me since I plan on getting the dual amd chip

I wasn't aware they said "instead of". I thought it was going to be an addition to type of deal. The highest end would get HBM, and everything else would get ddr5x. Of course it's all rumor as far as I know.

 

[alcoholbob]

A Nano is just a binned Fury X with custom BIOS that limits TDP. It's not like it was a chip that was engineered in particular for power efficiency. Hell all Nvidia would have to do is bin 80%+ ASIC 980 Tis and set them to 90% power limit and now you have a card faster than the Nano AND uses less power. It takes minimal effort for Nvidia to beat the Nano in every metric; it just makes no economic sense for them since the market is so small. It's honestly nothing special.

Pretty much anything card with a 150W TDP or more will run in circles around a R9 Nano at 14nm lmao.

 

[tential]

I wasn't aware they said "instead of". I thought it was going to be an addition to type of deal. The highest end would get HBM, and everything else would get ddr5x. Of course it's all rumor as far as I know.

What I meant is that hbm isn't ready for full volume from top to bottom.

I just don't think next year will be a 2x performance jump from top to bottom like he states, especially when he is using hbm as a factor which won't be available for every sku.

But well if it is great for me! Because I'll get the top sku. If an upgrade next year doesn't make sense for me I'll be annoyed even though I literally just bought the 290, I still need a real 4k card and monitor!

 

[Silverforce11]

A Nano is just a binned Fury X with custom BIOS that limits TDP. It's not like it was a chip that was engineered in particular for power efficiency. Hell all Nvidia would have to do is bin 80%+ ASIC 980 Tis and set them to 90% power limit and now you have a card faster than the Nano AND uses less power. It takes minimal effort for Nvidia to beat the Nano in every metric; it just makes no economic sense for them since the market is so small. It's honestly nothing special.

Pretty much anything card with a 150W TDP or more will run in circles around a R9 Nano at 14nm lmao.

How big is the mITX market?

Do you see major motherboard, case, cooling, PSU makers create products and compete in that market? Yes. I wonder why. Perhaps its not as small as you think.

 

[tential]

How big is the mITX market?

Do you see major motherboard, case, cooling, PSU makers create products and compete in that market? Yes. I wonder why. Perhaps its not as small as you think.

It's a growing market that's for sure.

 

[ShintaiDK]

How big is the mITX market?

It depends for what segment.
Looking on Geizhals for simplicity:
86 ATX, 74 mATX, 3 E-ATX and 17 mITX.

It has to be said tho that when using motherboard manufactors sites. Like Asus, MSI, Gigabyte and Asrock. You end up with something like ~180 boards and ~10 mITX.

However if we sort by Z170.
57 ATX, 8mATX, 3 E-ATX, 8 mITX. (Note that those 8 mITX comes from 5 makers only)

Its is a very slowly growing segment. But the trend is clear towards smaller. However expect many years before it really changes in numbers.

But the mITX segment is a lot more muddy than that. Because half the segment is roughly based on ultra slim that kills of any dGPU hopes for example or other standard components. While the other half is more or less open for full size components of all kinds. Then you have a niche in between that fits something else.

It's a growing market that's for sure.

Hopefully it will grow faster. Was a nightmare to find a good mITX board this round.

 

[Hugo Drax]

A Nano is just a binned Fury X with custom BIOS that limits TDP. It's not like it was a chip that was engineered in particular for power efficiency. Hell all Nvidia would have to do is bin 80%+ ASIC 980 Tis and set them to 90% power limit and now you have a card faster than the Nano AND uses less power. It takes minimal effort for Nvidia to beat the Nano in every metric; it just makes no economic sense for them since the market is so small. It's honestly nothing special.

Pretty much anything card with a 150W TDP or more will run in circles around a R9 Nano at 14nm lmao.

So how do you account the Nano having a TDP of 175 vs R9 285 with a tdp of 190 and the nano is twice the performance the the R9 285.



AMD had to have tweaked the architecture a bit to make it more efficient in addition to picking the top shelf binned chips.

Now its pretty obvious a 14nm 150w nano will outperform a 175w 28nm nano.

but I can see an 80w 14nm TDP nano2 that performs as good as this years 175w. This is were things get interesting.

 

[happy medium]

This is were things get interesting.

How do things get interesting?
AMD finally makes a gpu that's not a space heater after 3 years, you call that interesting?

I think its more interesting how Nvidia was best in performance using 100 watts less for such a long time while swallowing up more market share.

14nm Nano will not even be made after they see the sales figures of the non selling Nano. It was a big woops.

I think you guys better put down the pipe.

Warning issued for inflammatory language.
-- stahlhart

 

[IllogicalGlory]

AMD finally makes a gpu that's not a space heater after 3 years, you call that interesting?

The difference between a space heater and not a space heater is 20W? That's the difference in power draw between the 980 Ti and Fury X. How about 35W? That's the difference in power draw between the 290X (U) and 780 Ti.

Both are from AT's GPU 2015 load power consumption in Crysis 3.

 

[Techhog]

The difference between a space heater and not a space heater is 20W? That's the difference in power draw between the 980 Ti and Fury X. How about 35W? That's the difference in power draw between the 290X (U) and 780 Ti.

Both are from AT's GPU 2015 load power consumption in Crysis 3.

Protip: don't feed the troll.

 

[tential]

Ya shintai I mean I'm happy I don't care about mitx right now. But my next build will be an mitx wc build because I can lol. And because hedt is finally coming to mitx, now it's even more interesting.

I think amd is just too far ahead of the curve. Just like with multicore, apu, async, etc. Amd comes too early with new tech before it's mature enough to effectively use. Nvidia usually has better timing.

 

[ShintaiDK]

Ya shintai I mean I'm happy I don't care about mitx right now. But my next build will be an mitx wc build because I can lol. And because hedt is finally coming to mitx, now it's even more interesting.

I think amd is just too far ahead of the curve. Just like with multicore, apu, async, etc. Amd comes too early with new tech before it's mature enough to effectively use. Nvidia usually has better timing.

HEDT in mITX, good luck
I know boards exist, but its not something I would do.

Yep. It just turns into bad business trying to be technically pure. Everyone can do that, but not everyone can make money.

 

[lopri]

x2 performance jump from 28nm to 14nm for the GPUs seems entirely possible. There are different versions of 28nm processes, you know.

 

[tential]

HEDT in mITX, good luck
I know boards exist, but its not something I would do.

Yep. It just turns into bad business trying to be technically pure. Everyone can do that, but not everyone can make money.

It's why I'm waiting for skylake e or cannonlake e. Then getting the dual gpu wc cards.

It's like amd has a vision of how things should be, then tries to make that happen when no one is ready for it.

Another thing is even if the cards are nice, I need a new build around those small cards. There just is no inventive for a lot of high end gamers to get a small card, when it requires a build around it to make it useful. And you didn't even need that build in the first place.

That's why I'm waiting til skylake or cannonlake. Because by then a cpu upgrade is reasonable.

There are just a lot of hurdles towards widespread mitx builds and amd is just too early to be in a great position.