It is official AMD skips 20nm and jumps to 14nm

It is official AMD skips 20nm and jumps to 14nm from Guru3D.

Although this was a rumor for a long time now we now know that AMD skips 20nm and jumps onto a 14nm fabrication node for their 2016 GPUs. Yields in the 20nm have been a massive failure and as such many manufacturers are skipping that 20nm node.

The AMD Radeon 2016 line-up tagged under code-name Arctic islands as such will jump straight towards 14nm FinFET technology with stacked graphics memory dubbed HBM. as expreview reports. A huge jump considering the current products are still at 28nm. What this means is a smaller die and lower voltages. E.g. a Hawaii GPU would be precisely half the size of what it currently is.

Currently Intel and Samsung are the only nodes that can offer 14 nm production, so it will be interesting to see where the wafers will be baked. For example, Intel is manufacturing "Broadwell" CPUs on 14nm and Samsung is manufacturing the Exynos 7 SoCs on 14nm.

It will be interesting to see how the lineup of 2015 shapes up to be, but expect 28nm parts.

AMD Greenland will be the R9 400 Series Flagship in 2016.

From that new lineup, the Greenland GPU would be AMD’s first GPU to be built on the 14nm likely fabbed by Gloablfoundries. It would be the flagship GPU dubbed the R9 400 series which will be tied to second generation High Bandwidth Memory (stacked memory).

AMD to Skip 20 nm, Jump Straight to 14 nm with "Arctic Islands" GPU Family from TechpowerUp also.

So both AMD and Nvidia skipping 20nm. That also means the new GPUs from AMD will be 28nm.

And I see they name their next flagship after a bit colder country... So who knows, maybe these chips won't spew that much heat. :P

This also puts the R300 series from AMD in not being interesting for me. I've had a 28nm process GPU for 3 years now already, I can wait until 14nm arrives.

Honestly I want to hear more about AMD's cpu over anything else.. AMD Gpu's are at least still competitive in many price ranges.. Their cpu's on the other hand not so much, especially when you consider the numerous things their motherboard features lack compared to the intel equivalents.

To be honest R300 could bring a refresh in AMD's line but nothing more. They still using GCN.

They are so limited to what they could do with 28nm and GCN...

But at least it will be a move from AMD after 18!!! months.

Glad to hear it. It's a step in the right direction. I REALLY miss the days of competition. I really hope AMD becomes more competative. I've only bought intel/team green for years now. AMD just hasn't had anything on offer to interest me (except for a 270x I bought my kids...but they're my kids)

nice. I like the possibility of buying AMD again... i hope the power usage and performance is good. I don't get it though, if they were failing at doing 20nm, why would doing a smaller 14nm be easier?

'twill be a big year for AMD (2016, that is)

If they can last that long =(

Good news. Now, let's see if AMD can execute, which has been an issue for them lately. Seriously, no high end single-GPU launch in 18 months??

I know :-(

AH Bol*Beep*!

i was planning on building a console sized PC this year. New linux based gaming box.

Nvidia and intel were the only ranges i was looking at since AMD hardware does not play nice with small spaces at the moment.This could certainly change things though.

maybe i will hold out until next year. my current linux gaming rig will just have to stretch that bit further.

I did have an Intel/nVidia rig, but was not very happy with it, so I bought an all AMD rig. AMD just seems to work better for me, especially with linux.

Wasn't planning on buying a new GPU before 2016 anyways so this is good news. By that time maybe HBM will be available for the mid range-class cards as well. And hopefully those mid-range cards won't just be the high end of the previous series like they're apparently doing with the 300's.

They update the GCN architecture. Isn't 7970/280X GCN 1.0? 285 is 1.2 or 1.3, and they managed to lower the power draw some 50W or so.. Without sacrifise too much performance (I think). Mostly 2GB vram instead of 3 though. :(

Yeah, and the 285 actually uses compression to make its memory bandwidth more efficient. Benches indicate it's not as impressive as the 960's efficiency (what that does with 2GB 128-bit bus is really surprising), but if they made a 285x with 3-4GB it would be a pretty killer card, actually.

Its just the same as GF100 to GF110, GF104 to GF114 etc etc (GTX480 to GTX580, GTX460 to GTX560).

They cant make a good leap with just improving an old architecture... And GCN is already 3+ years old.

Exactly the same? I have the impression that the changes are bigger than what nVidia did.

Nope they are the same. Take R9 280 vs R9 285. Their performance is almost the same:

And the power consumption is a little lower than R9 280. Although I read at other sites about 40w difference, it seems that this depends on game:

Its like the same with what Nvidia did. Maybe a few percentages more or less but the general idea is the same...

Nvidia did a lot more to power consumption also: GTX570 vs GTX480 (same cores, same difference in performance with the R9 280 vs R9 285):

edit: Also look at this:

Under load we once again see NVIDIA’s design and TSMC’s process improvements in action. Even though the GTX 570 is over 20% faster than the GTX 470 on Crysis, power consumption has dropped by 5W. The comparison to the GTX 480 is even more remarkable at a 60W reduction in power consumption for the same level of performance, a number well in excess of the power savings from removing 2 GDDR5 memory chips. It’s quite remarkable how a bunch of minor changes can add up to so much.

Does this reminds you something? *whistles looking at the R9 280 vs R9 285.

:P

@Coseniath: Damn. Thought the difference was bigger than that. They seriously need to update their architecture...

This has caught my eye. I might keep my Trifire 7970 combo for another year and wait for the 490X series before I upgrade. They are still doing wonderful for 1440p/120fps. And Asynchronous Shaders and Multi-Threaded Command is supported by all GCN architecture cards and DirectX12 will finally support it, so my 7970s might live on even more.

Or I might fall for the temptation and buy a 390X (or maybe a 980 Ti). I'm having a bias for the AMD cards since they are going to support DirectX 12 Feature Level 3 and Nvidia's offerings only support DirectX 12 Feature Level 1.

I think AMD designs better cards than Nvidia. The only problem is that maybe they are too far ahead that the API and software don't utilize them to their potential. Nvidia on the other hand, designs efficient cards for "the moment"

The thing that's missing from those 280 vs 285 comparisons, though, is the fact that it has 30% less ram and 30% less memory bandwidth and yet outperforms the 280. Note also that its theoretical benefits aren't being captured here (e.g. http://www.anandtech.com/show/8460/amd-radeon-r9-285-review/16). Of course, what matters to us, the gamers, is real world performance, not synthetic, but the synthetic benches are indicative of what could be done if the hardware's resources are used right.

Not saying it's a miracle in raw numbers, but there's stuff going on behind the scenes. Plus it does have new features.

GTX570 had less VRAM and less memory bandwidth too compared to GTX480, that's why I didn't said it either.

Well then they have an other thing in common for their comparison.

And probably R9 280 wouldn't need such a big bandwidth so AMD just removed it from R9 285. Keep in mind that GTX960 delivers same performance as R9 285 with half memory bandwidth.

edit: Synthetic benchmarks are not bad, unless they are not testing real world scenarios.

Firestrike is a great example what a synthetic benchmark should test using your GPU/CPU.

Tessmark, Texel/Pixel Fill etc etc on the other hand, just show the potential of a GPU in just one situation.

GTX4xx series were scoring multiple times better than HD5xxx series in Tessmark. That didn't made them multiple times better in any game....