Thursday, April 23rd 2015

AMD to Skip 20 nm, Jump Straight to 14 nm with "Arctic Islands" GPU Family

AMD's next-generation GPU family, which it plans to launch some time in 2016, codenamed "Arctic Islands," will see the company skip the 20 nanometer silicon fab process from 28 nm, and jump straight to 14 nm FinFET. Whether the company will stick with TSMC, which is seeing crippling hurdles to implement its 20 nm node for GPU vendors; or hire a new fab, remains to be seen. Intel and Samsung are currently the only fabs with 14 nm nodes that have attained production capacity. Intel is manufacturing its Core "Broadwell" CPUs, while Samsung is manufacturing its Exynos 7 (refresh) SoCs. Intel's joint-venture with Micron Technology, IMFlash, is manufacturing NAND flash chips on 14 nm.

Named after islands in the Arctic circle, and a possible hint at the low TDP of the chips, benefiting from 14 nm, "Arctic Islands" will be led by "Greenland," a large GPU that will implement the company's most advanced stream processor design, and implement HBM2 memory, which offers 57% higher memory bandwidth at just 48% the power consumption of GDDR5. Korean memory manufacturer SK Hynix is ready with its HBM2 chip designs. Source: Expreview
Add your own comment

71 Comments on AMD to Skip 20 nm, Jump Straight to 14 nm with "Arctic Islands" GPU Family

#1
matar
Not an AMD Fan but I have to say that's a smart move AMD
Posted on Reply
#2
Noel.VSL
14nm NAND Flashmemory, really!?
Posted on Reply
#3
HumanSmoke
Can't say it comes as a surprise, but all those people yelling from the rooftops about the 390X being a 20nm design must be getting acid indigestion about now.
matar, post: 3272980, member: 96547"
Not an AMD Fan but I have to say that's a smart move AMD
20nm from either GloFo or TSMC is wholly unsuited for large power budget IC's. Bit of a no-brainer that both AMD and the completion would skip the process node. With Nvidia already looking to cover their bases by sourcing from TSMC and Samsung, AMD really couldn't commit to a late (and by all accounts, underperforming) 20nm.
Posted on Reply
#4
john_
HumanSmoke, post: 3272984, member: 98425"
Can't say it comes as a surprise, but all those people yelling from the rooftops about the 390X being a 20nm design must be getting acid indigestion about now.
I though the 20nm big GPU idea died months ago. I didn't knew there where still debates about that.
Posted on Reply
#5
micropage7
nice move, but seriously AMD should offer more, if they cant pass their competitor they should offer something new
Posted on Reply
#7
john_
HumanSmoke, post: 3273009, member: 98425"
Well, I wasn't pointing at TPU in particular, but some people still refuse to believe that both TSMC's CLN20SOC and GloFo's 20LPM were never intended for large power budget IC's that characterise GPUs, as their own respective literature spells out.
The only rumors I know they still persist about big chips at 20nm, have to do with Xbox One and PS4 APUs. Everything else about 20nm is either about Nolan/Amur or other ARM SOCs. But discrete GPUs on 20nm for desktops are abandoned as an idea many months ago from most people/publications. It would have been great if Granada was a 20nm Hawaii instead of just a rebrand, but it's not happening.
Posted on Reply
#8
Caring1
micropage7, post: 3273001, member: 82848"
nice move, but seriously AMD should offer more, if they cant pass their competitor they should offer something new
AMD are already comparable on performance, a power reduction is all they need to get the edge, combined with ease of overclocking cores and memory they should come out in front.
Posted on Reply
#9
GreiverBlade
micropage7, post: 3273001, member: 82848"
nice move, but seriously AMD should offer more, if they cant pass their competitor they should offer something new
passing? they are toes to toes ... technically

well power consumption is a letdown, but as for example my 290 still keep up the brand new 970 that a friend has in a quite similar setup to mine ... and the 980 is not too far above so that even a 290X can keep it in check

ok now there is the Titan X and the "upcoming" 980Ti but ... the T'X is a steal and the 980Ti is a "bend over, here it come again" scenario.

reducing the manufacturing node mean reduced power need if i am not mistaken? so maybe 14nm will be the feature that will enable Arctic Island to have a lowered power consumption

(even if Volcanic island will still be on 28nm it will still be a competitor for Maxwell and the upcoming nV cards until Arctic island is released in the wild)

i guess i will keep my 290 until the next next gen .... :roll:
Posted on Reply
#10
dorsetknob
"YOUR RMA REQUEST IS CON-REFUSED"
Noel.VSL, post: 3272982, member: 156821"
14nm NAND Flashmemory, really!?
Re read the OP 1 or

btarunr, post: 3272976, member: 43587"
Named after islands in the Arctic circle, and a possible hint at the low TDP of the chips, benefiting from 14 nm, "Arctic Islands" will be led by "Greenland," a large GPU that will implement the company's most advanced stream processor design, and implement HBM2 memory, which offers 57% higher memory bandwidth at just 48% the power consumption of GDDR5. Korean memory manufacturer SK Hynix is ready with its HBM2 chip designs.
btarunr, post: 3272976, member: 43587"
Samsung is manufacturing its Exynos 7 (refresh) SoCs. Intel's joint-venture with Micron Technology, IMFlash, is manufacturing NAND flash chips on 14 nm.
there was no mention of AMD using NAND Flashmemory
Posted on Reply
#11
micropage7
GreiverBlade, post: 3273076, member: 105443"
passing? they are toes to toes ... technically

well power consumption is a letdown, but as for example my 290 still keep up the brand new 970 that a friend has in a quite similar setup to mine ... and the 980 is not too far above so that even a 290X can keep it in check

ok now there is the Titan X and the "upcoming" 980Ti but ... the T'X is a steal and the 980Ti is a "bend over here it come again" scenario.

reducing the manufacturing node mean reduced power need if i am not mistaken? so maybe 14nm will be the feature that will enable Arctic Island to have a lowered power consumption

(even if Volcanic island will still be on 28nm it will still be a competitor for Maxwell and the upcoming nV cards until Arctic island is released in the wild)

i guess i will keep my 290 until the next next gen .... :roll:
yeah, the 14nm may reduce the power consumption, and as usual top of the line has basic problems:its heat and high power consumption

personally i wonder how about the performance of new 14nm on the road
Posted on Reply
#12
Ferrum Master
I just want a new GPU... quit the tease already... my 7970 is starting to cough up blood already...
Posted on Reply
#13
Lionheart
micropage7, post: 3273001, member: 82848"
nice move, but seriously AMD should offer more, if they cant pass their competitor they should offer something new
HBM memory is new... :cool:
Posted on Reply
#14
64K
Then we should see a considerable leap in performance from the 390X to the 490X. I also think AMD has code named it Arctic Islands because they have found a way to design a GPU that doesn't draw as much power as they currently do to compete with Nvidia. I hope their flagship is still around a 250 watt card design though. Using that much wattage on the improved efficiency of the 14nm process should be a beast.
Posted on Reply
#15
buggalugs
AMD should have dumped TSMC long ago although there isn't that many choices. AMD should try to do a deal with Samsung.
Posted on Reply
#16
the54thvoid
Caring1, post: 3273053, member: 153156"
AMD are already comparable on performance, a power reduction is all they need to get the edge, combined with ease of overclocking cores and memory they should come out in front.
Depends on how you define comparable. Perf/price - they're better than Nvidia. Perf/watt they are way behind. On pure performance you can argue, Titan X -> GTX980 -> Titan Black -> GTX 780ti ->(or =) R9 290X. All single GPU of course. 295X2 beats them all when crossfire scaling works.

AMD need the 390X to shift the market. By 2016 Nvidia will be on their Pascal designs. June can't come soon enough for me, let alone 2016.
Posted on Reply
#17
Naito
All I can say is that this will be a very interesting time; not that I'll be getting 4K any time soon, but will be keen to see what tech on such a fab is capable of. That, and DX12.
Posted on Reply
#18
JMccovery
buggalugs, post: 3273112, member: 56431"
AMD should have dumped TSMC long ago although there isn't that many choices. AMD should try to do a deal with Samsung.
Who was able to compete with TSMC 'long ago', hmm?

Absolutely no one.

Samsung wasn't even a blip on the radar until recently.
Posted on Reply
#19
jabbadap
Hmh there's no upcoming 14nm process on TSMC. TSMC:s road maps shows next node after 16nm ff+ is 10nm ff. So if amd is truly going to 14nm it must be GF/samsung nodes(I doubt intel will sell them its 14nm nodes).
Posted on Reply
#20
Kaotik
Actually GlobalFoundries has 14nm in production already, too.
Their biggest owner, Mubadala Development, announced early this month that GloFo is already ramping up 14nm production for a client (meaning it's not even test chips)
Posted on Reply
#21
Jorge
Kaotik, post: 3273152, member: 101367"
Actually GlobalFoundries has 14nm in production already, too.
Their biggest owner, Mubadala Development, announced early this month that GloFo is already ramping up 14nm production for a client (meaning it's not even test chips)
Exactly. And GloFo will be delivering a variety of 13 Nm chips - CPU, APU and GPU for said customers.
Posted on Reply
#22
Casecutter
btarunr, post: 3272976, member: 43587"
Named after islands in the Arctic circle, and a possible hint at the low TDP of the chips, benefiting from 14 nm, "Arctic Islands" will be led by "Greenland," a large GPU that will implement the company's most advanced stream processor design, and implement HBM2 memory, which offers 57% higher memory bandwidth at just 48% the power consumption of GDDR5. Korean memory manufacturer SK Hynix is ready with its HBM2 chip designs.
Whew... that was 3 breath sentence.

But it's that last one that surprises, all of a sudden rumors say 1st Gen HBM is constrained, even though SK Hynix indicated client shipments started in January 2015. While this says SK Hynix is "ready" for HBM2, sure not near production but appears on track.

What's more in question is where is TSMC with 16 nm FinFET? As from some of the rumors others have been "investigating options" or at "keeping open mind" for their next shrink. Some speculate TSMC might not have full production for large power budget IC's until Q3 2016. Such a lapse might give AMD the window to get Arctic Islands parts solidly vetted at GloFo and still be ready by this time next year.
Posted on Reply
#23
Bjorn_Of_Iceland
GreiverBlade, post: 3273076, member: 105443"
passing? they are toes to toes ... technically

well power consumption is a letdown, but as for example my 290 still keep up the brand new 970 that a friend has in a quite similar setup to mine ... and the 980 is not too far above so that even a 290X can keep it in check
So is my GTX780.. a 2 year old card.. and the 980 is not too far above so that even a 780ti can keep it in check.

AMD is lagging that much, they needed to skip a 20nm just to make them competitive.
Posted on Reply
#24
Casecutter
Bjorn_Of_Iceland, post: 3273240, member: 47228"
So is my GTX780.. a 2 year old card.
Could almost be... GTX 780 launched May 23, 2013
Posted on Reply
#25
alwayssts
Casecutter, post: 3273197, member: 94772"
Whew... that was 3 breath sentence.

But it's that last one that surprises, all of a sudden rumors say 1st Gen HBM is constrained, even though SK Hynix indicated client shipments started in January 2015. While this says SK Hynix is "ready" for HBM2, sure not near production but appears on track.

What's more in question is where is TSMC with 16 nm FinFET? As from some of the rumors others have been "investigating options" or at "keeping open mind" for their next shrink. Some speculate TSMC might not have full production for large power budget IC's until Q3 2016. Such a lapse might give AMD the window to get Arctic Islands parts solidly vetted at GloFo and still be ready by this time next year.
I get the impression it is the 2x1GB stacks that are constrained; everything points to that imho.

First, and for a long time, we heard 'Fiji' was only going to be 4GB (4x1GB). Then we heard murmurs AMD was internally battling with offering an 8GB design, even though it may hold up production and raise the price over $700. Then, we got that slide deck that included what appeared to be info fresh off the line about making 2x1GB stacks (likely meaning the bandwidth of a single 1GB stack with two connected stacks or 2x chips in a stack)...something that nobody really saw coming (HBM1 was going to be 4hi 1GB, HBM2 up to 8hi 4GB). I have little doubt this was a last-second addition/decision as they noticed peoples' concerns with 4GB per gpu (especially in crossfire) for such an expensive investment. This can be noticed by the frantic 'dx12 can combine ram from multi gpus into a single pool' coming across the AMD PR bow.

AMD really seems in a tough place with that. 4GB is likely (optimally) not enough for the 390x, especially with multi-gpu in the current landscape, but 8GB is likely a little too much (and expensive) for a single card (and I bet 390 non-x will be perfectly fine with 4GB aimed at 1440p)...it's the reason a 6GB similar-performance design from nvidia makes sense....that's just about the peak performance we can realistically expect from a single gpu on 28nm.

One more time with gusto: 28nm will get us ~3/4 of the way to 4k/8GB making sense on the whole. 14nm will pick up the slack..the rest is just gravy (in performance or power savings).

While I want 4k playability as much as anyone in demanding titles (I'm thinking a dual config on 14nm is in my future, depending on how single cards + dx12 handle the situation), I can't help but wonder if the cards built for 1440p60+ will be the big winners this go-round, as the value gap is so large. That is to say, 390 (non-x, 4GB), perhaps a cheaper gtx 980, and/or a similarly-priced salvage GM200.
Posted on Reply
Add your own comment