[SeekingAlpha] Cherry Trail not ramping until March 2015

[NTMBK]

http://seekingalpha.com/news/2015175-intels-cherry-trail-atom-cpus-reportedly-ramping-in-march

Supply chain sources tell Digitimes Intel's (NASDAQ:INTC) 14nm Cherry Trail Atom CPUs, once expected in 2014, will enter volume production in March 2015. Intel is also said to be prepping its existing 22nm Bay Trail (Atom Z3000) CPUs for use in 64-bit tablets.

Which probably means we won't see devices on shelves before Back to School 2015. Looks like this roadmap has seriously slipped!

 

[witeken]

You mean this roadmap:

There was (I think) also a roadmap with Willow Trail on it (CT Q3, WT Q4), but I couldn't find it, strangely enough. It was from the investor meeting.

But I found this slide:

I wonder how Broxton is affected by this. This also doesn't look like 6 months anymore... Initially, it would be 3 months (Q2->Q3), then it was confirmed to be 6 months (Q3->Q4), and now some products seems to be almost 1 year for certain products.

BTW, I remember some rumor saying that CT would be released at CES, right?

 

[NTMBK]

You mean this roadmap:

Yeah, those are a little more specific, thanks

BTW, I remember some rumor saying that CT would be released at CES, right?

Well, if the latest rumour is true then that's not going to happen. But I guess we might see some demo units with early samples of CT chips, even if it isn't in volume production yet- be interesting to see what OEM partners are planning.

 

[krumme]

Looks the only sane thing to do imho, considering the lack of interest in the market for the current solution. Its good to see Intel adapt. Nothing wrong with that, thats what makes a business survive and be profitable.

The question is also - if the performance, and hence process, is actually the weak link on the current 22nm solution be it either cpu or gpu?
Imho eg the lack of integration of LTE, better DSP capabilities, can easily be seen as more important as performance from the current solution seems pretty strong to me.
And as Intel have been tied to the dismal success of Windows, perfect 64bit Android implementation comes into mind as another important goal.

When the adaption have happened so late its imho because fighting QQ, SS & Apple is just an entire different ballgame than the usual competition - or lack thereoff - Intel is used to. These guys are just rough, cynical and have loads of cash and power. Add to that the extremely low cost Chinese arm solution wipe the rest of the floor, leaving Intel in the same position with Atom as AMD is with beema on the x86 side. Squeezed from all sides. Adapting cost by introducing more expensive nodes later than planned, is a good way to get a breathing room, while Intel sharpen up the design and solution in general.

 

[witeken]

Well, if the latest rumour is true then that's not going to happen. But I guess we might see some demo units with early samples of CT chips, even if it isn't in volume production yet- be interesting to see what OEM partners are planning.

If Intel is going to launch at CES, Cherry Trail must have been in production since mid-2014, which is much before this rumor saying they will only start production in March. I actually don't think this rumor is credible, but it isn't good if we only get bad news/rumors.

 

[NTMBK]

If Intel is going to launch at CES, Cherry Trail must have been in production since mid-2014, which is much before this rumor saying they will only start production in March. I actually don't think this rumor is credible, but it isn't good if we only get bad news/rumors.

Oh, certainly. Given Seeking Alpha's nature, this could even just be someone playing stock market games with information they know to be false. Take everything with a pinch of salt, as always!

 

[positivedoppler]

Oh, certainly. Given Seeking Alpha's nature, this could even just be someone playing stock market games with information they know to be false. Take everything with a pinch of salt, as always!

It's usually a shitty place to get news because of that

 

[frozentundra123456]

Given the delays in broad well, I would not be surprised if this is true. However , as krumme said, I am not sure better performance is really the solution. The bigger problem is lack of integrated LTE support and just breaking into the ARM ecosystem.

 

[Arachnotronic]

The original source is here:

http://www.digitimes.com/news/a20141003PD205.html

Looks like we'll see the 20nm AMD low power parts in Q3 2015.

 

[Homeles]

The question is also - if the performance, and hence process

The process is not wholly responsible for the performance of the chip. Regardless, Bay Trail is hands-down the best performer per watt on the tablet market (or at least it was prior to the A8 -- and it probably still is, given how tame A8 was).

If you were to port Apple A7/A8 to 22nm, its performance would increase significantly. Even TSMC's upcoming 16nm process lags Intel's 22nm in performance, as I'd proven in the Samsung 14nm thread.

Intel's problems with Bay Trail were first and foremost tied to its poor cost optimization. The spatial utilization of Bay Trail's die is quite poor, and its requirement of needing more PCB layers resulted in Intel paying dearly to have a product that was competitive on price.

This is fixed in Cherry Trail.

Other issues Intel has with Bay Trail are with the "extras" -- as you pointed out, which was the only part of your post that wasn't complete nonsense. There is no doubt that having a better ISP, modem integration, and something like Qualcomm's nifty little envelope tracker would be of great benefit to Intel. But if you're going to question its performance, you're just going to make a fool out of yourself.

Also to be fixed will be GPU performance. It's tough to say how well Cherry Trail will compete, but there's no doubt it'll be head and shoulders above Bay Trail.

But I found this slide:

I'm fairly certain that slide was fake.

And regards to the delay, it's a shame that it's been delayed so much. At this point, one wonders why Intel wouldn't just axe Cherry Trail to pull in Willow Trail.

 

[witeken]

Given the delays in broad well, I would not be surprised if this is true.

14nm has been qualified and is being shipped to customers since Q2.

http://seekingalpha.com/article/2091473-intel-14-nanometer-production-looks-like-a-go

 

[witeken]

I'm fairly certain that slide was fake.

And regards to the delay, it's a shame that it's been delayed so much. At this point, one wonders why Intel wouldn't just axe Cherry Trail to pull in Willow Trail.

I was fairly certain there was an original Intel slide from IM, but after some digging, apparently not. The rumor however did turn out to be correct (Cherry Trail initially planned for Q3 with 16 Gen8 EUs), so it isn't a big leap of faith to think the other information (including 14nm smartphone platform Morganfield in Q1'15) was correct as well, although obviously outdated by now due to Intel's 14nm failure.

http://www.digitimes.com/news/a20130821PD203.html

IM resources: http://investorshub.advfn.com/boards/read_msg.aspx?message_id=94321762
IM mobile keynote: http://intelstudios.edgesuite.net/im/2013/pdf/2013_IM_Eul.pdf
Just for fun: http://intelstudios.edgesuite.net/im/2013/archive/qa1/archive.html

 

[NTMBK]

14nm has been qualified and is being shipped to customers since Q2.

http://seekingalpha.com/article/2091473-intel-14-nanometer-production-looks-like-a-go

Does Intel's cost-per-wafer on a node tend to drop over time? (Genuine question, I don't know the answer.) Since Cherry Trail parts are going to have very slim profit margins compared to the likes of Broadwell-Y, it might need to wait until the process is more mature before release.

Just a guess.

 

[witeken]

Does Intel's cost-per-wafer on a node tend to drop over time? (Genuine question, I don't know the answer.) Since Cherry Trail parts are going to have very slim profit margins compared to the likes of Broadwell-Y, it might need to wait until the process is more mature before release.

Just a guess.

I don't know if this has anything to do with it, but anything seems better than a $1B/Q contra-revenue cost.

 

[ShintaiDK]

Does Intel's cost-per-wafer on a node tend to drop over time? (Genuine question, I don't know the answer.) Since Cherry Trail parts are going to have very slim profit margins compared to the likes of Broadwell-Y, it might need to wait until the process is more mature before release.

Just a guess.

I would guess Intels wafer price is pretty fixed through its node life.

TSMC and Samsungs 20nm only cost what they do due to supply and demand. However for 3rd party I am sure it gets cheaper over time for the same node. Just as TSMC and Samsung.

But for example with Samsungs 20nm chips, I bet their wafer cost is less than half of 3rd party.

 

[ShintaiDK]

In related news:
http://www.digitimes.com/news/a20141003PD205.html

AMD is launching both x86 and ARM for tablets in 2015.

 

[krumme]

Does Intel's cost-per-wafer on a node tend to drop over time? (Genuine question, I don't know the answer.) Since Cherry Trail parts are going to have very slim profit margins compared to the likes of Broadwell-Y, it might need to wait until the process is more mature before release.

Just a guess.

Yields is the most kept of all secrets, and we know it improves in time, but as Intel usually have a very steep ramp compared to TSMC i think its a fair asumption to say they have comparable high yields from day 1? So -marginal- cost doesnt lower much during time, but another way to be looking at it is what is the value of bringing newer nodes to the market faster? (cost/benefit)

I dont think Intel 14nm process is any failure in any way, and that it is on track with the problems new nodes always have - exactly like 22nm fets. I think its just as its supposed to be. If that is true the roadmap is more or less as it always was, but we will see a slower ramp, and shipments, because the anticipated demands for 14nm Atom is not going to make a huge difference for the discount they give now - as noted by Witeken. What you can do then is to save some $ by making the ramp slower - its not much, but its part of the bigger capex adaption.

So its not because of process difficulties, but because the solution is not fitting the market. I view it as a very deliberate decision and it make sure the credibility to execution is still there. Contrary to eg fantacy ppt slides from gf, that went completely crazy years ago, to a degree nobody actually reads it any more.

 

[jpiniero]

Does Intel's cost-per-wafer on a node tend to drop over time? (Genuine question, I don't know the answer.) Since Cherry Trail parts are going to have very slim profit margins compared to the likes of Broadwell-Y, it might need to wait until the process is more mature before release.

Absolutely, and I think that's why the second delay happened for everything but Core M. Like for instance if 14 nm hasn't passed 22 nm in cost/transistor but will once they start tamping up the other Broadwell Y and U models and the cost will drop as production increases.

 

[Ajay]

In related news:
http://www.digitimes.com/news/a20141003PD205.html

AMD is launching both x86 and ARM for tablets in 2015.

That's interesting. I couldn't, for the life of me, figure out why AMD was only doing a server oriented ARM CPU. Seems to make sense to that same basic IP, add an iGPU and throw their hat in the game for tablets (even though the competition is very tough).

 

[Shivansps]

Someone please tell AMD that no one cares about ARM, everyone uses ARM, nothing special on there.

 

[Ajay]

Someone please tell AMD that no one cares about ARM, everyone uses ARM, nothing special on there.

Well, the only thing special ATM, is getting x86 into ultra-mobile - and that's a slog even for Intel.

 

[Shivansps]

AMD @ARM is just one more of the bunch, its really not that interesting.

x86 is another thing, BT is right now making possible for everyone to get a good performing windows/android tablet at the price you can get a dual/quad A7.
Where are the tablet segment of mullins anyway?

and now, using both x86 and arm? x86 for windows and arm for android? thats not gona end good for the x86, and probably no good for arm prices either.

 

[escrow4]

Balls Intel. I have a Vivotab with a dual core Bay Trail and it gives around 8hrs or so purely for ebooks. Still looking for a Windows tablet that has 15hrs+ battery life for less than $500. And yes, you cannot compare Calibre to a Kindle.

 

[Shivansps]

There is not a dual core bay trail for tablets, they are all quads.

 

[krumme]

There is not a dual core bay trail for tablets, they are all quads.

So is a7. A single a7 core is 0,5mm2 including l1 for the cpu part (the entire soc is far bigger), but for the fun of it you can have aprox 55000 a7 cores on a 28nm wafer. And a 28nm wafer is probably cheap now - around 1500 usd perhaps. Do the math, its not even half a $ for quad core cpu part. Is no fun competing with that even if its quad or octo core - And rockchip whatever have dirt cheap organizations compared to Intel.

As for the high end, no matter how good their products is just they are good enough, both Apple, Samsung and QQ prefer their own tech - because it protects their own long term market and monopoly, and secondly they just dont want to pay want intel is used to.

Intel can not do the bt funding forever. Its not a viable long term strategy. The product or the market need to change radically, and 14nm/10nm whatever is not going to cut it for change.