Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
What benefit would there be to Intel to intentionally slow the rate of process introduction?
Margin is a function of manufacturing cost and ASP. If Intel were simply "slowing down" the 10nm ramp to boost margins, then they would be completely off their rockets. 14nm might give them better mfg costs right now, but their product competitiveness has been hurt pretty severely as a result, and it'll get worse if AMD is shipping 7nm EPYC against 14nm++ Cascade Lake.
The cost structure side of 14nm is OK, but if the products are much less competitive as a result, they will lose sales and they will see ASPs come down, which is far worse than having a higher-than-expected manufacturing cost structure for margin.
Anyway, Intel's slowing down the 10nm ramp not for any other reason than the 10nm technology is a total wreck right now and can't support viable products in the market.
Cobalt use is probably a main culprit, but Intel's 10nm process really seems like a total dumpster fire. This was a process that was supposed to be in HVM during 2016 and we're now looking at 2H 2019 most likely for mass production.
Whatever the issue is, it smells rotten.
They used cobalt wires at the lowest metal layers, something that the foundries didn't do because they weren't as aggressive on minimum metal pitch scaling. The cobalt use is likely what's tripping them up.
Intel has 6 core 14nm++ mobile chips for high perf, Cannon Lake-Y for ~5W, Cannon Lake-U for 15W to complement Kaby Lake-Refresh, and also Coffee Lake-U for 15W and 28W (4C with GT3e on 14nm++).
GloFo doesn't have anywhere close to the wafer or revenue scale of Intel. GloFo's only leading edge customer seems to be AMD, and AMD ships far fewer chips (probably an order of magnitude difference) than Intel does.
They are. Next Intel modem (XMM 7560) will support CDMA/EVDO, which will allow Intel-based phones to work on Sprint/Verizon and others.
I wonder if the TR cores are thermal throttling, which is why the MT score is so underwhelming.
Appears that the top SKU has an all-core turbo of 3.2GHz. That's really impressive.
There are also Intel versions of those systems with Skylake-X.
morrowinder, what I find hilarious is that the 12-18 core SKUs of Core X are viewed as a "panicked reaction" to Threadripper, but Intel has at least announced the various SKUs and price points. AMD hasn't given any SKU details about Threadripper.
Looks like they chose Intel.
Talk about an unfair comparison. You could construct a similar scenario with a homebrew Intel HEDT platform vs a Mac Pro, too (the latter of which is based on Ivy Bridge-E).
26% is transistor performance improvement from 14nm (original) to 14nm++. That's irrespective of the underlying architecture/circuit changes.
Kaby Lake Refresh and Coffee Lake will be on 14nm++, a third gen version of Intel 14nm that's ~26% faster than the first gen 14nm.
Cannon Lake will be on 1st gen 10nm and begin shipping at the end of this year/beginning of next, with serious availability in the first half of 2018 for certain mobile systems.
44mm^2 refers to the size of a quad core Zen CCX on the Zeppelin die, it is not the size of the full Zeppelin chip, which includes two CCX + uncore.
Look at the out-of-the-box frequencies of these chips. Turbo boost 2.0 of 4.3GHz for the top 8/10 core SKUs (which means all of the cores on the die need to be validated to run at that speed).
Aggressive binning.
Navin Shenoy is a good executive, he'll do well in this role.
Intel didn't introduce a 22nm FDSOI process, they introduced a low-cost 22nm FinFET process to compete with 22nm FDSOI/22nm planar from GloFo/TSMC, respectively.
I doubt it. Intel can't even publicly confirm that they're inside of the iPhone, just listen to the execs when anybody even mentions the iPhone.
I think the margins on these modems are low because Intel has to pay TSMC to build them as well as third party packaging & test, but I don't think they're zero or negative. And I suspect they will only get better with 7560 (internally built modem) and beyond.
IMO, this is a pretty decent business for Intel. I wouldn't write it off as just PR/publicity.
Stacy Smith used to be the strategy officer, IIRC, but now he's in charge of Sales, Mfg, and Ops, so I guess they needed to fill the Chief Strategy Officer role.
Evans is solid, she seems to have done a very good job getting Intel's modem business on a reasonable path after the previous head of the group (Hermann Eul) just couldn't deliver the design wins.
Intel Appoints Chief Strategy Officer
Intel at least generates positive revenue/gross profit from these modem sales.
Apple will never get into the server SoC business, it's just not big enough.
The Apple Watch generates more revenue than any of the non-Intel server chip makers can reasonably hope to see from server SoC sales. AirPods probably fit that description.
Apple is in a whole different league than Intel.
That SoFIA chip from Rockchip has an ARM based counterpart in 16FFC. The ARM one has slightly better CPU performance -- and it uses much smaller, more efficient A53 cores.
$15B for this is bonkers. Not even $400M in revenue last year and barely over $100M in profit.
Intel said to buy Israel’s Mobileye for up to $16 billion
Intel 14nm v1 had a 70nm GP and 52nm MMP, but Intel says they relaxed the gate pitch for 14nm+ (performance enhanced version currently being used).
Samsung 10nm is believed to have a 64nm GP and 48nm MMP. No figures available for TSMC 10nm, but I'd imagine same ballpark.
TSMC has said its 7nm process will have a 40nm MMP. Intel had said that its 10nm process will have a 54nm GP.
Intel itself admits in its marketing materials that both Samsung and TSMC 10nm are denser than its 14nm, FYI.