Intel's Future Chips: News, Rumours & Reviews

Page 126 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Status
Not open for further replies.
Meanwhile it seems that it's TSMC who has a noticeable lead over everyone else.

Quote fron Daniel Neni:
At the TSMC Symposium yesterday 7nm and 5nm were front and center. 10nm and 7 EUV not so much so my expectation is that most people will skip 7 EUV, including Apple. 7nm is in HVM today with more than 50 tape-outs expected in 2018 and 5nm will start risk production in 1H 2019. 5NM will use EUV and offer a +15% performance advantage or a +30% power advantage over 7nm. Density is 1.8X.
Click to expand...
 


Also explains this one, see Tapa Ghosh answer :ouch:
https://twitter.com/CDemerjian/status/989618558203236352
 


Charlie has been hammering away at Intel for years now, and there were people that dismissed him. Now, those people are as silent as the vacuum of outer space, or worst outwardly questioning everything Intel's says now! Their definitely is a change in mind share happening, because of Spectre/Meltdown/10nm. Intel has lost the trust of some of it's most avid followers.
 
Yeah that's correct..Jim Keller, who has been running Tesla's Autopilot hardware program since 2016, is leaving the company for Intel. The news, a major loss for Tesla, was first reported by Electrek.

Keller has had crucial roles in overseeing some of the major tech projects of the last decade. Before Tesla, he played a major role in Apple's development of A4 and A5 processors, the processors that powered many Apple mobile devices from 2010 to 2012. And before that, he worked at chipmaker Advanced Micro Devices (AMD), developing the Zen computer processor microarchitecture.
https://www.popularmechanics.com/cars/car-technology/a20075542/tesla-loses-its-self-driving-architect/
 
Tick, Tock, Boom! A Cannon has interrupted the Intel waltz once again
Subject: General Tech | May 8, 2018 - 02:52 PM | Jeremy Hellstrom
Cannon Lake and their 10nm process have proven a serious problem for Intel these past few years. It was in 2016 that they originally announced Cannon Lake would be delayed a year, which was then corrected to 2018. Barely a week ago we heard from Brain Krzanich that 2018 was too optimistic a date, and the 2016 CPU is now scheduled for some time in 2019.

Along with the Cannon Lake delay, DigiTimes also reports that the supply of the 14nm H310 chipset has completely dried up and we may not see more for a month or so, with July being the latest expected date. This means the only low teir Intel motherboard available for system builders, both professional and home, is the B360.

The next quarter's financials for both AMD and Intel should be very interesting.
Click to expand...
"Intel initially planned to launch 10nm Cannon Lake CPUs in July 2018, but its CEO Brain Krzanich unexpectedly disclosed at a meeting with financial analysts in late April that volume production of 10nm chips will be moving from the second half of 2018 into 2019 as it will take time to improve yield rates."
Click to expand...
https://www.pcper.com/news/General-Tech/Tick-Tock-Boom-Cannon-has-interrupted-Intel-waltz-once-again?utm_source=dlvr.it&utm_medium=twitter
 
Contra-revenue comes back in a big way at Intel
Officially nothing is happening….look a shiny thing over there!
May 8, 2018 by Charlie Demerjian
Guess what is making a comeback at Intel, yup, contra-revenue funding. SemiAccurate has learned that it never really went away, just changed targets, and the numbers appear to be pretty huge.

You might recall that Intel is trying everything it can to deflect that its 10nm process simply doesn’t work now, won’t ever work at a financially viable level, and in generally is going to cripple the company in the coming quarters. Don’t take our word for it, listen to Intel’s Q1 2018 call for the grim details.

You can hear the company trying to deflect to other new and emerging growth markets as a distraction. This has worked in the past, the investment community seems to actually believe that these new avenues are going to change Intel’s fortunes rather than simply being a distraction. Remember Intel’s phone SoC ambitions? Wearables? VR? IoT? Foundry? Shall I go on?

Back to the original point, SemiAccurate has been digging into how Intel is doing in their core markets, and how the outlook for adjacent markets is doing, and we found a whopper. That whopper is of course what looks to be a huge river of contra-revenue funding once again.

And again once again to torture the phrase, Intel seems to be trying very hard to keep this one quiet. We’re not sure what brings about this reticence though, if you were spending what appears to be roughly 10 digits of slush-like funding, wouldn’t you be proud of it? I sure know I would. So what is going on this time, and more importantly what does it say about the product line involved?
Click to expand...
https://semiaccurate.com/2018/05/08/contra-revenue-comes-back-in-a-big-way-at-intel/
 

Yeah, but it looks like they have never really stopped that practice considering there is still steady flow of articles about it.
How does Intel funnel contra-revenue funds?
Same old story with many more zeroes
Dec 29, 2014 by Charlie Demerjian
https://semiaccurate.com/2014/12/29/intel-funnel-contra-revenue-funds/
Contra-Revenues
The accounting definition of contra-revenue is best introduced by example. Say that in a given week, your shoe store sells 100 pairs of shoes for $30 each. They go on your income statement as $3,000 in revenue. But 20 customers decide they don't like their shoes, so they return them for $600 in refunds. When you're totaling up the week's sales, you could simply report $2,400 in sales, the equivalent of 80 pairs of shoes. But that misstates the volume of your business. You didn't sell 80 pairs. You sold 100 pairs, but 20 eventually came back. If you were planning your inventory based on selling only 80 pairs a week, you could have run out of shoes at some point during the week, which would lead to every retailer's nightmare: turning away customers. So instead of deducting returns from sales and reporting a flat $2,400 in revenue, you set up two separate entries in the revenue section of your income statement: gross sales revenue, which totals $3,000, and sales returns, which has a negative balance of $600. The sales returns entry is a contra-revenue account. It offsets gross revenue. It allows the company to accurately report its volume but also account for the portion of its revenue that won't make it into net income
Click to expand...
http://smallbusiness.chron.com/accounting-definitions-contrarevenues-vs-expenses-23874.html

Intel Corp.’s Contra-Revenue Strategy Was a Huge Waste of Money
Much ado about nothing.
Ashraf Eassa (TMFChipFool)
Apr 21, 2016 at 10:30AM
https://www.fool.com/investing/general/2016/04/21/intel-corps-contra-revenue-strategy-was-a-huge-was.aspx
 
DctHWjMXcAU1Nex.jpg

https://twitter.com/david_schor/status/993999812323348480

Interesting.
 


Reason is cost. The xeon Phi line is too expensive to produce for Intel. Xeon Phi are very big chips and aparently yelds are not good enough. This is the reason why Knights Hill, despite the design being complete since 2015, was cancelled. Well, this, and the 10 nm disaster.

In addition, for coprocessors, Intel has to add extra components (like HBM) that add a lot to the cost but don't produce any extra profit for Intel. Profit margin is lower for coprocessors than for other products. So, if processors are well accepted, no wonder they cancel the low margin line.

 
10nm! It's only an i3, but at least it's 10nm! This guys comments sound a bit critical, but Intel graph showed these 10nm parts were not going to be that much different from 14nm++.
The Infamous Cannonlake i3 8121U.
The Good, The Bad, The Ugly.
The Good. Intel is actually shipping 10nm processors but in low volume like they said in the Q4 2017 and Q1 2018 Earnings Calls and I am glad that they seem to be finally coming to market, at least soon.

The Bad. Its barely an improvement. Yes it supports LPDDR4, yes it supports SHA extensions and, yes it supports AVX512(on a mobile chip I do not know why) but all this should have come 2 years ago in 2016!

The Ugly. It is a 2.2GHz Dual Core with a boost up to 3.1GHz with no iGPU. That is disgraceful really, and it shows how much of a mess that Intel's 10nm really is. A chip that is 2 years late has no iGPU and it can't even surpass 3.5GHz in a 15w package like its 14nm brothers do? (See Kaby Lake R). That is both uninspiring and disappointing, Intel.
Click to expand...
http://tech-toniks.blogspot.co.uk/2018/05/intels-10nm-finally-yields-chip.html

Edit: Adding Intel process chart
intels_10%2B_and_10%2B%2B.png
 
10nm! It's only an i3, but at least it's 10nm! This guys comments sound a bit critical, but Intel graph showed these 10nm parts were not going to be that much different from 14nm++.
Click to expand...
What?

10nm parts are Cannonlake (and Tremont), while 14nm++ parts are Coffee Lake. That's a significant difference.
 


Let me clarify what I meant when I said,"10nm parts were not going to be much different from 14nm++." The performance illustrated in Intel's process chart shows 10nm will have less performance than 14nm++, but should be more power efficient. There are differences mentioned in the quote above shows the same 15W package like 14nm++, but operating at a lower frequency. I would have expected frequencies to be higher considering it's uses the same power(no iGPU either) as 14nm++, but they are actually lower. Intel's process chart isn't lining up with the finished product, which makes me wonder what is happening with 10nm manufacturing.
 
The chart specifically talks about clock speed. When you said "part", that meant a product made on 10nm, which is vastly different form talking just about the process node.

Kaby Lake-R chips have below 2GHz base clock. They're quad cores though. 3.2GHz max boost on a dual core is iffy, true.
 
That looks like a "bottom of the barrel" type of product... I'll give Intel the benefit of the doubt and think that they're under NDA with the OEM/Company actually buying the decent stuff (might even be Lenovo?) and can't really say they have more than just dual cores using 10nm.

If that is not the case, then they're in deep brown and shareholders will get angry soon.

Cheers!
 


Interesting:
Pairing with this CPU will be an AMD Radeon RX 540 2GB, as the Intel processor does not have integrated graphics - technically the silicon for them is there, but in this design it is disabled.
Click to expand...
Actually it's not disabled, it's defective due the the already known yield issues.
 


He seems to be very critical with intel 10 nm process
https://www.fool.com/investing/2018/05/08/3-reasons-intel-corp-should-shut-down-its-factorie.aspx

Anyway its inline with what i was told a couple weeks ago (had a few beers with some former Intel engineers).
I was a bit reluctant to believe them when they mentioned single digit yields, but apparently they knew what they where talking about ...

 


Yeah, he is a writer for The Motley Fool. He was a big supporter of Intel for a long time, and the more information he found out about Intel it's made him pretty mad. He was also joking about the single digits in a round about way a couple weeks ago...
 
That's because now they are not in a rush, ZEN+ is not as good as expected they closed the gap a bit but when a 6 core i7 8700K Coffee Lake still faster than The 8 core Ryzen 2700X, Intel is not worried and now they can take their time and play with the new 10nm even more . AMD's ZEN CPU's are using a 14nm process And the new ZEN+ are using a revised same 14nm process(AKA 12nm process). This isn't a new architecture; rather, it's a tweaked version of the first-generation Zen architecture and the die size and transistor count are the same for Zen and Zen+, The new chips are built on Global Foundries' 12nm revised process rather than the first-generation 14nm. " But just like AMD says that the overall die size and transistor counts are unchanged from the first generation: the GLOFLO isn't using the smaller process to pack the chip's transistors closer together. Rather, it's getting about another 250MHz and has reduced voltages by about 50mV.. The Problem here is that Intel's 14nm still offers more raw performance than GLOFLO's 14mm and 12nm.
 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom