At Intel's Technology and Manufacturing Day event in San Francisco there was a lot of talk about how Intel's 14nm process technology compares to the 16nm, 14nm, and 10nm offerings of their competitors. Investors and enthusiasts are curious if Intel can hold their lead in process tech as Samsung seems to be on track to release chips fabbed on 10nm process before Intel will. Intel rightly pointed out that not all process tech is measured the same way and that pitch measurements give only one part of the picture; meaning Samsung might not actually be smaller than them.
The Tech Report were present at that meeting and have written up an in depth look at what Intel means when they dispute the competitions claims, as well as their rationale behind their belief that the 14nm node still has a lot of life left in it.
"As process sizes grow smaller and smaller, Intel believes that the true characteristics of those technology advances are being clouded by an over-reliance on a single nanometer figure. At its Technology and Manufacturing Day this week, the company defended its process leadership and proposed fresh metrics that could more accurately describe what a given process is capable of."
Here is some more Tech News from around the web:
- Scientists Discover Way To Transmit Taste of Lemonade Over Internet @ Slashdot
- There's a Samsung Galaxy S8 Microsoft Edition, for some reason @ The Inquirer
- 'Trash-80' escapes the dustbin of history with new TRS-80 emulator @ The Register
- Beyond Zelda: The first month of Switch games acts as a promising crystal ball @ Ars Technica
- ZX Spectrum Vega Plus backers complain of months-long refund delays @ The Register
- Microsoft wants screaming Windows fans, not just users @ The Register
- GDC 2017 and NVIDIA Editor's Day Coverage @ Neoseeker
How about taking 14-22nm
How about taking 14-22nm process, and making a physically larger chip?
Why are we still stuck on a 2″ square silicon waffer? Aside from shaking up the ‘standard’ is there a reason for that particular size?
1.greed.
2.cost.
3.they got a
1.greed.
2.cost.
3.they got a cow to milk,too early to think of anything else.
4.they have to make space of IGP, so they can force sell you something that 90% of desktop users are never using.
Thought so. Was hoping for a
Thought so. Was hoping for a tech hurdle to overcome….
Though in my case, the iGPU gets used for twitch/YT streaming fairly regular. The VCE option on my R9 390 looks like hash.
If you look at the 7700k, it
If you look at the 7700k, it see that the IGP take more space then the 4 CPU cores.
Wouldn’t you prefer to have 4 CPU core VS an IGP/VCE if the cost was equal ?
$340 – I7-7700K : 4 core + IGP
or
$340 – i8-7700K : 8 core
Would not cost Intel any extra to fab/manufacture as the die size would be the same.
And you could run x264 on those 4 core for even better encoding quality…
Or get the R7 1700 for $330 that already comes with 8 core and no IGP 🙂
1 square inch of wood cost
1 square inch of wood cost alot less than 1 square inch of gold.
the point is not the space itself, but how expensive the stuff is in that space.
I wish Ryzen had a very basic
I wish Ryzen had a very basic tiny igpu just for purposes of keeping power consumption low, and card slots free for server purposes. A 2d igpu should only need 1-2mm2 at this point..
All motherboards used to
All motherboards used to provide that very thing before Intel and AMD started wasting everyone’s time and money with garbage integrated CPU graphics.
The reality is that the only
The reality is that the only users in the consumer market who need discrete graphics are PC gamers and we make up a minority part of the market. Its simpler for Intel to not deal with another product line for a small market segment.
There are also things like nvidia’s Optimus which also systems to save power by disabling the discrete GPU and using the onboard GPU for simple tasks.
Square is due to simple
Square is due to simple physical properties of the wafer that makes it easy to cut into those shapes.
It’s extremely likely to shatter if cut into other shapes.
As for the size, smaller and smaller means heat concentration grows.
Which is why energy efficiency is getting more and more important for chips.
Change square to rectangular.
Change square to rectangular.
What if we rotate it 45
What if we rotate it 45 degrees so we have diamonds.
Are you talking about the die
Are you talking about the die itself or the full cpu with heatspreader and all?
Nice, but now is the time to
Nice, but now is the time to make use of it.
Yes but does Intel use the
Yes but does Intel use the saved space for better graphics at a lower price point for any laptop SOCs or does Intel just use that space saved for more dies/wafer and no savings passed on to the consumer. And what off brand of toothpaste does Intel use on these denser packed CPU dies with their circuits packed closer together that need a much better thermal compound to properly dissipate the heat into the heat spreader. Intel’s ability to get with its 14nm process more densely packed circuits is great for Intel but maybe Intel should be passing some more cost savings on to the consumer or Intel’s should be giving its laptop SOC SKUs some better graphics at a lower price for the consumer and not milking the space savings for all its worth and giving the consumer nada in return.
AMD’s Samsung/GF 14nm process sure is nice for the price point for consumers from AMD on its R7 series SKUs and I’ll bet that any Zen/Vega Ryzen branded APU SKUs will have some very nice graphics at a much lower price point when they arrive in a short time with some nice solder between the die and the heat spreader. Intel sure has the funds to spaff away on contra revenue and marketing spin, but where is the results to show for billions wasted. AMD with just a fraction of the R&D funding compared to what Intel has wasted on Contra Revenue and marketing has produced some very good competition this time around and Intel can only rehash what is already known for some marketing spin to try and fool the slack jawed yokels out there.
Could… Intel, when moving
Could… Intel, when moving to 10nm, sell its 14nm fab capacity to nvidia/amd to fab GPU ? (VS just shutting it down)
I can understands Intel not wanting to open its fab to AMD to build Ryzen processors…
But what stop Intel from opening capacity to nvidia or AMD ?
Wouldn’t that alone add a few billion in revenue (low margin, but can intel be that picky at this point in time ?)
What do you mean! Intel’s
What do you mean! Intel’s 14nm lines don’t go poof when Intel enters 10nm tooling or risk production in 2018. Most CPU micro-architectures stay in production for a while after the newer micro-architectures at smaller nodes go into full production. And Intel’s server lines are usually behind in getting that latest process node/micro-architecture by some months time in getting the latest process node/micro-architectures updates for the server SKUs.
Intel will plan in advance the transition from 14nm to 10nm and that process is done over a number of months. The only thing that can effect Intel greatly is sudden loss of market share and Intel has in the past mothballed a new chip fab(Fab 42) until just recently. Most of the chip fab eguipment can be still used for any smaller processer node work except for some newer specilized 10nm eguipment that Intel has to get from the third party market that supplies the entire world’s chip fab equipment needs. There are about 6 chip fab line equipment suppliers that supply Intel, Samsung, GF, TSMC/others with the specilized chip fabrication equipment to create those 14nm/smaller node size chip plants(Fabs).
Intel has already sold some limited fab capacity for non competing device production, but AMD is locked into to GF for most of its production needs and Nvidia is already using TSMC’s 16nm node for Pascal and Nvidia will have had to already scheduled Volta for production with a fab partner by now as that has to be done months in advance for obvious engineering reasons.
If anything happens in the marketplace for Intel it takes a few years to get any fab capacity sold to any third party processor makers and the engineering needs plenty of lead time to pull that off, Months of lead time.
Great. So if nvidia / Intel
Great. So if nvidia / Intel where in talk 2 years ago about moving some 14nm production this could be announced in the month to come ?
Thank you, thats an exciting prospect !
So, how many other third party is Intel fabing for already ?
Also AMD did pay GF a fair amount ($300+m) to revise itsa contract so it can outsource some fab outside GF… So they are not bound to fully bound to GF. Could this indicate AMD might have a deal with Intel to fab at 10nm ?
Intel is not used to doing
Intel is not used to doing large discrete GPU dies and I think that Nvidia is already going with TSMC’s 12nm process(1) for Volta. And don’t you think that if Nvidia decided to go with Intel that such news would have already been reported on for a good while if that arrangement was made 2 years ago! Nvidia is a publicly traded company and can not hide such important info from their investors and the press as a shift to a new foundry partner.
Really Nvidia is not going to risk their big Volta supeercomputer contract with IBM on Intel who has no experience fabricating GPUs. And under AMD’s new agreement with GF AMD still has to pay GF for wafers produced at other other fab companies. AMD just renegotiated those contract terms in case GF really screwed up that licensed from Samsung 14nm proceess and AMD was left without any options. With that new wafer agreement AMD would still have to pay GF for any wafers produced at say Samsung, but at least AMD could get its CPUs made if GF really screwed up that 14nm node shift from from GF’s 28nm process.
“So, how many other third party is Intel fabing for already?”
I think there was only one customer for some 10nm contract for ARM SOCs maybe rumors of LG using Intel. But Intel may be using it’s fabs more for its own memory SKUs than CPUs, as really Intel still has not much third party fab contracts to speak of other than remors of contracts at this point in time.
(1)
“Nvidia Volta to be manufactured on TSMC 12nm process”
http://hexus.net/business/news/components/103456-nvidia-volta-manufactured-tsmc-12nm-process/
Cannot hide because they are
Cannot hide because they are publicly traded ?
Did you know 2 years ago that nintendo & nvidia where working together on the switch ?
Most deals are kept secret. For example, we know that AMD as a few deals done but we dont know with who.
Announcement, is private matter , even for publicly traded companies.
One for example might be the Intel GPU IP licensing.
nvidia told investor that the licensing deal will not be renewed,
but Intel did not say who they are going to license the tech from now on.. or if they even will do a licensing deal at all.
Anyways… my point is , if Intel fabs are so great, they could sell its services to no competing companies.
Look at TSMC, its their entire business model and they make billion upon billions in revenue. And as intel stated, “our process is so much better then TSMC”
It seem to be a win/win/win situation. Intel grows revenue by the billion, non competitive company get an edge over their competitors, consumer get higher end products.
So I wouldn’t be surprised if we hear in the next 12 month that Intel is opening its fab even more, because of being a publicly traded company.. it needs to raise its revenue.
Intel has been trying to get
Intel has been trying to get its third party fab business going for some years now, starting in about the same time frame as the empty fab 42 building was put into mothballs because of the PC market decline. Most of the OEM device makers without their own in-house CPU/GPU engineering resources are using ARM Holdings’ consulting services and are using some other Arm Holdings IP under arrangements with Arm Holdings’ chosen fab partner to make up complete working SOC systems for the OEM’s devices. So those OEM’s will contract with Arm Holdings system intergration services to design systems around Arm Holding’s refrence design cores.
So Only the top tier Arm Holdings ARMv8A ISA Architectural licensees(Apple, Samsung, Qualcomm, AMD/others) and maybe a few of the larger semi-custom ARM refrence design licensees have their own in-house SOC engineering and SOC IP integration services engineers in-house and can pick and choose their own fab partner. Most of the ARM Holdings smaller licensees are using some of Arm Holdings services that include a chip fab arrangement where that chip fab partner comes already prearranged as part of the total package of services from ARM Holdings. So Apple has Samsung/TSMC mostly and AMD has GF, and probably Samsung as a fallback, while Samsung has its own fab division and Qualcomm uses Samsung/TSMC. These third party chip fab companies represent total unit production scales that dwarfs Intel’s mostly in-house chip fab capacity and economy of scale.
I just do not see Intel gaining much market traction very quickly with the total amount of competition for fab services and the high volume sales that derive most of the income from very thin margins. Intel has never been a low margin player to that degree in its history.
Intel may become like IBM is now, with Big Blue(IBM) heading towards becoming a shell of its former self.
Thank goodness that the Power CPU IP is out there now from OpenPower so the Power8 and Power9 IP can be used by many third party OpenPower licensees, ditto for the OpenCAPI IP for the OpenCAPI consortium(AMD, IBM, other founding members).
This is one of those rare
This is one of those rare occasions where people are proud to have it smaller than others.
Its so tiny I can barely see
Its so tiny I can barely see it
Cool.. Now start making CPU’s
Cool.. Now start making CPU’s with more cores with a decent pricetag, instead of milking us
intel lies are epic !
intel lies are epic !
And Charlie the D. over at
And Charlie the D. over at S/A chimes in using some of that Intel history with process node marketing obfuscation. It’s a good long read.
“Is Intel’s Hyperscaling really a change?
http://semiaccurate.com/2017/04/04/intels-hyperscaling-is/