Samsung’s first 14nm SoC is a 64-bit, 8-core Exynos aimed at high-end phones

New manufacturing process extends to other Exynos and third-party chips soon.

Read the whole story

 

[spartak]

It's amazing how Samsung has brought back the arrearage with Intel to about half a year, with Intel only now starting to ship 14nm in mass.

This used to be two and a half years only two or three years ago... Samsung's semiconductor business and resources should scare the hell out of Intel (and TSMC).

 

[ethd]

It's strange that Samsung hasn't brought Exynos phones stateside a lot earlier than now, at least a year or two. Hasn't Samsung had LTE chips in their Exynos-powered phones since at least the S4? From what I understood, that was the reasoning behind using Snapdragons originally.

 

[semakindidepan]

spartak[/url]":gfkkawit]It's amazing how Samsung has brought back the arrearage with Intel to about half a year, with Intel only now starting to ship 14nm in mass.

This used to be two and a half years only two or three years ago... Samsung's semiconductor business and resources should scare the hell out of Intel (and TSMC).

Except that Samsung's "14 nm" is not 14 nm. So, no.

 

[bug77]

rumor has it that problems with Qualcomm's new Snapdragon 810 have prompted Samsung to drop the chip from the S6

I think you can safely stop repeating that, now that the chip has been tested and proven to run quite bit cooler than previous Snapdragons.

 

[ethd]

semakindidepan[/url]":12vavjuy]

spartak[/url]":12vavjuy]It's amazing how Samsung has brought back the arrearage with Intel to about half a year, with Intel only now starting to ship 14nm in mass.

This used to be two and a half years only two or three years ago... Samsung's semiconductor business and resources should scare the hell out of Intel (and TSMC).

Except that Samsung's "14 nm" is not 14 nm. So, no.

Elaborating on your No True Scotsman fallacy makes it more difficult to immediately tear apart.

 

[BullBearMS]

It's fun to watch the third party Fabs close in on Intel's process node now that they have more than enough business to afford large R&D budgets.

ARM really has changed things.

 

[dagamer34]

ethd[/url]":6y2sju55]It's strange that Samsung hasn't brought Exynos phones stateside a lot earlier than now, at least a year or two. Hasn't Samsung had LTE chips in their Exynos-powered phones since at least the S4? From what I understood, that was the reasoning behind using Snapdragons originally.

I've always thought that US carriers were more confident testing Qualcomm basebands once they got such an early lead and other legitimate competitors never really showed up. Intel was delayed by years, Nvidia dropped out of favor for phones, TI/ STErriccson both closed up shop making phone chips, so there was very little reason to use anything other than a Qualcomm chip in the US.

 

[BullBearMS]

ethd[/url]":eg8chkl1]

semakindidepan[/url]":eg8chkl1]

spartak[/url]":eg8chkl1]It's amazing how Samsung has brought back the arrearage with Intel to about half a year, with Intel only now starting to ship 14nm in mass.

This used to be two and a half years only two or three years ago... Samsung's semiconductor business and resources should scare the hell out of Intel (and TSMC).

Except that Samsung's "14 nm" is not 14 nm. So, no.

Elaborating on your No True Scotsman fallacy makes it more difficult to immediately tear apart.

Here you go. Read this:

Chenming Hu, the coinventor of the FinFET, began by mapping out the near future. Soon, he said, we’ll start to see 14-nm and 16-nm chips emerge (the first, which are expected to come from Intel, are slated to go into production early next year). Then he added a caveat whose casual tone belied its startling implications: “Nobody knows anymore what 16 nm means or what 14 nm means.”

It’s actually become a fairly common refrain among industry experts. The practice of attaching measurements to chip generations has “been hijacked by marketers to an enormous extent,” one chip-design expert told me. “A lot of it’s really smoke and mirrors,” says analyst Dan Hutcheson of VLSI Research in Santa Clara, Calif. It’s “spin,” he says, that’s designed to hide widening technological gaps between chip companies.

http://spectrum.ieee.org/semiconductors ... omplicated

 

[evan_s]

ethd[/url]":1i7dip4j]It's strange that Samsung hasn't brought Exynos phones stateside a lot earlier than now, at least a year or two. Hasn't Samsung had LTE chips in their Exynos-powered phones since at least the S4? From what I understood, that was the reasoning behind using Snapdragons originally.

I think that it was still largely the capabilities of the integrated modems that kept the Snapdragon chips in the us versions of the phones. Other players have had LTE modems but they've been behind Qualcom and it seems like they are just finally getting to the good enough as far as specific LTE versions supported and bands support to be useful in the US market. Other markets had less LTE deployments and on fewer bands so it was easier to support them with Samsung's own chips.

 

[theoilman]

ethd[/url]":13bw6tfs]It's strange that Samsung hasn't brought Exynos phones stateside a lot earlier than now, at least a year or two. Hasn't Samsung had LTE chips in their Exynos-powered phones since at least the S4? From what I understood, that was the reasoning behind using Snapdragons originally.

my tmobile note 2 had an exynos chip. their chromebooks have exynos chips. their use stateside isn't as ubiquitous but they have used them. not using them as much in the states, as I understand it, was an issue with LTE support.

 

[Carewolf]

BullBearMS[/url]":15g184sa]

ethd[/url]":15g184sa]

semakindidepan[/url]":15g184sa]

spartak[/url]":15g184sa]It's amazing how Samsung has brought back the arrearage with Intel to about half a year, with Intel only now starting to ship 14nm in mass.

This used to be two and a half years only two or three years ago... Samsung's semiconductor business and resources should scare the hell out of Intel (and TSMC).

Except that Samsung's "14 nm" is not 14 nm. So, no.

Elaborating on your No True Scotsman fallacy makes it more difficult to immediately tear apart.

Here you go. Read this:

Chenming Hu, the coinventor of the FinFET, began by mapping out the near future. Soon, he said, we’ll start to see 14-nm and 16-nm chips emerge (the first, which are expected to come from Intel, are slated to go into production early next year). Then he added a caveat whose casual tone belied its startling implications: “Nobody knows anymore what 16 nm means or what 14 nm means.”

It’s actually become a fairly common refrain among industry experts. The practice of attaching measurements to chip generations has “been hijacked by marketers to an enormous extent,” one chip-design expert told me. “A lot of it’s really smoke and mirrors,” says analyst Dan Hutcheson of VLSI Research in Santa Clara, Calif. It’s “spin,” he says, that’s designed to hide widening technological gaps between chip companies.

http://spectrum.ieee.org/semiconductors ... omplicated

According to that Intels 14nm is not 14nm either. So why did you object to saying Samsung is catching up?

 

[hertzsae]

bug77[/url]":2mdl2p2d]

rumor has it that problems with Qualcomm's new Snapdragon 810 have prompted Samsung to drop the chip from the S6

I think you can safely stop repeating that, now that the chip has been tested and proven to run quite bit cooler than previous Snapdragons.

Exactly, the rumors (no official statement that I'm aware of) came from Samsung who:
1. Sells a competing product that is finally viable worldwide (LTE modems that work in the US)
2. Likely needs to greatly increase production now that Apple is also using other non-samsung foundries

The rumor did sound soooo plausible when we first it though.

 

[theoilman]

dagamer34[/url]":3i1fxq5y]

ethd[/url]":3i1fxq5y]It's strange that Samsung hasn't brought Exynos phones stateside a lot earlier than now, at least a year or two. Hasn't Samsung had LTE chips in their Exynos-powered phones since at least the S4? From what I understood, that was the reasoning behind using Snapdragons originally.

I've always thought that US carriers were more confident testing Qualcomm basebands once they got such an early lead and other legitimate competitors never really showed up. Intel was delayed by years, Nvidia dropped out of favor for phones, TI/ STErriccson both closed up shop making phone chips, so there was very little reason to use anything other than a Qualcomm chip in the US.

Verizon and sprint might be different, but Tmo and AT&T just use GSM and LTE standards. I don't think they do any testing on standards qualified devices, they just care about loading bloatware.

 

[sonicmerlin]

evan_s[/url]":3bamdjmq]

ethd[/url]":3bamdjmq]It's strange that Samsung hasn't brought Exynos phones stateside a lot earlier than now, at least a year or two. Hasn't Samsung had LTE chips in their Exynos-powered phones since at least the S4? From what I understood, that was the reasoning behind using Snapdragons originally.

I think that it was still largely the capabilities of the integrated modems that kept the Snapdragon chips in the us versions of the phones. Other players have had LTE modems but they've been behind Qualcom and it seems like they are just finally getting to the good enough as far as specific LTE versions supported and bands support to be useful in the US market. Other markets had less LTE deployments and on fewer bands so it was easier to support them with Samsung's own chips.

Qualcomm's newest 9x45 modems, integrated into the Snapdragon 810, are still at minimum a generation ahead of their competitors. Samsung is willing to forego Category 10 LTE for the cost savings, perhaps feeling that so few carriers are even capable of aggregating 3 separate 20 MHz channels that no one will notice.

 

[Brakiel]

ethd[/url]":308bm2p5]

semakindidepan[/url]":308bm2p5]

spartak[/url]":308bm2p5]It's amazing how Samsung has brought back the arrearage with Intel to about half a year, with Intel only now starting to ship 14nm in mass.

This used to be two and a half years only two or three years ago... Samsung's semiconductor business and resources should scare the hell out of Intel (and TSMC).

Except that Samsung's "14 nm" is not 14 nm. So, no.

Elaborating on your No True Scotsman fallacy makes it more difficult to immediately tear apart.

Samsung's 14 nm production is a hybrid design. It uses 14 nm for front end of line, and 20 nm for back end of line. Intel's production is 14 nm throughout.

http://en.wikipedia.org/wiki/Front_end_of_line
http://en.wikipedia.org/wiki/Back_end_of_line

 

[Uxorious]

So, with Samsung creating chips with big.LITTLE technology and Intel with Turbo Boost, I wonder when we will see a mix of these technologies on one CPU die.

For example, an ultra-low power core to run background processes while the phone is not being used, a few medium power cores for housekeeping processing while the phone is being used and the remainder being variable-power cores that can spike well beyond the median clock speed of the cores at rest to handle demanding loads quickly and then race back to idle.

 

[sonicmerlin]

BullBearMS[/url]":3035vmtn]

ethd[/url]":3035vmtn]

semakindidepan[/url]":3035vmtn]

spartak[/url]":3035vmtn]It's amazing how Samsung has brought back the arrearage with Intel to about half a year, with Intel only now starting to ship 14nm in mass.

This used to be two and a half years only two or three years ago... Samsung's semiconductor business and resources should scare the hell out of Intel (and TSMC).

Except that Samsung's "14 nm" is not 14 nm. So, no.

Elaborating on your No True Scotsman fallacy makes it more difficult to immediately tear apart.

Here you go. Read this:

Chenming Hu, the coinventor of the FinFET, began by mapping out the near future. Soon, he said, we’ll start to see 14-nm and 16-nm chips emerge (the first, which are expected to come from Intel, are slated to go into production early next year). Then he added a caveat whose casual tone belied its startling implications: “Nobody knows anymore what 16 nm means or what 14 nm means.”

It’s actually become a fairly common refrain among industry experts. The practice of attaching measurements to chip generations has “been hijacked by marketers to an enormous extent,” one chip-design expert told me. “A lot of it’s really smoke and mirrors,” says analyst Dan Hutcheson of VLSI Research in Santa Clara, Calif. It’s “spin,” he says, that’s designed to hide widening technological gaps between chip companies.

http://spectrum.ieee.org/semiconductors ... omplicated

That's a pretty cool article. In fact that whole website is pretty and has lots of interesting articles. Definitely saved for future reads. Thanks.

 

[tipoo]

14nm FinFET fab, eh? I wonder if Nvidia or AMD couldn't tap into some of that.

Intels 18 month to 2 year fab advantage over the universe seems to have shrunk substantially in the last few years. Mostly due to 14nm delays I guess, I wonder if they can jump ahead again.

 

[toukale]

spartak[/url]":18gr5a3r]It's amazing how Samsung has brought back the arrearage with Intel to about half a year, with Intel only now starting to ship 14nm in mass.

This used to be two and a half years only two or three years ago... Samsung's semiconductor business and resources should scare the hell out of Intel (and TSMC).

Not sure if TSMC is right because they are suing samsung over trade secrets leak by a former employee as the reason why samsung caught up so fast.
http://www.eetimes.com/document.asp?doc_id=1325589

 

[issor]

Brakiel[/url]":7sp27y58]

ethd[/url]":7sp27y58]

semakindidepan[/url]":7sp27y58]

spartak[/url]":7sp27y58]It's amazing how Samsung has brought back the arrearage with Intel to about half a year, with Intel only now starting to ship 14nm in mass.

This used to be two and a half years only two or three years ago... Samsung's semiconductor business and resources should scare the hell out of Intel (and TSMC).

Except that Samsung's "14 nm" is not 14 nm. So, no.

Elaborating on your No True Scotsman fallacy makes it more difficult to immediately tear apart.

Samsung's 14 nm production is a hybrid design. It uses 14 nm for front end of line, and 20 nm for back end of line. Intel's production is 14 nm throughout.

http://en.wikipedia.org/wiki/Front_end_of_line
http://en.wikipedia.org/wiki/Back_end_of_line

That, and I imagine the yields are higher on smaller die. One imperfection in the footprint of an Intel die= a bad die, whereas in the same footprint it might mean 1 bad exynos die and 8 good ones. Of course this is simplified and ignores yield enhancing tricks like downbinning, but you get the idea.

 

[foreignreign]

Is there really a point to having 8 cores on a smartphone CPU? I can understand wanting to future proof your tech and pushing out enhancements in a SoC line earlier than needed, but even 4 cores seems overkill to me as it stands right now.

 

[BullBearMS]

Carewolf[/url]":3un2ohc2]

BullBearMS[/url]":3un2ohc2]

ethd[/url]":3un2ohc2]

semakindidepan[/url]":3un2ohc2]

spartak[/url]":3un2ohc2]It's amazing how Samsung has brought back the arrearage with Intel to about half a year, with Intel only now starting to ship 14nm in mass.

This used to be two and a half years only two or three years ago... Samsung's semiconductor business and resources should scare the hell out of Intel (and TSMC).

Except that Samsung's "14 nm" is not 14 nm. So, no.

Elaborating on your No True Scotsman fallacy makes it more difficult to immediately tear apart.

Here you go. Read this:

Chenming Hu, the coinventor of the FinFET, began by mapping out the near future. Soon, he said, we’ll start to see 14-nm and 16-nm chips emerge (the first, which are expected to come from Intel, are slated to go into production early next year). Then he added a caveat whose casual tone belied its startling implications: “Nobody knows anymore what 16 nm means or what 14 nm means.”

It’s actually become a fairly common refrain among industry experts. The practice of attaching measurements to chip generations has “been hijacked by marketers to an enormous extent,” one chip-design expert told me. “A lot of it’s really smoke and mirrors,” says analyst Dan Hutcheson of VLSI Research in Santa Clara, Calif. It’s “spin,” he says, that’s designed to hide widening technological gaps between chip companies.

http://spectrum.ieee.org/semiconductors ... omplicated

According to that Intels 14nm is not 14nm either. So why did you object to saying Samsung is catching up?

I just said they (and the other third party Fabs) were catching up in a prior post.

I was just pointing out that your "no true Scotsman fallacy" was not a fallacy at all.

You can't just go by the name of the process node anymore now that the name is often more about marketing than reality.

 

[Bengie25]

tipoo[/url]":2rtolybl]14nm FinFET fab, eh? I wonder if Nvidia or AMD couldn't tap into some of that.

Intels 18 month to 2 year fab advantage over the universe seems to have shrunk substantially in the last few years. Mostly due to 14nm delays I guess, I wonder if they can jump ahead again.

Intel's 2 year advantage has primarily been with how soon they actually start selling retail. Intel rarely has been 2 years ahead in announcements.

AMD and Nvidia are both still on 32nm(28nm half-node), while Intel has had 22nm for quite a while now.

 

[melgross]

Is this SoC ready for use, or has it just been announced? From what I've been reading, it's hard to tell.

 

[panton41]

issor[/url]":381zxl5j]

Brakiel[/url]":381zxl5j]

ethd[/url]":381zxl5j]

semakindidepan[/url]":381zxl5j]

spartak[/url]":381zxl5j]It's amazing how Samsung has brought back the arrearage with Intel to about half a year, with Intel only now starting to ship 14nm in mass.

This used to be two and a half years only two or three years ago... Samsung's semiconductor business and resources should scare the hell out of Intel (and TSMC).

Except that Samsung's "14 nm" is not 14 nm. So, no.

Elaborating on your No True Scotsman fallacy makes it more difficult to immediately tear apart.

Samsung's 14 nm production is a hybrid design. It uses 14 nm for front end of line, and 20 nm for back end of line. Intel's production is 14 nm throughout.

http://en.wikipedia.org/wiki/Front_end_of_line
http://en.wikipedia.org/wiki/Back_end_of_line

That, and I imagine the yields are higher on smaller die. One imperfection in the footprint of an Intel die= a bad die, whereas in the same footprint it might mean 1 bad exynos die and 8 good ones. Of course this is simplified and ignores yield enhancing tricks like downbinning, but you get the idea.

An imperfection on a Core series CPU results in a lower binned CPU. It's not just the market shift to mobile driving Intel to release mobile chips first. You can deal with yield problems by cutting off the damaged sections with a laser and selling it as a lower end CPU with fewer cores and less cache. They've been doing it since the 486 and it's pretty much an industry standard practice these days.

Edit: I should have read closer, the OP already talked about downbinning.

 

[melgross]

foreignreign[/url]":1hirpaqd]Is there really a point to having 8 cores on a smartphone CPU? I can understand wanting to future proof your tech and pushing out enhancements in a SoC line earlier than needed, but even 4 cores seems overkill to me as it stands right now.

If we look at Anandtech's tests of the current chip, it seems as though it's very difficult to determine where an instruction should go, and it's very inefficient when using the LITTLE cores, so it's even using more power than using the BIg cores would, at times, because it's taking more time to complete the instruction. Then there's the overhead on the chip for the circuits that have to decide where those instructions go.

If they were eight of the same cores, it's would be different. I don't really understand what doing it this way is supposed to accomplish other than to market eight cores.

 

[BullBearMS]

Bengie25[/url]":1kdgfwiy]

tipoo[/url]":1kdgfwiy]14nm FinFET fab, eh? I wonder if Nvidia or AMD couldn't tap into some of that.

Intels 18 month to 2 year fab advantage over the universe seems to have shrunk substantially in the last few years. Mostly due to 14nm delays I guess, I wonder if they can jump ahead again.

Intel's 2 year advantage has primarily been with how soon they actually start selling retail. Intel rarely has been 2 years ahead in announcements.

AMD and Nvidia are both still on 32nm(28nm half-node), while Intel has had 22nm for quite a while now.

Global Foundries, IBM and Samsung share the same process node and R&D costs, so AMD won't be too far behind Samsung.

http://www.commonplatform.com/

 

[melgross]

panton41[/url]":1nz0xxhd]

issor[/url]":1nz0xxhd]

Brakiel[/url]":1nz0xxhd]

ethd[/url]":1nz0xxhd]

semakindidepan[/url]":1nz0xxhd]

spartak[/url]":1nz0xxhd]It's amazing how Samsung has brought back the arrearage with Intel to about half a year, with Intel only now starting to ship 14nm in mass.

This used to be two and a half years only two or three years ago... Samsung's semiconductor business and resources should scare the hell out of Intel (and TSMC).

Except that Samsung's "14 nm" is not 14 nm. So, no.

Elaborating on your No True Scotsman fallacy makes it more difficult to immediately tear apart.

Samsung's 14 nm production is a hybrid design. It uses 14 nm for front end of line, and 20 nm for back end of line. Intel's production is 14 nm throughout.

http://en.wikipedia.org/wiki/Front_end_of_line
http://en.wikipedia.org/wiki/Back_end_of_line

That, and I imagine the yields are higher on smaller die. One imperfection in the footprint of an Intel die= a bad die, whereas in the same footprint it might mean 1 bad exynos die and 8 good ones. Of course this is simplified and ignores yield enhancing tricks like downbinning, but you get the idea.

An imperfection on a Core series CPU results in a lower binned CPU. It's not just the market shift to mobile driving Intel to release mobile chips first. You can deal with yield problems by cutting off the damaged sections with a laser and selling it as a lower end CPU with fewer cores and less cache. They've been doing it since the 486 and it's pretty much an industry standard practice these days.

Binning is more than finding defective cores. It's also that some chips will run faster, and some slower, on the same wafer.

The problem is that chip manufacturing isn't perfect, and at these small sizes, just a fraction of a micron is enough to change the characteristics of parts of a chip.

 

[issor]

panton41[/url]":3lkbqkt5]

issor[/url]":3lkbqkt5]

Brakiel[/url]":3lkbqkt5]

ethd[/url]":3lkbqkt5]

semakindidepan[/url]":3lkbqkt5]

spartak[/url]":3lkbqkt5]It's amazing how Samsung has brought back the arrearage with Intel to about half a year, with Intel only now starting to ship 14nm in mass.

This used to be two and a half years only two or three years ago... Samsung's semiconductor business and resources should scare the hell out of Intel (and TSMC).

Except that Samsung's "14 nm" is not 14 nm. So, no.

Elaborating on your No True Scotsman fallacy makes it more difficult to immediately tear apart.

Samsung's 14 nm production is a hybrid design. It uses 14 nm for front end of line, and 20 nm for back end of line. Intel's production is 14 nm throughout.

http://en.wikipedia.org/wiki/Front_end_of_line
http://en.wikipedia.org/wiki/Back_end_of_line

That, and I imagine the yields are higher on smaller die. One imperfection in the footprint of an Intel die= a bad die, whereas in the same footprint it might mean 1 bad exynos die and 8 good ones. Of course this is simplified and ignores yield enhancing tricks like downbinning, but you get the idea.

An imperfection on a Core series CPU results in a lower binned CPU. It's not just the market shift to mobile driving Intel to release mobile chips first. You can deal with yield problems by cutting off the damaged sections with a laser and selling it as a lower end CPU with fewer cores and less cache. They've been doing it since the 486 and it's pretty much an industry standard practice these days.

Exactly. That's what I was referring to with my last sentence. It is still just a stopgap though, because at some point you still have to be able to have a good yield. They don't want to downbin 100% of a wafer. Every CPU downbinned is a loss even though it ia mitigated to be usable, whereas a smaller die could still yield a few 100% good die in that same footprint.

 

[Brakiel]

Bengie25[/url]":2twx866x]

melgross[/url]":2twx866x]Is this SoC ready for use, or has it just been announced? From what I've been reading, it's hard to tell.

If we use history as a guide, give it 1-2 years. Intel doesn't have any 14nm devices yet, so I doubt anyone else is near ready to ship.

Broadwell is shipping. We've already seen a few 5000-series systems come out, such as the Gigabyte Brix 5500 and Dell XPS 13.

 

[Ushio]

tipoo[/url]":35jknkou]14nm FinFET fab, eh? I wonder if Nvidia or AMD couldn't tap into some of that.

Intels 18 month to 2 year fab advantage over the universe seems to have shrunk substantially in the last few years. Mostly due to 14nm delays I guess, I wonder if they can jump ahead again.

Doubtful since next is 11/10nm and no one has even mentioned having even viable test chips of that size where as with the 16/14nm process the first test chips were shown off in 2009 with retail last year.

 

[HeidiMann]

I remember all the Android fanboys talking smack when Apple went to 64-bit CPUs in phones.

 

[aexcorp]

Samsung's process is 20nm BEOL; the interconnect are all still 20nm, but transistors are in 14nm. Intel's process has the entire chip in 14nm, interconnect included. Intel still retains a marked advantage in density, and is shipping 14nm chips NOW (Broadwell is in end users' hands, Cherry Trail is already in OEMs' hands and soon to be in end users' hands).

In the case of Samsung, I would be utterly shocked if this exact chip (the "14nm" design mentioned here) shows up in products in the next 6 months. I don't think their yields will be sufficient to accumulate the inventory they need for the S6 for example.

Of course, time will tell...

 

[BullBearMS]

issor[/url]":5uog2x4s]a smaller die could still yield a few 100% good die in that same footprint.

I think that the fact that Intel has only put out the very smallest Broadwell versions (Y and dual core U) shows that they have not been able to work out the flaws spread out across a wafer and are relying on the small part size to get more usable parts off the wafer.

 

[Bengie25]

Brakiel[/url]":3oxc35jf]

Bengie25[/url]":3oxc35jf]

melgross[/url]":3oxc35jf]Is this SoC ready for use, or has it just been announced? From what I've been reading, it's hard to tell.

If we use history as a guide, give it 1-2 years. Intel doesn't have any 14nm devices yet, so I doubt anyone else is near ready to ship.

Broadwell is shipping. We've already seen a few 5000-series systems come out, such as the Gigabyte Brix 5500 and Dell XPS 13.

I forgot that Broadwell is shipping on non-desktop platforms. Thanks for the correction.

I'm just waiting for a quad core 3ghz+ Broadwell for the desktop,

 

[BullBearMS]

aexcorp[/url]":1mb9g2v7]I would be utterly shocked if this exact chip (the "14nm" design mentioned here) shows up in products in the next 6 months. I don't think their yields will be sufficient to accumulate the inventory they need for the S6 for example.

Of course, time will tell...

Apple has the largest SOC's on the market (being willing to trade off die size for power reductions) and the word is that Samsung has managed to steal the A9 away from TSMC because they were getting a better yield.

http://recode.net/2015/02/04/apple-taps ... hone-chip/

 

[redleader]

aexcorp[/url]"5bf78rx]Samsung's process is 20nm BEOL; the interconnect are all still 20nm, but transistors are in 14nm. Intel's process has the entire chip in 14nm, interconnect included. Intel still retains a marked advantage in density, and is shipping 14nm chips NOW (Broadwell is in end users' hands, Cherry Trail is already in OEMs' hands and soon to be in end users' hands).

In the case of Samsung, I would be utterly shocked if this exact chip (the "14nm" design mentioned here) shows up in products in the next 6 months. I don't think their yields will be sufficient to accumulate the inventory they need for the S6 for example.

Of course, time will tell...

Yeah the phrasing of the ars article is quite different than the phrasing of the source, which merely says that they'll have 14 nm SOCs eventually, not that they're all that close to shipping.