What was their original schedule? Seems like they're on track: 2011 - Sandy Bridge, 2012 - Ivy Bridge, 2013 - Haswell, 2014 - Broadwell. Were you expecting it earlier?
This isn't too far off from their earlier announced schedule. But 14nm may be the last one in which we'll see the two year move. If they do go to 10nm, it may not be for three years. It's thought that moving down to that will be difficult, and moving below that may be almost impossible.
10nm isn't difficult, in a relative sense. You can either use quad patterning or EUV. It may be difficult to make economical, but Intel's 10nm R&D is finished, IIRC.
7nm is where things get hairy, however there's a fair bit of time between now and then.
Mike Mayberry, director of component research at Intel Corp. "Mayberry said that everything up to the 10-nm CMOS node – which is in development at Intel and will ramp production in 2015 - is effectively done. However, he said his job depends on being able to continue to double density and performance every two years beyond that, something for which the way forward is much less clear."
This is his quotes from the eetimes article Intel pushes for more research beyond 10-nm May 23, 2013 // by Peter Clarke
I'm sick and tired of Intel being focused on and caring more for 'battery longevity' improvements that would cater only to mobile/handheld markets rather than their DT audiences.
I'm going to be very upset if Broadwell turns out to be another Haswell where DT enthusiasts were seemingly neglected amidst of AMD's lack of presence in the performance segments.
Why would they focus on desktop? The mainstream consumer doesn't want desktops. Intel knows this from sales figures and customer orders. Why would they spend excess money supporting a market segment that is small and shrinking?
I argue it's a self-fulfilling prophecy. The more they target battery life improvements, the less desktop and (to a lesser extent) laptop products they'll sell. As long as they're fighting ARM on ARM's home turf, they're going to keep losing because their chips have more performance in a "good enough" fight and never have enough battery life and/or cheap enough cost to keep up with ARM.
If Intel wants to fight in a "good enough" world where cheap and battery life are the most important factors, then they need to work something out with MS to get Windows free. Otherwise, they're doomed to lose because Intel and Android is not going to do much.
Meanwhile, I'd argue they're alienating their own targetbase, which is why you're going to see them announcing next year or the year after that Haswell did so poorly they're focusing on Atom-like CPU's. The truth is Haswell is such a poor upgrade for enthusiasts they're ignoring it. IVB-E took two years to show up with ancient chipsets (no change) and old technology with little but downgraded overclocking to show for the trouble. The only thing Haswell-E has going for it is the no brainer 8-core instead of 6 core, so things look dire for the LGA20xx socket in general, but at least it's coming within a year.
Meanwhile, last I checked Intel is just doing another refresh (not even a dieshrink) of Haswell to suit the enthusiast market that already rejected Haswell as pathetic.
Perhaps if Intel had focused their resources on making a truly high end chip for the desktop/laptop markets while making a great battery life mostly high performance option for the thin laptops and tablets with Atom, they could use the nVidia strategy of having an expensive high end sell the medium to low end.
Instead, they've convinced even the enthusiasts at this point not to upgrade because they're obsessed with a market they aren't winning and aren't even close to winning. If they want to win it, they need to focus on it entirely and get costs down to the levels that would let them win it. As long as they think they can charge high margins, they're going to keep failing, great tech or not.
Well said. Desktop Haswell was crap for enthusiasts and gave no reason to upgrade over Ivy or even Sandy. I fear the days of Wolfdale and Sandy Bridge overclocks are behind us.
Then you're going to be disappointed for a long time as DT has peaked a few years ago.
There are more people buying laptops than DTs and that's where Intel is heading toward to make money. And since the mobile devices are increasing quicker than laptop/DT, Intel is also focusing on making mobile SoCs as well.
zeock9 -- Get a grip. Broadwell is a SoC (System-On-a-Chip) in the MOBILE marketing segment where the focus is on smaller die size, lower power and lower cost. The Desktop & Server segments don't use SoCs because they target higher performance. The have bigger die sizes than Mobile as well as higher power and higher cost. Intel learned long ago that one size doesn't fit all in computer chips
A 30% improvement in power efficiency would be very welcome for us BOINC crunchers! I don't neccessarily need it in a socket (as long as the mainboard doesn't break under sustained load), but give me 4 cores with a GT3 GPU and eDRAM.. with unlocked multipliers, power etc. and I'll upgrade from Ivy!
I doubt you will see a 30% gain in efficiency at full load, that is probably quoted during a 'tpical use' load, which will include lots of idle time, and thus clock gating/etc.
Considering how the PC market is in a downward spiral, I wonder if it's wise to have such massive capital spending on new nodes. They already had to delay the Irish 14nm fab because of lack of demand.
PC market isn't in a "downward spiral". It's stagnant, which is quite a bit different.
Also, if you didn't notice they're using the new nodes for mobile-focused SoCs, and making their desktop/laptop parts more power-efficiency focused in order to put them into smaller and more mobile platforms.
IDC says it's down 7% - 10% this year. I think eventually it'll pick up, but I still don't think now is the time to be burning captial on new process for chips that no one's buying. It's good that they finally have Atom SoCs on 22nm, but the margin for tablet SoCs is quite low - nowhere near the 60% they are used to. It won't make up for the slack in the PC market.
IDC data does have PC data down this year, but then flat-ish beginning next year IIRC. The key is if they have to revise it down or not. And they still have the desktop market pegged at 134M units a year, and AFAIK IDC only tracks vendor shipments. Jon Peddie Research tracks gaming hardware if you're looking for that kind of data.
It does not say that at all. You obviously didn't even read the report that you linked to.
"The market as a whole is expected to decline through at least 2014, with only single-digit modest growth from 2015 onward, and never regain the peak volumes last seen in 2011."
This doesn't offer complete clarity about some conflicting roadmaps stating Broadwell would be pushed on the desktop and replaced by Haswell refresh. The statement about Broadwell's power saving features at 14nm indicate to me it may still launch in the mobile space, but Haswell refresh will be the desktop parts for 2014.
Does anyone know (seen a slide or something), when we'll see 8core Intel CPUs? And I mean for the High End market, not those "E" types, where the tech is like 2 years old. As it's gonna be with Haswell-E, which is coming fall 2015, right?
Intel doesn't feel the desktop/mobile market wants or needs that. So no, its not gonna happen anytime soon. Lots of new stuff just dropped from Intel and it will be at least 6-12 months before this changes.
Case in point, For the Server Market, Intel just released: Ivy EP with up to 12c per CPU Avoton (Atom) with up to 8 cores for the Microserver market (Moonshot, SeaMicro, etc)
Meanwhile Haswell Desktop is still only 4 cores and Ivy E maxes out at 6 even though its VERY similar to Ivy EP which can go to 12.
Airmont at best will arrive for tablets (the "tablet chip" - i.e. the less optimized one). It won't really be available for smartphones until 2015, just like Merrifield won't be ready until next year. In 2015, ARM chip makers will be using 14nm/16nm (and FinFET), too. Intel is going to lose its advantage in process technology.
Ermmm :). Obviously you know the track record of Foundries, in my knowledge 28nm was a disaster in the first year of life. Global Foundries is in volume on 28nm only from a quarter or two. TSMC boss, more honestly, has said about a very "limited shipment" of 16nm devices in 2015, the bulk of will be on 20nm planar. ARM cpus will not be the firsts things built in 16nm/14nm, more likely high margins GPUs and FPGA. All foundries need to wait better yelds to give away their capacity for inexpensive Socs. My bet is ARM on 16/14nm in 2016.
Also the Broadwell power improvements are not surprising, considering Intel will not improve performance anymore. If anything, they'll go even more backwards in performance with Broadwell for "mobile" than they're doing with Haswell this year.
This is really one of their three-pronged misleading marketing campaign to achieve this. The first trick is to lower performance as much as possible, and hope nobody will notice, while promoting the "turbo-boost" speed as the "real" speed of the device - another misleading trick they've been using for Atom for smartphones/tablets.
The 3rd trick will be using SDP numbers instead of TDP, and hope the idiots in the tech media will start forgetting to add the SDP name everytime they mention wattage, and using the number alone, like this" Intel's Broadwell XYZ is a 4 watt chip". Period. No mention of SDP. I've already seen it happen.
Seems to conflict with all late June info. To me it's late if I can't get DESKTOP Broadwell chips in 2014. It seems some things are on time (or moved up? BGA or whatever) but new i-7's etc on broadwell will be late. All we'll be getting is the refresh. This is probably more a response to a lacking AMD than anything. Their cpus/servers are not in any trouble from AMD so that stuff is being pushed off. Even steamroller appears to be far behind even if you believe AMD's own numbers (10-20% won't catch a haswell even, let alone broadwell, heck wouldn't catch Ivy either probably in much).
So they can say broadwell in some form will ship in 2014, but not the form I'm after. Can't say I blame them really, as the competition is ARM now not AMD. http://hothardware.com/News/Intel-Broadwell-Report... Another one showing it's desktops delayed (which is the story pretty much everywhere). Google broadwell desktop delayed.
It is refreshing to see somebody on the internet back up an argument with links, but those you refer suggest a Haswell "refresh". Just as IB was a SB refresh or a "tick" to SB's "tock", so too could Haswell be the "tock" and Broadwell be the "tick" refresh.
Otherwise, what would a refresh be? Assuming it is Broadwell, 30% less wattage at the same clock would translate into ??? 10-15% more clock ??? at the same wattage? Or would 6 core chips become mainstream?
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
36 Comments
Back to Article
Soleron - Tuesday, September 10, 2013 - link
On schedule for 2014. Which is a completely different schedule to that originally announced.sorten - Tuesday, September 10, 2013 - link
What was their original schedule? Seems like they're on track: 2011 - Sandy Bridge, 2012 - Ivy Bridge, 2013 - Haswell, 2014 - Broadwell. Were you expecting it earlier?cmikeh2 - Tuesday, September 10, 2013 - link
If I remember correctly they were actually behind schedule and on pace for a 2015 release.pxc - Tuesday, September 10, 2013 - link
You remember a bad rumor. This official information should trump what that idiot Charlie vomits out on his click-bait web site.Homeles - Tuesday, September 10, 2013 - link
It's been 2014 for years, with 2013 being the year they'd be going into manufacturing.Coincidentally, that's exactly what this article says is going to happen.
CharonPDX - Tuesday, September 10, 2013 - link
They first announced it at IDF in 2011. Using that exact same slide: http://www.youtube.com/watch?v=VeCJd3CfeNostickmansam - Thursday, September 12, 2013 - link
Maybe thats why Haswell is listed as a "future product"melgross - Tuesday, September 10, 2013 - link
This isn't too far off from their earlier announced schedule. But 14nm may be the last one in which we'll see the two year move. If they do go to 10nm, it may not be for three years. It's thought that moving down to that will be difficult, and moving below that may be almost impossible.Homeles - Tuesday, September 10, 2013 - link
10nm isn't difficult, in a relative sense. You can either use quad patterning or EUV. It may be difficult to make economical, but Intel's 10nm R&D is finished, IIRC.7nm is where things get hairy, however there's a fair bit of time between now and then.
jfinely - Tuesday, September 10, 2013 - link
Mike Mayberry, director of component research at Intel Corp."Mayberry said that everything up to the 10-nm CMOS node – which is in development at Intel and will ramp production in 2015 - is effectively done. However, he said his job depends on being able to continue to double density and performance every two years beyond that, something for which the way forward is much less clear."
This is his quotes from the eetimes article
Intel pushes for more research beyond 10-nm
May 23, 2013 // by Peter Clarke
Casper42 - Tuesday, September 10, 2013 - link
"fact that they’re seeing significant power >gains< with 14nm Broadwell"Wouldn't it be easier to read if you said significant power SAVINGS with 14nm Broadwell?
I had to read that twice and the next line cleared it up for me.
Specifically Intel is already seeing a “30 percent power improvement” ...
zeock9 - Tuesday, September 10, 2013 - link
I'm sick and tired of Intel being focused on and caring more for 'battery longevity' improvements that would cater only to mobile/handheld markets rather than their DT audiences.I'm going to be very upset if Broadwell turns out to be another Haswell where DT enthusiasts were seemingly neglected amidst of AMD's lack of presence in the performance segments.
Fuck this shit.
AMD, you guys really need to step it up please.
stadisticado - Tuesday, September 10, 2013 - link
Why would they focus on desktop? The mainstream consumer doesn't want desktops. Intel knows this from sales figures and customer orders. Why would they spend excess money supporting a market segment that is small and shrinking?x347 - Tuesday, September 10, 2013 - link
Because wahhh they need to cater to MY needs, not what would be profitable for themHisDivineOrder - Wednesday, September 11, 2013 - link
I argue it's a self-fulfilling prophecy. The more they target battery life improvements, the less desktop and (to a lesser extent) laptop products they'll sell. As long as they're fighting ARM on ARM's home turf, they're going to keep losing because their chips have more performance in a "good enough" fight and never have enough battery life and/or cheap enough cost to keep up with ARM.If Intel wants to fight in a "good enough" world where cheap and battery life are the most important factors, then they need to work something out with MS to get Windows free. Otherwise, they're doomed to lose because Intel and Android is not going to do much.
Meanwhile, I'd argue they're alienating their own targetbase, which is why you're going to see them announcing next year or the year after that Haswell did so poorly they're focusing on Atom-like CPU's. The truth is Haswell is such a poor upgrade for enthusiasts they're ignoring it. IVB-E took two years to show up with ancient chipsets (no change) and old technology with little but downgraded overclocking to show for the trouble. The only thing Haswell-E has going for it is the no brainer 8-core instead of 6 core, so things look dire for the LGA20xx socket in general, but at least it's coming within a year.
Meanwhile, last I checked Intel is just doing another refresh (not even a dieshrink) of Haswell to suit the enthusiast market that already rejected Haswell as pathetic.
Perhaps if Intel had focused their resources on making a truly high end chip for the desktop/laptop markets while making a great battery life mostly high performance option for the thin laptops and tablets with Atom, they could use the nVidia strategy of having an expensive high end sell the medium to low end.
Instead, they've convinced even the enthusiasts at this point not to upgrade because they're obsessed with a market they aren't winning and aren't even close to winning. If they want to win it, they need to focus on it entirely and get costs down to the levels that would let them win it. As long as they think they can charge high margins, they're going to keep failing, great tech or not.
superjim - Wednesday, September 11, 2013 - link
Well said. Desktop Haswell was crap for enthusiasts and gave no reason to upgrade over Ivy or even Sandy. I fear the days of Wolfdale and Sandy Bridge overclocks are behind us.MikhailT - Tuesday, September 10, 2013 - link
Then you're going to be disappointed for a long time as DT has peaked a few years ago.There are more people buying laptops than DTs and that's where Intel is heading toward to make money. And since the mobile devices are increasing quicker than laptop/DT, Intel is also focusing on making mobile SoCs as well.
Hector2 - Wednesday, September 11, 2013 - link
zeock9 -- Get a grip. Broadwell is a SoC (System-On-a-Chip) in the MOBILE marketing segment where the focus is on smaller die size, lower power and lower cost. The Desktop & Server segments don't use SoCs because they target higher performance. The have bigger die sizes than Mobile as well as higher power and higher cost. Intel learned long ago that one size doesn't fit all in computer chipsMrSpadge - Tuesday, September 10, 2013 - link
A 30% improvement in power efficiency would be very welcome for us BOINC crunchers! I don't neccessarily need it in a socket (as long as the mainboard doesn't break under sustained load), but give me 4 cores with a GT3 GPU and eDRAM.. with unlocked multipliers, power etc. and I'll upgrade from Ivy!jwcalla - Tuesday, September 10, 2013 - link
Wouldn't a GPU be more power-efficient for BOINC?extide - Wednesday, September 11, 2013 - link
I doubt you will see a 30% gain in efficiency at full load, that is probably quoted during a 'tpical use' load, which will include lots of idle time, and thus clock gating/etc.lefty2 - Tuesday, September 10, 2013 - link
Considering how the PC market is in a downward spiral, I wonder if it's wise to have such massive capital spending on new nodes. They already had to delay the Irish 14nm fab because of lack of demand.kyuu - Tuesday, September 10, 2013 - link
PC market isn't in a "downward spiral". It's stagnant, which is quite a bit different.Also, if you didn't notice they're using the new nodes for mobile-focused SoCs, and making their desktop/laptop parts more power-efficiency focused in order to put them into smaller and more mobile platforms.
lefty2 - Tuesday, September 10, 2013 - link
IDC says it's down 7% - 10% this year. I think eventually it'll pick up, but I still don't think now is the time to be burning captial on new process for chips that no one's buying.It's good that they finally have Atom SoCs on 22nm, but the margin for tablet SoCs is quite low - nowhere near the 60% they are used to. It won't make up for the slack in the PC market.
jljaynes - Tuesday, September 10, 2013 - link
IDC data does have PC data down this year, but then flat-ish beginning next year IIRC. The key is if they have to revise it down or not. And they still have the desktop market pegged at 134M units a year, and AFAIK IDC only tracks vendor shipments. Jon Peddie Research tracks gaming hardware if you're looking for that kind of data.http://www.idc.com/getdoc.jsp?containerId=prUS2429...
CyberAngel - Wednesday, September 11, 2013 - link
Either way going to 14nm AND deploying the 450mm wafers cuts down the cost of a single chip and that is welcome!lefty2 - Wednesday, September 11, 2013 - link
It does not say that at all. You obviously didn't even read the report that you linked to."The market as a whole is expected to decline through at least 2014, with only single-digit modest growth from 2015 onward, and never regain the peak volumes last seen in 2011."
chizow - Tuesday, September 10, 2013 - link
This doesn't offer complete clarity about some conflicting roadmaps stating Broadwell would be pushed on the desktop and replaced by Haswell refresh. The statement about Broadwell's power saving features at 14nm indicate to me it may still launch in the mobile space, but Haswell refresh will be the desktop parts for 2014.dlder - Wednesday, September 11, 2013 - link
Does anyone know (seen a slide or something), when we'll see 8core Intel CPUs? And I mean for the High End market, not those "E" types, where the tech is like 2 years old. As it's gonna be with Haswell-E, which is coming fall 2015, right?Casper42 - Wednesday, September 11, 2013 - link
Intel doesn't feel the desktop/mobile market wants or needs that. So no, its not gonna happen anytime soon. Lots of new stuff just dropped from Intel and it will be at least 6-12 months before this changes.Case in point, For the Server Market, Intel just released:
Ivy EP with up to 12c per CPU
Avoton (Atom) with up to 8 cores for the Microserver market (Moonshot, SeaMicro, etc)
Meanwhile Haswell Desktop is still only 4 cores and Ivy E maxes out at 6 even though its VERY similar to Ivy EP which can go to 12.
Krysto - Wednesday, September 11, 2013 - link
Airmont at best will arrive for tablets (the "tablet chip" - i.e. the less optimized one). It won't really be available for smartphones until 2015, just like Merrifield won't be ready until next year. In 2015, ARM chip makers will be using 14nm/16nm (and FinFET), too. Intel is going to lose its advantage in process technology.Gondalf - Wednesday, September 11, 2013 - link
Ermmm :). Obviously you know the track record of Foundries, in my knowledge 28nm was a disaster in the first year of life. Global Foundries is in volume on 28nm only from a quarter or two.TSMC boss, more honestly, has said about a very "limited shipment" of 16nm devices in 2015, the bulk of will be on 20nm planar. ARM cpus will not be the firsts things built in 16nm/14nm, more likely high margins GPUs and FPGA. All foundries need to wait better yelds to give away their capacity for inexpensive Socs. My bet is ARM on 16/14nm in 2016.
Krysto - Wednesday, September 11, 2013 - link
Also the Broadwell power improvements are not surprising, considering Intel will not improve performance anymore. If anything, they'll go even more backwards in performance with Broadwell for "mobile" than they're doing with Haswell this year.This is really one of their three-pronged misleading marketing campaign to achieve this. The first trick is to lower performance as much as possible, and hope nobody will notice, while promoting the "turbo-boost" speed as the "real" speed of the device - another misleading trick they've been using for Atom for smartphones/tablets.
The 3rd trick will be using SDP numbers instead of TDP, and hope the idiots in the tech media will start forgetting to add the SDP name everytime they mention wattage, and using the number alone, like this" Intel's Broadwell XYZ is a 4 watt chip". Period. No mention of SDP. I've already seen it happen.
TheJian - Wednesday, September 11, 2013 - link
http://www.pcper.com/news/Processors/Intel-Not-Rel...http://www.dailytech.com/Report+Intel+Delays+14+nm...
Seems to conflict with all late June info. To me it's late if I can't get DESKTOP Broadwell chips in 2014. It seems some things are on time (or moved up? BGA or whatever) but new i-7's etc on broadwell will be late. All we'll be getting is the refresh. This is probably more a response to a lacking AMD than anything. Their cpus/servers are not in any trouble from AMD so that stuff is being pushed off. Even steamroller appears to be far behind even if you believe AMD's own numbers (10-20% won't catch a haswell even, let alone broadwell, heck wouldn't catch Ivy either probably in much).
So they can say broadwell in some form will ship in 2014, but not the form I'm after. Can't say I blame them really, as the competition is ARM now not AMD.
http://hothardware.com/News/Intel-Broadwell-Report...
Another one showing it's desktops delayed (which is the story pretty much everywhere). Google broadwell desktop delayed.
purerice - Thursday, September 12, 2013 - link
It is refreshing to see somebody on the internet back up an argument with links, but those you refer suggest a Haswell "refresh". Just as IB was a SB refresh or a "tick" to SB's "tock", so too could Haswell be the "tock" and Broadwell be the "tick" refresh.Otherwise, what would a refresh be? Assuming it is Broadwell, 30% less wattage at the same clock would translate into ??? 10-15% more clock ??? at the same wattage? Or would 6 core chips become mainstream?
everex11 - Monday, September 16, 2013 - link
my uncle recently purchased an awesome six month oldAudi A6 by working online... visit site......www.pick85.com/