Intel Confirms Devil's Canyon, Haswell-E, Broadwell Rumors

[InvalidError]

I don't get the continual quest for lower power in a high-end CPU.In my eyes, high-end should be high power consumption, but high performance, increasing with each cycle.

There is one major problem with that: it would require a completely different hardware architecture and the high-end market that MIGHT be interested in chips with much worse performance per watt but higher overall performance is much too small to justify that.

 

[RedJaron]

I don't get the continual quest for lower power in a high-end CPU.In my eyes, high-end should be high power consumption, but high performance, increasing with each cycle.

There is one major problem with that: it would require a completely different hardware architecture and the high-end market that MIGHT be interested in chips with much worse performance per watt but higher overall performance is much too small to justify that.

Not only that, but more power also equals more heat, which usually leads to lower efficiency. It wouldn't matter how much extra performance you crammed into a chip if it was the size of a Post-It note and had a 400W TDP.

The lower power consumption of late is partially tied to a focus on mobile and sharing architectures across product lines. However it's also a by-product of better fab processes. Even if lower power consumption isn't big on everyone's mind, it's surely appreciated nonetheless by people who are paying the electric bill and don't want to spend a bundle on CPU cooling.

 

[InvalidError]

The lower power consumption of late is partially tied to a focus on mobile and sharing architectures across product lines.

I would say it is mostly about most people not needing or even wanting high-power chips since today's low-power chips are already overkill for most people's needs.

As for better fab process, until the turn toward power-efficiency, process gains were usually overwhelmed by increased transistor count and faster clocks, yielding higher net TDP. CPU manufacturers could certainly have continued on this path if they wanted to - replace the IGP running at ~900MHz by four cores using a similar surface area but running at ~3.5GHz and you just increased the chip's TDP by ~50W.

 

[RedJaron]

I would say it is mostly about most people not needing or even wanting high-power chips since today's low-power chips are already overkill for most people's needs.

As for better fab process, until the turn toward power-efficiency, process gains were usually overwhelmed by increased transistor count and faster clocks, yielding higher net TDP. CPU manufacturers could certainly have continued on this path if they wanted to - replace the IGP running at ~900MHz by four cores using a similar surface area but running at ~3.5GHz and you just increased the chip's TDP by ~50W.

Yes, I meant to imply that ( obviously I didn't succeed. ) Since there wasn't a big need/demand to continue cranking chip performance, they had the option to consolidate architecture research and make something that was more scale friendly for both mobile and desktop. Since performance didn't need to climb so fast, why not save on the thermal headroom?

For those niche users that still want as much performance as possible, power and cooling requirements be damned, we have the -E series.

 

[InvalidError]

For those niche users that still want as much performance as possible, power and cooling requirements be damned, we have the -E series.

And for cases where price is not a major concern either, there are multi-socket Xeons with up to 16 cores per socket.

 

[southernshark]

The reason they do 10% upgrades every year is because there is no one to compete. Even when AMD tried to match Intel they were not a real threat. The GPU market is VERY competitive at the moment, look at rebranded $450 GTX 680's (770's) selling for $330. Its incredible. If we had that in the CPU market, things might actually go DOWN in price every year.

Maybe but the fact is that sales are way down and I believe that a big reason for that is the lack of meaningful difference. Pondits like to blame tablets and mobile computing for the decline in the PC, but I'm convinced that a big reason people don't buy new PCs is because the new ones don't offer meaningful difference. I used to upgrade every 2-3 years, but now feel zero need to do so.If we were talking about a 6GHZ CPU, I'd be the first in line. But a theoretical 10 percent improvement in certain applications, with little real world difference hardly makes me break out my wallet. If Intel wants people to buy new PCs, Intel needs to offer them a reason to do so.

 

[knowom]

The reason they do 10% upgrades every year is because there is no one to compete. Even when AMD tried to match Intel they were not a real threat.

Yeah they weren't a threat and that's why Intel resorted to using monopoly tactics? Right because I'm certain that had no adverse effects on AMD what so ever and that's why they lost the lawsuit against them.AMD wasn't competing with them though that's why AMD64 crippled the hot running P4? I don't particularly care for AMD they turned their backs on Nvidia who was also the victim of Intel's evil monopoly ways instead of embracing them.Nvidia helped AMD grow it's CPU market share with it's GPU's and SLI tech in the AMD64 days back before crossfire came about.

 

[RedJaron]

Maybe but the fact is that sales are way down and I believe that a big reason for that is the lack of meaningful difference. Pondits like to blame tablets and mobile computing for the decline in the PC, but I'm convinced that a big reason people don't buy new PCs is because the new ones don't offer meaningful difference. I used to upgrade every 2-3 years, but now feel zero need to do so.If we were talking about a 6GHZ CPU, I'd be the first in line. But a theoretical 10 percent improvement in certain applications, with little real world difference hardly makes me break out my wallet. If Intel wants people to buy new PCs, Intel needs to offer them a reason to do so.

It's not Intel's job to make other companies' software. It's also not Intel's fault that recent software hasn't required any massive jump in computing power. Mid- to -upper-range CPUs have had processing power to spare for years now.

First off, your example of a 6GHz CPU is flawed since clock speed is only a small part of overall compute power. If you were to take into account architecture improvements, instruction set acceleration, IPC, etc, I think you'd find that a 3.5GHz Haswell performs just as well ( if not better, ) as a 6GHz Penryn. So in that sense, yes, Intel ( and AMD by and large, ) has delivered.

Second, suppose you could have a quad-core 6GHz Haswell right now. What would you do with it? What do you have that would possibly tax it? The only applications right now that could top out such a chip are the well-threaded professional apps that have been hungry for more resources as long as they've been around. Users that need such gargantuan power get better served with the multi-socket server crowd ( or even GPU assistance. ) No OS, productivity software, game, or home office use has come close to taxing a mid-range or better CPU before its EOL.

Bottom line, if you're not not hurting for power right now, why are you bitching that you don't have more?

 

[InvalidError]

Pondits like to blame tablets and mobile computing for the decline in the PC, but I'm convinced that a big reason people don't buy new PCs is because the new ones don't offer meaningful difference.

If you have to choose between spending $1000 on a new PC or laptop that is only ~30% faster than the one you bought 2-3 years ago and spending $400 on a new tablet or smartphone that is 150-250% faster than models from two years ago, I suspect most people who have grown fond of tablet/smartphone computing would choose to upgrade their mobile device without giving it much of a second thought.

 

[IQ11110002]

Pondits like to blame tablets and mobile computing for the decline in the PC, but I'm convinced that a big reason people don't buy new PCs is because the new ones don't offer meaningful difference.

If you have to choose between spending $1000 on a new PC or laptop that is only ~30% faster than the one you bought 2-3 years ago and spending $400 on a new tablet or smartphone that is 150-250% faster than models from two years ago, I suspect most people who have grown fond of tablet/smartphone computing would choose to upgrade their mobile device without giving it much of a second thought.

I'll still take the pc upgrade thanks! Tablet for web browsing doesn't need the speed, Where as my pc games and photoshopping loves any cpu/gpu love I can throw at it.

 

[Alec Mowat]

Honestly; I'm far more interested in what AMD is doing. Hopefully AMD adopts DDR4, otherwise it might go the way of RAMBUS.

 

[an]

Meh, I'll stick with my FX-8320. Improved TIM doesn't mean its as reliable/overclockable as solder, and it could also be a relative improvement which means nothing.

 

[an]

Meh, I'll stick with my FX-8320. Improved TIM doesn't mean its as reliable/overclockable as solder, and it could also be a relative improvement which means nothing.

 

[InvalidError]

I'll still take the pc upgrade thanks! Tablet for web browsing doesn't need the speed, Where as my pc games and photoshopping loves any cpu/gpu love I can throw at it.

How many people would bother upgrading their PC for a 30% performance gain? For most mainstream games and applications, this would make only a barely perceivable difference aside from benchmarks, which is why you see so many people with i5/i7-2xxx skipping Ivy Bridge, Haswell.and probably Broadwell too. Enthusiasts and prosumers represent less than 10% of the total PC market; the rest will be mostly fine with 4-5 years old PCs. If XP and Vista's combined market shares are to be believed, about 35% of PCs are over eight years old.

As for tablets, going from the original N7 to the N7v2 is a pretty substantial responsiveness improvement and the N7v2 still has a fair amount of room for improvement, even more so for lower-end tablets. For many people, interest in tablets only started with the multiple options under $200 (tablet sales boomed after the flood of $170-200 devices) and it will take at least two more years for performance below $200 to reach a point where people might stop ogling annual refreshes the same way they do with desktop CPUs today.

 

[Impatienc3]

Only thing with Iris Pro is that *if* it's for higher end CPUs, those buying them will more than likely have a beefy GPU making it pointless.

The slide does say "unlocked desktop processors" and that does usually mean the highest-end of the range.Unlocked desktop processors YOU say... Unlocked Pentium THEY say.. Is this Intel's answer to AMD's APU's? An overclockable budget chip with decent graphics..? *Hopes*IIRC, there was a leaked slide a few months ago that showed the eDRAM "L4 cache" being standard across the Broadwell lineup. I would not be too surprised if Intel's hyped air gestures used Iris Pro GPGPU.

 

[Master467]

Intel resorted to using monopoly tactics

You seriously need to Google "Monopoly" and find the definition. If there is competition its not a monopoly, one company cannot use "Monopoly tactics" when there is another company unless they try to buy them out. Infact, Intel WANTS AMD to stay alive, or they will get screwed by the gov't.

AMD wasn't competing with them though that's why AMD64 crippled the hot running P4? I don't particularly care for AMD they turned their backs on Nvidia who was also the victim of Intel's evil monopoly ways instead of embracing them.Nvidia helped AMD grow it's CPU market share with it's GPU's and SLI tech in the AMD64 days back before crossfire came about.

Thanks for the history lesson, but pretty much irreverent to everything else said in this thread.

 

[southernshark]

Second, suppose you could have a quad-core 6GHz Haswell right now. What would you do with it? What do you have that would possibly tax it? The only applications right now that could top out such a chip are the well-threaded professional apps that have been hungry for more resources as long as they've been around. Users that need such gargantuan power get better served with the multi-socket server crowd ( or even GPU assistance. ) No OS, productivity software, game, or home office use hasote come close to taxing a mid-range or better CPU before its EOL.

You are the one who is confused. Software comes after hardware, not the other way around. No one is going to write code for a program that can't run on existing computers.And I suppose that no it isn't Intel's job to sell software. But it is their job to sell hardware. And what they are pushing right now, the public ain't buying.Anyway, you must work for Intel. Rather than produce something that the public wants to buy, just call the public a bunch of whiners. The customer is always wrong, right?Time to SHORT the stock.

 

[Crashman]

Intel resorted to using monopoly tactics

You seriously need to Google "Monopoly" and find the definition. If there is competition its not a monopoly, one company cannot use "Monopoly tactics" when there is another company unless they try to buy them out. Infact, Intel WANTS AMD to stay alive, or they will get screwed by the gov't.

The US gov't groups anti-competitive practices together and includes things like threatening not to sell to a company if they do business with a competitor. Intel lost a suit over something along those lines. VIA destroyed SiS 735 sales in the same way, but did so outside of US control.

 

[AdmiralDonut]

I am seriously looking forward to this next gen of chips, even though I will be staying with my Ivy Bridge for the foreseeable future. I am especially intrigued by the unlocked Pentium and am now wondering what, if anything, AMD are going to bring to the table.

 

[RedJaron]

You are the one who is confused. Software comes after hardware, not the other way around. No one is going to write code for a program that can't run on existing computers.And I suppose that no it isn't Intel's job to sell software. But it is their job to sell hardware. And what they are pushing right now, the public ain't buying.Anyway, you must work for Intel. Rather than produce something that the public wants to buy, just call the public a bunch of whiners. The customer is always wrong, right?Time to SHORT the stock.

Those are some lovely blind assumptions you're making about me. No, I don't work for Intel. And I'm not confused about anything in this regard.

Of course software usually follows hardware ( a few games are sometimes the exception that can stress a high-end machine at launch. ) But we've had excess CPU power available for how long now? Dual cores have been readily available for over ten years and have been the standard for at least six. Four-, six-, and eight-core CPUs are readily available now, and also have been for years. So why are multi-threaded applications still so relatively rare? Are you saying software is ten years behind hardware?

The public isn't buying for a lot of reasons. You're conveniently forgetting a rather drastic economic downturn about five years ago. People had a lot less disposable income so they just stuck with what they had for a few extra years than normal. Meanwhile hardware has improved so that a lot of basic computing needs can now be met by phones, tablets, and laptops. Now when they look for a new computer, what do they get? They can either spend $500 on a basic desktop system, or $500 for a laptop that meets their needs and is portable to boot. And if they just want basic communication ability and nothing more, they get a tablet for under $300.

Fact is the tech geek community is actually a very small slice of the market pie. Companies like Intel, nVidia, and AMD can't stay in business on us alone. Nor can the software developers. They rely on the mainstream, and the mainstream is a few years behind the tech geeks. Software is developed for the mainstream or enterprise level because that's where the money is. That software has to run on hardware that's penetrated those markets.

So once again I say there is relatively no software right now that would demand even an i7-4770, let alone your hypothetical 6GHz chip. And no chip maker is going to pour billions into research and manufacturing for such a chip when it wouldn't turn a profit.

 

[InvalidError]

No one is going to write code for a program that can't run on existing computers.

That is a completely meaningless statement: software may fail to meet a certain performance standard on a certain hardware configuration but most of it will run. Whether or not it will meet your specific performance expectations or not is a completely subjective thing because everyone has different definitions of how good is good enough and everyone has different limits on how much they are willing to spend on incremental improvements beyond that.

For most software, performance is elastic - there is no intrinsic minimum performance requirement; it just runs slower on slower hardware and faster on faster hardware. How slow is too slow and how fast is fast enough are highly subjective. If you compare the minimum system requirements for a given game against the hardware (mainly GPUs) hardcore enthusiasts rave about to power their multi-display 144Hz setups, there is something like a 50:1 performance spread to cover.

Some hardware configurations are behind mainstream, some are (way) ahead. Games are written to accommodate some range between those two extremes.

 

[LHR Dovah]

Bah. Why can't Intel do a pure 8 core? No hyperthreading. No igp.

 

[LHR Dovah]

Bah. Why can't Intel do a pure 8 core? No hyperthreading. No igp.

 

[Novuake]

Bah. Why can't Intel do a pure 8 core? No hyperthreading. No igp.

Because that is not what sells well. The money is in MOBILE PRODUCTS.

 

[InvalidError]

The money is in MOBILE PRODUCTS.

With most mobile products moving to single-chip SoC solutions in the $25-50 range, I would not be so confident about mobile products being where the money (as in profit) is even if that is undeniably where the sales are going: $50 for a high-end SoC with everything including the kitchen sink on-chip is a far cry from the $150-180 Intel used to charge even for their low-end conventional laptop CPUs on top of their ~$30 chipset and other Intel-branded support components. Companies like Intel which used to sell multi-chip system solutions are going to have to get used to making $100-150 less revenue per low-end system sale.

The next two years should be quite entertaining in terms of market share shifts between (ultra-)mobile/low-power/embedded and more conventional/high(er)-power PC/laptop.