Intel plans to reduce performance of next-generation CPUs

[crashtech]

It's not low power, it's lower clock speed.

What is the purpose of this distinction re this discussion? Clock speeds get maximized within the power/thermal envelope required, the only reason for clock speed regression is power/thermal concerns.

 

[poofyhairguy]

Haven't you thought of maybe, just maybe there's a technological limit of how fast they can go?

That is the real problem. We have a need for faster computers TODAY, but the technology isn't ready and might never be.

I mean, a lot of the "neato" stuff people are doing nowadays takes a lot of computing power- more than any single person could own. Personal assistants like Siri or Google Now will probably never be powered at the level they are currently at on a desktop PC, going forward so many "advancements" will rely on us all just sharing time on the server farm.

 

[jpiniero]

Clock speeds get maximized within the power/thermal envelope required, the only reason for clock speed regression is power/thermal concerns.

That's why they brought it up, this is new. Post-Silicon can't hit the clock speeds that Silicon can.

 

[dark zero]

Intel's problem in phones is that its chips are complete and total garbage.

Nope. The real reasons is that they coose the wosrt posible OS for them. They needed to go to Windows Mobile path

 

[sm625]

The future of CPU performance scaling is in programmable logic. Programmable logic linked tightly to the actual execution units of the CPU core. Smaller blocks, probably only a square millimeter or perhaps even less. But many of them. Just like Skylake has 6 execution units. One of these programmable blocks would be only about the same size as one of those existing execution units. They would have direct access to the prefetcher and scheduler and instruction/data caches. They would be power gated. Applications that want to make use of these blocks could program them on the fly. Any blocks of code that repeat the same sequence of instructions would be a candidate for massive IPC increases. I'm thinking about javascript specifically. Such a core would be so advanced that it would really unlock the software developers to develop extremely fast and efficient code. These blocks could be reprogrammed very quickly every time there is a context switch. I view this as the ultimate bare metal programming. You could literally write an ASIC to execute your code. In about 10 years I expect the CPU to construct its own programmable logic blocks in real time based on what is running; to, in essense, profile and optimize itself to achieve the highest possible IPC.

 

[poofyhairguy]

Nope. The real reasons is that they coose the wosrt posible OS for them. They needed to go to Windows Mobile path

Uh...I don't think Intel has the ability to dictate which mobile OS the market uses/wants. Focusing on almost dead Windows Phones instead of Android would have ensured they would never be relevant in mobile.

 

[USER8000]

Where has single thread speed or ipc decreased? Single thread ipc continues to increase, all be it at a slower rate. Clockspeed has increased as well until haswell, and is stable with skylake, and intel has said kaby lake will bring performance increases. These off hand statements about what may happen at some distant point in the future are being blown way out of proportion.

Yeah,but the issue is that at the rate it is increasing you will need to wait a few generations if you want a decent improvement in single threaded performance. Things like reductions in power consumption and IGP performance are increasing much more rapidly and seem to be the focus of things now. Just look at how tiny the Core i7 6700 die is?? Most of it is the GPU and chipset functionality.

It would not surprise me that after Kaby Lake Intel moves to making its mainstream chips all SOCs,

People are just not upgrading that much for performance reasons anymore,even in the DIY segment.

Last year,total motherboard shipments to the DIY sector dropped over 20% it appears:

http://www.customstoday.com.pk/tag/global-diy-motherboard-shipments-to-drop-21-7-in-2015/

 

[tenks]

I've seen 100s of articles like these over the years about this that or the other. And all the knee jerk reactions they produce (this thread).. It's always the same and it's always hilarious. I remember when skylake was supposed to be socketed and it was the end of replacement CPU's and enthusiast computing as we knew it! ZOMG run for the hills!


haha, nothing changes in tech 'journalism'.

 

[crashtech]

Doom and gloom is always more exciting than business as usual, that goes for pretty much everything in life.

 

[ShintaiDK]

/me looks at his BGA soldered 6700K.

Oh wait...

 

[krumme]

The future of CPU performance scaling is in programmable logic. Programmable logic linked tightly to the actual execution units of the CPU core. Smaller blocks, probably only a square millimeter or perhaps even less. But many of them. Just like Skylake has 6 execution units. One of these programmable blocks would be only about the same size as one of those existing execution units. They would have direct access to the prefetcher and scheduler and instruction/data caches. They would be power gated. Applications that want to make use of these blocks could program them on the fly. Any blocks of code that repeat the same sequence of instructions would be a candidate for massive IPC increases. I'm thinking about javascript specifically. Such a core would be so advanced that it would really unlock the software developers to develop extremely fast and efficient code. These blocks could be reprogrammed very quickly every time there is a context switch. I view this as the ultimate bare metal programming. You could literally write an ASIC to execute your code. In about 10 years I expect the CPU to construct its own programmable logic blocks in real time based on what is running; to, in essense, profile and optimize itself to achieve the highest possible IPC.

I like your thinking.
One can hope a company like Intel can do stuff like that. If there is anyone who can its them. We have seen some failed attempts like that before and the went titanic but hey lets hope they will try again.

 

[tenks]

/me looks at his BGA soldered 6700K.

Oh wait...

exactly

 

[Phynaz]

/me looks at his BGA soldered 6700K.

Oh wait...

Hehe.

 

[superstition]

Haven't you thought of maybe, just maybe there's a technological limit of how fast they can go?

LOL

If VR does take off, they'll probably find out using GPU to offload more stuff was a much better idea anyway.

GPUs are not suited for AI. Maybe they can do some floating point calculations or something but they're no substitute for a CPU.

Holodeck is a fantasy!

The only fantasy part is the transition from images to matter. The rest is pretty much possible and coming.

And, if you don't like to think about holodecks then think about Watson.

 

[MarkizSchnitzel]

VR

If Intel doesn't go to arsenide transistors with terahertz speeds then China will, eventually.

Any "more focus on just low power applications" is going to be temporary.

You don't power a Holodeck with Sandy Bridge.

Those will be ultra high end, no? Like HEDT is, maybe?

I don't see how massive VR/AR revolution will come in the next 10 years..

Hololens, with larger FOV, would not require such local raw power.

Maybe holodeck will be cloud powered.

 

[NTMBK]

Maybe holodeck will be cloud powered.

No way. It's hard enough getting the latency down for VR on a local system, never mind streaming it over the World Wide Web. And putting hardware in someone else's datacenter doesn't make it any cheaper.

 

[dud]

In an effort to perhaps "stir up the pot" a bit I will ask ... I wonder if the future of PC computer would have looked brighter if AMD had remained more competitive with Intel? Competition is almost always a good thing for the customer as it encourages innovation and a better product and lower cost to the consumer.

 

[ShintaiDK]

In an effort to perhaps "stir up the pot" a bit I will ask ... I wonder if the future of PC computer would have looked brighter if AMD had remained more competitive with Intel? Competition is almost always a good thing for the customer as it encourages innovation and a better product and lower cost to the consumer.

Nope. People seem to forget what the 99% crowd demands.

And competition isn't always good, far from. Competition rarely creates innovation, just drives the pace.

Not to mention the issue with cash flow in the segment. Higher costs and lower volume...

 

[TheELF]

Competition is almost always a good thing for the customer as it encourages innovation and a better product and lower cost to the consumer.

Nope,in the best case you'll end up with similar products at similar prices.
If AMD could bring out the skylake line up right now alongside intel's skylake line up they would charge just as much as intel does,nobody wants to drive down the prices on the product they produce.
That's the way for any company/product,you always see "xx% more product" or "better formula" you never see a company lowering prices(permanently) .

As a matter of fact we should wish for intel to not lower their prices since that would drive AMD to bankruptcy even faster.

 

[krumme]

Another midrange 14nm mobile soc. No mm2 for cpu. Damn.
http://www.anandtech.com/show/10045/samsung-announces-new-exynos-7870-midrange-14nm-soc