How much of an improvement will cannonlake be over kaby lake?

>>55663534
It's going to be a 5% improvement in speed.

>>55663534
The only improvements will come through IPC >>55663674
From this point on Intel will be focusing on energy efficiency instead of raw speed. We won't see faster single-core speeds potentially for many years, unless a major breakthrough in processor design comes to market.

>>55663534

And of course, prices for Skylakes will stay the same and Kaby Lake wil cost 50$ more and Cannonlake - 100$ more?

I kinda hoped for like 200$ for Skylake i7s, but I guess Intel dont care about price dropping, they just remove old models from shops.

>>55663534
As an AMD owner I can only advise you to start saving up and wait for quantum computing.

>>55663534
Right now they're pushing better motherboard features(or gimmicks) rather than pure speed.
Native USB 3.1, thunderbolt 3, HEVC main10 decoding, more PCI lanes, Intel Optane SSD support, etc.

>>55663534
the same as the past 5 years, nothing

Processor speed has barely increased at all int eh past, what, 7? Years. Its seriously retarded. Maybe 10% increase for general users?

>>55664428
Idiot

The only thing you can advise is deez nuts.

>>55663534
They gotta pull those AVX1024 (or AVX2048/AVX4096 whatever is in fashion nowadays) instructions off. Add HMC and a bunch of asymetric smallers cores and CPUs will kill GPUs.

>>55664881

Yeah, the CPU trend is 10% increase a year.

Isnt cannonlake going to have 6 cores?

>>55668849
...Depends on efficiency gains. Intel really do not want a mass-market CPU hitting over 100w, since the average person is not going to drop even more money on a cooler just so that their new CPU can work. Once you start going over that amount the cooling requirements skyrocket, with a typical 140w 2011-3 CPU needing a decent aftermarket cooler, and if you OC or get a 165w Xeon, you'll already want something on the level of a D15 or an AIO. This is a problem since consumers often want small and quiet systems.
The way it's currently set up is fine, X99 is not much more expensive than a similar Z170 setup, and the people who actually genuinely NEED multi-core processing can afford the difference.

when will we see kabylake laptops?

>>55663794

We're now within 1 die shrink of running into electron tunneling issues. Are we witnessing the end of Moores law on the x86 design after 50 years?

Or is there some possibility of some work around I'm not thinking of like gpu coprocessing?

>>55663534
Aside from gayming there isn't really much demand or need for faster CPUs. It's all about power efficiency and feature set nowadays. Hopefully DX12/Vulkan will help fix those CPU bottlenecks in the mean time.

>>55665156
t. intel engineer

>>55669409
I've seen news regarding working CPUs that channel light, potentially being orders of magnitude faster than current technology on a per-"transistor" basis, although the first chips will likely have very few of them. I'm guessing it's a good decade away from being a lab experiment and being an actual product in some form. With a photon-based processor we could pull back on shrinkage and logic gate count, which would offer potential for future expansion. Since light is vastly faster than electricity, CPUs won't be as constrained in physical size either, allowing chip makers to simply increase chip size outwards to add more gates and not worry about desync.

>>55663534
>Does going from 14mn to 10mn really make that much of a difference?

yes it will dramatically speed up electromigration

>>55664881
More like 25-30%

But still.


Dennard scaling is dead, get over it you fags.

I'm personally pretty happy with the way things are right now. New processors offering nothing but lower power consumption + heat generation sounds great to me.

I've used a computer with as little as 2 cores 1.9GHz and it ran perfectly fine. My main machine has 4 cores and overclocks to around 3.6 or so, and I can't even really notice a difference.

It's a good time to be alive.

File: ayy.jpg (108KB, 681x960px) Image search: [Google]

108KB, 681x960px

>>55669409
They'll uset the stashed ayy lmao tech.

whats gonna happen when we hit the 2-4 nanometer limit?

>>55663794
>unless a major breakthrough in processor design comes to market.

You mean, unless Zen is competitive.

Oh who am I kidding. Even if Zen is competitive, Intel has the entire industry in their pockets, nobody will risk ordering Zen Opterons for fear of Intel shitcanning them on the long term.

>>55670302
Retard. Hitting the physical limits of material has nothing to do with either amd or intel.

>>55670369
Intel probably has discovered new materials that work much better than silicone and are just milking what's left of silicone before they switch

>>55670422
>silicone

>>55663534
Who cares?
10nm is a half node shrink from 14nm, die area scaling won't be as dramatic, not that it matters for intel.

>>55668849
Cannon Lake is dual core or quad. Its also mobile only.
Intel's desktop platform for 2018 is 14nm Coffee Lake which comes with up to 6 cores and an Iris Pro IGP with slab of eDRAM.

>>55669409
>HURRRF MUH ELECTRON TUNNELING GUYS
>I READ IT ONE TIME SO I'M GOING TO REPEAT IT FOREVER IN EVERY THREAD
>NEVER MIND THE SIMPLE FACT THAT ITS LITERALLY IMPOSSIBLE INSIDE OF AN INSULATED SiGe CHANNEL

>>55670978
intel will have all 10nm processors by 2018

>>55670264
More gimmicks to keep you buying new cpus.

>>55663534
Maybe 5% improvement per clock, more of the die dedicated to integrated graphics compared to previous gen.

>>55670422
>>55670596
Silicon is what CPUs are made of.

Silicone is what fake breasts have in them.

I'm just waiting for the day when shrinking stops, IPC improvement stops, and cpu makers start printing stupid huge dies with 60 cores and gigabytes of cache because lol fuck it why not

>>55670422
>>55675702
is there any groundbreaking research on it yet? i've been thinking about it for years, it's about time we moved to new materials.

In 10 years we have gone from 90nm to 14nm. Below 7nm quantum tunneling will make things very challenging. I don't think we will get much smaller than 5nm even with new materials. This is extremely close to atomic scale.
For perspective, 5nm is about the distance spanned by just 15 carbon atoms in solid graphite.

kaby lake and cannonlake will not present are performance increase over skylake, only perf/watt and platform upgrades.

intel's new selling point is the platform and not the cpu's performance, we won't see any tangible performance increase until 2018 AT THE EARLIEST. there is still a chance the new architecture after cannonlake could be pushed back if 10nm is delayed any further.

File: AOCQubitMechanicalResonator_EAL1.jpg (3MB, 4790x3193px) Image search: [Google]

3MB, 4790x3193px

>>55669450
Demand is there, but there is a cost factor as well as efficiency.
Probably still way into the future, and possibly too narrow of useful applications for widespread adoption, but Qubits could be a huge leap forward.

>>55675922
People have been talking about using graphene for ages, but a lot of scientists say that it isn't feasible.

http://www.extremetech.com/extreme/175727-ibm-builds-graphene-chip-thats-10000-times-faster-using-standard-cmos-processes

https://www.maketecheasier.com/why-cpus-wont-be-made-of-graphene/

Didn't they say Win10 only?

>>55679569
You won't be able to use new instructions like avx512 because microdick won't push software support on older os. Shit's akin to DX12 gpu on windows 7.

(1/2)
For software we are still transitioning to fully multithreaded programs in many cases. When a process shrink comes out, we get better density for everything on processors. More cores, more cache, and less power usage for similar workloads. Physical sizes of the processors themselves becoming smaller results in benefits for all consumers. The only place where we haven't seen huge levels of improvement is desktop. Otherwise, laptops, servers, and mobile have all seen uninterrupted progress in overall performance. Power usage in the server and mobile markets is often the constraining factor for what a company or consumer can do, which simply isn't true for desktop. This freeze in performance gain is literally because of intel dragging their ass on offering more than 4 cores for the mainstream desktop environment, which has not been needed with AMD having been out of the game. If you want to look at some hard numbers to back up that statement, you don't have to look any further than desktop i7 die size from intel.


Sandy bridge i7 = 216mm2 die size
Skylake i7 = 122mm2 die size (skylake is also a thinner chip too)

>>55679769
(2/2)
Despite a skylake chip being literally half the size of sandy bridge, intel still charged consumers roughly the same price. Intel could have easily added extra cores to their desktop line, but opted to only add cores for their HEDT lineup which has exploded in price without competition from AMD (10 cores for 1700 MSRP). Intel's research budget hasn't moved much, and they try to balance it close to 20% of their yearly revenue. When you have the whole market, you just need to meet the supply requirements for growth. Even if they offered 20% improvement for the desktop market at the same price, it wouldn't be likely that intel would see a noticeable increase in sales from it. It's more than likely intel is sitting on top of some viable improvements to their processors, but has shifted more towards overall research to computing hardware instead of continuing their focus on processors. That theory makes even more sense when you consider intel's recent announcement of shifting focus to compete in cloud computing and hosting.

Probably one of the most interesting things that could possibly be on the horizion is intel releasing server chips combining multiple types of processing cores which are more than just CPU+GPU. Possibly some bizarre combination of full x86 cores that have access to a variety of ASIC bundled onto the die and a host of reduced functionality x86 processors similar to Phi, but all on a single package.

Does AMD even make CPUs anymore?

>>55679782
AMD hasn't been relevant since Intel ditched netburst. All they've accomplished is ruinig ATi.

>>55663534
apparently it's going to be laptop chip

first 10nm desktop coming in 2019
kaby-cofee(mainstream 6core\2018)-icelake(10nm)

nothing to look forward to, pc market is dying