[WCCF] Intel Skylake 2015 Platform Details Revealed

[ShintaiDK]

What would happen if there was actually competition on the desktop? x86 is a barrier. And someone can't just roll in with MIPS or something to replace x86 either, which further solidifies their position.

It wouldnt change the core count. If more cores actually made a difference we would already have it. Its software lacking behind.

And this is why AMD also went down to 2M/4T as highest count and focus on IGP instead.

 

[IntelUser2000]

GMA X3000 is Gen4. Skylake GT4 Gen9 is just a continuation of Intel's focus on IGPs.

Gen 4 or not, ALL their GPUs from then to NOW is based on the GMA X3000.

They are obviously trying to make actual money from the iGPUs rather than continue to offer them for nearly free. Hence the availability of GT2 OR GT4e on LGA processors. GT2 at x price vs. GT4e at x+ price.

How will they be able to fit 96EUs within 220mm²? 177 + (260-177)*3.8 is more than 230mm².

The current Iris Pro with GT3e has GPU portion taking LESS than half of the 260mm2 die: http://pc.watch.impress.co.jp/img/pcw/docs/613/951/8_s.jpg

I'm counting it as 36% or 95mm2. Obviously simplistic calculations on a device complicated as a modern CPU does not work.


Haswell U is at 130mm2 but Broadwell U is at 80mm2. Despite that it increases EU count by 20%, and moves to Gen 9, and has CPU and other unmentioned changes, the chip size is at 61%.

I wouldn't be surprised GT3e scales to ~60% of the size and ends up at 160mm2. PLENTY of room for the EU to double for GT4. Besides, Intel usually leaves 250mm2+ die sizes for the Tock generation. And in this case, that's Skylake.

Another thing that puts rumors to rest is the fact that it still uses MCP, or Multi Chip Package. Meaning integrated PCH is still few years after Skylake.

 

[AtenRa]

Well, eventually Intel will have to start to price its SKUs on the iGPU performance because the iGPU will occupy more than 50% of the die.

For example, Haswell 2+ GT3 the iGPU is larger than the CPU cores. The iGPU in 2+ GT3e Skylake will be even bigger in relation to the CPU cores and it may be more than 60% of the die. The more they invest in iGPU die the more they will have to price their SKUs according to that and not the CPU portion.

 

[mikk]

Assuming that a 96 EU Gen8 @1 Ghz produces 2 Tflop of shader performance, a 112 EU Gen9 @1.3 Ghz could come close to 3 Tflop assuming that Gen9 doesn't improve per EU performance. Maybe too optimistic but 2.5 Tflop should be feasible for 4+4 Skylake-LGA.

 

[ShintaiDK]

Another thing that puts rumors to rest is the fact that it still uses MCP, or Multi Chip Package. Meaning integrated PCH is still few years after Skylake.

Everyone hopefully already knew that 6 months ago with the information of the Sunrise Point PCH. The optnal network is also the I219LM

 

[Fjodor2001]

Mainly due to Samsung. And the reason why is the marketing budget. Apple used around 1B$. Samsung used 14B$ in 2013. Yes, you read right, 14 billion$ just on marketing. Thats 3 times more than their 2012.

Samsung sells many more types of consumer products than Apple, that's why they have to spend more on marketing. We're talking refrigerators, TV:s, stereos, computer monitors, washing machines, you name it. It's probably easier to mention what type of consumer electronic products Samsung does not make.

But regardless, the simple reason Android phones are doing better than iPhones is that people do like them better. It's not like people will buy a crap product just because marketing tells them to.

 

[Ajay]

Samsung sells many more types of consumer products than Apple, that's why they have to spend more on marketing. We're talking refrigerators, TV:s, stereos, computer monitors, washing machines, you name it. It's probably easier to mention what type of consumer electronic products Samsung does not make.

But regardless, the simple reason Android phones are doing better than iPhones is that people do like them better. It's not like people will buy a crap product just because marketing tells them to.

The articles I read said that Samsung's *smartphone* marketing budget was $14B up from $4B the year before. So unless several sources made the same mistake - sammy is playing to win.

I think that there are allot of reason that Android phones are doing better
- cheaper models available
- multiple hardware vendors == more choices (screen size, cameras, etc.).
- Apple has lost some of it's 'cool' factor (so my nephews tell me).

 

[exar333]

The articles I read said that Samsung's *smartphone* marketing budget was $14B up from $4B the year before. So unless several sources made the same mistake - sammy is playing to win.

I think that there are allot of reason that Android phones are doing better
- cheaper models available
- multiple hardware vendors == more choices (screen size, cameras, etc.).
- Apple has lost some of it's 'cool' factor (so my nephews tell me).

The iPhone is the top smartphone choice of the 55+ age bracket. That has to be worth something.

 

[rtsurfer]

The articles I read said that Samsung's *smartphone* marketing budget was $14B up from $4B the year before. So unless several sources made the same mistake - sammy is playing to win.

I think that there are allot of reason that Android phones are doing better
- cheaper models available
- multiple hardware vendors == more choices (screen size, cameras, etc.).
- Apple has lost some of it's 'cool' factor (so my nephews tell me).

Still you can't argue that Samsung sells a lot of different phones, unlike Apple's 1 phone.

 

[Fjodor2001]

The articles I read said that Samsung's *smartphone* marketing budget was $14B up from $4B the year before. So unless several sources made the same mistake - sammy is playing to win.

Reuters which is the original source of the information says:

"Samsung Electronics Co is expected to spend around $14 billion - more than Iceland's GDP - on advertising and marketing this year"

I.e. it is for the complete company Samsung Electronics Co, and there is no mentioning of it being restricted to smartphones.

Also, the $1 billion figure for Apple is not comparable to the $14 billion, since the $14 is for both advertising and other types of marketing, where as the $1 billion figure is for advertising alone. So apples (no phun intended) and oranges are being compared. As the article also said:

"But, while Samsung has become the world's biggest advertiser, spending $4.3 billion on ads alone last year, its global brand value of $39.6 billion is less than half that of Apple, which spent only $1 billion on advertising"

 

[SlickR12345]

2015, that is a long way off. 2014 has barely begun guys and by 2015 its probably over a year until its released, then about a month for channels to be fully supplied, then a month or two until prices stabilize, so you may be seeing late 2015, even possibly 2016 if there is short demand and full inventory in party channels.

So a lot of time, I'm wondering if Intel will have something sooner like a summer refresh with small clock upgrades on models!

 

[Arachnotronic]

Reuters which is the original source of the information says:

"Samsung Electronics Co is expected to spend around $14 billion - more than Iceland's GDP - on advertising and marketing this year"

I.e. it is for the complete company Samsung Electronics Co, and there is no mentioning of it being restricted to smartphones.

Also, the $1 billion figure for Apple is not comparable to the $14 billion, since the $14 is for both advertising and other types of marketing, where as the $1 billion figure is for advertising alone. So apples (no phun intended) and oranges are being compared. As the article also said:

"But, while Samsung has become the world's biggest advertiser, spending $4.3 billion on ads alone last year, its global brand value of $39.6 billion is less than half that of Apple, which spent only $1 billion on advertising"

2/3rds of Samsung's profit comes from smartphones. TVs and the like are very low margin businesses.

I'd imagine most of the marketing spend is due to phones.

 

[Fjodor2001]

2015, that is a long way off. 2014 has barely begun guys and by 2015 its probably over a year until its released, then about a month for channels to be fully supplied, then a month or two until prices stabilize, so you may be seeing late 2015, even possibly 2016 if there is short demand and full inventory in party channels.

So a lot of time, I'm wondering if Intel will have something sooner like a summer refresh with small clock upgrades on models!

Yeah, Broadwell Refresh as mid-life kicker in 2015Q2-Q4, Skylake actual availability sometime in 2016Q2 or later...

 

[AtenRa]

I could see High-End Desktop and low power Laptop Skylake in H2 2015, it is feasibly but it all depends on the market and the competition.

 

[Sweepr]

According to this leak the only LGA Broadwell is a 4C+GT3e chip. Again, all other models (Haswell Refresh dual-cores and ''non-K'' quad-cores) will be replaced by Skylake. Anything could happen till launch (including delays) but all current leaks point to 2015 launch for desktop Skylake.

 

[Fjodor2001]

So only a few months after Broadwell-K has been released? What OEM would jump on the B-K wagon then...

 

[IntelUser2000]

What may happen is that the K releases later than other desktop chips, or maybe the entire desktop line is at Q2 while mobile chips are at Q4 of the previous year.

I just noticed that GT4 isn't available with Broadwell. So I guess my expectation of 96EU doesn't exist anyway.

 

[AtenRa]

So only a few months after Broadwell-K has been released? What OEM would jump on the B-K wagon then...

hmm forgot about Broadwell-K is H1 2015. Then i dont see a Desktop Skylake until H1 2016. Laptop may come in H2 2015 though.

 

[Fjodor2001]

What may happen is that the K releases later than other desktop chips, or maybe the entire desktop line is at Q2 while mobile chips are at Q4 of the previous year.

So mobile Skylake in 2015Q4, desktop Skylake in 2016Q2? Maybe...

 

[Enigmoid]

I'm going to go out on a limb and say that I expect some delays.

 

[jdubs03]

I'm going to go out on a limb and say that I expect some delays.

I don't expect it to be delayed in the sense of yield issues, as neither do you I'm assuming. I also don't think the IA designers will have much issue there. If anything a delay would come down to too much Broadwell inventory prevalence.

Cannonlake on the hand could be delayed, but that's not til 2017.

 

[Enigmoid]

I don't expect it to be delayed in the sense of yield issues, as neither do you I'm assuming. I also don't think the IA designers will have much issue there. If anything a delay would come down to too much Broadwell inventory prevalence.

Cannonlake on the hand could be delayed, but that's not til 2017.

I have no doubt that intel could launch 14 nm NOW. Its yields and inventories and selling enough 22nm to pay for its development.

That said almost everything in the computer industry is delayed, looking at predictions from a few years ago. Intel's tick-tock cycles are getting longer, AMD is delaying their products (kaveri and kaveri mobile), Nvidia with pascal/volta, TSMC with their nodes. I strongly expect this to continue into the future. Stuff is still happening but much more is being pushed back than being bumped up.

 

[witeken]

Intel's tick-tock cycles are getting longer

I will only believe that when it's confirmed that 10nm production will not start in 2015, which is when Intel claim they will start volume production. If they can do that, I hope their Tick-Tock model will be closer to 1 year in the future again, because that isn't what happened with Ivy Bridge, Haswell and soon Broadwell. Skylake, if it launches in 2015, will be about 1 year and Cannonlake too when it launches mid-2016.

 

[jdubs03]

I will only believe that when it's confirmed that 10nm production will not start in 2015, which is when Intel claim they will start volume production. If they can do that, I hope their Tick-Tock model will be closer to 1 year in the future again, because that isn't what happened with Ivy Bridge, Haswell and soon Broadwell. Skylake, if it launches in 2015, will be about 1 year and Cannonlake too when it launches mid-2016.

switching to 10nm i think will be manageable. but that is the last node before switching to iii-v materials for the transistor channels and possibly new transistor structures, . intel has a good shot at implementing it first like finfet when they announced it in 2011. but its gonna be really hard, it'll be harder to achieve than the transition to finfet. so that tick cycle may take longer, but i don't think the tock cycle will lengthen.

 

[Idontcare]

I will only believe that when it's confirmed that 10nm production will not start in 2015, which is when Intel claim they will start volume production. If they can do that, I hope their Tick-Tock model will be closer to 1 year in the future again, because that isn't what happened with Ivy Bridge, Haswell and soon Broadwell. Skylake, if it launches in 2015, will be about 1 year and Cannonlake too when it launches mid-2016.

I will personally be surprised to see Intel 10nm products available and in stock on Newegg in 2016, let alone 2015.

Personally I think 10nm retail availability (the only kind that matters to me) won't happen until sometime 2017.

IMO 10nm will be delayed for an economic reason, not a technology reason, same as what has happened with the 14nm ramp.