Intel Skylake / Kaby Lake

[dark zero]

I guess the AMD crowd has to pick on something now that Fury X failed to beat NVIDIA at the same price point and there's nothing interesting from the CPU camp till Q4-2016 (plenty of PowerPoint slides though).

If Skylake is 10% faster than Devil's Canyon across the board at similar/lower power then it's going to be a much better product than Zambezi (Bulldozer) compared to its predecessors.

Also poor Carrizo failed to get significant design wins and will go head to head with Skylake-U after Windows 10 launch.

If the makers goes to the U tier and NOT to the H tier, they won't complain later if the sales continues to drop even more. The stupidity of the makers must be stopped and if this means the death of the notebook industry... so be it. Still... it would be the 2nd mobile market killed by Intel.
Pine Trail killed the Netbook one.

Meanwhile Fury failed hard if goes alone.... however going Dual or Tri becomes even better than Titan X. To add insult to the injury, the Titan Z is better than X on DP.

 

[Sweepr]

Meanwhile Fury failed hard if goes alone.... however going Dual or Tri becomes even better than Titan X. To add insult to the injury, the Titan Z is better than X on DP.

You mean this?
I'd never choose Fury X for a multi-GPU system, it's plain obvious 4GB VRAM is going to be a problem sooner or later and I usually keep my systems for a few years before upgrading.

 

[dark zero]

You mean this?
I'd never choose Fury X for a multi-GPU system, it's plain obvious 4GB VRAM is going to be a problem sooner or later and I usually keep my systems for a few years before upgrading.

Actually you've chosen nVIDIA despite the competition improvement ¬¬.

BTW, seems that Maxwell is not scaling well. That is bad news. That's why they are moving to HBM too. GDDR5 are on it's limits, so... HBM is the future after all.

 

[rtsurfer]

Seems Skylake LGA1151 is PCIe 3.0 and Skylake-E LGA20xx is PCIe 4.0.

According to this article that won't happen
http://www.kitguru.net/components/g...es-and-new-connector-to-be-finalized-by-2017/

 

[Sweepr]

Actually you've chosen nVIDIA despite the competition improvement ¬¬.

BTW, seems that Maxwell is not scaling well. That is bad news. That's why they are moving to HBM too. GDDR5 are on it's limits, so... HBM is the future after all.

Actually I have only owned ATI/AMD GPUs over the past decade. Fact is, looking at the performance per watt Maxwell delivers (28nm) it becomes clear why NVIDIA is dominating mobile. Intel is also covering all fronts. They fixed Bay Trail's slow iGPU with Cherry Trail/Braswell and they will have a brand new low-power x86 core next year which should (hopefully) improve Atom performance per clock. They have the best iGPU on the market right now with Broadwell GT3e (best all around CPU+iGPU chip) and an agressive roadmap ahead (higher clocked desktop Broadwell GT3e parts this year, desktop Skylake GT4e and mobile/server Kabylake GT4e 128+128 MB next year).

 

[TidusZ]

I used to get excited about new processors launching, this time I'm just like maybe in a few more new architectures I'll buy one, maybe.

 

[Fjodor2001]

Apparently Intel has hailed Skylake as "its most important chip architecture in a decade"!

Does anybody know how they can arrive at that conclusion given the minuscule improvements we're seeing?

 

[Sweepr]

ASUS OK to import the Alpine Ridge wafer, Z170 motherboard with Thunderbolt daughter card exposure

http://benchlife.info/asus-will-launch-alpine-ridge-thunderbolt-card-in-future-07132015

 

[LTC8K6]

Apparently Intel has hailed Skylake as "its most important chip architecture in a decade"!

Does anybody know how they can arrive at that conclusion given the minuscule improvements we're seeing?

The leaks are inaccurate?

There is new technology in Skylake that is successful, and will allow Intel to move on to better stuff? That is, Skylake has proven a few things for Intel, which can now be fully utilized in later chips.

 

[dahorns]

Apparently Intel has hailed Skylake as "its most important chip architecture in a decade"!

Does anybody know how they can arrive at that conclusion given the minuscule improvements we're seeing?

Theories:

(1) the leaks are wrong;

(2) the platform improvements brought by Skylake are critical in the markets that matter (i.e., not desktop).

 

[Fjodor2001]

(3) Intel is exaggerating the architectural improvements to hype Skylake in hope of selling more CPUs?

 

[frozentundra123456]

@ dahorns: I suspect the latter. Or at least the improvement is directed towards igp and performance per watt, as in pretty much every generation since Ivy Bridge, whether we like it or not.

In any case, that linked article seems pretty poor. But I dont know why the poster is bent out of shape about it. Every company does it. Already the hype train is running full throttle for Zen, and it is 1+ years away. At least Skylake should be available soon, so we can get some hard numbers.

 

[dahorns]

(3) Intel is exaggerating the architectural improvements to hype Skylake in hope of selling more CPUs?

Sure, it is also possible that Intel is an evil corporation that is trying to get away with zero effort on Skylake since it has no competition. Is that what you want to hear?

I expect Skylake to be good for mobile/laptops and to be just another processor for desktop. If you want more desktop heft you'll have to go to the e series.

 

[Fjodor2001]

Sure, it is also possible that Intel is an evil corporation that is trying to get away with zero effort on Skylake since it has no competition. Is that what you want to hear?

I expect Skylake to be good for mobile/laptops and to be just another processor for desktop. If you want more desktop heft you'll have to go to the e series.

Nah, not evil. Just marketing I'd say, like many other companies that hype their product in hope of selling more. Quite common unfortunately.

But I guess we'll see if mobile Skylake is better than desktop Skylake, so that explains Intel's hailing of the Skylake architecture. I have my doubts though.

 

[Kallogan]

i don't know what you guys expect, skylake will be all about igpu, so yeah good for mobile not desktop

 

[Fjodor2001]

Sure, but still don't you think Intel calling Skylake "its most important chip architecture in a decade" is over-hyping it quite a bit?

Is the Skylake arch really more important than Sandy Bridge and Conroe was? If so, in what way?

 

[Sweepr]

Intel skips Broadwell-EP/EX, Purley (Skylake-EP/Skylake-EX) launching sooner

Source (italian): www.bitsandchips.it/50-enterprise-b...x-ep-scompaiono-dalle-ultime-roadmap-di-intel
Translated to english: http://translate.google.com/transla...x-ep-scompaiono-dalle-ultime-roadmap-di-intel

As we reported our two sources who work in major companies operating in the Enterprise, the CPU Broadwell-EX / EP would have disappeared from the latest Intel roadmap.

Our sources tell us that Intel did not send them any official notification regarding the cancellation of the actual CPU Broadwell-EX / EP, but still strange that these have disappeared from official documents under NDA. At this point it seems that Intel has decided to leapfrog these CPUs to go directly to Skylake-EX / EP.

In fact, this choice may not be so unrealistic as Broadwell, compared to Haswell, not actually brings no significant new developments, but a slight increase in the CPI. Only Skylake we will use the new AVX instructions and, perhaps, of some important new features that can make a difference within the multitasking / multithreading.

After the sudden appearance of the platform Kaby Lake, between Skylake and Cannonlake, it seems plausible that Intel is scrambling the cards as much as possible to maximize the effectiveness of their future product lines, avoiding both the marketing of CPU not too profitable (eg , Broadwell-D appeared only in two variants) is the preco money in the departments of R & D projects of limited interest.

At this point we just have to wait any official communication from Intel, to know whether Broadwell-EX / EP was officially ousted or not.

That's good news IMHO. Skylake-E would be awfully late compared to Skylake-S (2 years, 2017 launch) and Haswell-E is more than good enough for LGA2011 users. Maybe now there's hope we'll get AVX-512 enabled Skylake-E enthusiast goodness next year?

 

[Sweepr]

Broadwell-EP didn't look like the kind of leap we're used to from Intel server 'Ticks':

Only 2 extra cores for servers and the same 6-8 cores in desktops.

I'm wondering how that roadmap looks without Broadwell-EP/EX:

 

[rtsurfer]

Apparently Intel has hailed Skylake as "its most important chip architecture in a decade"!

Does anybody know how they can arrive at that conclusion given the minuscule improvements we're seeing?

From what I've heard from Pro overclockers

1) Significant performance improvements.
2) 4000+MHz JEDEC (or X.M.P) memory support.
3) Bclk Overclocking ability not seen on recent Intel platforms. Already achieved 200Mhz Bclk on air, but its probably only stable for benchmarking.

 

[Fjodor2001]

From what I've heard from Pro overclockers

1) Significant performance improvements.
2) 4000+MHz JEDEC (or X.M.P) memory support.
3) Bclk Overclocking ability not seen on recent Intel platforms. Already achieved 200Mhz Bclk on air, but its probably only stable for benchmarking.

Ok, sounds interesting, but what actual performance improvements can we see from that? In the benchmarks presented so far I'm only seeing single digit average performance improvements compared to two CPU generations ago (i.e. Haswell).

 

[Arachnotronic]

Ok, sounds interesting, but what actual performance improvements can we see from that? In the benchmarks presented so far I'm only seeing single digit average performance improvements compared to two CPU generations ago (i.e. Haswell).

You do realize that the PC market is bigger than just tower desktops, and that Skylake will potentially bring lots of interesting "non-CPU" advancements such as hardware HEVC decode, probably significantly better power management (AT user PaulIntellini said in a post that Intel is adding "duty cycling" to the Skylake CPU core), and other platform-level goodies.

If you want raw desktop CPU perf, x99 + 5820K or better should be a very viable option.

 

[rtsurfer]

Ok, sounds interesting, but what actual performance improvements can we see from that? In the benchmarks presented so far I'm only seeing single digit average performance improvements compared to two CPU generations ago (i.e. Haswell).

To be honest, personally, I'm not really optimistic about very significant performance improvements (read IPC), but the Benches that are leaked (the Techbang ones) are not accurate, from what I've been told.

 

[dahorns]

You do realize that the PC market is bigger than just tower desktops, and that Skylake will potentially bring lots of interesting "non-CPU" advancements such as hardware HEVC decode, probably significantly better power management (AT user PaulIntellini said in a post that Intel is adding "duty cycling" to the Skylake CPU core), and other platform-level goodies.

If you want raw desktop CPU perf, x99 + 5820K or better should be a very viable option.

Pushing traditional desktop users to the enthusiasts platforms makes a lot of sense for Intel. I wouldn't be surprised if Intel kills its current desktop line at some point. I assume the only reason they still exist is because businesses are easily upsold to buy performance they don't need for their workers. But, that leads to really long upgrade cycles.

Instead, Intel could continue its focus on mobility and small form factor platforms (which is where the market is moving anyway). Businesses will either continue to buy more than they need and purchase enthusiast platforms--which would result in increased income per unit for Intel--or they will buy the mobility/SFF platforms--which would likely result in shorter business upgrade cycles.

 

[Fjodor2001]

Yeah, perhaps. But we're not seeing much performance increase on the HEDT enthusiast platform either. So it would be a one time performance jump from mainstream desktop -> enthusiast platform for most people. And primarily only for MT workloads, for ST it would actually be a regression compared to 4790K. Also, once those enthusiasts have done that one time migration to HEDT they won't need to upgrade for another 10 years on that platform either...

 

[tenks]

Intel skips Broadwell-EP/EX, Purley (Skylake-EP/Skylake-EX) launching sooner

Source (italian): www.bitsandchips.it/50-enterprise-b...x-ep-scompaiono-dalle-ultime-roadmap-di-intel
Translated to english: http://translate.google.com/transla...x-ep-scompaiono-dalle-ultime-roadmap-di-intel



That's good news IMHO. Skylake-E would be awfully late compared to Skylake-S (2 years, 2017 launch) and Haswell-E is more than good enough for LGA2011 users. Maybe now there's hope we'll get AVX-512 enabled Skylake-E enthusiast goodness next year?

While I want this very much to be true, I'm highly skeptical. Reading the article gives me the impression that they jumped to some conclusions. They were told by sources Broadwell-E was not on recent roadmaps. That does not automatically mean SKY-E was pulled in. It's a safe assumption, but its just that, an assumption.

Skylake-E was planned for later in 2017? I can't see them bringing it in a year, unless Intel regularly finishes chips and shelves them for a year before they sell them. Which it sometimes feels like.

"Skylake-E isn't ready, we're dropping it in 2017. Oh no more Haswell-E? Ok JK Skylake-E is ready now""

Also I don't know how credible that site is, (im not familiar with it) but this roadmap was floating around here in May reported by numerous sites with no broadwell-e on it but it was deemed fake. I wouldn't be surprised if they got a hold of the same one..