Intel RoadmapSkylake-XKaby Lake-X and Skylake-W Processors in Q2 2017

[Drazick]

What are the chances seeing SKylake - X with Iris Pro?

I really want to have 6 / 8 Cores + IGP + eDRAM + 140W TDP.

It would be great workstation for MATLAB / Python / Julia.

 

[Arachnotronic]

What are the chances seeing SKylake - X with Iris Pro?

I really want to have 6 / 8 Cores + IGP + eDRAM + 140W TDP.

It would be great workstation for MATLAB / Python / Julia.

Probably pretty much zero. Iris Pro isn't something most gamers are likely to pay a premium for. I doubt that 5775C/5675C were particularly popular.

 

[Drazick]

What about other target market?

People who need a string workstation for Scientific Development (MATLAB / Python / Julia).
Machine Learning, Optimization and Simulations are very popular these days.

Moreover, video editors, image editors (Photoshop, GIMP, etc...) they all would love ot have this kind of CPU.

Mainly because many of the operations on these worlds are Memory Limited and Highly Parallelizable.
Hence 4 Memory Channels + eDRAM will do wonders (See how fast SkyLake is with decent memory bandwidth, many of the operations today are memory limited).
The Iris Pro is good enough GPU for those use cases.

I wish Intel will become aware of this market.

 

[ZGR]

Well there is this.

But for 6+ cores? Not anytime soon it seems.

edit: From the article

From a revenue and profit perspective, Intel’s goal is to sell high-end, high margin E5/E7 parts which can do similar things but cost up to 10x. By offering server level eDRAM parts at consumer prices when there is no competition in that space could drive potential customers for cheaper options, lowering Intel’s potential, and why we only see Iris Pro on quad core ECC-enabled processors at this time. There are a few Iris Pro enabled SKUs at the consumer space, but as mobile parts or for mini/all-in-one machines, rather than full blown gaming systems or workstations.

 

[Arachnotronic]

Indeed, no reason to do a separate die for something so trivial as using a different socket.

Just remembered this slide:

Looks like Kaby for enthusiasts is an LGA-1151 part.

 

[mikk]

Looks like Kaby for enthusiasts is an LGA-1151 part.

Benchlife says it's a 4 core SKU which doesn't make sense if true. Imho it only makes sense if this version offers at least 6 cores or alternatively a GT4e GPU so that Intel could ask for a premium price. Otherwise such SKU placed in the enthusiast segment doesn't make sense.

 

[Khato]

Just remembered this slide:

Looks like Kaby for enthusiasts is an LGA-1151 part.

Well, image doesn't show up for me. But I don't see any reason why there wouldn't be.

Let's put it another way with an example from current products - there are 'K' series 6000-level SKUs based on SKL-S, SKL-H, and BDW-E. So why must we draw the conclusion that the presence of a KBL-X would negate the possibility of 'K' series KBL-S and KBL-H SKUs?

 

[2blzd]

The way I read the roadmap was that the unlocked Kabylake lga-115x 6700k replacement was coming out in Q2 of next year, around the same time as the HEDT Skylake chips on Socket R (2061?)..

Skylake-X is the new name for Skylake-EP/P, merged for 1S using the KabyLake PCH
Kabylake-X is the new name for the unlocked 6700k Kabylake replacement, also on Kabylake PCH.


Intel made this very confusing, but there is no "Kabylake-E". After Skylake-X (EP/E) comes Cannonlake-X (EP/E). This also confirms that whatever Kabylake chips are launching this Fall with the new 200 chipset series will not have unlocked or performance skus.

The new PAO cycle, (Performance, Achitecture, Optimize) is allowing the HEDT to catch up with DT cores. So for those who were complaining about that, things are changing. We're getting 3 DT core generations for every 2 HEDT cores, with the 3rd DT and 2nd HEDT being the same Core Arch. So Cannonlake and Cannonlake-X should be out at around the same time, eliminating the gap.

 

[witeken]

To be honest, the CPU space just isn't exciting anymore. Innovation is basically zero. It's a mature and declining market.

 

[jpiniero]

The way I read the roadmap was that the unlocked Kabylake lga-115x 6700k replacement was coming out in Q2 of next year, around the same time as the HEDT Skylake chips on Socket R (2061?)..

I guess the question is whether Kabylake-X is also (only) on Socket R.

 

[VirtualLarry]

Innovation is basically zero.

And H1B's are increasing, by leaps and bounds. Coincidence? Hmm...

 

[DaveSimmons]

Sigh. It's cool that they keep improving the mobile and atom chips, but I keep waiting and waiting for a good reason to replace my i5-2500 for gaming.

+20% for buying a new motherboard, CPU and RAM is not very exciting. I can easily afford them, but for the effort I'd like to get at least +50% and preferably +100%. How many more iterations will that take

 

[RussianSensation]

Hmmm... they are suggesting Kaby Lake-X would have to serve the mainstream desktop market with quad cores with unlocked multiplier, and TDP of 95W? Q2 2017 for barely an improvement over the 6700K? That doesn't look good for Icelake-K (now rebranded X) coming out in 2H of 2018 because where would Cannonlake-X slot in then for LGA115x?

I hate to say it but IDontCare seems to be more right with each passing month since August 2015 Skylake launch. It seems there won't be any viable upgrade from 6700K for 5 years. With this news, there isn't much point for Sandy/Ivy i5 users to keep waiting unless they specifically want SKL-E and to see how Zen stacks up against it.

As mentioned earlier, we could even see KBL-X come out at $379-399 as Intel easily sold out of 6700K for a good 5-6 months at > $400 retail prices. Looks like the CPU space is dying a slow uneventful death. Intel is practically giving i7 Sandy/Ivy/Haswell users no reason to upgrade until Icelake at this rate. Soon the average life-cycle of an i7 CPU will go beyond 5 years as it's hard to imagine IPC confinuing to keep improving at the same rate. It somewhat explains why Intel raised prices on BW-E. It smells like a start of $430-$2000 roll-out of 6/8/10/12/14 core CPUs in the future across this entire spectrum. Since Intel is struggling with IPC gains as an incentive to spur upgrades, the next best option is selling MOAR cores as a differentiating factor.

Good to see that SKL-E may stick with 2011-3 socket so that coolers can be reused.

 

[2blzd]

I guess the question is whether Kabylake-X is also (only) on Socket R.

The sources say its a lga 1151 chip...Socket-R would indicate an HEDT Kabylake chip, which has been stated many times, is not happening.

Good to see that SKL-E may stick with 2011-3 socket so that coolers can be reused.

The article says lga 2061. We've know for a long time SKY-E would get a new socket and chipset

 

[Arachnotronic]

As mentioned earlier, we could even see KBL-X come out at $379-399

I have no doubt that they are going to increase prices of KBL-X. The chip is now in a totally different tier, and will be priced as such.

My bet is on $399.

 

[IndyColtsFan]

Hmmm... they are suggesting Kaby Lake-X would have to serve the mainstream desktop market with quad cores with unlocked multiplier, and TDP of 95W? Q2 2017 for barely an improvement over the 6700K? That doesn't look good for Icelake-K (now rebranded X) coming out in 2H of 2018 because where would Cannonlake-X slot in then for LGA115x?

I hate to say it but IDontCare seems to be more right with each passing month since August 2015 Skylake launch. It seems there won't be any viable upgrade from 6700K for 5 years. With this news, there isn't much point for Sandy/Ivy i5 users to keep waiting unless they specifically want SKL-E and to see how Zen stacks up against it.

As mentioned earlier, we could even see KBL-X come out at $379-399 as Intel easily sold out of 6700K for a good 5-6 months at > $400 retail prices. Looks like the CPU space is dying a slow uneventful death. Intel is practically giving i7 Sandy/Ivy/Haswell users no reason to upgrade until Icelake at this rate. Soon the average life-cycle of an i7 CPU will go beyond 5 years as it's hard to imagine IPC confinuing to keep improving at the same rate. It somewhat explains why Intel raised prices on BW-E. It smells like a start of $430-$2000 roll-out of 6/8/10/12/14 core CPUs in the future across this entire spectrum. Since Intel is struggling with IPC gains as an incentive to spur upgrades, the next best option is selling MOAR cores as a differentiating factor.

Good to see that SKL-E may stick with 2011-3 socket so that coolers can be reused.

If that article is to be believed, the 7700K not being released in a year is an incredible fail IMO. At this stage, I am just waiting until Zen is released to see how it stacks up and then I'll pick my next platform.

I think X99 with a 6-core CPU might be the best route at this point - as you state, it is clear that rapid IPC growth isn't something we can depend on so the ability to upgrade core count in the future without buying an entire new platform might be important. Of course, if everyone is still buying quads due to these high prices, maybe core count won't matter for a long, long time.

At any rate, bring on Zen and let's see what happens.

 

[IndyColtsFan]

I have no doubt that they are going to increase prices of KBL-X. The chip is now in a totally different tier, and will be priced as such.

My bet is on $399.

I'm a bit confused - isn't the Kabylake X just the 7700K?

 

[VirtualLarry]

I'm a bit confused - isn't the Kabylake X just the 7700K?

Yes, but Intel is shuffling deck chairs so that they can charge even more "Premium" for those chips. After all, now they'll be "X" class, and not just "K" class.

 

[IndyColtsFan]

Yes, but Intel is shuffling deck chairs so that they can charge even more "Premium" for those chips. After all, now they'll be "X" class, and not just "K" class.

It would seem that Intel is trying to make up for shrinking revenue by slapping the "premium" label on as much stuff as they can rather than innovating.

 

[Arachnotronic]

Yes, but Intel is shuffling deck chairs so that they can charge even more "Premium" for those chips. After all, now they'll be "X" class, and not just "K" class.

We should wait to see what they offer before writing it off. If the chip comes at very high clock speeds out of the box and/or are very specially binned, then I could see that being worth a premium.

If they literally ship a Skylake + 200Mhz and try to sell it at $399+ then that's not cool.

 

[bigi]

To be honest, the CPU space just isn't exciting anymore. Innovation is basically zero. It's a mature and declining market.

True()

The architecture is getting killed my Moore's Law. CPUs are slow and run hot.

This trend will not change with die shrinking and such.

Unless we get Quantum CPU out of big super-cooled labs and make them usable for 'normal' stuff current CPUs will end.

Also, there are plenty of computational issues that current architecture can't even touch.

 

[StrangerGuy]

Hmmm... they are suggesting Kaby Lake-X would have to serve the mainstream desktop market with quad cores with unlocked multiplier, and TDP of 95W? Q2 2017 for barely an improvement over the 6700K? That doesn't look good for Icelake-K (now rebranded X) coming out in 2H of 2018 because where would Cannonlake-X slot in then for LGA115x?

I hate to say it but IDontCare seems to be more right with each passing month since August 2015 Skylake launch. It seems there won't be any viable upgrade from 6700K for 5 years. With this news, there isn't much point for Sandy/Ivy i5 users to keep waiting unless they specifically want SKL-E and to see how Zen stacks up against it.

As mentioned earlier, we could even see KBL-X come out at $379-399 as Intel easily sold out of 6700K for a good 5-6 months at > $400 retail prices. Looks like the CPU space is dying a slow uneventful death. Intel is practically giving i7 Sandy/Ivy/Haswell users no reason to upgrade until Icelake at this rate. Soon the average life-cycle of an i7 CPU will go beyond 5 years as it's hard to imagine IPC confinuing to keep improving at the same rate. It somewhat explains why Intel raised prices on BW-E. It smells like a start of $430-$2000 roll-out of 6/8/10/12/14 core CPUs in the future across this entire spectrum. Since Intel is struggling with IPC gains as an incentive to spur upgrades, the next best option is selling MOAR cores as a differentiating factor.

Good to see that SKL-E may stick with 2011-3 socket so that coolers can be reused.

I'm not complaining since this trend saves me $$$ on CPU upgrades.

Not just CPUs....RAM, HDDs, SSDs, I/O interfaces are already in deep real-world diminishing returns territory.

 

[jpiniero]

The sources say its a lga 1151 chip...Socket-R would indicate an HEDT Kabylake chip, which has been stated many times, is not happening.

OK, maybe it's simply a branding move to start the transition to really push enthusiasts over to the X/W platform. Maybe if you're lucky the X/W will get a high clockspeed optimized process but that probably won't start until Cannonlake.

 

[2blzd]

OK, maybe it's simply a branding move to start the transition to really push enthusiasts over to the X/W platform. Maybe if you're lucky the X/W will get a high clockspeed optimized process but that probably won't start until Cannonlake.

I think thats probably whats going on. Intel is now branding the unlocked K processors in the same segment as HEDT chips. Which makes sense because in ST and other workloads, its similar or better in performance (than HEDT).

 

[RussianSensation]

I think X99 with a 6-core CPU might be the best route at this point - as you state, it is clear that rapid IPC growth isn't something we can depend on so the ability to upgrade core count in the future without buying an entire new platform might be important.

If I had a 2600K @ 4.5Ghz, I'd wait until SLK-E. You waited this long, might as well wait 1 more year and get a 6-core SLK-E and an all new socket/chipset that will have a long life and one a newer architecture than BW-E.

Of course, if everyone is still buying quads due to these high prices, maybe core count won't matter for a long, long time.

At any rate, bring on Zen and let's see what happens.

I think they looked at their long-term roadmap and realized they are near the end of the road as far as CPU clocks and IPC goes. By slowing down the mainstream platform and allowing the workstation platform to catch up, when Sandy/Ivy/Haswell owners are up for upgrades, Intel may offer them a $434-450 6-core or a $399 4-core. They'll win either way since the workstation platform's mobo/chipset will have a premium while on the mainstream platform they'll have huge margins due to small die sizes.

If Intel teamed up with AMD and worked with PC developers to have more AAA DX12 games that benefit from 8-12 cores, they could spur upgrades. However, they a dilemma because if they do this, they are walking right into Zen's strength (8C 16T). I guess that's why they announced a change of strategy away from the traditional PC space, but remain focused on the PC enthusiast/overclockers space by raising prices on us with BW-E and now we might see more price hikes with SLK-X and KBL-X.

I'm not complaining since this trend saves me $$$ on CPU upgrades.

Not just CPUs....RAM, HDDs, SSDs, I/O interfaces are already in deep real-world diminishing returns territory.

Ya true, but JHH realized gamers have more descritionary income to spend since the main platform from Intel now lasts 5+ years. We save $ on the CPU/platform side but GPU side's prices are going way up. I am not even sure what Intel's strategy will be to get i7 3770K/i7 4770K/i7 4790K/i7 6700K users to upgrade. For sure the latter 3 of those CPUs will last until the end of PS4/XB1 generation (I am thinking 2019-2020).

Looking at this chart, A9X is only as fast as a Intel Core i3-3217U. I looke up the scores of C2D T9800 2.93Ghz and it beats the i3-3217U in 3DMark06, CB10 single and multi. I have a couple friends using older C2D 2.0Ghz T5800/6400 laptops. Going to order T9800 online, upgrade them to W10 and put the SSD into the laptops. Those laptops will probably be good to go for another 2-3 years, enough until Icelake. Once my friends upgrade to Icelake in 2018-2019, I bet that laptop will last them 10+ years considering those C2D laptops are from early to mid-2009. :sneaky:

Intel needs to do something by working with software developers, or sooner or later the only sales of CPUs they will get will be the top 5% enthusiasts who want the latest and greatest, or only newer generations of people going to high school/university. Even firms will likely shift from a 2 year laptop replacement to 4-5 years.