Intel Skylake / Kaby Lake

[R0H1T]

Z2x0/H2x0 chipset will be forward compatible with Cannonlake.

Z97/H97 chipset were launched in the same year as Devil's Canyon & Haswell Refresh and is fully forward compatible with Broadwell.

Actually what I meant was that all chipsets based on LGA 1151 socket should be forward, & backward, compatible especially since we're in the post tick-tock era & Kaby would barely qualify as an upgrade from Skylake, same goes for CNL chipsets.

 

[jpiniero]

Z2x0/H2x0 chipset will be forward compatible with Cannonlake.

Z97/H97 chipset were launched in the same year as Devil's Canyon & Haswell Refresh and is fully forward compatible with Broadwell.

I'd say it's unlikely now that Kabylake (and Skylake-C) is here. Besides FIVR.
Broadwell only had the Broadwell-C release and not the typical full desktop lineup. It's a sales/marketing thing to limit it to 2 gens more than anything else. And at the rate they are going at, mainstream will be BGA only before you know it.

 

[pantsaregood]

The use of 7xxx part numbers should finally bring the mainstream and HEDT parts back into line.

Skylake-E will be 7xxx and Kaby Lake will be 7xxx. It's unlikely that we'll see a HEDT Kaby Lake, so Cannonlake and Cannonlake-E will both fall under 8xxx.

 

[dark zero]

There's already a precedent, Devil's Canyon, though I did expect something more than just a feature or two being added to the next gen (9.5?) IGP. If the 2xx series chipsets aren't compatible with Cannonlake, & skylake, then Kaby would be the biggest ripoff in recent memory & that'd be some achievement even for Intel.

INBF they add again the DLC for the processors lol... that would be hilarious...

 

[frozentundra123456]

Who cares about the model number. HEDT is always at least a generation behind, unfortunately. So you have to make the choice between more cores and the latest architecture.

Still looking for that mainstream hex core.

 

[crashtech]

Mainstream hex will kill HEDT sales, so until/unless Zen forces their hand, ain't gonna happen.

 

[Arachnotronic]

Mainstream hex will kill HEDT sales, so until/unless Zen forces their hand, ain't gonna happen.

HEDT sales are pretty minimal as it is. The reason that there currently isn't a mainstream Hex is that it would have a very limited market -- desktop "K" SKUs only. It would not sell well as a server chip nor would it do well in power constrained notebooks.

 

[frozentundra123456]

Mainstream hex will kill HEDT sales, so until/unless Zen forces their hand, ain't gonna happen.

That is the point, isn't it? Get the highest performance chip on the latest process node and best architecture. A sale is still a sale, and it might even encourage more people to upgrade, increasing sales. And with Broadwell, they could continue the niche 8 and 10 core models on the HEDT platform. If anything, it would threaten the quad i7 more than the HEDT lineup.

 

[crashtech]

So then the only plausible rationale is that Intel doesn't think hexacores are relevant to current power user's needs.

 

[IntelUser2000]

I think HEDT is Intel's Titan X and Radeon Pro Duo. They fall under the pretense of "workstation chip at consumer pricing" but really just unfortunate product positioning.

The reason they don't do hexcore on LGA115x is otherwise they wouldn't sell HEDT at all. If they increase cores on HEDT then they risk cannibalizing -EP sales.

 

[DrMrLordX]

HEDT sales are pretty minimal as it is. The reason that there currently isn't a mainstream Hex is that it would have a very limited market -- desktop "K" SKUs only. It would not sell well as a server chip nor would it do well in power constrained notebooks.

Funny how the market has changed? Flagship products used to be worth it despite low volume of sales. Now Intel can't stand to have a mainstream hexcore that has a very limited market? I tend to agree with frozentundra: Intel needs something that can energize the buying public somehow, someway. Obviously usage trends have changed but still, hexcore Skylake would still be quite an upgrade from quadcore Haswell.

 

[nerp]

Funny how the market has changed? Flagship products used to be worth it despite low volume of sales. Now Intel can't stand to have a mainstream hexcore that has a very limited market? I tend to agree with frozentundra: Intel needs something that can energize the buying public somehow, someway. Obviously usage trends have changed but still, hexcore Skylake would still be quite an upgrade from quadcore Haswell.

I bet something like this is coming.

 

[skipsneeky2]

I bet something like this is coming.

A mainstream 6 core without ht would be nice,could have one at like 3.4Ghz that is locked for like $200 that can drop into a cheap micro atx motherboard.Unlocked k model sitting closer to 3.8Ghz in the $230-$250 range.All these improvements in IPC and energy efficiency,i bet there is a market for a 90w Hexacore as well.

I know i would certainly drop my i5 2500 in a second for a affordable Hexacore that can drop into a affordable Micro atx. Just slap on USB 3.0 and Sata 6 native and there is a upgrade i would buy.Extreme chips like a 8 core with ht can still sell with their given platforms.

 

[poofyhairguy]

A mainstream 6 core without ht would be nice,could have one at like 3.4Ghz that is locked for like $200 that can drop into a cheap micro atx motherboard.Unlocked k model sitting closer to 3.8Ghz in the $230-$250 range.All these improvements in IPC and energy efficiency,i bet there is a market for a 90w Hexacore as well.

I know i would certainly drop my i5 2500 in a second for a affordable Hexacore that can drop into a affordable Micro atx. Just slap on USB 3.0 and Sata 6 native and there is a upgrade i would buy.Extreme chips like a 8 core with ht can still sell with their given platforms.

That would be awesome. I too would drop Sandy for that. Let's all hope AMD's Zen forces some more aggressive behavior like that.

 

[pantsaregood]

Who cares about the model number. HEDT is always at least a generation behind, unfortunately. So you have to make the choice between more cores and the latest architecture.

Still looking for that mainstream hex core.

I care about the naming convention because it makes no sense.

Nehalem was all "first generation."
Westmere was just forgotten.
From Sandy Bridge on, we've seen the HEDT parts incremented one "generation" above what they should be.

This makes no sense in any capacity. Generation 8 could potentially be a good thing - 8xxx on LGA 1151 would denote Cannonlake, and 8xxx on the HEDT socket would indicate Cannonlake-E. There's no nonsense where you're claiming an old node (or even worse, old architecture) is part of the "current" generation.

I also believe Intel is trying to bring E/EP/EX release schedules closer to mainstream releases. They've kind of uncontrollably spiraled at this point. At this rate, we're going to see Kaby Lake before we see Broadwell-E.

 

[ShintaiDK]

They've kind of uncontrollably spiraled at this point. At this rate, we're going to see Kaby Lake before we see Broadwell-E.

Broadwell-E is out, you just didn't see the HEDT yet because they wait for Computex. Its purely artificial due to the gaming factor of Computex. The SKUs are stocked etc.

 

[frozentundra123456]

Well, i was referring to the current situation where HEDT is one or two generations behind. All I am saying is that the cpu "is what it is" no matter how you label it. Now if the lengthening of the tic/toc cycle allows HEDT to catch up in process or architecture to mainstream, that would be a good thing.

But if they just brought out a mainstream hex core, it would solve all the problems. And it would even allow moving uo from a dual or quad core on the same motherboard.

 

[ShintaiDK]

The problem with server/client is that a new core is developed as a client first, then stripped its uncore and the server uncore gets finished. Then you have extra server validation time. So no matter what, there will still be 9-12 months minimum between the 2 even in the best utopian case. Unless you start to delay the client product.

 

[Sweepr]

Non K it seems.

3.6Ghz base, 4.2Ghz turbo. Seems to replace 3.4Ghz base, 4Ghz turbo 6700.

Apparently clickbait websites missed your post.

 

[ShintaiDK]

Apparently clickbait websites missed your post.

It sure seems so. But then they can make another article with just as many clicks to correct it

 

[DrMrLordX]

The problem with server/client is that a new core is developed as a client first, then stripped its uncore and the server uncore gets finished. Then you have extra server validation time. So no matter what, there will still be 9-12 months minimum between the 2 even in the best utopian case. Unless you start to delay the client product.

Server uncore huh. Do the HEDT platforms have the server uncore on their CPUs?

 

[ShintaiDK]

Server uncore huh. Do the HEDT platforms have the server uncore on their CPUs?

What else would they have?

Client uncore is what you see on LGA115x.

 

[frozentundra123456]

Again, same old tune, if they made a mainstream hex core, would not have to worry about any of this.

 

[DrMrLordX]

What else would they have?

Client uncore is what you see on LGA115x.

That's what I thought, just wanted to be sure about it. What are the performance implications between client and server uncore?

 

[ShintaiDK]

That's what I thought, just wanted to be sure about it. What are the performance implications between client and server uncore?

Just compare the products?