Why is Kabylake-X a downgrade from Skylake-X?

It goes from 44 lanes of PCIe to 16. I thought those rumours were typos, but it's getting so close to the launch of x299, there's gotta be some validity now. The only thing I'm thinking of is they're trying to compete with ryzen by having entry level quad cores on an enthusiast platform and there will be higher end Kabylake-X chips later.

Did you see the "up to 24 PCIe Gen3" lanes from the PCH? Doesn't that add up to 68?

1 Like

They basically take the die from the 7700k and put it onto a 2088 pin substrate.
It still only supports 2 channel memory, 16 lanes and so on.

I think they're doing this for people that want higher clockspeeds, they might push 5ghz stock on those chips.
One thing I'm curious about though is how they're going to do the power delivery, X299 probably has FIVR like haswell-broadwell, but kaby lake die doesn't have that.
Maybe just use the normal power delivery on those boards as normal vcore instead of vinput.

1 Like

I don't think there will be any Kaby Lake-X processors with higher core counts. That's because the Intel's HEDT chips were almost always based off of server dies, and Intel doesn't plan to introduce Kaby Lake to servers. So the X299 platform will likely be stuck with the 6+ core SKUs still on the Skylake architecture (no process improvements from Kaby Lake).

THats the only plausible reason i could think off, for that 4 core Kabylake X chip.
But then still i dont see any point for it to exists.

1 Like

If I was trolling I'd say: Because 4 cores ought to be enough for everyone, duh.

Maybe it's the poor mans E7-8893v4. 4 cores with a metric-but ton of cache.

Or maybe it's something to do with yields?

What I do like about that thing is the equal number of physical cores and memory channels. That's nice.

So KabyLake-X will be a 7700k but with a bigger DMI to support more pcie lanes through the chipset, but no quad channel memory and only 16 lanes off the cpu? That's the dumbest part I've ever heard of. What is the target audience for that? Why is the quad-core Kabylake but the rest are skylake? Why not just label them all Kabylake? Kabylake is skylake but taking advantage of process refinements, so it implies that they're implementing process improvements on the quad core, but not on any chips above 4 core.

Nobody buy the Kabylake-X chip if that's all it is. That's a dumb part.

The dmi on X299 will still be 4x gen3, the same as mainstream Skylake and Kabylake i think?
Basiclly i dont really see much of a point for that Kabylake-X chip.
However on X79 they basiclly did the same thing, with the i7-3820 and the i7-4820K
But those 2 particular i7´s were the entry parts of the X79 platform.
So i dont really see where a 4 core 8 threads Kabylake-X cpu,
would fit any current market in terms of pricing.

this seems like its going to fail

Intel has no clue what it wants to compete against on ryzen

More cores?
More PCIe lanes?
Less Cost?
Better IPC?

Seems like intel is throwing everything at the wall and see what sticks

6 Likes

Well they have to do something with pricings definitelly.
Allthough Skylake-X will be faster then Ryzen.
But launching a 7900K 8 core / 16 threads cpu for $1000,- all over again,
isnt gonne convince manny buyers most likely.

2 Likes

I think your right about the DMI stuff, my bad. I misinterpreted the diagram in the top article. The KabyLake-X part is basically identical to the Kabylake mainstream 7700k. I initially viewed the diagram as saving that the PCH would add an additional 24 lanes to a KYLK-X cpu's 16, but I think it means the total would be 24 lanes with 8 provided by the chipset. The only real difference I see is that the Kabylake-X part has a higher 112 watt tdp vs the 91 watt tdp of the 7700k, leading me to think that the stock clock on this chip may be closer to 5ghz. That would theoretically make sense since it will also be axing the iGpu, meaning a reduction in tdp from the lack of that piece.

I really don't get this move including the quad core part. If Coffee Lake is supposed to bring the 6 core to the mainstream, why now bring a 4 core to the HEDT sector?


I'll put my price predictions with core counts and names.

7950x 10 core $999
7900k 8 core $499
7850k 6 core $349
7800k 4 core $299 and a delayed release/availability.

I think Coffee Lake will start with a 6 core I7 at $309 and work down from there, with quad core + HT at the I5 level and quad no HT at the I3 level, maybe with some dual core + HT I3's as well.


What REALLY is going to be interesting is how the Xeon pricing is structured in a preemptive response to Naples and the Ryzen Server platform.

1 Like

There are always usage cases for higher clocked chips, I think intel wants to push out something like that.

The mainstream die on a HEDT substrate does a few things.

1st, increases the cooling potential with the larger heatspreader, likely soldered as well to further increase cooling, thus higher clocks can be sustained with lower power draw.
2nd, using a 2088pin socket that has a lot more pads for ground and power decreases the impedance(resistance) of the power delivery, yielding lower voltage sag which can help with higher clocks.
3rd, the stronger power delivery will also help a fair amount.

I mainly think intel wants to push something that can sustain 5ghz+ clockrates for those needing high single thread, and I also think we're going to see Xeon equivalents like we have seen of some really highly clocked Xeons before.

One thing to take note of is that this implies that skylake-x wont clock as high as kaby-lake, I'd expect something that clocks slightly worse than a 6700k.

Still, who's going to buy it? The ONLY way I see this making some sense is if motherboard makers purpose make motherboards only for the KBYLK-X part. 2088 pin socket but with lesser pcie slots, and only 4 DDR4 Dimm slots. Still even the cpu might be priced more than the Kabylake mainstream parts, and the chipset cost of the HEDT platform means the motherboards will be more expensive than the mainstream boards. It'll be a weird qwazzi middle-ground between the mainstream and the HEDT, where at that point your probably best off either spending a bit more for the 6 core skylake or saving a bit and going to the mainstream platform.

The only reason I've been able to think of why as to Intel's doing this is that they don't want to completely skip having a Kabylake chip on the HEDT line, but they don't want to put out an entire generation of Kabylake chips on the HEDT platform, so this is there way of at least having one chip on the platform. This way they can skip having a processor generation's worth of time of waiting before they drop Coffeelake-X on the HEDT.

I think they are really just running around trying to come up with something because their 10nm process is not going so well for them. It will be a while before they can get something out there for the consumer so this and S-X are just an attempt to tide us over.

Will probably be a failure IMO.

Yeah the pricings are really gonne be interesting here.
I mean intel has to do something with their pricings.
I´m not sure if true, but i also have heard some rumors about a 12 core i7 on the new platform.
So if i look at your price predictions, that would totally fit.

The Kabylake X part, will indeed be similar to the 7700K, but without a igpu,
and probablly more overclock potential.
If i look at it from a vrm topology standpoint, then that makes totally sense.
You dont have to be bottered with the additional input rails (powerphases) needed on the vrm side of things for the igpu.
Which makes things allot easier.
But a larger socket with more pins and a larger chip could all lead to higher overclock potentials.
Like @Castellorizon allready pointed out.
Also given the rumor that the Kabylake X part only to support dual channel memory, which is probablly also done with an eye on overclock potential.
As we all know, quad channel memory causes allot more stress on the imc.
Thats why on X99 you have a pretty hard time to reach high memory overclock on a quadchannel configuration.
Especially if you also populate all slots.
Thats one of those reasons why Gigabyte with their X99 Soc champion only has 4 dimm slots.

There is of course allways a scenario thinkable in which a stable high clocked quadcore could be usefull.
But still from a price perspective, i totally dont really see where it would fit.
Especially not with their current main stream line.

all i hear is a waste of moneeyyyyyyyy

more ways to milk.money from u... why bjy a ything from intel

I wonder if it will be soldered to the IHS, as I wonder how much cache it will have. Leaked numbers say only 8MB? Same amount as ordinary desktop 7700K, sounds small to me. The silicon die would be tiny without an IGP. Soldering a very small die to a heat-spreader isn't easy, there are quite serious problems with micro cracks. If it has a lot of cache the die could be quite a bit larger, making it easier to solder. Cache is expensive though.

I agree that it doesn't make much sense this way. A lot more cache would make more sense. And/or insane clockspeed. Not that earlier four-core CPUs on the HEDT platform has made a lot of sense either, at least the old Sandy and Ivy four cores had plenty of PCIe lanes.

That's b/c Intel are some greedy money grubbing scumbags who provide minimal gains yr after yr w/ their supposed upgrade lineup offerings. I hope AMD takes them to the cleaners w/ the next chip they release. Hope that'll help Intel get their act straight

I thought the reason why the 7xxx CPUs aren't soldered is that soldering puts too much stress on smaller dies. Ryzen dies are bigger, so AMD can solder them, but even AMD uses thermal paste for their smaller dies.