Discussion Intel current and future Lakes & Rapids thread

[Yotsugi]

So we now know Cascade Lake is basically a bug-fixed Skylake as some have speculated(CD for example).

And Apache Pass support, probably, hopefully.

For example, allow 28 core part to be available for $4-5K, rather than $10-13K.

I can already audibly hear shareholders screeching at BK for mere suggestion of that.

 

[Kaloi48]

According to two reliable posters on Taiwan PTT forum (in Chinese), Kaby Lake-U would be succeeded by Coffee Lake-U, and then Whiskey Lake-U (WHL-U 4+2) next year. 10nm Cannonlake is still missing in action.

 

[DrMrLordX]

Cascade Lake has 28 cores.

That's um . . . hmm. Well that's very interesting, when you consider what <insertcompanynamehere> is doing at <insertdatehere> that may make 28c Xeons seem a bit dated.

Hell the stuff Cavium and Qualcomm are doing ought to raise a few eyebrows. Just a few.

 

[Yotsugi]

That's um . . . hmm. Well that's very interesting, when you consider what <insertcompanynamehere> is doing at <insertdatehere> that may make 28c Xeons seem a bit dated.

Hell the stuff Cavium and Qualcomm are doing ought to raise a few eyebrows. Just a few.

Intel's only way to push more cores right now is a node shrink or a bigger die.
XCC is already 698mm^2, and it's pretty stingy for Intel's margins even that way.

 

[Dayman1225]

Cascade Lake has 28 cores.

Thanks to Ashraf
https://twitter.com/TMFChipFool/status/937245082561974274

So we now know Cascade Lake is basically a bug-fixed Skylake as some have speculated(CD for example). Since they are not increasing core count at all, for their sake I hope the market positioning is far better than they have now. For example, allow 28 core part to be available for $4-5K, rather than $10-13K.

The process fallout continues.

Yes, it's simply a refresh on 14nm++ with optane dimms support. I doubt "MSRP" prices will change because they don't charge companies nearly that much. Like I mentioned in another thread, Cascade Lake SP/X is lameo.

 

[Dayman1225]

According to two reliable posters on Taiwan PTT forum (in Chinese), Kaby Lake-U would be succeeded by Coffee Lake-U, and then Whiskey Lake-U (WHL-U 4+2) next year. 10nm Cannonlake is still missing in action.

First gen 10nm is DOA and won't receive a proper ramp. CNL is DOA too because it's 2/4 and will have such a short time on the market. 10nm+ is the "real" 10nm ramp IMO.

 

[Ajay]

Yes, it's simply a refresh on 14nm++ with optane dimms support. I doubt "MSRP" prices will change because they don't charge companies nearly that much. Like I mentioned in another thread, Cascade Lake SP/X is lameo.

I’m interested in seeing how much improved Cascade Lake in perf/watt. 14++ will help a fair bit, but if the 'bug fixes' address power consumption then CL's market position would be improved vs The competition. Obviously, there is no fundamental change, such as core counts, as we have seen in the past - yields would have tanked if Intel pushed to the reticle limits.

 

[Dayman1225]

According to two reliable posters on Taiwan PTT forum (in Chinese), Kaby Lake-U would be succeeded by Coffee Lake-U, and then Whiskey Lake-U (WHL-U 4+2) next year. 10nm Cannonlake is still missing in action.

Do you have a link to it?

 

[Kaloi48]

Do you have a link to it?

https://www.ptt.cc/bbs/PC_Shopping/M.1511170137.A.94F.html
WHL-U 4+2: Production window ww27-34'18

 

[Dayman1225]

https://www.ptt.cc/bbs/PC_Shopping/M.1511170137.A.94F.html
WHL-U 4+2: Production window ww27-34'18

Cheers!

 

[Yotsugi]

https://www.ptt.cc/bbs/PC_Shopping/M.1511170137.A.94F.html
WHL-U 4+2: Production window ww27-34'18

Wait, STILL no godforsaken Cannonlake?
Is this a joke?
What will be the next excuse for BK and his boys to make?

 

[Dayman1225]

Wait, STILL no godforsaken Cannonlake?
Is this a joke?
What will be the next excuse for BK and his boys to make?

"I'm stepping down as Intel's CEO" Well hopefully

 

[Yotsugi]

"I'm stepping down as Intel's CEO" Well hopefully

Isn't BK of fab background?
Anyway, Intel's execution on 10nm is beyond woeful.
TSMC will be fabbing A11x when Intel STILL won't have any Cannonlake on the market.
Why Intel, why.
Just why.

 

[Dayman1225]

Isn't BK of fab background?
Anyway, Intel's execution on 10nm is beyond woeful.
TSMC will be fabbing A11x when Intel STILL won't have any Cannonlake on the market.
Why Intel, why.
Just why.

Can anyone hear that..? It's called Bad Management! and yes I believe BK is of a Fab backround.

 

[IntelUser2000]

I’m interested in seeing how much improved Cascade Lake in perf/watt. 14++ will help a fair bit, but if the 'bug fixes' address power consumption then CL's

The "bug fixes" I referred likely includes Optane DIMM support.

One I'd like to see is Loop Stream Detector enabled again. They had to disable it on Skylake-SP/X. That should bring some power consumption reduction. I wouldn't be surprised if few % performance improvements happen as well.

"I'm stepping down as Intel's CEO" Well hopefully

I won't go that far, yet. Mistakes of predecessors get carried over. Nothing exists in a vacuum(well computer benchmarks try).

I doubt "MSRP" prices will change because they don't charge companies nearly that much.

They should. Because MSRPs aren't totally irrelevant. You can buy them in stores for close to MSRP pricing. Volume discounts are likely to discount from MSRP as well. I looked at Dell and HP system configurations and Xeon Platinum CPUs add few thousand dollars over Xeon E5 V4 ones.

Either way, raising prices from $4k to $13k is a price increase for those that come from E5. A proper successor to E5 does not exist. That's for AMD to take unfortunately.

Wait, STILL no godforsaken Cannonlake?

Cannonlake is early next year.

Whiskey Lake is Icelake so that's even better, if they can do this. I doubt it'll come before holiday 2018.

 

[DrMrLordX]

Wait. So Cannonlake is going to be a paper release on mobile, and a complete non-starter on desktop. 10nm is dead. Stick a fork in it, it's done.

Whiskey Lake is Icelake so that's even better, if they can do this. I doubt it'll come before holiday 2018.

So 10nm+ there? You sure about that? I thought 10nm+ was going to be server-first.

 

[IntelUser2000]

So 10nm+ there? You sure about that? I thought 10nm+ was going to be server-first.

That never made sense. Servers have a long validation cycle. "Server-first" can be true even if technically the chips arrive a bit later than client.

https://www.nextplatform.com/2017/04/28/intel-moves-xeons-moores-law-leading-edge/

Data Center Group would be a fast follower on 10 nanometer process node and would come out ahead of the client chips with the 7 nanometer node.

So 7nm will do the first thing. Even then I wouldn't be surprised if it comes couple of months later than client.

So 10nm+ there? You sure about that? I thought 10nm+ was going to be server-first.

I may be wrong on this. Updates as its developed.

Update:
→ chrisdar : CFL-U 4+3e Prod ww11-16'18
→ chrisdar : WHL-U 4+2 Prod ww27-34'18

So Whiskey Lake is a platform, and may be a Coffeelake-U on a 300-series PCH. Based on Prod dates it makes it a summer 2018 launch.

If its true they are having problems on a 10nm process as a whole. It means *not* 10nm not okay and 10nm+ being ok.

You'd think they could wait 3-4 months and launch Icelake. I have hard time seeing how we'll get any Icelake-S parts in 2018 if this is true. We might see a 8790K with 8 cores instead.

 

[Dayman1225]

Not 100% confirmed but Whiskey Lake seems to be just Coffee Lake-U but in a 4+2 configuration instead of 4+3e for mass market notebooks. 14nm++, Cannon Lake PCH. Ouch. - Ashraf Eassa on Twitter

 

[Yotsugi]

Not 100% confirmed but Whiskey Lake seems to be just Coffee Lake-U but in a 4+2 configuration instead of 4+3e for mass market notebooks. 14nm++, Cannon Lake PCH. Ouch. - Ashraf Eassa on Twitter

Ayy, it's nothing.
Where's Cannonlake?

 

[Dayman1225]

Ayy, it's nothing.
Where's Cannonlake?

I honestly do not see the point in releasing Cannonlake at this point. Last investors call they still said EOY. So expect stuff at CES I guess. By Stuff I mean a second wafer and half a laptop

 

[Dayman1225]

That never made sense. Servers have a long validation cycle. "Server-first" can be true even if technically the chips arrive a bit later than client.

https://www.nextplatform.com/2017/04/28/intel-moves-xeons-moores-law-leading-edge/



So 7nm will do the first thing. Even then I wouldn't be surprised if it comes couple of months later than client.



I may be wrong on this. Updates as its developed.

Update:
→ chrisdar : CFL-U 4+3e Prod ww11-16'18
→ chrisdar : WHL-U 4+2 Prod ww27-34'18

So Whiskey Lake is a platform, and may be a Coffeelake-U on a 300-series PCH. Based on Prod dates it makes it a summer 2018 launch.

If its true they are having problems on a 10nm process as a whole. It means *not* 10nm not okay and 10nm+ being ok.

You'd think they could wait 3-4 months and launch Icelake. I have hard time seeing how we'll get any Icelake-S parts in 2018 if this is true. We might see a 8790K with 8 cores instead.

I assume you missed the roadmap?





Stop with the multiple posting of images across multiple CPU threads.


esquared
Anandtech Forum Director

 

[LTC8K6]

Not 100% confirmed but Whiskey Lake seems to be just Coffee Lake-U but in a 4+2 configuration instead of 4+3e for mass market notebooks. 14nm++, Cannon Lake PCH. Ouch. - Ashraf Eassa on Twitter

Boy that guy doesn't seem to like Intel at all...

I wish I'd known that before bothering to read what he says.

 

[IntelUser2000]

I assume you missed the roadmap?

You don't think they'll release it at all because nothing is on the roadmaps for -S series beyond 6 core CFL in 2018?

They can put whatever they want on these presentations. That's how they kept the Radeon effort as a secret for so long.

Skylake-X was known as 10 core tops until a short time before release and 12 core tops until very short time before release. 18 cores was a big secret.

Boy that guy doesn't seem to like Intel at all...

It has nothing to do with liking Intel. I read his posts and articles for some time. I guess at some point he was a "fanboy" but his articles started turning neutral, probably because he became an analyst for an investor community.

You can like something but still admit that its sucking.

 

[Yotsugi]

Boy that guy doesn't seem to like Intel at all...

At least he's not trying too (https://twitter.com/FPiednoel/status/937798460895412224) hard.

 

[LTC8K6]

At least he's not trying too (https://twitter.com/FPiednoel/status/937798460895412224) hard.

I don't like to refer to such sources, regardless of who they appear to cheer for.

But a sample size of one is pretty silly to make a judgement with...