Intel Comet Lake Thread

Page 5 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

jpiniero

Lifer
Oct 1, 2010
14,599
5,218
136
It gives a good idea when Rocket Lake-S will be released a year later. A little surprised it's not earlier.
 

Ajay

Lifer
Jan 8, 2001
15,454
7,862
136
It gives a good idea when Rocket Lake-S will be released a year later. A little surprised it's not earlier.
I thought it would be earlier too. Intel's manufacturing and CPU divisions seem to have two left feet lately. I suppose it’s allot harder to turn a cargo vessel than a sail boat.
 

jpiniero

Lifer
Oct 1, 2010
14,599
5,218
136
I thought it would be earlier too. Intel's manufacturing and CPU divisions seem to have two left feet lately. I suppose it’s allot harder to turn a cargo vessel than a sail boat.

Intel might be doing it intentionally for various reasons, like if they thought Rocket Lake-S wasn't likely to be ready until sometime like May 2021. By earlier I was thinking like January.
 

Ajay

Lifer
Jan 8, 2001
15,454
7,862
136
Intel might be doing it intentionally for various reasons, like if they thought Rocket Lake-S wasn't likely to be ready until sometime like May 2021. By earlier I was thinking like January.
Last I recall, from a while ago, CML was listed as 2019-2020. I suppose I thought that was for the full stack and that sooner was better than latter. Obviously, U/Y came first.
 

jpiniero

Lifer
Oct 1, 2010
14,599
5,218
136
Last I recall, from a while ago, CML was listed as 2019-2020. I suppose I thought that was for the full stack and that sooner was better than latter. Obviously, U/Y came first.

I believe you are correct. The other reason they would hold it back is of course that it would reduce the number of dies per wafer they would get with the additional 2 cores. So no early K release anymore, and they decided to release the 9900KS instead.
 

inf64

Diamond Member
Mar 11, 2011
3,698
4,018
136
Do we have any reliable source about Rocket Lake core? Will it be good old Skylake or back-ported Sunny/Willow Cove?
 

jpiniero

Lifer
Oct 1, 2010
14,599
5,218
136
Do we have any reliable source about Rocket Lake core? Will it be good old Skylake or back-ported Sunny/Willow Cove?

No good rumor on what Core it is, although there's been questionable speculation that it's Sunny. I tend to believe it's Skylake TBH but it could be any of the three.
 
  • Like
Reactions: inf64

TheGiant

Senior member
Jun 12, 2017
748
353
106
No good rumor on what Core it is, although there's been questionable speculation that it's Sunny. I tend to believe it's Skylake TBH but it could be any of the three.
my bet its willow cove or skylake
if they can backport sunny cove, the can do willow cove too- any of them will be expensive but if it must be done.....

my guess skylake 60%/willow cove 35%/sunny cove 5%

about that lineup fo comet..it comes down to the voltage/freq curve

if Intel managed to bend it somehow and 10C/20T comet is power wise on the level of previous 9900K, then its super competitive with 3900X
if you need water no matter what then ....
 

DrMrLordX

Lifer
Apr 27, 2000
21,632
10,845
136
my bet its willow cove or skylake
if they can backport sunny cove, the can do willow cove too- any of them will be expensive but if it must be done.....

Problem is we don't know transistor counts on Willow Cove at all. Sunny should already result in a larger core than CoffeeLake if implemented in 14nm. How much more die area will Intel have to sacrifice to move to their latest design on an old node? Will that hold them to 6c dice? Or will they do some kind of weird EMIB monstrosity with two sizeable dice and then a 10nm iGPU on a separate die?
 

inf64

Diamond Member
Mar 11, 2011
3,698
4,018
136
If it's indeed some of the "Coves" it will be rather big die, unless they do the chiplet thing. In that case it might just pay off to do a back port of the core as yields will be much better. Power draw on such a core done on 14nm(+++) is another thing. Fun times ahead as AMD is not stopping with Zen. I heard from one guy that Zen3 is at least 10% IPC uplift, so lower than Zen+ -> Zen2 but still very respectable jump. It will be tough for intel to compete using 14nm node but they have no choice in the near term.
 
  • Like
Reactions: lightmanek

jpiniero

Lifer
Oct 1, 2010
14,599
5,218
136
If it's indeed some of the "Coves" it will be rather big die, unless they do the chiplet thing. In that case it might just pay off to do a back port of the core as yields will be much better.

It'd likely be only one CPU chiplet max, and at 10 cores just the CPU chiplet would be at least 150, maybe 200.
 
  • Like
Reactions: inf64

Ajay

Lifer
Jan 8, 2001
15,454
7,862
136
It'd likely be only one CPU chiplet max, and at 10 cores just the CPU chiplet would be at least 150, maybe 200.
I think HardwareLuxx suggested RocketLake was still Skylake 14nm. Other leaks seem to hint that the iGPU will be on it's own chipset - saving some die space. That would make sense if Intel was trying to push the core count up to 12, which would be the practical limit for Skylake's ring bus.

Intel has pushed back hard on the idea that there will be no Desktop 10nm part. The current rumor is 10nm -S part in 2021/2022! With an iGPU chiplet, and perhaps using Golden Cove, Intel could deliver a substantial enough increase in IPC to make it worth something. Though, if true, I wonder if it will be restricted to OEMs.
 

Asterox

Golden Member
May 15, 2012
1,026
1,775
136
Starting to look more likely that the entire lineup will see HT.


i3:4/8
i5: 6/12

i7: 8/16 (though it could be 10/10)
i9: 10/20

Yes great news, especially for people who paid a ton of money for the "old Intel 6/12 CPU".

Now i3 4/8, but in the past it was a very expensive flegship CPU again great news.


Hm what to buy, i3 4/8 for 130$ or "6/12 Ryzen 5 2600 for 120$ a difficult choice".
 
Last edited:

jpiniero

Lifer
Oct 1, 2010
14,599
5,218
136
I think HardwareLuxx suggested RocketLake was still Skylake 14nm. Other leaks seem to hint that the iGPU will be on it's own chipset - saving some die space. That would make sense if Intel was trying to push the core count up to 12, which would be the practical limit for Skylake's ring bus.

It's been pretty consistent on Rocket being 10 cores.
 

TheGiant

Senior member
Jun 12, 2017
748
353
106
Problem is we don't know transistor counts on Willow Cove at all. Sunny should already result in a larger core than CoffeeLake if implemented in 14nm. How much more die area will Intel have to sacrifice to move to their latest design on an old node? Will that hold them to 6c dice? Or will they do some kind of weird EMIB monstrosity with two sizeable dice and then a 10nm iGPU on a separate die?
well Intel is manufacturing already 28C cascade lake x monolithic monstrosities...
if 10nm partially faulty Xe GPU + 10C willow cove is done....with all that IPC uplift and voltage/freq curve it can be done
9900K does 95W full throttle avx2 at 4,1GHz and that is the 14nm++, not the more+
it cant be related directly but gives some insight IMO
 

DrMrLordX

Lifer
Apr 27, 2000
21,632
10,845
136
well Intel is manufacturing already 28C cascade lake x monolithic monstrosities...

Two things

1). Those dice sell at a premium
2). Those dice are packaged on large PCBs that go into sockets with a lot of pins.

LGA1200 - assuming Intel sticks with that after CometLake - isn't going to feature CPUs that are either one of those things. They may run out of room under the IHS. Or they can dump LGA1200 and move to a larger socket that can feature a larger package.
 

mikk

Diamond Member
May 15, 2012
4,140
2,154
136
Sunny Cove is also in the mix for Rocket Lake, RedGamingTech says it.

We also have news that rocket lake is based on the sunny cove architecture, when a source of mine confirmed this to me via email.

And some time ago Intel hinted that only the first Sunny Cove products will start on 10nm.