Leaked Intel Rocket Lake tests show double digit perf gains

[HEXUS]

And it is claimed that the Intel 11th Gen Core family will mix RKL and CML-Refresh CPUs.

Read more.

[[GSV]Trig]

As much as I like my AMD stuff, I really wanna see what Intel would do on 7nm TSMC wafers..

[Tabbykatze]

As much as I like my AMD stuff, I really wanna see what Intel would do on 7nm TSMC wafers..

It would be nice because then we have an apples to apples comparison.

[spolsh]

Have I got the wrong end of the stick .. or are they actually reducing the number of cores ?

[Tabbykatze]

Have I got the wrong end of the stick .. or are they actually reducing the number of cores ?

They are, they backported the cove cores from 10 to 14nm and it seems the dies were too big and toasty meaning bad yields and hot cores so they had to reduce the top end core count.

[ik9000]

wait what, so this is another 14nm design? WTF happened to "10nm is finally ready" (after all those pushbacks before it)

[Tabbykatze]

wait what, so this is another 14nm design? WTF happened to "10nm is finally ready" (after all those pushbacks before it)

Apparently they just have no capacity between doing the mobile designs and the fact it's still (somewhat) garbage on yields so they ported it back to 14nm because they'll just migrate capacity.

Not a good look no matter how you spin it because of how much money and time they would have spent porting from one node back to a predecessor node (likely less than the expected because of how experienced they are at 14nm) but it tells a damning story about 10nm.

[ik9000]

do you think for desktop they'll jump from 14 straight to their 7nm (as apparently they had two different teams working on each process simultaneously). That would be a weird turn of events. If their old claims about their 7nm (not to be confused with TSMC 7nm) using a different fab technique altogether with less multi-layering woes it's presumably possible to end up with that working before their 10nm ever gets there.

[Tabbykatze]

do you think for desktop they'll jump from 14 straight to their 7nm (as apparently they had two different teams working on each process simultaneously). That would be a weird turn of events. If their old claims about their 7nm (not to be confused with TSMC 7nm) using a different fab technique altogether with less multi-layering woes it's presumably possible to end up with that working before their 10nm ever gets there.

Well 7nm is meant to be Intels first foray into EUV lithography so shouldn't be using multi-patterning, at least initially. Whereas 10nm still uses traditional UV masking with multi-patterning and I remember some very in depth articles from a few years ago that the biggest hurdle of cost and yield was due to the multi-patterning. Every single pass has to be pitch perfect, the tiniest deviation can make whole swathes of the plate unusable.

I would expect that 7nm is the big target but if they keep having these issues, trust and faith will be problematic in the node. 10nm performs well but is expensive and still seeming to be leaky at high frequencies which have gotten quite a bit better with the SuperFIN tech but it's not enough because they really need to get out of the multi-patterning mode.

But by they time they have EUV working, both Samsung and TSMC will have had many years with the technology in production and that's why at least TSMCs roadmap for new fabrication shrinkages seem to be so aggressive, it's nuts. If Intel doesn't ramp up and TSMC keeps their rate of acceleration going, Intel 7nm will be competing with TSMC 3nm of which Intels 7nm was meant to be feature similar to TSMCs 5nm.

I also fully expect for Intel to be pushing to go the next node down with 5nm PDQ to at least match, there was talk about that earlier in the year but not much talk about it now.

[edmundhonda]

Apparently they just have no capacity between doing the mobile designs

If the yields were good, they would have already switched capacity. I just don't buy the conclusion that they're designing new 14nm chips to launch in 2021 for any other reason.

do you think for desktop they'll jump from 14 straight to their 7nm

The Intel 7nm process got pushed to late 2022 at the earliest, Alder Lake is going to be on 10nm next year.

[Tabbykatze]

If the yields were good, they would have already switched capacity. I just don't buy the conclusion that they're designing new 14nm chips to launch in 2021 for any other reason.

Absolutely, I'm willing to bet bottom dollar there are many internal politic fires still being put out over the absolute crapshow that Murthy created and is still ongoing.

The Intel 7nm process got pushed to late 2022 at the earliest, Alder Lake is going to be on 10nm next year.

That was a big narker when they announced that, Sapphire Rapids will be 7nm iirc but they're still not looking good according to industry insiders.

[Iota]

14nm+++

This is a meme by now, surely.

[3dcandy]

They are tanking stocks wise as well at moment (Intel that is)
And after a new CEO and writing off debt and bad investments

If you believe it - 10nm is so utter garbage they've binned it for 10 nm+++++ FinFET+%$&*

[bae85]

Do they come with liquid nitrogen kits as standard for the 400 watt power draw?

[bae85]

14nm+++

This is a meme by now, surely.

I thought they were on 14nm+++++++++++++++++++++++++++++++

[ik9000]

14nm(+)^n is an easy thing to remember and type. Intel I'm trademarking that. You can use it under licence subject to terms and fees.