Intel Rocket Lake Preview: A New Architecture

>> " Why is Intel adding 10nm graphics to a 14nm processor, and how?"

Through chiplets, that's how.

With Rocket Lake at best a minor upgrade (with the possible exception of those Xe integrated graphics), and still likely half a year from release, it's hard for me to get terribly excited over it.
 
Going to be a tough sell to get 10900K owners to 'upgrade' from a 10 core to an 8 core.

But let's be honest, the 10xxx CPU's & these new Rocket Lake CPU's are nothing more than a short term stop-gap and Intel probably don't care whether current i9/i10 owners make the move to them, as they will be quickly forgotten.

I suspect Alder-Lake in Q3/Q4 2021 is where & when Intel will make their stand and go on the offensive against Ryzen.
 
Last edited:
Well, Rocket Lake in March 2021 will be late... it might have been sense in December, to fight against zen 3 (I’m expecting a 10/15% performance improvement on comet lake), but in March it will be late. Zen 3 will be widely available and prices will start to drop. How can you sell a rocket lake knowing that alder lake will come before the end of the year ? Unless alder lake has been delayed too...
 
Going to be a tough sell to get 10900K owners to 'upgrade' from a 10 core to an 8 core.

But let's be honest, the 10xxx CPU's & these new Rocket Lake CPU's are nothing more than a short term stop-gap and Intel probably don't care whether current i9/i10 owners make the move to them, as they will be quickly forgotten.

I suspect Alder-Lake in Q3/Q4 2021 is where & when Intel will make their stand and go on the offensive against Ryzen.
I9 10900K had no sense since the beginning... it was a marketing move, to have some benchmarks where Intel still prevails.
 
Wow, Intel's people finally got off of their butts and DID SOMETHING. It only took three years of AMD Zen to do this. However, what they've done is make hypocrites of themselves:

Hypocrisy #1: Now they're using chiplets, a method pioneered by AMD that Intel belittled calling Ryzen CPUs "Glued Together".

Hypocrisy #2: In that recent unveiling of their latest mobile APUs (the one where they talked more about the AMD R7-4800U more than their own offerings), they referred to AMD as "imitators" and now they're not only imitating AMD64 but also AMD's effective chiplet strategy.

Intel's hubris and hypocrisy is so overt that it's almost funny but it ensures that enthusiasts will be all too happy to tell Intel to "stick it".
 
AVX-512 on a standard desktop part? Either that is going to run hot or they'll down-clock when running AVX-512 instructions (as my HEDT does). Interesting move but is it worth the silicon (guess it will help with the compression/decompression benchmarks - but it's not about benchmarks is it intel)?
 
Wow, Intel's people finally got off of their butts and DID SOMETHING. It only took three years of AMD Zen to do this. However, what they've done is make hypocrites of themselves:

Hypocrisy #1: Now they're using chiplets, a method pioneered by AMD that Intel belittled calling Ryzen CPUs "Glued Together".

Hypocrisy #2: In that recent unveiling of their latest mobile APUs (the one where they talked more about the AMD R7-4800U more than their own offerings), they referred to AMD as "imitators" and now they're not only imitating AMD64 but also AMD's effective chiplet strategy.

Intel's hubris and hypocrisy is so overt that it's almost funny but it ensures that enthusiasts will be all too happy to tell Intel to "stick it".

Having a separate CPU and GPU on an MCM (Multi-Chip Module) is not the same thing as splitting the functional units of a single processor into multiple chiplets as AMD has done with Zen 2 and 3. MCMs, specifically for CPUs, have been around for quite a long time now.
 
I9 10900K had no sense since the beginning... it was a marketing move, to have some benchmarks where Intel still prevails.
It was faster in games, by a measurable margin, then the 9900k. The 7700k was a marketing move. The 10900k was a genuine improvement, even if it came at the cost of rampant power consumption and wasnt all that impressive.
 
So the Rocket lake chipset will equal the B550 but not x570 when it comes to features.

And the IPC claims need to be verified when we have a shipping product.

"Intel claims that Sunny Cove has an 18% IPC advantage over Skylake. Consequently, Rocket Lake should also have an ~18% IPC increase"

I suspect Alder-Lake in Q3/Q4 2021 is where & when Intel will make their stand and go on the offensive against Ryzen.

Just to run into Zen4 Q1 2022.
 
A poetic name. It's probably gonna fall out of the sky in pieces like two space shuttles did before ever reaching space.
 
Having a separate CPU and GPU on an MCM (Multi-Chip Module) is not the same thing as splitting the functional units of a single processor into multiple chiplets as AMD has done with Zen 2 and 3. MCMs, specifically for CPUs, have been around for quite a long time now.
Notably, intel did this before AMD did both with slot 1 putting L2 cache ont he same PCB as the CPU and with the pentium D having 2 processor dies on 1 PCB.
 
Alder Lake may be better than Rocket Lake. But if Rocket Lake is better than Ryzen, then wouldn't people holding off on buying Rocket Lake because Alder Lake is coming... also hold off on buying Ryzen because Alder Lake is coming?
Given that Dennard Scaling is no longer the case, being on 14nm shouldn't have to make Intel chips all that inferior, so if they've found a way to provide performance without being quite so power-hungry on the 14nm node with their new architecture, maybe the 14nm++++ derision will die down.
 
Going to be a tough sell to get 10900K owners to 'upgrade' from a 10 core to an 8 core.

But let's be honest, the 10xxx CPU's & these new Rocket Lake CPU's are nothing more than a short term stop-gap and Intel probably don't care whether current i9/i10 owners make the move to them, as they will be quickly forgotten.

I suspect Alder-Lake in Q3/Q4 2021 is where & when Intel will make their stand and go on the offensive against Ryzen.

Like we are going to see Alder-Lake in Q3 or even Q4 of 2021 is a best a joke. This will be a 1H 2022 Launch, otherwise it will be nothing more than a paper launch Q4 2021. Intel's 10nm for the Desktop is just not ready for the scale needed. It will go head to head with Zen 4, and will probably do very little to close the gap.

Rocket Lake is just the 14nm version of Tiger Lake, it isn't impressive. While 14nm will allow it to clock higher, even with high clock speeds it will be behind Zen 3 in single thread performance. Zen 3 is just has that much more IPC compare to the best Intel has ATM. And with Intel's chips requiring a lot of power, you'll need a good cooling solution. Rocket Lake is packing more transistors, so don't expect these to sip power at load.

IMO Skip Rocket Lake, if you're dying for an upgrade Zen3 is here. Otherwise wait for the Zen4 vs Alder-Lake fight.
 
It was faster in games, by a measurable margin, then the 9900k. The 7700k was a marketing move. The 10900k was a genuine improvement, even if it came at the cost of rampant power consumption and wasnt all that impressive.
that was exactly my point. It was faster IN GAMES and its existence was justified only by that, at the cost of a price and power consumption almost ridiculous. Not to speak about very limited availability on stores...
 
Why is Intel adding 10nm graphics to a 14nm processor, and how?

Hypocrisy #1: Now they're using chiplets, a method pioneered by AMD that Intel belittled calling Ryzen CPUs "Glued Together".
As nice as Ryzen's are, let's not spread FUD. Splitting CPU + iGPU across different dies made on different process nodes but within one CPU package aren't "chiplets", nor is it even new or "invented" by Ryzen. For those who can't remember, here's two 2010-era Clarkdale i3-530 reviews by Anandtech and TechPowerUp illustrating how (32nm CPU + 45nm iGPU on separate die but same package) is exactly how Rocket Lake (14nm CPU + 10nm iGPU graphics on separate die but same package) will work (as well as a photo of one delidded). I'm honesty surprised Techspot had to ask readers "how" it could be done...
 
Intel grabbing at 14nm straws. The war is over, they have lost for now. Alder Lake better be something else or AMD is going to continue eating away Intel.
 
Considering Intel's track record of delivering products on time I would be shocked if Rocket Lake has any availability before next summer, and there's no way Alder lake is actually launching next year and it'll likely be a paper launch as well.
 
It's just a throwaway generation especially since it tops out at 8 cores. I expect low availability and weird regional pricing.

Intel is praying that its 12th gen will be able to fight Zen4 in 2022.
 
" Fourth generation of Ryzen, launching next month, will put the final nail in Intel's coffin". I beg to differ, Intel is alive and well and probably immortal anyway (;p). Completely dismissing Intel on some inside AMD benches is just dumb. If, god forbid, they can pull another Core2duo moment with Alder Lake, then you guys better be prepared to nail another coffin.
 
I don't buy into the Alder Lake hype for gaming at least, the combo Strong core - efficient core seems interesting for laptops - mobile devices mostly. If rocket lake has both decent IPC and clock improvements, and is priced competitively, it could probably be a better choice for gaming than Alder Lake.

Intel current lineup is fine for gaming (GPU is still the limit) and if the info released so far is true, so will be AMD 5xxx. It will be mostly a question of pricing and availability for the gaming market share.

The real question is when intel is releasing new HEDT and what It would actually be. AMD's 5950x seems to be destroying both Intel HEDT and AMD Threadripper at the same time.
 
Back