News Leaked Meteor Lake Mobile Chips Reportedly Have Lower Clock Speeds Than Raptor Lake

Status
Not open for further replies.
New Chips with disappointing characteristics ! ? ?
No, all that matters is IPC vs watts for meteor lake. Until we have that its pure speculation.

This could be a superior chip in multiple ways to raptorlake mobile which while could clock to the moon, was massively inefficient once you applied turbo.

Meteorlake according to the multiple releases/leaks brings on a host of welcome new innovations:
•First tile based consumer chip from Intel consisting of four tiles: Graphics, SoC, CPU, and IOE. The LP E-cores reside inside the SoC tile,
•Larger and more advance iGPU (128EU based alchemist+ vs raptors 96EU)
•VPU capabilities (low power dedication AI)
•P cores go from Raptor Cove to Redwood Cove (Expect higher IPC)
•E cores go from Gracemont to Cresmont (IPC and efficiency improvements?)
•Enabled on-package DRAM which brings massive memory performance uplift

Until we have them released though its any's guess how it will perform. Looking forward to Intel joining AMD on being able to release glued together computer chips :)

I do find it very ironic that AMD is still releasing monolithic mobile computer chips and does chiplets for desktop and server. While Intel has just released its first chiplet server product(Sapphire Rapids) and is starting its chiplet approach only on Mobile (Meteor Lake) while desktop will stay monolithic for another generation(Raptor Lake Refresh).

There was a power penalty to go chiplet and that is why AMD is wise to keep it monolithic on mobile, where something as small as 100mW is significant. Interested to see how Intel has compensated for this chip to chip power penalty.
 
While this might be a new leak, I think this had been leaked before. It certainly makes sense, in the context of desktop Meteor Lake being cancelled - the two obvious reasons for doing so would be uncompetitive clocks and cost. The fact that core counts are similar to Raptor Lake's suggests it's not a serious problem with yield.

The clock speed disparity should be expected with a brand new architecture like Meteor Lake
No, not if you look back at Intel CPUs from the past decade+. In fact, if you do that, the pattern that emerges is not one with new architectures, but with new process nodes:
  • Broadwell, the first on 14 nm, couldn't clock high enough, so its desktop version was essentially cancelled and Intel had to issue Haswell Refresh.
  • Cannon Lake and Ice Lake, the first on 10 nm, had such problems with clock speed that Intel had to make Whiskey Lake - yet another Skylake derivative - for the high-end laptop segment.
  • Meteor Lake is the first on Intel 4, and like with the Gen 10 laptop CPUs, there's rumored to be a Gen 14 high-end mobile part that's actually a Raptor Refresh.

'Intel 4' node, which has twice the transistor density of Alder Lake designs and features 21.5% higher frequencies at the same power level.
Kinda funny that the whole point of the article is that this claim doesn't seem to hold up. I guess we can't know for sure, since it's meant to apply to the same design being fabbed on both Intel 7 and Intel 4, whereas Redwood Cove and Crestmont are new cores.
 
This could be a superior chip in multiple ways to raptorlake mobile
If it weren't, then it probably would've been canceled, outright.

•P cores go from Raptor Cove to Redwood Cove (Expect higher IPC)
•E cores go from Gracemont to Cresmont (IPC and efficiency improvements?)
The leaked benchmarks of single- & multi- threaded tasks from Arrow Lake (see below) suggest the bigger improvement is going to be in the E-cores, which makes sense. Other analysis suggests Redwood cove didn't change much from Raptor Cove.

•Enabled on-package DRAM which brings massive memory performance uplift
Nope, that's not actually new. Raptor Lake already did this.


Also, I'm not really sure that you can't do LPDDR5X-7400 off-package. AMD's Phoenix supports that standard, and I've never heard of an on-package memory version of it.

Until we have them released though its any's guess how it will perform. Looking forward to Intel joining AMD on being able to release glued together computer chips :)
Leaked benchmarks, supposedly from Intel themselves of Raptor Refresh and Arrow lake - the generation after Meteor Lake:


I do find it very ironic that AMD is still releasing monolithic mobile computer chips and does chiplets for desktop and server.
Yup, but don't forget about Lakefield - that was Intel's first foray into EMIB and Foveros - also on a mobile SoC. While not terribly successful, we can hope that Intel learned a lot from it.
 
Status
Not open for further replies.