MLID says Meteor Lake is in trouble. Desktop pushed to 2024, only lower core count configurations, likely not able to compete on the high end.
Notebook should still come out in 2023.
Tom seems to be hinting at some potential problems with "Intel 4" and casts some doubt on "Intel 3" as well, but it may just be speculation, and the problems may be on the design end of things, and not on process technology side of things.
IF this ends up true, I just want to add, Raichu leaked this 7 months ago, and Kopite Kimi wasn't far behind at 5 months ago.
Lets do a deeper analysis of when this most likely would have been cancelled.
Based on Intel's Raptor Lake development cycle:
7 months ago we heard that MTL and ARL will both compose a one generation product. Lets use the worst case scenario and say that was exactly when it was cancelled, though likely the decision probably would have been made a couple months before that even since leaks don't immediately match up to when events occur.
Analysis of when MTL-S could have been cancelled
Based on an 1Q 2024 launch date, the rumors of cancellation would be ~ T-22 months in the development cycle. This would place it at the very end of the design phase, right before a A0 Tape-In.
However, it is important to remember that A) the official decision to cancel likely happened a bit before the rumors caught wind of it B) MTL-S might launch later than 1Q 2024 and most importantly, C) the RPL development cycle was shorter because it shared the same basic architecture AND platform as ADL. 6 months shorter according to Intel themselves. Since MTL is supposed to be an improvement of Golden Cove/ Raptor Cove, I doubt the dev cycle would be closer to ADL cycle of 36 months, but still, I would add a couple of months because of the different platform.
Does this match with Intel's other statements about MTL? Yes. If MTL mobile is to launch in end of 2023, and MTL-P powered on T-18 months ago, then it would fit in the dev cycle as RPL powered on 15 months before it launched. So it seems like MTL has an extra dev cycle time of +3 months compared to RPL, and -3 months compared to ADL.
So overall, I strongly believe that MTL-S got cancelled in the middle of the design aspect of the development cycle. Meaning that Tom saying that Intel's execution of MTL sucked which is why they cancelled, I believe, is incorrect. Maybe they didn't want to risk creating such a big compute tile on Intel 4, or they didn't want to spend more on developing more compute tiles for desktop, or maybe they didn't want to limit mobile stock, but considering that it was cancelled so early in the design process, I don't think it's because of execution problems.
But was MTL-S ever seriously planned?
At the end of July 2021, Intel announced Meteor Lake going up to the 125 watt power range. This is the main justification that Tom, and many others, use that Intel had always planned for a MTL-S lineup. And while that is true, we have to think about the time frame of these events. Even if we say Intel, at the time, planned for MTL-S to release at the end of 2023 (with no RPL-R refresh in mind), that would put MTL-S a good 30 months away. And keep in mind, when this event took place, the fact that Intel thought they would have a meteor lake desktop variant had probably already been in mind for a couple of months as well, considering that announcements publicly aren't when the decision in company had been made. What this means was that MTL was probably in the very VERY early stages of design, if even that....
And if you need any further confirmation of this timeline, keep in mind other aspects of Intel products that would be released in around the same timeframe were ALSO changed. Granite Rapids was announced to be on Intel 4 in that event, and MTL was also supposed to scale up to 192 EUs, which also doesn't seem like it's happening anymore.
What I think ACTUALLY happened
So, Tom had been insisting about a MTL-S complete lineup in his past 3 major Intel leak videos. This is despite other leakers like Kopite and Raichu having differing views. However the way and reason Tom claimed MTL-S, or atleast the highest end of it, isn't coming out, gives him a win-win situation. If MTL high end DOES come out, he can point to his old videos and claim that Intel fixed whatever recent execution issues they had with MTL, and are therefore launching a HCC MTL-S model. However, if they don't launch a HCC MTL-S model, he can point to this video and claim that he always knew Meteor Lake-s was facing issues and that's why they cancelled.
And the reason why he released THIS video when he did, I believe, is not because of any inside info about execution issues with the HCC models THIS late in the development cycle, but rather because we are ~12-15 months away from a 1Q 2024 MTL-S launch, but he also must have seen a striking lack of ES-1 and early validation of those HCC MTL-S models.... which SHOULD have been happening IF they existed.
Facebook
Twitter