News Intel Rumored to Pull-in 2nm Arrow Lake-S Launch to H1 2024

[Admin]

Report says Intel's Arrow Lake-S may replace Meteor Lake-S in desktops.

Intel Rumored to Pull-in 2nm Arrow Lake-S Launch to H1 2024 : Read more

 

[MBOO7]

They dont seem to learn from their old mistakes...
Still wanting to put too much into one product all at once, causing high dependencies on multiple levels...
If one level fails the entire product goes through purgatory, waiting for a final release (only getting delayed and delayed).

 

[zecoeco]

More of the same.
In my opinion, the only rumors that are hard to believe is Intel being ahead of schedule.

 

[cyrusfox]

In my opinion, the only rumors that are hard to believe is Intel being ahead of schedule.

🤣 Too true

With the move to dissaggregation though, should be faster to iterate (Keep same IO & GPU and change out CPU part).
As Meteor Lake is the first iteration, still premature speculation that Intel can execute.

I don't see how Raptor lake Refresh can compete if Meteor Lake Desktop is shelved. Reminds me of Comet to Rocket lake, they pushed out new CPU that was no better than prior gen... I don't get doing a Raptor Lake refresh, makes more sense to me to just release i5 meteor lakes, especially if it can best the 7800X3D.

 

[JamesJones44]

The timing doesn't seem right. It's rare you start production on a node and start shipping volume on that node all at the same time. Maybe Intel 4 is way way ahead and volume ramp would not be at the same time as the product launch, but just based on timing I think odds seem to favor a rumor for clicks.

 

[Arbie]

I'm trying to recall when I last believed or cared about an Intel report. I think it was around 2017.

 

[ikjadoon]

Why is Intel not giving any concrete news about this year's node, Intel 4?

A lot of "silence on 2023, shiny new thing in 2024" statements from Intel.

Intel 4 is apparently manufacturing ready, so it should have a lot more concrete news to share by now. CPU launches, sure, you're focused on Raptor Lake sales.

But the node? Jeez, Intel, the whole tech world is basically waiting for the "under-performing" or "delayed launch" or "slower ramp" report from IFS.

 

[Howardohyea]

I thought 20A isn’t even taped out yet and would only be production ready in H1 23, if that’s the case then the soonest product we can see using 20A is probably in H2 23 or even 24

 

[SiliconFly]

I thought 20A isn’t even taped out yet and would only be production ready in H1 23, if that’s the case then the soonest product we can see using 20A is probably in H2 23 or even 24

The article states it's based on a unverified leak/rumor. 20A test chips already taped out. But thats just proof of concept. ARL cpu tile hasn't taped in yet in 20A (probably within 3 to 6 months). We can expect ARL tape out only after that.

20A volume ramp in 1H 2024 is unlikely. Q4 2024 is more likely (but that too is a hard bet).

 

[SiliconFly]

Why is Intel not giving any concrete news about this year's node, Intel 4?

A lot of "silence on 2023, shiny new thing in 2024" statements from Intel.

Intel 4 is apparently manufacturing ready, so it should have a lot more concrete news to share by now. CPU launches, sure, you're focused on Raptor Lake sales.

But the node? Jeez, Intel, the whole tech world is basically waiting for the "under-performing" or "delayed launch" or "slower ramp" report from IFS.

Ya it's kinda strange that there isn't much news/leaks about Intel 4 & meteor lake at this point considering it's in the later stages of the cycle. It's very unlike Intel.

Either something has changed abt how they do things there (unlikely). Or the state of affairs for MTL/Intel 4 is actually worse than thought. Already there are way too many rumors and/or predictions about MTL clock-speed regression due to Intel 4.

At this point, MTL might match RPL. Or might DOA. No one knows. Got to wait for some concrete info.

 

[bit_user]

With the move to dissaggregation though, should be faster to iterate (Keep same IO & GPU and change out CPU part).

I think that's more of a side-benefit. The main benefit should be their ability to use better cost/perf-optimized nodes for each tile.

I don't see how Raptor lake Refresh can compete if Meteor Lake Desktop is shelved.

Compete against what? I thought AMD isn't slated to launch Zen 5 until 2024. Raptor Lake already holds its own against (non-3D) desktop Zen 4 products.

It probably won't move the needle much vs. Ryzen 7000X3D, unless they're doing it to add their own 3D stacked cache.
🤯

Reminds me of Comet to Rocket lake, they pushed out new CPU that was no better than prior gen... I don't get doing a Raptor Lake refresh, makes more sense to me to just release i5 meteor lakes, especially if it can best the 7800X3D.

That's a lot to unpack.

• Rocket Lake did advance lightly-threaded workloads - especially those benefiting from AVX-512.
• Rocket Lake brought Xe-LP to the desktop, which was a pretty big step up from Gen 9.5 HD graphics. More of a footnote, but worth mentioning.
• Haswell Refresh was pretty successful (i.e. i7-4790K), though maybe CPUs like the i9-13900KS already reaped most of the gains left to be squeezed out of Raptor Lake.
• The refresh could include some architectural tweaks that address various bottlenecks, such as L3 cache.
• Even an i5 Meteor Lake might not be viable on the desktop, if they're facing issues like Intel had with Ice Lake - where it couldn't clock high enough to beat prior-generation desktop CPUs with the same core-count. It was still viable for low-power laptops, which is the market Meteor Lake is set to pursue.

 

[bit_user]

I thought 20A isn’t even taped out yet and would only be production ready in H1 23,

Tape-out refers to a specific chip design, not a manufacturing node.

 

[Howardohyea]

Tape-out refers to a specific chip design, not a manufacturing node.

My bad, I got them confused. Regardless, this is way too fast for Intel or any development in general and I don’t see how they can get this on the shelves in one year

 

[bit_user]

My bad, I got them confused.

In your defense, I have occasionally seen a mention of taping out a test chip, in discussions about development of new process nodes. In order to do that, I think they probably need to have completed their first version of the cell libraries... or perhaps the test chip happens even earlier and is used to develop the cell libraries.