Vendors Share Raptor Lake Refresh Non-K CPU Specs

Metal Messiah. said:

Nope. The entire 14'th gen "refesh" lineup on the LGA 1700 socket platform is over, and Intel has already dropped the Pentium and Celeron branding/naming scheme . They will only use the CORE or CORE ULTRA nomenclature from now onwards, be it a desktop or a Mobile SKU. Like 300T.

So no more models are coming out under the RPL-refresh family.

These two things are not mutually exclusive, they can still release celeron and pentium tier CPUs on 14th gen and just use the new names for them.
https://www.intel.com/content/www/us/en/processors/processor-numbers.html

Starting in 2023, these affordable computing products will be simply branded as Intel® Processor, instead of Intel® Pentium® and Intel® Celeron® processor.

domih said:

@bit_user is not erring much on this topic. As a "detached" aspect of this RPL refresh, the question is:

Will INTEL move ahead from the Alter Lake N-series to a Raptor Lake-N series?

See https://ark.intel.com/content/www/us/en/ark/products/codename/232598/products-formerly-alder-laken.html
This series include low power embedded processors massively used in TV boxes and SBCs. The N100 for example, is all over the place in news sites like cnx-software, liliputing, etc and you can find numerous models on AliExpress. INTEL notably bumped up the series with the N-300/305 providing 8C8T, compared for instance to the Gemini Lake or Jasper Lake series that were topping at 4 cores. The thing is that through the fauna of architecture names and processor core names, one can easily get lost.

INTEL has a definitive $$$ interest to continue upgrading this series, whatever the name they'll use. Especially because AMD is in comparison way too costly for these embedded solutions. Small SBC makers can't deal with AMD up front cost (unit OEM price X minimum qty). INTEL is doing a much better job at allowing small SBC makers to create sub-$200 mobos or boxes. IMHO, AMD is plain dumb in this story, it could produce low cost processors that would easily compete and beat the INTEL models. One may guess that AMD is content with selling its silicon to the major game boxes makers.

I don't know who brought up Alder Lake-N, but I think I can answer that.

First off, we have seen a "Twin Lake" leak from MLiD. Supposedly, it will be a minor refresh of Alder Lake-N, with no Crestmont cores. But I can think of one thing Intel could do with the same silicon that would make it much more interesting: upgrade the N200 to 6 cores. Alder Lake-N has 2-core N50, 4-core N95/97/100/200, and 8-core N300/305. I don't see any technical barrier to introducing a 6-core in there if they can make a 2-core. The only thing N200 has going for it is the full 32 EUs iGPU. Adding two more cores would make it more compelling and waste less silicon.

Rather than AMD being dumb, I think they are focusing on premium CPUs/GPUs/APUs that have much better margins. Intel is supposedly making razor thin margins on some of its consumer chips, but it keeps their fabs busy and market share high.

AMD theoretically has a good response to Alder Lake-N: Mendocino. It has similar CPU/GPU performance and AV1 video decode capability in a small die. But they clearly aren't going to challenge the N100 tsunami. They will continue to make various premium laptop APUs, console APUs, and chiplets for Ryzen/TR/Epyc.

If AMD wanted to go for the low-end again, I think they would need to tap Samsung. Just make a single Mendocino-like chip there, and spam it out. But it might not help them make much money if it ends up in the bargain bin.

domih said:

@bit_user is not erring much on this topic. As a "detached" aspect of this RPL refresh, the question is:

Will INTEL move ahead from the Alter Lake N-series to a Raptor Lake-N series?

See https://ark.intel.com/content/www/us/en/ark/products/codename/232598/products-formerly-alder-laken.html
This series include low power embedded processors massively used in TV boxes and SBCs. The N100 for example, is all over the place in news sites like cnx-software, liliputing, etc and you can find numerous models on AliExpress.

I'm well familiar with that series. I have seen no information about a refresh of the Alder-N die, but I'm just reading news sites and not even as plugged into the rumor mill as some.

My gut says "no", for the simple reason that Intel didn't truly launch Alder-N until sometime around early this year (those ARK database entries match my recollection, listing Q1 '23 as the launch date). I think those dies are pretty big, relative to what Intel can sell them for, resulting in a limited upside to any further investment Intel puts into that range.

I think I found a die size for them that indicated they're about 60% as big as the full fat 8P + 8E Alder-S die. Furthermore, I believe all N-series are made from the same die, since they all have the same amount of L3 cache. This leads me to wonder if they vary in the amount of L2 cache. For instance, some of the higher-spec 2-core and 4-core models might split the cores between clusters. That should also result in less thermal throttling, by better distributing the heat.

Finally, consider that Intel already has a new E-core design (Crestmont) they're launching in Meteor Lake. They'd get more performance & efficiency from focusing their efforts on getting a Meteor-N launched using the Intel 4 or Intel 3 nodes. I guess it depends somewhat on how the costs and production volumes of the new nodes are looking.

domih said:

INTEL is doing a much better job at allowing small SBC makers to create sub-$200 mobos or boxes. IMHO, AMD is plain dumb in this story, it could produce low cost processors that would easily compete and beat the INTEL models. One may guess that AMD is content with selling its silicon to the major game boxes makers.

AMD makes some embedded SoCs, but those are using obsolete Zen cores and GCN graphics. For mini-PCs, people want newer than that, which puts them into the AMD laptop range. That's a little higher-end than Intel's E-core based SoCs. There's the problem.

AMD has announced some lower-end products, seemingly aimed at the Chromebook market, but the AMD-based mini-PCs I see don't seem to embrace those SoCs. I've seen plenty of AMD-based mini-PCs announced over the past couple years, but why they don't seem to target the same entry-level market as Intel's N100 I can't really say. It could be that the cost of TSMC's N7 node just hasn't fallen enough.

I think the SBC market is pretty niche. I don't mean to cast aspersions, but the SBC market are "bottom-feeders", in marketing lingo. They basically eat the scraps left over by other markets. Traditionally, you don't see SoCs targeted at them, aside from maybe the Broadcom chips being made specifically for the Raspberry Pi series.

bit_user said:

I think I found a die size for them that indicated they're about 60% as big as the full fat 8P + 8E Alder-S die. Furthermore, I believe all N-series are made from the same die, since they all have the same amount of L3 cache. This leads me to wonder if they vary in the amount of L2 cache. For instance, some of the higher-spec 2-core and 4-core models might split the cores between clusters. That should also result in less thermal throttling, by better distributing the heat.

https://locuza.substack.com/p/info-snack-alder-lake-m-raptor-lake
More importantly, it's about 75% the size of 2P + 8E + 96 EU Alder Lake die that has much better single/multi, triple the size of the graphics, more cache, and a better memory controller.

I don't know if it's accurate. Maybe they used a different cell library or something to make Alder Lake-N even smaller, but it's disturbing if true. I think even a small reduction can have a big impact on yields and wafer usage, but then we see seemingly the majority of these chips being disabled from 8-core, 32 EUs down to 4-core, 24 EUs. And yes, they all use the same die, so if you see a dual-core N50, it has 6 cores disabled.

I wonder if the economics would change for the better if "N"/Atoms eventually make the move to "tiles", instead of being monolithic dies.