First Gracemont laptop available for sale.
Anybody got disposable $500 to buy and test this laptop?
i3 1215u is better than that in both single and multi core and has better GPU.AMD built a chip (Mendocino) that is pretty good match for this low power, low cost market. But it seems no one is using it for this purpose. Only chromebooks. E.g. 7520U compared to N305 I think I would rather have the 7520U due to its slightly superior single-thread performance and dual-channel DDR5 support.
Yeah it's better than the N305 too. Is it price competitive with Mendocino and Alder Lake N? Should it be with its die size?i3 1215u is better than that in both single and multi core and has better GPU.
A17 Pro is on N3B. That's an EXPEN$IVE chip.Yeah it's better than the N305 too. Is it price competitive with Mendocino and Alder Lake N? Should it be with its die size?
If price and availability wasn't an issue then the A17 'Pro' is better still in both single and multi core and has a better GPU while having a smaller die size. And I'd go for that.
Smaller than ADL-M and that's my point. Mendocino and ADL-N are small and cheap by virtue of design. 1215U might be cheap by virtue of spare parts. That it too is better than ADL-N is about as relevant as A17 'Pro' blowing them all out of the water.A17 Pro is on N3B. That's an EXPEN$IVE chip.
What's the die size of Alder Lake N?
You can add ease of use across software platforms to that list too.If price and availability wasn't an issue then the A17 'Pro' is better still in both single and multi core and has a better GPU while having a smaller die size. And I'd go for that.
Yes but are they low power and small as these ones?If they cost below $150 yes but from $200 and above plenty of options. Those 6-8 core Zen 2/3 machines offer lot of value.
N305 use as much as 40W, same as 1215u hereYes but are they low power and small as these ones?
You'd be smacking a lot of people at Intel. Their product lineup is significantly curtailed compared to even a few years ago when they released multiple products based on Tremont (and only Tremont). You'll notice that since then:I want to smack the person around a bit who is responsible for the decision to feature darkmont cores in server and not tested in the consumer space first with the N series of CPUs. No one can even begin to imagine how infuriated I feel. It is unfair to let this market segment stagnate with older technology since it has a lot of volume in developing markets.
I wouldn't mind that at all!You'd be smacking a lot of people at Intel.
I wouldn't mind that at all!
Intel is beginning to see what happens when you lay off all your talent.You'd be smacking a lot of people at Intel. Their product lineup is significantly curtailed compared to even a few years ago when they released multiple products based on Tremont (and only Tremont). You'll notice that since then:
1). We've only had one tepid Gracemont-only product: Alder Lake-N.
2). The only Crestmont we've seen so far is in Meteor Lake. Eventually we'll get it in Sierra Forest, but that is likely to be a limited release to select clients that really want it.
Standalone "atom"/e-core SoCs for the consumer space do not appear to be Intel's priority at this time.
You aren't entirely wrong. Alder Lake-N isn't . . . a great product, but at least in theory, there should be fertile ground for more embedded/comm gear SoCs in the Ridge lineup as well as consumer SoCs they could release with 4c-8c (or more) based on Crestmont and Darkmont.They could have done something cool and made some money with the mont cores.
Why does it need to be tested?I want to smack the person around a bit who is responsible for the decision to feature darkmont cores in server and not tested in the consumer space first with the N series of CPUs. No one can even begin to imagine how infuriated I feel. It is unfair to let this market segment stagnate with older technology since it has a lot of volume in developing markets.
The greater volume for the consumer market should help them fix errata quickly and get to a very stable stepping which they can then use in servers.Why does it need to be tested?
Sierra Forest's Sierra Glen core is also slightly lower performing per clock than Crestmont due to not having the 6-wide rename/retire. Raichu said it's due to them wanting more clocks out of SRF.1). We've only had one tepid Gracemont-only product: Alder Lake-N.
2). The only Crestmont we've seen so far is in Meteor Lake. Eventually we'll get it in Sierra Forest, but that is likely to be a limited release to select clients that really want it.
Clearwater Forest is sometime in 2025, meaning it comes earlier than any client implementation, which is Pantherlake. They are overlapping product lines, process, designs to speed up introduction, which will take a lot of effort, but if they can do it, let them.The greater volume for the consumer market should help them fix errata quickly and get to a very stable stepping which they can then use in servers.
Uhh 2025? Or time machine confirmed?!?!?Clearwater Forest is sometime in 2015
ADL-N will get smacked by A78 cores very soon, if Intel delays it will be BayTrail all over again. But they dont care much as WOA is no where near to be a threat to their small cores, yet, and that will take at least one more year, or two.You aren't entirely wrong. Alder Lake-N isn't . . . a great product, but at least in theory, there should be fertile ground for more embedded/comm gear SoCs in the Ridge lineup as well as consumer SoCs they could release with 4c-8c (or more) based on Crestmont and Darkmont.
It's 2024 and there is no refresh of that using the newer Crestmont. It's like they think this part of the market is totally dumb and doesn't care about updated stuff. I'm sure the engineers who worked on Crestmont would really like the N series of processors to be updated so more people can benefit from their hard work but management is like, oh no way. We are fighting for our life here. Our server stuff is abysmal. Desktop CPUs are getting overvolted to the gills to win benchmarks. No resources for you for your little Crestmont cores. Go back into your corner and work on Skymont and they better come with higher cinememe scores or you'll be in trouble!ADL-N
Seems more like they don't have the budget or team available to do the minimal work required to get a product like this functioning. As I pointed out in an earlier post, it doesn't seem like they have any successors to their Ridge products (based on Tremont), despite those having an available niche market that should be easy money. If they DO have immediate successors to those products, they certainly flew under my radar.It's like they think this part of the market is totally dumb and doesn't care about updated stuff.
Someone posted Geekerwan video in MTL thread I repost it here,ADL-N will get smacked by A78 cores very soon, if Intel delays it will be BayTrail all over again. But they dont care much as WOA is no where near to be a threat to their small cores, yet, and that will take at least one more year, or two.
They do care about the low power server, so what they are doing it makes sense.
Uhhhh no they are are not really efficient.Someone posted Geekerwan video in MTL thread I repost it here,
Check around 9:50
Lpe Core match Apple E cores inside M3, there is no data on power consumption but judging battery life of MTL laptop when using Lpe cores only they are really efficient too.
At same time you can see A78 data here
D1200 mid core run around 2.6ghz almost same frequency as 2.5Ghz Lpe core, but perfomance is much lower.
Meanwhile E cores is better than Arm mid cores and can match or slightly surpass older X cores like X2 and behind X3.
So core design E and Lpe are good enough, it's their scheduler which lag behind as that video show.
10% different per clock is noticeable but not big. It could be also inferred from the fact that Golden Cove in Integer is around X2 in performance, and Raptor Cove is 25% faster per clock compared to the Gracemont variant in the same chip.D1200 mid core run around 2.6ghz almost same frequency as 2.5Ghz Lpe core, but perfomance is much lower.
Meanwhile E cores is better than Arm mid cores and can match or slightly surpass older X cores like X2 and behind X3.
So core design E and Lpe are good enough, it's their scheduler which lag behind as that video show.
Quoting myself to issue a correction: there IS an Atom-based low power server CPU based on Gracemont: Arizona Beach. Today was my first day of hearing about it. It's in the C-series of Xeon CPUs.Seems more like they don't have the budget or team available to do the minimal work required to get a product like this functioning. As I pointed out in an earlier post, it doesn't seem like they have any successors to their Ridge products (based on Tremont), despite those having an available niche market that should be easy money. If they DO have immediate successors to those products, they certainly flew under my radar.
155H, uncharged uncore 2.5-4W,charged uncore 5-6w,WHY？Any test is under charge or uncharge？Uhhhh no they are are not really efficient.
View attachment 96866
They’re terrible, but the key is they can shut off the compute tile intermittently and leave it to the SoC tile which is always on and only has 2 LPE cores. The compute tile due to Intel’s design has a lot of idle fabric overhead so then they can use the LP E Cores for a cursor movement or background tasks or until they realize some some performance bar — and they need to go back to the compute tile when needed.
Actual performance under these conditions can still suffer too, even if battery life for some non-fixed task is “good”.
MTL basically needs this intricate setup to sometimes match AMD’s Monolithic, homogenous Phoenix on battery life, because of how bad low-load overhead is on the compute tile, and how inefficient the actual cores are vs AMD.
And keep in mind those A7x and E Core measurements are for the whole platform minus the display and statics, above is package power for these Intel/AMD cores which is directionally in the right place, but also still underestimating it slightly due to DRAM and power delivery (also was probably software estimated but theirs is more accurate than Apple’s).