Discussion Intel current and future Lakes & Rapids thread

Page 663 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

igor_kavinski

Diamond Member
Jul 27, 2020
4,710
2,964
106
Anything is possible between now and MTL's launch. Who knows? Maybe Intel switches the compute tile from N5 to N3 (miracles happen) and you get two tiles. If not, MTL may not be the high end performance part. That might end up being Arrow Lake.
 

Timmah!

Golden Member
Jul 24, 2010
1,043
272
136
Anything is possible between now and MTL's launch. Who knows? Maybe Intel switches the compute tile from N5 to N3 (miracles happen) and you get two tiles. If not, MTL may not be the high end performance part. That might end up being Arrow Lake.
So you think they would release a generation with a significantly weaker performing top product than the previous generation´s top product, or AMD competition?
I find it unlikely.
 

igor_kavinski

Diamond Member
Jul 27, 2020
4,710
2,964
106
They might release a weaker product just to gain experience with this new way of doing stuff. Then they come back in full force with Arrow Lake. All speculation, of course. By the way, they've had weaker products before with Comet Lake and Rocket Lake.
 

JasonLD

Senior member
Aug 22, 2017
479
440
136
Arrow Lake as a product I can't see it releasing earlier than Q3 2024, that will leave Intel witnout any highend part release for 2 years. I think Meteor Lake will get higher end performance part. Last year's slide did say Meteor Lake SKUs up to 125W.
 

shady28

Platinum Member
Apr 11, 2004
2,520
396
126
So you think they would release a generation with a significantly weaker performing top product than the previous generation´s top product, or AMD competition?
I find it unlikely.
I think they are going to lose something, going from a monolithic single die to tiles. In the end the tiles will probably perform well, but I imagine Intel has some things to learn in that area. Keep in mind it took AMD four generations of tiles to beat monolithic Skylake, and much of the final win can be attributed to 7nm vs 14nm.

Raptor Lake may actually be the last and best of the monolithic dies, and it's likely going to be better at some things than any tile based design will ever be.
 
  • Like
Reactions: igor_kavinski

Exist50

Golden Member
Aug 18, 2016
1,134
1,134
136
No they claim 1H24... and that's "manufacturing ready", not production shipments. And that's before the inevitable delays.
Ah, typo, thanks. Yeah, just meant that it's almost certainly Q2, not Q1, so a March launch for 20A anything is unlikely.

Good point I forgot about that. I don't really care about monster iGPUs (or tGPUs in this case) in the desktop space, but it is a big deal for many.
Arrow Lake-H was also rumored to use N3 for a compute tile, but we haven't heard anything about that in a while.

Well, that was my initial question. And IntelUser2000 claimed more compute tiles are not happening in case of MTL.

Something does not compute here :)
There's definitely no way we're seeing more than one compute tile per package with Meteor Lake or Arrow Lake.
 

Timmah!

Golden Member
Jul 24, 2010
1,043
272
136
There's definitely no way we're seeing more than one compute tile per package with Meteor Lake or Arrow Lake.
You may be right, but then again, unless you have some insider info, you dont know that for sure. Anyway, if what you claim is true, and there is no bigger tile design than the 6+8 they showcased, after getting murdered in server space, they will stop being competitive even in desktop. Imagine AMD released Zen4 refresh with that third chiplet, thus 24 cores - and Intel would decide to compete with that with 6+8 chip....

If they dont intend to do actual "tiling" with those tiles, i dont see what is the point of it at all.
 
  • Like
Reactions: igor_kavinski

dullard

Elite Member
May 21, 2001
24,082
2,319
126
You may be right, but then again, unless you have some insider info, you dont know that for sure. Anyway, if what you claim is true, and there is no bigger tile design than the 6+8 they showcased, after getting murdered in server space, they will stop being competitive even in desktop. Imagine AMD released Zen4 refresh with that third chiplet, thus 24 cores - and Intel would decide to compete with that with 6+8 chip....If they dont intend to do actual "tiling" with those tiles, i dont see what is the point of it at all.
Intel is simultaneously (A) discussing lots of information about Meteor Lake and (B) holding the important Meteor Lake cards close and undisclosed. They are trying to drum up excitement for their technology (Intel 4, Foveros, etc) while not revealing too much for competition reasons. More is scheduled to be announced at Hot Chips 34 on Aug 23, 2022: https://hotchips.org/advance-program/ That hopefully will clarify a lot about Meteor Lake.

First people claimed Meteor Lake would be mobile only with a 2 + 8 tile (even though Intel kept saying 125 W chips would be made and listed desktop power supplies for meteor lake). Then the 6 + 8 tile was revealed. Unless you have inside information, no one knows how many other compute tiles Intel has (if any). For example, Moore's Law Is Dead claims an unconfirmed 8 + 16 tile is in the works for Meteor Lake. I think that 8 + 16 tile would make sense as a backup in case Arrow Lake is delayed, but might not see the light of day if Arrow Lake is doing well. That is the point of tiling, they can much more easily swap in a new tile if needed.

Now some people are claiming that there are additional cores in the large SOC tile on Meteor Lake. There is a hint at that if you read between the lines on a presentation, but that really is stretching the imagination until we get more concrete info.

My reading of the tea leaves is that for desktop, Raptor Lake will be a bit better than many people expect and Meteor Lake will be worse that many people expect. I think Arrow Lake is the one to really wait for. Although I think it will likely be a bit later than @IntelUser2000 estimate of March 2024. Maybe June 2024?
 
  • Like
Reactions: igor_kavinski

pakotlar

Senior member
Aug 22, 2003
718
175
116
Intel is simultaneously (A) discussing lots of information about Meteor Lake and (B) holding the important Meteor Lake cards close and undisclosed. They are trying to drum up excitement for their technology (Intel 4, Foveros, etc) while not revealing too much for competition reasons. More is scheduled to be announced at Hot Chips 34 on Aug 23, 2022: https://hotchips.org/advance-program/ That hopefully will clarify a lot about Meteor Lake.

First people claimed Meteor Lake would be mobile only with a 2 + 8 tile (even though Intel kept saying 125 W chips would be made and listed desktop power supplies for meteor lake). Then the 6 + 8 tile was revealed. Unless you have inside information, no one knows how many other compute tiles Intel has (if any). For example, Moore's Law Is Dead claims an unconfirmed 8 + 16 tile is in the works for Meteor Lake. I think that 8 + 16 tile would make sense as a backup in case Arrow Lake is delayed, but might not see the light of day if Arrow Lake is doing well. That is the point of tiling, they can much more easily swap in a new tile if needed.

Now some people are claiming that there are additional cores in the large SOC tile on Meteor Lake. There is a hint at that if you read between the lines on a presentation, but that really is stretching the imagination until we get more concrete info.

My reading of the tea leaves is that for desktop, Raptor Lake will be a bit better than many people expect and Meteor Lake will be worse that many people expect. I think Arrow Lake is the one to really wait for. Although I think it will likely be a bit later than @IntelUser2000 estimate of March 2024. Maybe June 2024?
It sure looks like Raptor Lake is going to be about 30% better in perf/watt, and more like 50% for PL4 workloads (burst), compared to Alder Lake. At the same time AMD will have only a tiny perf/w advantage, with 230W PPT, compared to 253W PL2 for Raptor Lake, with Raptor Lake being a bit faster. AMD seems to have made a great processor, but not to Zen 3’s level, unfortunately. Had they given us 24-32 cores, it would have killed, as it stands, a bit underwhelming for those who don't need AVX-512. Going on rumor anyway. I suspect its going to be quite a good desktop part.

Meteor Lake and beyond seems too speculative at this point, and maybe more focused on power efficiency for mobile.
 
Last edited:

JasonLD

Senior member
Aug 22, 2017
479
440
136
Although I think it will likely be a bit later than @IntelUser2000 estimate of March 2024. Maybe June 2024?
That sounds still too early for something that is only being manufacturing ready in H1 2024 (talking about 20A). Earliest date I can see is close to the end of 2024.
 

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
23,359
12,281
136
It sure looks like Raptor Lake is going to be about 30% better in perf/watt, and more like 50% for PL4 workloads (burst), compared to Alder Lake. At the same time AMD will have only a tiny perf/w advantage, with 230W PPT, compared to 253W PL2 for Raptor Lake, with Raptor Lake being a bit faster. AMD seems to have made a great processor, but not to Zen 3’s level, unfortunately. Had they given us 24-32 cores, it would have killed, as it stands, a bit underwhelming for those who don't need AVX-512. Going on rumor anyway. I suspect its going to be quite a good desktop part.

Meteor Lake and beyond seems too speculative at this point, and maybe more focused on power efficiency for mobile.
I think 230 is the socket max. I thought that the biggest Zen4 chip, the 7950x was 170 watt.
 
  • Like
Reactions: Drazick

pakotlar

Senior member
Aug 22, 2003
718
175
116
I think 230 is the socket max. I thought that the biggest Zen4 chip, the 7950x was 170 watt.
170W is the TDP, 230W is PPT, which is socket max (core + uncore); PL2 is the corresponding value on Intel CPUs, although not as hard a limit (see https://www.anandtech.com/show/17047/the-intel-12th-gen-core-i912900k-review-hybrid-performance-brings-hybrid-complexity/4). Ryzen 5950x will pin 142W under load, just as 12900K will pin ~225-253W. Raptor Lake PL2 (~PPT) will be only slightly larger than 7950X PPT.

edit: it looks like AMD processors can exceed TDP with core-only power draw, but under PPT. https://www.anandtech.com/show/14605/the-and-ryzen-3700x-3900x-review-raising-the-bar/19
 
Last edited:

DrMrLordX

Lifer
Apr 27, 2000
19,941
8,910
136
First people claimed Meteor Lake would be mobile only with a 2 + 8 tile (even though Intel kept saying 125 W chips would be made and listed desktop power supplies for meteor lake). Then the 6 + 8 tile was revealed. Unless you have inside information, no one knows how many other compute tiles Intel has (if any). For example, Moore's Law Is Dead claims an unconfirmed 8 + 16 tile is in the works for Meteor Lake. I think that 8 + 16 tile would make sense as a backup in case Arrow Lake is delayed, but might not see the light of day if Arrow Lake is doing well. That is the point of tiling, they can much more easily swap in a new tile if needed.
Couldn't Intel use two 6+8 compute tiles on a single package for desktop?
 
  • Like
Reactions: igor_kavinski

Timmah!

Golden Member
Jul 24, 2010
1,043
272
136
Couldn't Intel use two 6+8 compute tiles on a single package for desktop?
IMO it would be a logical choice and such product (with 12P + 16E) could be competitive with potential AMD 24 core Ryzen. 8 + 16 will not, as it seems it will more or less match 16 core 7950x im MT performance. And 6+8 would be obviously significantly worse.
 

dullard

Elite Member
May 21, 2001
24,082
2,319
126
Couldn't Intel use two 6+8 compute tiles on a single package for desktop?
It is possible, yes. Advisable? I don't think so. Three main problems. (1) Intel is already probably stretched thin on their Intel 4 production capability. Doubling up on the CPU tile might not be possible due to that limitation. (2) There will be a lot of added latency between the tiles. That latency is possible to partly overcome but I don't think that is what Intel wants to see with this chip. (3) More P cores isn't advisable. If you had 28 cores in 125 W, then each core gets 4.5 W of power (and that assumes absolutely no power goes to the iGPU or the uncore). Maybe Intel could shift it so the P cores get ~6 W and the E cores get ~3 W, but even 6 W each would be a stretch. And at 6 W each, the E core is tied with or outperforms the P cores. So, why add more P cores that will perform worse?

1656525363892.png
1656525379103.png
 
  • Like
Reactions: Vattila

ASK THE COMMUNITY