• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."

Discussion Intel current and future Lakes & Rapids thread

Page 223 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

uzzi38

Golden Member
Oct 16, 2019
1,042
1,523
96
From what I hear, the situation is the exact opposite. The Atom team is very, very proud of Gracemont. My source has been hyping it up without much subtlety, though I've been unable to weasel out numbers. I've also gotten the impression that they feel they've had to fight tooth and nail to get the recognition they deserve, and Gracemont is their opportunity to flaunt their potential a little. Supposedly Keller himself is responsible for pushing the Atom team into the limelight, partly due to some frustration with the Core team.

More practically speaking, why would they be using Gracemont at all if it were so bad? Hybrid is certainly a non-trivial amount of work to support.
See that's what gets me. As I noted above, Gracemont seems like a seriously impressive product to me, actually it fixes the one thing I'd have thought to be a major limitation with Atom so far in the lack of AVX support. That's one of the reasons I called my original post weird and why I was bamboozled, however decided to post it anyway because I know some people here have a lot of faith in what Sharkbay posts.

I can't think of a reason in the slightest to view Gracemont negatively myself, which is why that latest leak made no sense to me. Then on top of that there was the insinuations of MCM and so on... yeah, thanks for that confirmation on Gracemont at least.

Probably best to just disregard this one.
 

IntelUser2000

Elite Member
Oct 14, 2003
7,092
1,622
136
Wouldn't be surprised if there was some resistance from C2DG towards Atom.
Seems almost like evidence for articles saying Intel is divided because they pit their teams against each other

Competition works for short term, but I believe harmony along with true inspiration and hard work is better. Competition not only creates redundancy and waste but causes social issues as the idea becomes #1 or nothing and leaves everyone/everything not number 1 as trash when oftentimes there are useful knowledge to be shared even if its not the best implementation.

@uzzi38 For me I take credit in Sharkbay's information not his opinions. Intel's Lakefield presentation is already an indicator that their hybrid implementation isn't just slapping Gracemont cores and say "we're done". They're doing WAY more work to make sure they work asynchronously. This also is a strong indicator why putting the two core types in a separate chiplet would invalidate that because there are things you can't do unless its on-die.
 
Last edited:

Exist50

Member
Aug 18, 2016
115
147
116
Seems almost like evidence for articles saying Intel is divided because they pit their teams against each other
I swear, if even half the stories I've heard are true, it would absolutely explain some of Intel's stagnation despite their resources. Teams that outright throw out their counterparts' work because they either don't care about it or don't trust it, platform turmoil over different priorities (FIVR was given as an explicit example), rampant NIH syndrome (where "Here" means their group, not Intel), and even some project casualties along the way. Apparently at one point they were just going to split it outright, with C2DG handling the high power market and DDG the low power one, but it's been a while since I've heard anything about that.

Anyway, back to the topic of Alder Lake.

This also is a strong indicator why putting the two core types in a separate chiplet would invalidate that because there are things you can't do unless its on-die.
I highly, highly doubt that Alder Lake will have an 8c Atom-only chiplet for cost reasons alone. That said, I could theoretically see them splitting the CPU portion into two chiplets, along Core/Atom lines or symmetrically, and I see no fundamental problem with that. Sure they'd have to change up their cache structure a bit (maybe forbid allocation between chiplets?), but Ryzen works fine with significantly worse latency than Intel would have with Foveros or EMIB.

Edit:

Come to think of it, having separate Core/Atom chiplets could enable some cool possibilities. They could use one or two 16, 24, or even 32 core Atom chiplets for low-end, throughput-oriented servers (storage, maybe networking?), pair a big core chiplet and an Atom chiplet for mobile, and pair two big core chiplets for certain desktop segments (with full ISA compatibility). Could be really neat.
 

Cardyak

Member
Sep 12, 2018
30
28
61
Given the rate of improvement in the Atom cores, and the rumoured addition of AVX in Gracemont, I wouldn't be surprised if future Lakefield-esque products discard the "Large" core and simply have more atom cores.

A future Lakefield like product could potentially consist of 8 Gracemont cores, as opposed to 4 Gracemont cores and 1 Golden Cove core.
 

DisEnchantment

Senior member
Mar 3, 2017
615
1,258
106
Given the rate of improvement in the Atom cores, and the rumoured addition of AVX in Gracemont, I wouldn't be surprised if future Lakefield-esque products discard the "Large" core and simply have more atom cores.
Banias encore? Sounds like familiar circumstances :nomouth:
 

jpiniero

Diamond Member
Oct 1, 2010
8,280
1,355
126
Given the rate of improvement in the Atom cores, and the rumoured addition of AVX in Gracemont, I wouldn't be surprised if future Lakefield-esque products discard the "Large" core and simply have more atom cores.

A future Lakefield like product could potentially consist of 8 Gracemont cores, as opposed to 4 Gracemont cores and 1 Golden Cove core.
ISA compatibility is one thing... performance is another. You'd still want the big core for single thread workloads.
 

coercitiv

Diamond Member
Jan 24, 2014
3,855
3,822
136
Yeah, let's go full Atom, especially considering Apple already started their transition towards small cores. /s This new found faith in the power of Atom is weird to say the least.

A quick recap on performance numbers from Intel:

lkf.jpg

So LKF 1+4
  • 100% ST perf relative to SNC core
  • 300% MT perf relative to SNC core /w SMT if running 6 threads
Hypothetical LKF 2+0
  • 100% ST perf
  • 200% MT perf if running 4 threads
Hypothetical LKF 0+8
  • almost 70% ST performance
  • 400% MT perf if running 8 threads
  • 300% MT perf if running 6 threads
  • 200% MT perf if running 4 threads
This is what 8 Atom cores deliver: a 30% drop in ST performance relative to LKF 1+4 in exchange for the possibility of 30-40% higher MT performance if running more than 6 threads.

Good luck running 8 threads in consumer workloads often enough to compensate for the 30% loss in ST performance.
 
Last edited:

lobz

Golden Member
Feb 10, 2017
1,345
1,502
106
Label it as speculation the next time and not as a fact, otherwise it's more a troll posting. You have nothing beside some comical 3dmark leaks, everyone with some knowledge should be aware that the current leaks can't be representative. You didn't taken into account the memory which is crucial for a fast iGPU,. Even worse you don't even know what DDR4 memory speed the 1185G7 was running with, this is a joke. Comparing premium Icelake devices with faster LPDDR4x and possibly higher TDP, your postings are really low lately. Tigerlake supports LPDDR4x-4266! One more thing for you: You might have noticed that the newer Renoir-U scores are quite a bit better than the leaks we got some weeks or months ago. Performance can vary from device to device and one or two 3dmark entries without system details can't be representative, the sample base is way too low. As I said everyone with some knowledge is aware of this.



If your observation is based on 3dmark and nothing more which seems the case then I have more than you.
Wow, you're not even talking to me, yet you've earned a full fat blocking here. This full blown pompous juanrga style of arguing can only achieve 2 things: either take away your recipient's every intention to even write here again or stir up flaming. Both of them are toxic to a forum.
 
  • Love
Reactions: spursindonesia

DrMrLordX

Lifer
Apr 27, 2000
16,365
5,273
136
It's not a question of how much better Gracemont gets vs Tremont, but a question of how Gracemont stands vs. Golden Cove.
I guess? Allegedly IPC uplift for Tremont over Goldmont+ was 30%. If intel can pull that off again for Gracemont, it might not be that bad.

A J5005 @ 2.8 GHz gets somewhere around 82 in CBR15 ST. Tremont should be ~106 and maybe Gracemont would score ~138? Assuming 2.8 GHz. An R5 1400 @ 3.4 GHz scores 139:


(according to one source)

That's all speculation though! I don't know if the Atom team can pull another IPC increase like that, or even if Tremont is as good as folks say it is . . .
 

ondma

Golden Member
Mar 18, 2018
1,469
406
106

uzzi38

Golden Member
Oct 16, 2019
1,042
1,523
96
Intel has to adopt to Atom-derivates:


SNC performance/watt is ridiculously low. 25W Icelake has trouble of keeping up with 15W Picasso in some MT workloads. They have to adopt big.little approach because their big cores are so awful.
If what has leaked of Tiger Lake so far - for example the 1185G7 being capable of maintaining 3GHz all-core base clock in 15W as per Intel spec for base clock - is accurate, then I don't believe this will be an issue at all for Tiger Lake.
 
  • Like
Reactions: french toast

naukkis

Senior member
Jun 5, 2002
353
183
116
It is more like the 10 nm process is bad. Very little if any power savings from 14 nm.
Icelake idle power consumption is fine, as is iGPU performance. And problem isn't little power savings from 14nm but enermously worse power efficiency on cpu core load - even AMD Picasso is better.

And if problem is process nothing can save it. But if problem is cpu cores being power hogs cure is easy - replace those with ones with better efficiency. Isn't that just what Intel is doing?
 

ksec

Senior member
Mar 5, 2010
420
116
116
I am going to ask here instead of opening another thread.

What is the current status of Intel 7nm CPU ?

Intel promised in one of their easier investor meeting that 7nm will arrive in 2021. While we can all sort of guess that means "late" 2021, I try looking up and it dont see any lake or cove launches that arrives in that timeframe.

Do we have details on ANY Intel 7nm CPU?
 

Hitman928

Platinum Member
Apr 15, 2012
2,765
2,087
136
I am going to ask here instead of opening another thread.

What is the current status of Intel 7nm CPU ?

Intel promised in one of their easier investor meeting that 7nm will arrive in 2021. While we can all sort of guess that means "late" 2021, I try looking up and it dont see any lake or cove launches that arrives in that timeframe.

Do we have details on ANY Intel 7nm CPU?
From what I know, 7nm in 2021 is their enterprise GPU which they are under a lot of pressure to deliver by the end of the year because there is a supercomputer that needs it delivered by then in order for the builders to meet their own schedule. More 7nm products are supposed to come at some point in 2022 but it is unknown when, what volumes, and the scope of products that will launch. There's a lot of rumors out there about timelines and such but nothing firm outside of the 7 nm enterprise GPU at the end of 2021.
 
  • Love
Reactions: spursindonesia

jpiniero

Diamond Member
Oct 1, 2010
8,280
1,355
126
I am going to ask here instead of opening another thread.

What is the current status of Intel 7nm CPU ?

Intel promised in one of their easier investor meeting that 7nm will arrive in 2021. While we can all sort of guess that means "late" 2021, I try looking up and it dont see any lake or cove launches that arrives in that timeframe.

Do we have details on ANY Intel 7nm CPU?
Late 2021 is just Ponte Vicchio IIRC. Granite Rapids is the first CPU but that's more like early 2022.
 

dmens

Golden Member
Mar 18, 2005
1,966
285
126
From what I know, 7nm in 2021 is their enterprise GPU which they are under a lot of pressure to deliver by the end of the year because there is a supercomputer that needs it delivered by then in order for the builders to meet their own schedule. More 7nm products are supposed to come at some point in 2022 but it is unknown when, what volumes, and the scope of products that will launch. There's a lot of rumors out there about timelines and such but nothing firm outside of the 7 nm enterprise GPU at the end of 2021.

Nah Intel isn’t going to use their own busted process to satisfy a must-deliver contract, they will use something that actually works.

 

ksec

Senior member
Mar 5, 2010
420
116
116
Ok, that means there is some possibility that intel repeat their mistake again prioritising 7mn for other products or contract and completely missed capacity planning for others.

The Apple ARM transition is increasing likely not because of anything, but Intel simply cant deliver, and not willing to compromise with price.
 

Hitman928

Platinum Member
Apr 15, 2012
2,765
2,087
136
Nah Intel isn’t going to use their own busted process to satisfy a must-deliver contract, they will use something that actually works.

I remember seeing this rumor, I have a hard time believing it but if it is true, that says that Intel's problems will only be getting worse over the next few years.
 

Ajay

Diamond Member
Jan 8, 2001
7,257
2,484
136
I am going to ask here instead of opening another thread.

What is the current status of Intel 7nm CPU ?

Intel promised in one of their easier investor meeting that 7nm will arrive in 2021. While we can all sort of guess that means "late" 2021, I try looking up and it dont see any lake or cove launches that arrives in that timeframe.

Do we have details on ANY Intel 7nm CPU?

Nope. Well, aside from Intel saying it's on track; which is worth the cost of the bits required to display this post.
 

dmens

Golden Member
Mar 18, 2005
1,966
285
126
I remember seeing this rumor, I have a hard time believing it but if it is true, that says that Intel's problems will only be getting worse over the next few years.
Thankfully I don't work at that clown show any more so I am totally unburdened to comment on publicly available rumors. But it's true, bank on it.
 

uzzi38

Golden Member
Oct 16, 2019
1,042
1,523
96
You might have noticed that the newer Renoir-U scores are quite a bit better than the leaks we got some weeks or months ago.
Fun fact. Renoir's iGPU scores roughly the same whether it has, say, a 35W cTDP and DDR4-3200, or a 25W cTDP and LPDDR4X-4266. The difference in scores is within margin of error.

I don't really think the configuration is limited by either of those, if anything the 8CUs@1850mhz is the real bottleneck for Renoir.

I'm fairly confident the Renoir numbers you see now are the best they'll ever get.

So then... what about Tiger Lake huh? How much of an uplift do you expect from the best score using LPDDR4X-4266 that comes in at just under 7% higher over these scores, and what do you think will cause such a drastic uplift?

DDR4 Renoir + 35W cTDP: https://www.3dmark.com/spy/11758228

LPDDR4X Renoir + 25W cTDP:
 
Last edited:

ASK THE COMMUNITY