Discussion Intel current and future Lakes & Rapids thread

Page 288 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

SAAA

Senior member
May 14, 2014
541
126
116
Ideally in 2020 Q4, but Intel punctuality means probably 2021 Q1 or Q2.

Rocket Lake will use Sunny/Willow Cove, already proven in Ice/Tiger Lake to have on average 1.18x IPC of Skylake. Rocket Lake has been seen to run 5.0 GHz (Geekbench), which suggests frequency won't decline too much from Comet Lake's.

I was going through Tom's Hardware GeForce RTX 3080 CPU scaling article to see how "IPC" relates to gaming performance:
Uqr5v4KLCp69twSmfuF78i.png
Comparing Core i3-10100 (4.1 GHz all-core) to i7-4770K OC (4.3 GHz), same config of 4 cores + HT, the i3-10100 gets 1.237x the frame rate of i7-4770K. Now, I don't remember Skylake having greater than 1.23x performance per clock than Haswell does.

Individual games:
Game​
i7-4770K 4.3 GHz​
i3-10100 4.1 GHz​
Ratio​
Borderlands 3​
120.0​
141.2​
1.177​
The Division 2​
122.0​
157.1​
1.288​
Far Cry 5​
95.0​
125.0​
1.316​
Final Fantasy XIV​
171.4​
187.9​
1.096​
Forza Horizon 4​
120.4​
157.4​
1.307​
Metro Exodus​
87.7​
109.1​
1.244​
Red Dead Redemption 2​
102.5​
118.5​
1.156​
Shadow of the Tomb Raider​
95.5​
134.6​
1.409​
Strange Brigade​
250.9​
310.7​
1.238​
That's the infamous Skylake's "gaming" IPC. It held so well for 5-6 years because of great gaming results improved by higher cpu and memory clocks over time.
The same happened for other niche softwares (3d cad) were it displays well over ~5% IPC that was measured on average vs Haswell.

As for rocket lake it will probably have slightly less IPC than Tiger or Ice lake if the rumors are true, still above Skylake and definitely with similar clockspeeds.

Now what I haven't found yet is purely gaming comparisons between Icelake and Skylake. I know there's only mobile CPUs out but some test with the same discrete GPU and memory could work. Even more interesting would be Tiger with the wider caches...
 

TheGiant

Senior member
Jun 12, 2017
748
353
106
My first research paper was a literature review.
mine too, then I got a question from my father "and now what?" and I didn't know what to say

The 'what is next' is the 10nm Enhanced SuperFin.
that is Intel's next step
what is anandtech next step? They got cannonlake, icelake and tigerlake laptop
maybe 10nm atom, tons of measurements

so the next is like
  • voltage/freq/power curve
  • predicted enhanced superfin/10nm+++/wtfFIN curve
  • variability...
    • my contact from the biggest retailer here told me, that icelake notebugs have liek 300MHz all core turbo variability within the same power and model
  • icelake/tigerlake overclocking/undervolting?
  • whatever
that is how to separate from the reposters of reposterd and collectors of collected IMO
 

Dave2150

Senior member
Jan 20, 2015
639
178
116
Any news on Alder Lake desktop release date? Any sign it will be delayed by DDR5 availability? I'm hoping it releases within 2021, as my 6700k is going to struggle a bit (even at 4k) when my 3090 arrives.
 

mikk

Diamond Member
May 15, 2012
4,111
2,105
136
Intel said H2 2021 and according to rumours the first ADL version will be ADL-S not ADL-P. Also ADL-S might come as a Hybrid which supports both DDR4 and DDR5. Actually I think they may have to release ADL-S with DDR4 support only followed by DDR5 in 2022 if the DDR5 ecosystem isn't ready similar to TGL-U LPDDR5. But it's too early to say.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
Any news on Alder Lake desktop release date? Any sign it will be delayed by DDR5 availability? I'm hoping it releases within 2021, as my 6700k is going to struggle a bit (even at 4k) when my 3090 arrives.

Even if it arrives by December of next year, they'll be pushing it hard, and there will likely be internal resistance from the OEMs. Systems need 1-year lifecycle otherwise its not optimal for revenue.

We don't even have Rocketlake and we're expecting Alderlake desktop in just a year from now? Rocketlake is actually expected in Q1 of next year.

I assume first Alderlake will be mobile and in Fall just like Tigerlake.
 

PlanetJosh

Golden Member
May 6, 2013
1,815
143
106
Android stopped with the alphabetic names and now it's just Android 11, so Intel can stop with the water names and call it Intel xx. That said I'm waiting for Rocket Lake and its pcie4.
 

coercitiv

Diamond Member
Jan 24, 2014
6,151
11,677
136
my contact from the biggest retailer here told me, that icelake notebugs have liek 300MHz all core turbo variability within the same power and model
It's already been discussed on the forums that ICL had serious variability issues based on performance scaling we saw going from i7 to i5 (sustained perf took a serious dive). I expect TGL to do a lot better in this regard, since the frequency delta going down from i7 to i5 and i3 respectively has increased considerably. If we ignore the (rare) top i7 bin of each gen, the i7- > i3 max turbo frequency delta for ICL was 500Mhz, in TGL it has increased to 1Ghz. The lower quality i5 and i3 have a much easier performance target this time. (realtive to top bin)
 

borandi

Member
Feb 27, 2011
138
117
116
so the next is like
  • voltage/freq/power curve

A test for when we get frequency control over the CPU. Sub-binning notwithstanding (a detail we've noted in smartphone reviews).
Not relevant to a discussion about process node naming

  • predicted enhanced superfin/10nm+++/wtfFIN curve

This doesn't make sense. List filler.

  • variability...
    • my contact from the biggest retailer here told me, that icelake notebugs have liek 300MHz all core turbo variability within the same power and model

Hard to test without multiple versions of the same hardware, which is something we never get.
But also, remember that Intel doesn't guarantee any turbo frequency. Base frequency at TDP is all they guarantee. So variability on turbo is just luck.
OEMs deciding to put a wider range of bins in a singular product is something they would have built for. More an OEM problem than an Intel problem.
Certainly not relevant to a topic talking about node naming.

  • icelake/tigerlake overclocking/undervolting?

See point 1.



More list filler.



Seriously though, you're talking about testing. The article was purely about node naming. Try to keep it on-topic perhaps.
The Tiger Lake review article was here:
21000 words for you to read.

that is how to separate from the reposters of reposterd and collectors of collected IMO

I think my brain had a stronk reading this
 
  • Wow
Reactions: Grazick

piokos

Senior member
Nov 2, 2018
554
206
86
Android stopped with the alphabetic names and now it's just Android 11, so Intel can stop with the water names and call it Intel xx. That said I'm waiting for Rocket Lake and its pcie4.
Android is a single software product and naming releases by numbers is how almost everyone does it.

Intel makes a lot of stuff so there are multiple "Lakes", "Peaks", "Points", "Falls", "Trails" and "Creeks" circulating (and more). But it's actually quite elegant and introduces natural categorization.

Are you against geographical naming as such? Why?

My previous employer named his internal IT systems with Pokemon names. At the end of the day it's just a string that should be memorable and easy to use in a sentence.
 
Last edited:
  • Like
Reactions: pcp7

TheGiant

Senior member
Jun 12, 2017
748
353
106
A test for when we get frequency control over the CPU. Sub-binning notwithstanding (a detail we've noted in smartphone reviews).
Not relevant to a discussion about process node naming



This doesn't make sense. List filler.



Hard to test without multiple versions of the same hardware, which is something we never get.
But also, remember that Intel doesn't guarantee any turbo frequency. Base frequency at TDP is all they guarantee. So variability on turbo is just luck.
OEMs deciding to put a wider range of bins in a singular product is something they would have built for. More an OEM problem than an Intel problem.
Certainly not relevant to a topic talking about node naming.



See point 1.




More list filler.



Seriously though, you're talking about testing. The article was purely about node naming. Try to keep it on-topic perhaps.
The Tiger Lake review article was here:
21000 words for you to read.



I think my brain had a stronk reading this
this entire thread is about ....lakes...
parameters/power/impact/performance
manufacturing process can be measured via products, there is no other method
and yes my brain got hurt too, no content in your post
no analysis, link, proposal, just no mr wrong
teh win
 

ondma

Platinum Member
Mar 18, 2018
2,718
1,278
136
Personally, I think they should give Tiger Lake 2xxxxx series numbers to better differentiate it from Ice Lake/Comet Lake. Very confusing to have 3 generations with 1xxxxxx series numbers. And these are not just +++ "generations", but really are quite different.
 

mikk

Diamond Member
May 15, 2012
4,111
2,105
136
I assume first Alderlake will be mobile and in Fall just like Tigerlake.


According to Raichu ADL-S comes earlier than ADL-P. There are several other indications, the first Alderlake benchmark leak was based on ADL-S (on Sisoft), the first ADL version on Intels technical database was ADL-S based - ADL-P came later. First leaked socket/TDP/core count info was ADL-S based (leaked from sharkbay) - ADL-P came later. In the big DG2 driver inf leak from last year there was ADL-S but no other ADL version. On github ADL-S and ADL-P both are listed (currently A0 stepping), so there is no sign that mobile has priority over desktop this time. They may have a similar schedule but even in this case ADL-S is expected to come earlier than mobile, the time to market is much shorter for a standalone CPU. There are reasons why in the old Sandy Bridge, Haswell, Skylake days desktop versions were available before mobile and same for AMD, it takes more time for a notebook version.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
@Zucker2k Awesome.

10th Gen ICL U/10th Gen U 2C/10th Gen Y 4C/11th Gen U UP3/11th Gen U 2C UP3

SoC IccMax: 70A/55A/49A / 65A/35A

VccIn_Aux: 32A/32A/32A / 27A
Memory VDDQ/VDD2: 3.5A U/3A U for LPDDR4x, 3.5A U for DDR4 / 1.5A/1.5A for LPDDR4x

Vcc Sustain: 800mA U, 300mA Y / 500mA U
Vcc Sustain Gated: 150mA U, 60mA Y / 300mA U

First laptop with 11th Gen Core Mobile: XPS 13 and XPS 13 2-in-1


September 30th.
 
Last edited:

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
There are reasons why in the old Sandy Bridge, Haswell, Skylake days desktop versions were available before mobile and same for AMD, it takes more time for a notebook version.

This is true if technical considerations and what Intel wants are the only things that matter.

OEM considerations are just as important. I can believe the -P(likely a successor to H) coming a bit later, but how much of a difference?

Always expect later. Earlier is a rare exception. Heck, being on time should be celebrated! These are super complicated things to make. If they want to hype it up like Tigerlake, we could see a launch in January 2022, but do a triple paper launch with CPU availability in March.
 

piokos

Senior member
Nov 2, 2018
554
206
86
I have to say... the rate of Tiger Lake product announcements is impressive even by Intel standards. Good job.
And it finally seems like NUCs will be positioned as mainstream PCs, not some niche industrial (or home server) solution.

Funny how new XPS generation comes so quickly after the last one - often called the best ever - which is probably still selling pretty well. A lot of people will wish they waited for TB4 and the extra oomph.
 

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
I'm still wondering a bit at TigerLake-H coming out in Q2 2021. Why is it going to be so late? How does that position it in Intel's product lineup? What does that tell us about Alder Lake-P?
 

cortexa99

Senior member
Jul 2, 2018
318
505
136
I'm still wondering a bit at TigerLake-H coming out in Q2 2021. Why is it going to be so late? How does that position it in Intel's product lineup? What does that tell us about Alder Lake-P?
TGL and ADL is not same architecture, but yeah I have similar feeling that Intel's schedule/naming is weird by now......Seems intel decide to have seperate development for both desktop&mobile, but it's getting more&more puzzled to understand the difference(at least for me). This may be a result of 10nm's yield problem and being so late to the market.
 

uzzi38

Platinum Member
Oct 16, 2019
2,565
5,571
146
I'm still wondering a bit at TigerLake-H coming out in Q2 2021. Why is it going to be so late? How does that position it in Intel's product lineup? What does that tell us about Alder Lake-P?
That it's later.