Discussion Intel current and future Lakes & Rapids thread

Page 134 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

beginner99

Diamond Member
Jun 2, 2009
5,208
1,580
136
but is it enough to move product?

Probably yes as most persons even in IT don't really know about the changes or lack thereof. especially in companies if you are entititled to a new laptop after 3 years for example. And often that is done because by then the battery will be pretty bad due to being mostly docked (100% charge) which is terrible for li-ion batteries. ideally IT would configure laptops such that they only charge up to 80% by default and only start charging again at 20% without user intervention. But that would require too much training...
 

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
Probably yes as most persons even in IT don't really know about the changes or lack thereof. especially in companies if you are entititled to a new laptop after 3 years for example. And often that is done because by then the battery will be pretty bad due to being mostly docked (100% charge) which is terrible for li-ion batteries. ideally IT would configure laptops such that they only charge up to 80% by default and only start charging again at 20% without user intervention. But that would require too much training...

Just replacing the batteries might make the corporate types happy with their "new" laptops. Otherwise there's so little to be gained in raw power or battery life from these 14nm refresh ad nauseam laptops . . . meh. You would think someone would notice eventually.
 

beginner99

Diamond Member
Jun 2, 2009
5,208
1,580
136
Just replacing the batteries might make the corporate types happy with their "new" laptops. Otherwise there's so little to be gained in raw power or battery life from these 14nm refresh ad nauseam laptops . . . meh. You would think someone would notice eventually.

Replacing the not user replacable battery in all these ultrabooks usually requires sending the laptop to the supplier. Yeah it would make sense that this is somehow uncluded in support contracts (we can only get HP stuff) but never seen it happen.
 

mikk

Diamond Member
May 15, 2012
4,112
2,108
136
Looks like some OEMs can do better than Intels Icelake dev platform.


1ygkqj.png


2bdjrv.png

https://medium.com/performance-at-intel/real-world-performance-ifa-without-compromise-9f2dff21c277
 

Bouowmx

Golden Member
Nov 13, 2016
1,138
550
146
Not sure why dev platform was sandbagged when it would be the first (and will be lasting) impression of Ice Lake.
 
  • Like
Reactions: Lodix

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
Not sure why dev platform was sandbagged when it would be the first (and will be lasting) impression of Ice Lake.

NV kind of does the same thing, no? Their Founder's Edition cards are often slower (and more expensive than) AIB cards that come out later. Granted, sometimes you have to wait for availability with the AIB cards. I don't think anyone will be able to buy a reference Icelake dev system.
 

Dave2150

Senior member
Jan 20, 2015
639
178
116
NV kind of does the same thing, no? Their Founder's Edition cards are often slower (and more expensive than) AIB cards that come out later. Granted, sometimes you have to wait for availability with the AIB cards. I don't think anyone will be able to buy a reference Icelake dev system.

Nope - the last few Nvidia generations have seen the founders cards have better binned chips and thus run faster than than most AIB cards. This has been exacerbated with the twin axial fans of the 2000 series, further improving the founders edition performance.
 

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
Nope - the last few Nvidia generations have seen the founders cards have better binned chips and thus run faster than than most AIB cards. This has been exacerbated with the twin axial fans of the 2000 series, further improving the founders edition performance.

So the 10x0 generation was the last one to have slow FE cards? I didn't even know.

Hmm not sure why Intel did what they did with the dev Icelake platform then. Must have been some reason for it.
 

yeshua

Member
Aug 7, 2019
166
134
86
Seems like Intel's troubles with 7nm EUV and 10nm nodes are far from over:

Jason Grebe

So we're really excited right now because our – frankly, our 10 nanometer yields are ahead of where we thought they would be. So we're extremely bullish on the node at this point and it's really just having us trying to finish out two CPUs in a calendar year.

Unidentified Analyst

That’s good news on 10 nanometer. And any confidence that second half 2020, we feel pretty good, line in the sand that it's coming out?

Jason Grebe

Yes. We feel really good about our second half guide there.

Unidentified Analyst

Sure. And then a lot has been talked about the issues with 10 nanometer. Why do you think the yields are ahead of plan? Is it just we finally figured it out? Or…

Jason Grebe

Yes. So the quick history on 10 nanometer was typically when we go for a scaling benefit on node to node, it's been a 1.5 to 2x type range. When we went from 22 to 14, we were – we did 2.42x scaling, which was very aggressive when we pulled it off. On the next generation from 14 to 10, we went for even more scaling from 14 to 10, and in order to do that, we had to do a bunch of quad patterning, a very complicated silicon development. That caused a lot of delays and a lot of the yield issues. We've overcome all those.

The yield curves that we've been watching every week coming out of the factory, both in our development factories and our HPM factories are right on track and actually slightly ahead. So that we've kind of gotten over the learnings that we have to get over and we're ramping the process now.

Unidentified Analyst

Did EUV or lack of EUV at 10 nanometer had anything to do with the issues?

Jason Grebe

Yes, I would say the delays in EUV had us move down a multi-patterning path, which caused some of the delays for sure and now we're moving to EUV when we get to 7 nanometer.

Unidentified Analyst

Yes. That was my next question. So how do we feel about EUV 7 nanometer, in the bag, stone cold pipe lock?

Jason Grebe

Yes. It's done. It’s basically done. No. I feel really good about 7 nanometer. We are on a path right now to have a two-year cadence from 10 to 7. Everything from a development perspective on a power and a performance and a timing cadence perspective on 7 nanometer look really good. So we're really bullish on a 2021 product launch on 7 nanometer.

Unidentified Analyst

By the way, we do have some time left and I'd like to open it up to the audience if anybody has any questions or else I'll keep going. Anybody out there? Going once, going twice. Just to dig in a little more on EUV, what were the issues with the – at 10 nanometer to the extent you can…

Jason Grebe

I think we just weren't ready from a learning curve perspective to put it in the process development or a timing perspective and we were going to delay till 7.

This might mean we won't see Ice Lake/Tiger Lake desktop parts before 2021 which is a bummer.
 

extide

Senior member
Nov 18, 2009
261
64
101
www.teraknor.net
Yeah I mean he says that the trouble is far from over and then quotes Intel saying that basically the trouble IS over and things are back on track now. ...
 

TheGiant

Senior member
Jun 12, 2017
748
353
106
well I cant buy icelake notebook ATM so it is not that hot with the 10nm status
Intel is doing what they can with 14nm more+ and bow to them what they can achieve but its getting scary
 

ondma

Platinum Member
Mar 18, 2018
2,718
1,278
136
The problem is, is the performance or battery life really improved. I know the integrated graphics if much better, but not sure how important that is. I have not yet seen whether 10 nm gives lower power consumption. If the battery life was clearly better, then 10 nm might be worth it.
 

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136

Review of the Dell XPS i3 model.
 
  • Like
Reactions: trivik12

TheGiant

Senior member
Jun 12, 2017
748
353
106

Review of the Dell XPS i3 model.
cooling is dissapointing

performance seems nice, the i7 will be ok, but with another build

it looks like my surface pro 4 will wait for ryzen 7nm mobile parts.....or tigerlake
 

mikk

Diamond Member
May 15, 2012
4,112
2,108
136
The cooling looks pretty good, it's more a bios issue. It should throttle if it exceeds a certain temperature threshold.

Prime 95 3.1-3.3 Ghz
Prime95+Furmark 1.6-2.4

These clock speeds are outstanding in prime95 which explains the temperature. On a 14nm ULV Quadcore for such a thin device the CPU clock speed would be roughly at 2 Ghz for Prime 95 and 1 Ghz for Prime95+Furmark. Btw 4GB RAM is a joke, especially for this price.
 

mikk

Diamond Member
May 15, 2012
4,112
2,108
136
And it's only a dual core


It should be better with a Quadcore because 4 cores would prevent this high clock due to power limits, this effect can be seen in the Furmark test where the CPU temps are much lower due to much lower clocks. And also such a device is not designed to run Prime95 nonstop, temps would be more meaningful in a realistic scenario, for example a gaming session.
 

mikk

Diamond Member
May 15, 2012
4,112
2,108
136
Intel revealed the slice and core specs for Tigerlake.

ITGL_LP_1x6x16_UNKNOWN_SKU_F0_ID_5 0x9A49 // Remove this once newer enums are merged in OpenCL. Added this to avoid build failure with Linux/OpenCL.
ITGL_LP_1x6x16_ULT_15W_DEVICE_F0_ID 0x9A49 // Mobile - U42 - 15W
ITGL_LP_1x6x16_ULX_5_2W_DEVICE_F0_ID 0x9A40 // Mobile - Y42 - 5.2W
ITGL_LP_1x6x16_ULT_12W_DEVICE_F0_ID 0x9A59 // Mobile - U42 - 12W
ITGL_LP_1x2x16_HALO_45W_DEVICE_F0_ID 0x9A60 // Halo - H81 - 45W
ITGL_LP_1x2x16_DESK_65W_DEVICE_F0_ID 0x9A68 // Desktop - S81 - 35W/65W/95W
ITGL_LP_1x2x16_HALO_WS_45W_DEVICE_F0_ID 0x9A70 // Mobile WS - H81 - 45W
ITGL_LP_1x2x16_DESK_WS_65W_DEVICE_F0_ID 0x9A78 // Desktop WS- S81 - 35W/65W/95W

16 EUs per subslice for TGL GT2. There is also this

Maximums which bound all supported GT
GT_MAX_SLICE (4)
GT_MAX_SUBSLICE_PER_SLICE (8)
GT_MAX_SUBSLICE_PER_DSS (2) // Currently max value based on Gen12
GT_MAX_DUALSUBSLICE_PER_SLICE (6) // Currently max value based on Gen12LP


dualsubslice per slice....all 6 subslices contain another mini dual subslice of 2x8?
 
  • Like
Reactions: Dayman1225

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
Wonder if the H/S parts mentioned is basically Rocket Lake or the original plan for Tigerlake on desktop, which is obviously canned.