Intel "Ice Lake" (10 nm)

with slide above i think more reasonable to illustrate intel 10nm cpu area scaling.

anandtech made manual counting for 10nm cannon lake(intel will launch laptop first) die size around 70.5mm2. refer to dual core broadwell with 15.85 Mtr/mm2(82mm2 and 1.3billion transistors), intel 10nm will have 36.86mm2( 70.5mm2 and 2.59billion transistors).

with almost 2 times more transistor i think no doubt 4-core and graphics improvement still a question how good intel gen10.

 

Was included in HWInfo update

 

"Intel kills 10nm"

https://semiaccurate.com/2018/10/22/intel-kills-off-the-10nm-process/

Does this mean just Cannon Lake was scrapped and they're bringing forward Ice Lake on 10nm+, or does this have negative consequences for Ice Lake too?

 

There was an update to the article.

So, it depends on whether you believe SA or Intel.

Regards,
SB

 

They could both be true: Intel might just relax 10nm constraints and still call it 10nm.

 

Or get this node finally working, then rename the optimized '10nm+' node to '7 nm'.

 

Or rename "10nm+" to "10nm" and simply claim future Cannon Lake SKUs are cancelled to give way to Ice Lake.

 

To avoid lawsuits (or to simply be on the safe side) prolly Intel has to keep the name of their first ("7nm class") 10nm process "10 nm" . Same for the Cannonlake name. Internally, these products may be of course changed to something completly different

It would avoid/simplify some discussions/explanations to shareholders. And why not, to their customers as well.
It will thus only be a question of launch date, availability and of course performance

 

Cannlonlake won't be changed to something else, I don't think that would make sense. I mean, after all it's already in "mass production" and shipping - it even works, sort of... So pretty sure it will remain what it is, and there won't be any other Cannonlake SKUs, it has fulfilled its purpose.

 

I agree, it won't.

I was mearly mentioning it due to the post above, which suggested it's outright cancellation.

Anyway CannonLake could be sorta exiting due to AVX-512 for the masses (less ISA fragmentation). With some decent clockspeed, it's welcome

LE : Oh you said no other SKUs, sry. I thought there would be some more

 

https://www.anandtech.com/show/13699/intel-architecture-day-2018-core-future-hybrid-x86

 

Nice, adaptive sync looks like it's finally coming to Intel.

That means no more need to be limited to G-sync with an NV GPU if you have an Intel CPU that supports Adaptive Sync. Wonder if NV will finally pull the stick out of its backside and support it?

Regards,
SB

 

More likely nvidia will disable your discrete GPU if you try to use it like they did for using a nvidia card for physx with an AMD GPU.

 

So Cannon Lake is gone, together with its specific 10nm implementation.

Looks like semiaccurate was semi-accurate.

 

Yep. Ice Lake is supposed to use 10nm+ iteration, if you consider Intel roadmaps from last year:
https://www.anandtech.com/show/11722/intel-reveals-ice-lake-core-architecture-10nm-plus
https://www.guru3d.com/news-story/intel-talks-10nm-process-ice-lake.html
etc.
Since SemiAccurate didn't say anything about 10nm+ and 10nm++, their news were semi-accurate

 

At this point I don't get why Intel does not match the CPU name and the number of core.
They can't figth AMD 4 cores with 2 cores aymore.

Having three cores in their lower end CPU would also opens more performant option for their not so sexy pentium line (living atom derivatives aside).
Then 5, 7 and 9 cores.

They have to go away from their dual cores line ad they can afford not to jump to quad straight ahead.

 

Intel has little to worry about in the Y and U series market. As long as AMD doesn't support LPDDR in their APUs they just won't get any traction in low-power mobile applications.

 

https://pcper.com/2019/09/ice-lake-benchmarks-1065g7-vs-3700u/

 

I know its OT but it was mentioned above, but it looks like the Renoir APU will support LPDDR4, so ice lake vs Zen2 4core? + 20CU Vega is going to be really interesting to see.

 

If Renoir is coming with LPDDR4X 4266MT/s and the current Picasso vs. Ice Lake results are anything to go by, Renoir will be similar to slightly slower in CPU performance but will blast the G7 Iris Pro in GPU performance.

Intel keeps talking about Gen11 being so much faster than Gen9, but how Intel managed to keep the same ~400 GFLOPs 24 EU GPU from Broadwell-ish as standard for 6 whole years is the real white elephant in the room here, IMO.


This should be a real eye-opener on the importance of competition.