Discussion Intel current and future Lakes & Rapids thread

Page 133 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
Just the opposite - I guess the point is that Icelake is extremely limited volume because of the awful yield. Tigerlake will likely just be limited.

Putting aside 10nm though, we now have several generations of fairly redundant 14nm updates for buyers to contend with. Kaby, Coffee, Whiskey, Comet . . . yeah the latest products are somewhat of an improvement, but is it enough to move product?
 

scannall

Golden Member
Jan 1, 2012
1,944
1,638
136
Putting aside 10nm though, we now have several generations of fairly redundant 14nm updates for buyers to contend with. Kaby, Coffee, Whiskey, Comet . . . yeah the latest products are somewhat of an improvement, but is it enough to move product?
This kind of upgrade cycle is what made Apple toss the Power series off the cliff, that and IBM decided to not make a low power one for laptops. For an OEM, 4+ years of no actual upgrade is a disaster.
 
  • Like
Reactions: ozzy702

mikk

Diamond Member
May 15, 2012
4,112
2,108
136
Just the opposite - I guess the point is that Icelake is extremely limited volume because of the awful yield. Tigerlake will likely just be limited.


Tigerlake should come roughly 1 year after Icelake, so it should be much better than now. According to the Tweakers Roadmap Icelake is limited while Tigerlake isn't. Icelake isn't even in the stable platform program roadmap in contrast to Tigerlake which is a statement and the first ES leaks have much better clocks. Also it looks like Intel is very keen to update its entire product lineup with a Gen12 graphics as fast as possible.
 

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
Tigerlake should come roughly 1 year after Icelake, so it should be much better than now. According to the Tweakers Roadmap Icelake is limited while Tigerlake isn't.

It'll be more available I imagine but Rocket Lake is what people will be buying.
 

mikk

Diamond Member
May 15, 2012
4,112
2,108
136
Here is a first userbenchmark result from Icelake i7-1065G7: https://www.userbenchmark.com/UserRun/19625737

Only one sample, so not sure if this is accurate, it's a Dell XPS 13 7390 2-in-1.

Avg speed 3.05 Ghz for the Icelake entry, the Tigerlake entries were running with avg 3.6 Ghz despite being an early ES.

https://www.userbenchmark.com/UserRun/19310722
https://www.userbenchmark.com/UserRun/19310543

I think the "new transistor optimization" is a result of this. It would make sense if Tigerlake is on 10nm++
 

Ajay

Lifer
Jan 8, 2001
15,332
7,792
136
I think the "new transistor optimization" is a result of this. It would make sense if Tigerlake is on 10nm++
I just hope the xtor optimizations produce higher yields and not just higher clocks. I long for the days when Intel was transparent because they were proud of their technology.
 

majord

Senior member
Jul 26, 2015
433
523
136
Here is a first userbenchmark result from Icelake i7-1065G7: https://www.userbenchmark.com/UserRun/19625737

Only one sample, so not sure if this is accurate, it's a Dell XPS 13 7390 2-in-1.

Avg speed 3.05 Ghz for the Icelake entry, the Tigerlake entries were running with avg 3.6 Ghz despite being an early ES.

https://www.userbenchmark.com/UserRun/19310722
https://www.userbenchmark.com/UserRun/19310543

I think the "new transistor optimization" is a result of this. It would make sense if Tigerlake is on 10nm++

So theres no huge IPC uplift for tigerlake after all, they both just do well in this bench?
 

mikk

Diamond Member
May 15, 2012
4,112
2,108
136
So theres no huge IPC uplift for tigerlake after all, they both just do well in this bench?


Icelake IPC doesn't look that good over Wiskey Lake if it was running with the proper 1C Boost clock which we don't know from 1 sample, although the Dell 7390 should boost to 3.9 Ghz, we have Geekbench results from this.

Icelake 3.9 Ghz 129
Whiskey Lake 4.6 Ghz 135-140
Tigerlake xx Ghz 146


Updated with a better I7-1065G7 result.
 
Last edited:

Abwx

Lifer
Apr 2, 2011
10,852
3,298
136
Here is a first userbenchmark result from Icelake i7-1065G7: https://www.userbenchmark.com/UserRun/19625737

Only one sample, so not sure if this is accurate, it's a Dell XPS 13 7390 2-in-1.

Avg speed 3.05 Ghz for the Icelake entry, the Tigerlake entries were running with avg 3.6 Ghz despite being an early ES.

https://www.userbenchmark.com/UserRun/19310722
https://www.userbenchmark.com/UserRun/19310543

I think the "new transistor optimization" is a result of this. It would make sense if Tigerlake is on 10nm++

ICL low clock is likely not due to the process alone but also to the design, they had to downscale frequency to make room for the IPC improvement, said otherwise a CFL 10nm shrink would clock way higher than ICL, i guess that the improvement in TGL is to increase the frequency capabiity of the pipeline.
 
  • Like
Reactions: Nothingness

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
ICL low clock is likely not due to the process alone but also to the design, they had to downscale frequency to make room for the IPC improvement, said otherwise a CFL 10nm shrink would clock way higher than ICL,

Can't see how you can say that when there is no Coffee 10 nm. And everything Intel has done so far has only reinforced the idea that yields really are that bad, and they are doing what they can to maximize what they do get. Which includes reducing frequency and jacking up the voltage.

I expect Rocket Lake to run at 5 Ghz, even assuming it's using Willow Cove cores. The power draw will be a big problem though and could prevent it from doing 5 Ghz on all cores on a theoretical K part, even at unlimited TDP.
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,683
1,218
136
Don't know where I got it from... but supposedly there will be a special Tigerlake die specifically for Apple. Which might have lead to Apple killing off a next-gen undisclosed SoC.
 

Dave2150

Senior member
Jan 20, 2015
639
178
116
Apologies, I haven't been following this thread closely. How close are we to the 15% IPC increase Icelake (Sunnycove) desktop 8C/12C/16C? CPU's that will run close to 5ghz?

Had my 6700k@4.7Ghz since 2015, Ryzen 3000 series seem great but the 15% IPC increase from Icelake should be worth waiting for.
 

Yotsugi

Golden Member
Oct 16, 2017
1,029
487
106
How close are we to the 15% IPC increase Icelake (Sunnycove) desktop 8C/12C/16C? CPU's that will run close to 5ghz?
Any fatter cores will never run at 5GHz without funky (IBM-tier), totally unsuitable for mainstream cores logic.
 

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
Apologies, I haven't been following this thread closely. How close are we to the 15% IPC increase Icelake (Sunnycove) desktop 8C/12C/16C? CPU's that will run close to 5ghz?

Good luck getting one of those. Intel doesn't have any 10nm desktop processors on their roadmaps in all of 2020. They may never produce a 10nm desktop CPU. Rocketlake will be a 14nm CPU that will allegedly use Willow Cove cores, but due to the process, I doubt you'll see more than 8c there, and the heat. Ugh. You're not going to be a happy camper if you keep waiting for Intel.
 

Ajay

Lifer
Jan 8, 2001
15,332
7,792
136
Good luck getting one of those. Intel doesn't have any 10nm desktop processors on their roadmaps in all of 2020. They may never produce a 10nm desktop CPU. Rocketlake will be a 14nm CPU that will allegedly use Willow Cove cores, but due to the process, I doubt you'll see more than 8c there, and the heat. Ugh. You're not going to be a happy camper if you keep waiting for Intel.
Yeah, I agree. Though, I think Intel will still push a 10 core RocketLake, if "IPC" goes up, base clocks will go down - Intel may pull some tricks to eek out high turbo boosts.
 
  • Like
Reactions: coercitiv

Dave2150

Senior member
Jan 20, 2015
639
178
116
Good luck getting one of those. Intel doesn't have any 10nm desktop processors on their roadmaps in all of 2020. They may never produce a 10nm desktop CPU. Rocketlake will be a 14nm CPU that will allegedly use Willow Cove cores, but due to the process, I doubt you'll see more than 8c there, and the heat. Ugh. You're not going to be a happy camper if you keep waiting for Intel.

My 6700k @ 4.7Ghz still runs well at 1440P, I'll soon be moving to 4K, where the CPU will have even less impact. That said, I'd like to place my I7 920 'server' with my 6700k setup, and upgrade to either Intels' next architecture or AMD's. Will see how Rocket lake does, though if it's 14nm then it should clock close to 5Ghz, coupled with 15% IPC will be the clear CPU to go for, hopefully PCI-E V5/v6 and DDR5 as well, in 2020/2021. We'll see!
 

DrMrLordX

Lifer
Apr 27, 2000
21,582
10,785
136
My 6700k @ 4.7Ghz still runs well at 1440P, I'll soon be moving to 4K, where the CPU will have even less impact. That said, I'd like to place my I7 920 'server' with my 6700k setup, and upgrade to either Intels' next architecture or AMD's. Will see how Rocket lake does, though if it's 14nm then it should clock close to 5Ghz, coupled with 15% IPC will be the clear CPU to go for, hopefully PCI-E V5/v6 and DDR5 as well, in 2020/2021. We'll see!

Yes, we will. Just try not to be too optimistic. Intel seems to be focusing almost exclusvely on 4c and 2c mobile parts right now. I think their most exciting upcoming product is Tigerlake.
 

extide

Senior member
Nov 18, 2009
261
64
101
www.teraknor.net
Any fatter cores will never run at 5GHz without funky (IBM-tier), totally unsuitable for mainstream cores logic.
That's just speculation, though. People said the same thing when Ivy Bridge came out and couldn't hit 5Ghz like Sandy Bridge could. I would say hitting 5Ghz again is going to be tough, but definitely not impossible, especially with node progression slowing down a lot means that they have a lot more time to fine tune a given node over the years.
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,683
1,218
136
You know what will be weird, Intel saying PCIe-slot/discrete graphics are dead for consumer desktop.

Then, coming out with a socket xyL-G w/ LGA1899. Two clusters(CPU/GPU) w/ their own HBM2E local-memory. Then an I/O hub which interconnects them with on-package CXL x.y. Which also contains quad-channel DDR5. However, the motherboard has no external PCIe slots(+SATA might be gone as well). Instead, its free space for wild stuff; like a motherboard integrated 32-bit/768 KHz DAC+Advanced DSP soundcard. Or, for PCIe 5.0 Optane buffered raids with PCIe 5.0 NVMes. Up to the motherboard manufacturer, mostly.
 
Last edited:

Ajay

Lifer
Jan 8, 2001
15,332
7,792
136
Just the opposite - I guess the point is that Icelake is extremely limited volume because of the awful yield. Tigerlake will likely just be limited.
Maybe not. With all their experience gained with Icelake, I wouldn’t be surprised if the physical implementation of Tigerlake (cell implementations) have been altered to help improve yields. That and Tigerlake well may have been designed to slightly different xtor parameters (process change) that could also improve yields. I’m sure Intel has thrown the kitchen sink at Tigerlake considering that Renoir will be out soon.
 

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
Maybe not. With all their experience gained with Icelake, I wouldn’t be surprised if the physical implementation of Tigerlake (cell implementations) have been altered to help improve yields. That and Tigerlake well may have been designed to slightly different xtor parameters (process change) that could also improve yields. I’m sure Intel has thrown the kitchen sink at Tigerlake considering that Renoir will be out soon.

I still think it's chiplets. After all, they decided to increase the EU count by 50% which means a theoretical monolithic chip would be really big for 10 nm yield.
 

Ajay

Lifer
Jan 8, 2001
15,332
7,792
136
I still think it's chiplets. After all, they decided to increase the EU count by 50% which means a theoretical monolithic chip would be really big for 10 nm yield.
That would make sense. Intel has certainly spent enough time on 10nm designs to accomplish that. And they have already shown a variety of technologies at work in their 10nm FPGA products. I really hope Tiger will be available in significantly higher quantities than Icelake, and rumors so far seem to be trending that way. My comments were just to show that there are possible engineering solutions to improve yields based on a variety of lessons learned. Multi-chip designs is another, very tangible, one.
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,683
1,218
136
Palm Cove/Sunny Cove via Cannonlake/Icelake = 10nm FinFET+Metal Stack
Willow Cove via Tigerlake = 10nm++ FinFET+Metal Stack

Willowcove via Tigerlake will include the 10nm+ and 10nm++: Faster and lower power Vts, finer CPP, denser metal stack, denser std cell & bit-cell/memory compilers. Which were not used on Sunnycove/Icelake and Palmcove/Cannonlake. Skylake moves to Cannonlake's core and Sunnycove's core(OG-Icelake had the same AGU cluster as CNL/SKL) with Willowcove being a next-gen core design(increased ST performance is not guaranteed). With that Goldencove will have a ST performance over Willowcove/Sunnycove, but will be on the 7nm node.
 
Last edited: