Intel Cannonlake, Ice Lake, Tiger Lake & Sapphire rapid thread

Page 248 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

uzzi38

Senior member
Oct 16, 2019
894
1,104
96
There are enough differences between version 5.0 and 5.2 such that comparing the two is not appropriate. I recall reading about it in the Geekbench release notes.
Then here's a similar 9900K score from 5.2.2:


Not much has changed with my comparison

Well, aside from the fact the difference in navigation scores is now noticably smaller, that is. Rocket Lake-S still falls behind, but not to the same degree it did before.

As an aside, I'm not sure about the rest of you but I'm expecting an increase to memory latency after all for two reasons. Firstly, Ice Lake's memory controller isn't as good in terms of memory latency as Skylake's. And if you refer back to Sharkbay's leaks, you'll remember he had RKL-U rated for LPDDR4X-3733, not 4266 like TGL-U.

Secondly, there is a serious possibility that Rocket Lake is MCM with a 10nm iGPU+I/O die and 14nm CPU cores. Not only has it appeared on a roadmap that is almost certainly legitimate (contained a reference to CML-U only bringing LPDDR4 support with a future steppng, though this roadmap only covered RKL-U, not RKL-S), the aforementionned point about LPDDR4X support makes it possible Intel did not backport the IMC as well.

So what you're seeing as being a consequence of atrocious memory latency, I'm seeing as completely in-line with expectations.

BTW, I'm not sure what kind of memory you're expecting. While it's normal for OEMs and manufacturers to pop JEDEC spec memory into a system while testing, you don't normally see worse than that. Only in weird situations such as the Acer Swift 3 do you get memory that is notably worse than rated JEDEC speeds.
 
  • Like
Reactions: lightmanek

lobz

Golden Member
Feb 10, 2017
1,250
1,244
106
14nm strikes again.
It does indeed. Because GB results are all over the place for the 10900K, this 5.0 GHz RKL sample achieves anywhere between 4 and 12% better ST result than the wide variety of 10900K results @ 5.1 GHz (multicore will obviously be worse than a 10900K, no question there). While this is nice, The compromises that had to be made for 14nm mean that Rocket Lake's IPC gains are nowhere near the hoped for 20%.

In this sense, you are perfectly correct, 14nm strikes again.
 

lobz

Golden Member
Feb 10, 2017
1,250
1,244
106
I know, everyone with some serious thinking expected 10nm from ADL, however Charlie and some other usual suspects tried to defend the 14nm hypothesis till the end: https://www.reddit.com/r/AMD_Stock/comments/hr38ko/_/fy2dji2
My dear dude, in Charlie's call with that company, he had information about the GPU part being mfg'd at 14nm, not the CPU cores.
That's also why you don't just copy/paste things from reddit. So, care to change your previous ranting about the 'usual suspects' ?
 

cortexa99

Member
Jul 2, 2018
77
102
76
Wow, this thread is over 3 years old, and we still don't have a 10nm desktop chip.
It's an awkward situation right now intel is suffering.

If there wasn't any Zen1 in 2017, intel might had released 10nm desktop chip SKL-based i7-8700k at then which is only has 4C8T@4Ghz to continue fooling comsumer.

But it didn't happened. Going over past 3 years' history, it's very clear that Zen1's arrival & crappy 10nm messing up intel's whole original plan.

Which even more awful is when AMD release newer-gen Zen, intel has to refresh 14nm+++++(CFL, CFL-R, 9900K 9900KS,) to remain competitive, mainly because 10nm yielding is not able to roll out high-perf chip in time, now everytime Intel refreshing 14nm+++++, it would raise whole product lineup's performance baseline, that forcing 10nm to '+++++' at the meanwhile, otherwise it wouldn't make sense to roll out any 10nm SKUs cuz it is not advanced enough even against intel's own 14nm+++++.

By this logic, if AMD's Zen3,4,5 continue to improve step by step, it's possible that Intel has to passively '+++++' its old process step by step. Unless AMD make mistake at a or few generation of Zen, Intel would have a chance to breath.

My expectation. 10nm would be on time, 2021, but 7nm is pessimistic. Not sure whether intel would have to roll out 10nm++++++ SKUs by then to face newer Zen, and 7nm would have to '+++++' at the meanwhile which means vastly postoned like initial 10nm sitiation. Unless AMD mess up.....

(right now I even cannot understand difference between icelake, rocketlake, alderlake, tigerlake *facepalm*)
 
Last edited:

Antey

Member
Jul 4, 2019
40
39
51
If there wasn't any Zen1 in 2017, intel might had released 10nm desktop chip SKL-based i7-8700k at then which is only has 4C8T@4Ghz to continue fooling comsumer.
by that time? nah, yields were too awful for that. remember cannon lake. if ryzen didnt happen intel could have waited to 2018 and release another 4 cores 8 threads 14nm+ 8700k but with 5ghz single core turbo, and in 2019 another 4 cores 8 threads 14nm++ 9700k with 5,2GHz sc, and 5GHz all core turbo... and maybe a 4C/8T 10700K tiger lake-s in 2020...
 
Last edited:
  • Like
Reactions: lightmanek

DrMrLordX

Lifer
Apr 27, 2000
16,020
4,970
136
Yeah but you're talking about the same Raja who let a Vega tease video through with 'POOR VOLTA' in it.
It's alright so long as he is a god in his own mind. Seriously though, if Intel starts to falter as an IDM and if their TSMC-made dGPUs are at least moderately successful, that kinda puts Raja in the driver's seat.

I'm really curious, who claimed ADL would be on 14nm?
Honestly I don't know, Charlie-but-not-Charlie?
 

inf64

Platinum Member
Mar 11, 2011
2,977
1,462
136
It looks like Raja will be gone within a quarter or two... There is a power struggle at intel and Murthy is winning(according to both Jim and Charlie). 7nm is a worse disaster then 10nm and 10nm is seriously under performing. It all looks like AMD will be pummeling intel next 2 generations at least, which is sad for us consumers.
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,031
612
136
I think Michael C. Mayberry will be the next Intel CEO because he looks just like Tim Cook and that would make investors feel better ;)
With the CTO becoming Gary Patton, finally getting the position he originally had at GlobalFoundries.

Thus begins Intel's work away from leading edge nodes and placing Graphene on Si. Thus, beginning the era of Netburst 3000 and 1 THz cores w/ 300 GHz 3D-RAM CNTs.
 

DrMrLordX

Lifer
Apr 27, 2000
16,020
4,970
136
It looks like Raja will be gone within a quarter or two... There is a power struggle at intel and Murthy is winning(according to both Jim and Charlie). 7nm is a worse disaster then 10nm and 10nm is seriously under performing. It all looks like AMD will be pummeling intel next 2 generations at least, which is sad for us consumers.
Why Raja though? He appears to have steered DG1, DG2, and Ponte Vecchio to outside fabs successfully, which is probably the direction Intel needs to take anyway. You would think anyone pushing hard to keep manufacturing in-house would be losing power.
 
  • Like
Reactions: yuri69 and mikk

KompuKare

Senior member
Jul 28, 2009
594
90
91
It looks like Raja will be gone within a quarter or two... There is a power struggle at intel and Murthy is winning(according to both Jim and Charlie). 7nm is a worse disaster then 10nm and 10nm is seriously under performing. It all looks like AMD will be pummeling intel next 2 generations at least, which is sad for us consumers.
Well, while tit-for-tat between Intel and AMD would be exiting, currently Intel still make over 98% of the x86 profit (2019 net income: Intel $21,000 million, AMD $341 million).

So I wouldn't worry about them yet, plus I think for AMD to truly be able to deliver multi-year projects and more variants, AMD would have to grab at least 20% of the x86 profit for a few years. That would be good for consumer and PC enthusiasts, not so good for shareholders.

In the end it is money and how many (few) employees they have which determine that how many die variations the do - with more employees we might not get the near one year delay between Zen2 and the APUs, or get proper 4C dies for the low end, try-outs of chiplets design combining CPU and GPU parts, etc.)

You would think anyone pushing hard to keep manufacturing in-house would be losing power.
Yes, in a rational world you would think so but then who says a power struggle is rational?
While I doubt that Intel is as colourful as Apple at one stage (I'm thinking of the infamous pink vs blue when deciding on the next OS in the later 80s: https://en.wikipedia.org/wiki/Copland_(operating_system)#Pink_and_Blue), I'm sure with Intel's manufacturing being so huge there are lots of egos, perks, and positions at stake if they were to fab stuff elsewhere.
 

mikk

Platinum Member
May 15, 2012
2,733
603
136
After the Geekbench v5 result it is easy to say this. He should have posted this before it appeared. Pretty sure this is just based on the Geekbench v5 result which is risky to conclude based on only one benchmark. TDP is known as well and 5 Ghz lol, after the 5 Ghz Geekbench leak not so difficult to predict this.
 

uzzi38

Senior member
Oct 16, 2019
894
1,104
96
After the Geekbench v5 result it is easy to say this. He should have posted this before it appeared. Pretty sure this is just based on the Geekbench v5 result which is risky to conclude based on only one benchmark. TDP is known as well and 5 Ghz lol, after the 5 Ghz Geekbench leak not so difficult to predict this.
Bad assumption.
 

jpiniero

Diamond Member
Oct 1, 2010
7,924
1,206
126
Intel, in their 10-Q report said they took a 1.1B hit from ramping Tiger Lake. IOW they are burning a helluva lot of wafers to try to make up for the bad yield. It also explains the QC TIger Lake-H rumor (since that die would be smaller than the U I think due to having 32 EUs). The CCG margins are getting completely wrecked but I guess they are at the point where they don't have much of a choice versus Renior.

As it stands, there's also no way they can realistically fab Ice Lake-SP in any kind of volume.
 

ajc9988

Senior member
Apr 1, 2015
266
159
116
Well, while tit-for-tat between Intel and AMD would be exiting, currently Intel still make over 98% of the x86 profit (2019 net income: Intel $21,000 million, AMD $341 million).

So I wouldn't worry about them yet, plus I think for AMD to truly be able to deliver multi-year projects and more variants, AMD would have to grab at least 20% of the x86 profit for a few years. That would be good for consumer and PC enthusiasts, not so good for shareholders.

In the end it is money and how many (few) employees they have which determine that how many die variations the do - with more employees we might not get the near one year delay between Zen2 and the APUs, or get proper 4C dies for the low end, try-outs of chiplets design combining CPU and GPU parts, etc.)


Yes, in a rational world you would think so but then who says a power struggle is rational?
While I doubt that Intel is as colourful as Apple at one stage (I'm thinking of the infamous pink vs blue when deciding on the next OS in the later 80s: https://en.wikipedia.org/wiki/Copland_(operating_system)#Pink_and_Blue), I'm sure with Intel's manufacturing being so huge there are lots of egos, perks, and positions at stake if they were to fab stuff elsewhere.
You are a fool. Mercury research has had AMD at 14.5% of the x86 market in 2018 and 15.5% in 2019. They are making headway in OEMs. And Intel lost a large customer in losing Apple. Added together, you are wrong in the way you calculate market share AND in your estimate of what AMD needs to do.

On servers, AMD needed to prove itself. It did with Rome. With Milan, it's not even near fair anymore. It's like stomping on an infant.

Also, this isn't about ego, it's about ignorance and bad leadership at this point. Everyone saw this coming. This is how the cutting edge leadership competition has gone for the past 15 years. A stunted node could knock you out. IBM sold their fab to GloFo. GF then dropped due to costs for 7nm and lack of guaranteed contracts, but is doing great stuff with FD-SOI.

Now, of the remaining three, Intel is failing on 10nm and 7nm. They are dropping. This is whether they like it or not.

Samsung, although having a 7nm and working on smaller, is fabbing for Nvidia on 8nm, which is a 10nm refined node IIRC. Samsung won't drop because they make too many products. And even if they slip, they'll stay right behind.

So, anyone not blinded can already see TSMC has won the miniaturization/fab race. It's over and 3nm around 2023 will seal it.
 
Last edited:

KompuKare

Senior member
Jul 28, 2009
594
90
91
You are a fool. Mercury research has had AMD at 14.5% of the x86 market in 2018 and 15.5% in 2019.
Thanks!
But what does marketshare have to do with profit?

I said that taking in under 2% of the net income in the x86 market, was not enough to have a competitive AMD long-term.
 

ajc9988

Senior member
Apr 1, 2015
266
159
116
Thanks!
But what does marketshare have to do with profit?

I said that taking in under 2% of the net income in the x86 market, was not enough to have a competitive AMD long-term.
Also absurd. Intel cannot maintain their pricing structure moving forward, meaning there are hard realities coming to investors with their heads in the sand like you.

For example, Milan will have 20% IPC uplift on integer. AVX512 workloads are less than 8% of workloads and AMD is potentially adding that on Genoa. Intel charges $10K per top CPU that outside of AVX512, it takes 2 to amount to a single 64 core chip.

Now, after taking into account 18 month test needs, having to change software and optimize the change over, AMD delivering on Zen 3 will directly be a greater threat and take more market share, just like Zen 2 challenged intel's mobile department more than was thought. They also did stupid stuff on pcie lane giving Intel a slight reprieve.

Also, AMD's margins are now over 50%. That means even though Intel is getting revenue, there will be even more shift until 2023 when they possibly won't deliver on 7nm and have to compete for fab time against AMD at TSMC.

Now, if you are talking of exit strategies, sure there might be time to milk Intel. But if you ignore the writing on the wall, do so at your own peril. Intel will be losing market share for the next 2 to 3 generations, if not longer.
 

ASK THE COMMUNITY