Ajay
Diamond Member
- Jan 8, 2001
- 7,629
- 2,776
- 136
-
Community Question: What makes a good motherboard?
LTC8K6
Lifer
- Mar 10, 2004
- 28,523
- 1,569
- 126
Not a chance, imo.
We'll see some form of desktop 10nm before 2 years.
Having said that, if they make Ice Lake on 14nm+++ and it's better than Coffee Lake, I really won't care about the node.
Ajay
Diamond Member
- Jan 8, 2001
- 7,629
- 2,776
- 136
It's a year before Icelake mobile. Desktop Icelake would be at least a year after that. A backport of Icelake would means Intel knew 10nm was in trouble in the R&D stage!
Dayman1225
Golden Member
- Aug 14, 2017
- 1,000
- 576
- 106
The 20 Desktop is probably still Tigerlake on the Fake 10 nm, even if it doesn't clock that high/worse than AMD. The new consoles will be out then, so I reckon Intel will want to juice the core counts above 8.
It's going to be Comet Lake (up to 10 cores) according to a leaker on a Chinese message board.
Kind of silly. First we are stuck with quads for way too long and then they go the other way and put on "too many" cores?
LTC8K6
Lifer
- Mar 10, 2004
- 28,523
- 1,569
- 126
Well, they demoed a 10nm 2 in 1 laptop in January of 2017. They certainly knew of the problems with the node before then.
Dayman1225
Golden Member
- Aug 14, 2017
- 1,000
- 576
- 106
Intel has released a product brief for the XMM 7660 Modem, it is based on 14nm unlike 10nm as previously reported by SemiAccurate(Charlie)
https://www.intel.com/content/dam/www/public/us/en/documents/product-briefs/xmm-7660-brief.pdf
https://www.intel.com/content/dam/www/public/us/en/documents/product-briefs/xmm-7660-brief.pdf
IntelUser2000
Elite Member
- Oct 14, 2003
- 7,249
- 1,839
- 136
Yea, this is why big sites like Anandtech don't cover speculation and rumor. You really cannot guarantee the accuracy. The most Anandtech does is show roadmaps leaked by the companies themselves.
Word of mouth is the worst. At least with leaks there's some data. Relying on someone means you are relying on the person leaking to not lie to you, or be inaccurate.
Word of mouth is the worst. At least with leaks there's some data. Relying on someone means you are relying on the person leaking to not lie to you, or be inaccurate.
Another example for the incredibility of Semiaccurate/Charlie.
mopardude87
Diamond Member
- Oct 22, 2018
- 3,348
- 1,563
- 96
lol the confusion with these intel chips is just great.My buddy and me tried to get his i7 7700k and this Z370 board up and running for hours and i was about to call it doa. Then i remember how certain chips simply refuse to work in certain series boards and sure enough the dang blasted 7700k being Kaby Lake was not compatible with the Z370 which demands Coffee Lake after a quick google search.
Intel calm down with all the chips and dang blasted chipsets plz? lol i blink and there is just YET one more around the corner.
Intel calm down with all the chips and dang blasted chipsets plz? lol i blink and there is just YET one more around the corner.
IntelUser2000
Elite Member
- Oct 14, 2003
- 7,249
- 1,839
- 136
Intel's process delays. How bad is it really?
Original 10nm - Late 2016.
Is it really? Let's backtrack.
Late 2016 - Cannonlake
Late 2014 - Broadwell
Late 2012 - Ivy Bridge. Reality = Mid 2012
Late 2010 - Westmere(Oops!). Reality = Early 2010, very early, like January.
Just looking at 10nm delays, we're looking at 3 years, because first products are planned for late 2019. Even if we took Intel's claims at face value they were 3.5 years ahead(because they had FinFET 3.5 years before everyone, not that it mattered in reality but let's ignore that...), nearly all advantage was erased.
However, that ignores the delays that happened after 32nm. Let's recap. We know 14nm had serious troubles as well. That's why we ended up with them missing some of the markets and a staggered launch, along with the disappointment AKA Core M. But Skylake sorta redeems it by coming out in Late 2015 in full. There were sayings Skylake had to have its architecture modified to account for the underperforming 14nm process, so maybe not everything was rosy as we could have got few % faster uarch.
What's less known are the delays to 22nm. While it wasn't publicly known, it was definitely not in line with 12 month Tick/Tock cadence. Ivy Bridge came after 16-17 months in April, or 4-5 months delay. The performance and overclockability was a disappointment especially after the media fanfare that surrounded 22nm's TriGate process. We'll get to know later that the 22nm was really made for Silvermont Atom, which failed to have any impact anyway.
Haswell pushed that further to June. So 1-2 months delay. And finally, we see Broadwell coming in extremely limited quantities with a crippled chip after 16-18 months or 4-6 months delay.
Why did 22nm get delayed? Probably for mobile.
Why did 14nm get delayed? Mobile.
Why did 10nm get delayed? Density ePeen against ARM competitors, so again Mobile.
The real original roadmap:
Early 2010 - Westmere
Early 2012 - Ivy Bridge
Early 2014 - Broadwell
Early 2016 - 10nm Cannonlake
Early 2018 - 7nm Tigerlake
4 years delay. That's how we end up from having a lead in nearly every metric to a loss in just 8 years(Sandy Bridge in 2011 to Icelake in 2019). It's interesting how if instead they didn't directly pursue mobile but indirectly competed by making better PC chips, it would have been much better for Intel. By pursuing a side business they are losing the main. Many big businesses actually do this.
Original 10nm - Late 2016.
Is it really? Let's backtrack.
Late 2016 - Cannonlake
Late 2014 - Broadwell
Late 2012 - Ivy Bridge. Reality = Mid 2012
Late 2010 - Westmere(Oops!). Reality = Early 2010, very early, like January.
Just looking at 10nm delays, we're looking at 3 years, because first products are planned for late 2019. Even if we took Intel's claims at face value they were 3.5 years ahead(because they had FinFET 3.5 years before everyone, not that it mattered in reality but let's ignore that...), nearly all advantage was erased.
However, that ignores the delays that happened after 32nm. Let's recap. We know 14nm had serious troubles as well. That's why we ended up with them missing some of the markets and a staggered launch, along with the disappointment AKA Core M. But Skylake sorta redeems it by coming out in Late 2015 in full. There were sayings Skylake had to have its architecture modified to account for the underperforming 14nm process, so maybe not everything was rosy as we could have got few % faster uarch.
What's less known are the delays to 22nm. While it wasn't publicly known, it was definitely not in line with 12 month Tick/Tock cadence. Ivy Bridge came after 16-17 months in April, or 4-5 months delay. The performance and overclockability was a disappointment especially after the media fanfare that surrounded 22nm's TriGate process. We'll get to know later that the 22nm was really made for Silvermont Atom, which failed to have any impact anyway.
Haswell pushed that further to June. So 1-2 months delay. And finally, we see Broadwell coming in extremely limited quantities with a crippled chip after 16-18 months or 4-6 months delay.
Why did 22nm get delayed? Probably for mobile.
Why did 14nm get delayed? Mobile.
Why did 10nm get delayed? Density ePeen against ARM competitors, so again Mobile.
The real original roadmap:
Early 2010 - Westmere
Early 2012 - Ivy Bridge
Early 2014 - Broadwell
Early 2016 - 10nm Cannonlake
Early 2018 - 7nm Tigerlake
4 years delay. That's how we end up from having a lead in nearly every metric to a loss in just 8 years(Sandy Bridge in 2011 to Icelake in 2019). It's interesting how if instead they didn't directly pursue mobile but indirectly competed by making better PC chips, it would have been much better for Intel. By pursuing a side business they are losing the main. Many big businesses actually do this.
Last edited:
14 nm was at least a full year of a delay; hence Haswell Refresh.
The Skylake Architect in an interview admitted they basically redid Skylake 3 times (I think) during development. What they actually changed beyond the removal of the FIVR and AVX-512 support from Client is unknown.
IntelUser2000
Elite Member
- Oct 14, 2003
- 7,249
- 1,839
- 136
No, not when you look as a whole. While Broadwell was seriously delayed, Skylake brought most back. Haswell refresh was only for desktops so it doesn't really count.
14nm delay is 4-6 months not 12. But, I guess you didn't read the whole post.
There was a speculation by Hiroshige Goto at PC Watch that Skylake likely had to be designed for higher frequencies(likely with higher number of pipeline stages and circuitry tweaks) because the 14nm wasn't up to par.
It could explain why Skylake had a mediocre perf/clock gain.
That's not true, Haswell Refresh launch in Q2 14 included new desktop and laptop parts. Now they did paper launch Broadwell in Q3-Q4 in mobile and a full launch with laptops in Q1 15.
Looking at it that way, yes the 14 nm delay was only 6 months.
IntelUser2000
Elite Member
- Oct 14, 2003
- 7,249
- 1,839
- 136
The Haswell "Refresh" wasn't even a refresh for mobile. It included a 100MHz clock bump. The fact that they introduced Broadwell a quarter or two later further solidifies that.
I don't really count the 2014 14nm launch. The systems were available in extremely limited quantities with a shoddy ship called the Core M at latter end of Q4. If you look at mass availability of mobile it came to be in Q1 of 2015. That's why I look at Skylake instead, because it was a mass launch.
Yeah that's why I am saying 6 months.
I should point out that some launches might not have been a year because of inventory reasons and not because of fab issues.
Dell XPS 7390 2-in-1
Genuine Intel(R) CPU 0000 @ 1.60GHz (4C 8T 1.69GHz/1.6GHz 5% OC, 4x 512kB L2, 8MB L3)
Intel(R) UHD Graphics, Gen11 LP (512SP 64C 900MHz, 1MB L2, 3.1GB) (OpenCL)
http://ranker.sisoftware.net/show_system.php?q=cea598ab9cac9bae9bbddae7cafbddaf92a385ecd1e1c7af92a284fcc1f1d7b2d7eadafc8fb28a&l=en
From these entries we can see that OEMs received fully enabled 4+2 ICL-U samples and are working on it.
Genuine Intel(R) CPU 0000 @ 1.60GHz (4C 8T 1.69GHz/1.6GHz 5% OC, 4x 512kB L2, 8MB L3)
Intel(R) UHD Graphics, Gen11 LP (512SP 64C 900MHz, 1MB L2, 3.1GB) (OpenCL)
http://ranker.sisoftware.net/show_system.php?q=cea598ab9cac9bae9bbddae7cafbddaf92a385ecd1e1c7af92a284fcc1f1d7b2d7eadafc8fb28a&l=en
From these entries we can see that OEMs received fully enabled 4+2 ICL-U samples and are working on it.
Arachnotronic
Lifer
- Mar 10, 2006
- 11,719
- 2,003
- 126
Glad they're making some progress. It's really time for them to move off of Skylake.
Can someone please translate the results comparing to current skylake architecture with the same frequency?
Thanks
Arachnotronic
Lifer
- Mar 10, 2006
- 11,719
- 2,003
- 126
Nice, as long as it can be properly cooled and clocks to the moon, I'll take it.
TRENDING THREADS
-
Question New Apple SoC - M1 - For lower end Macs - Geekbench 5 single-core >1700- Started by Eug
- Replies: 2K
-
***OFFICIAL*** Ryzen 5000 / Zen 3 Launch Thread REVIEWS BEGIN PAGE 39- Started by ElFenix
- Replies: 2K
-
Question Speculation: RDNA2 + CDNA Architectures thread- Started by uzzi38
- Replies: 5K
-
-
Question 'Ampere'/Next-gen gaming uarch speculation thread
- Started by Ottonomous
- Replies: 5K
Facebook
Twitter