What do server CPUs have in common with mobile CPUs?
Non-leading-edge clock speeds.
It looks like Intel has fixed the yield issue on 10nm, and can make larger dies now (Ice Lake U with 1 TFLOPS GPU, and future Xeons with higher core counts).
But they clearly have not got anywhere with the performance issue (whereby 14nm is still higher performance), which means no 10nm Desktop processors. Straight to 7nm I guess, in 2022?
This "leak" appears to be a fabrication. The tables are pretty good but not good enough to be genuine. Besides if Intel wait until... 2022(!) to upgrade the full range of their CPUs to 10nm they are completely toast.
I believe an article in AnandTech any day over an article in WCCFtech. I would agree leak in WCCFtech appears to be fabrication because it is on WCCFtech.
What if Intel has figured out way to use same chips for desktop as in mobile. Have you ever thought of that.
WCCFtech is not a reliable source of technical information - they never provide there source. But they are fun to watch because they jump at news ( real or not ) quickly
Still waiting for the SEC investigation into that. (Or the acknowledgement by Intel fans that there is a long term pattern of lying that didn't just start with 10nm.)
Those "leaks" are based on enterprise timelines, which is slower to adopt new desktop CPUs, so even if those are true and still up to date, it may not necessarily say much about actual 10nm desktop parts being released in 2020.
Well, WTF does "our velocity is increasing" mean? “On the [10 nm] process technology front, our teams executed well in Q1 and our velocity is increasing,”
I think I'd vote for the liar being the guy who chooses to say word salad...
I know it’s management speak but it’s positive management speak. The roadmaps were so conservative that he really didn’t need to say those things. He could have said we’re shipping a few parts and moved on. I take it as some recent good news.
There's no such thing as negative management speak, though - the entire point is to say nothing concrete while sounding positive. Your response is effectively to say "yeah but I want to feel good about this so I made up a reason". That's fine, but it's also completely illogical :)
The WCC article was rumor, this is a release by the company. This is much much better news in general. I am no more interested in seeing Intel nosedive than I am in seeing AMD nosedive. What is happening now is almost perfect, AMD is gobbling up a sizable chunk of traditionally intel product segments (server, enthusiast, etc) and hopefully gaining R&D budget to spend on GPUs to pressure Nvidia. The same goes for Intel. I want to see them succeed with their GPU and pressure Nvidia. Right now Intel is being forced to offer competitive products at more competitive prices. Nvidia now needs that same pressure. Right now in almost all market segments Nvidias only competitor is Nvidia. Thats bad for innovation and bad for customers.
Radeons have only need to "optimize" just like Nv has done since the 600 series, AMD "more or less" have not been ABLE to do such in all honesty, now they are able, if something like an RX680 with optimizations became eventually the GTX 1080, titans and such, AMD did Vega and Radeon 7, but they are NOT "cut to the wick" to get them pure game grunt, Nv did exactly that and the "cost" was for some of the more advanced things that not everyone wants/needs to do anymore, such as "mining"
if Nv could "tame a hot beast, such as RX 480/580/680 than AMD given the same flexibility can very very easily do this same thing, they have lots of experience and many generations to back it up, 4870 "base example, became the 7870 by far one of the best gpu released for the price, power, temps etc.
Anyways, this is a very very rare occurrence for AMD, last time this happened was Athlon days, but, they could not keep up with demand THEN (even if Intel did nothing) unlike now where they can push tends of millions product per quarter when "fully ramped up", seems w/e problems Intel was.is having AMD has been able to dodge such since Ryzen release, and with GPU wise, they can see what they must face, Nv did a slight stumble, if AMD does not do such a stumble with Navi, then likely will be a repeat of 4870 vs 280 days, 280 for more grunt but that much hotter etc or 4870 for near the same play but less cost and power etc.
it can happen, it did for other companies and these days AMD has a massive amount of raw talent with tons of manufacturing space unlike a decade or so back ^.^
Intel have been lying a ridiculous amount about their nodes through the past, well 6 years ago now is when 14nm was supposed to be a year away and on track, then every quarter it got pushed back another quarter, then it finally launched a year late. Throughout that delay they said 10nm had zero issues, was on track and the delays at 14nm don't effect 10nm at all. Then throughout the 4 years since 14nm launched and the 3 or so years since the originally planned 10nm launch they've lied about how ready it is. They were lying right up to the quarter before they launched that 10nm chip that ended up in two products in tiny volume then officially delayed the node an extra 18 months. Yields don't change overnight, they had to know pretty much latest mid 2017 that there was absolutely no chance for desktop or server chips in 2018 but they lied up to and including with the launch.
Intel is not a better source of news in general, they've proven themselves to have effectively no credibility on news on their nodes at all. Until what they start saying is matched by actual events in the real world over a period of time such that they re-establish their credibility, then realise that these are messages to keep the stock prices going because that is their only real goal.
They can get around SEC lawsuits by both keeping the share prices high as they have done through the last 3 years, and by fudging the truth. Hey at that time we totally thought it was on track, but 3 months out from launch when we delayed by another quarter, then again, then again, that was new information.
I personally do want Intel to nosedive, they need less power, we need more AMD marketshare so they have a future and Intel have so much money and influence right now even with Zen being very good AMD's future is still precarious. We could do with a few bad years to level the playing field and let both companies survive long term without either one completely dominating the market and being able to use undue influence to prevent the other succeeding.
AMD fans would rather believe Intel rumors on WCCFtech than real information direct from the source - Intel reaching goals of IceLake / Sunny Cove means that AMD must work to provide another solution - it not just Intel has gone to 10nm - that is small part of IceLake, Sunny cove is major enhancement.
could be old info. plus arent fiscal quarters different than normal quarters? plus the wccftech stuff is rumored. as others say here enterprise timelines are different
Considering that roadmap places Whiskey Lake U - which has been out for quite some time - in Q2 2019, it seems safe to assume that the roadmap isn't accurate.
Wccftech is probably the worst place for technical forecast and they never provide there actual source. The comment are the worst of all - mostly people showing off their rigs and in politics.
I would look else where for your information, but it can be fun to see what they state.
I would disagree with above, they are better than WCCFtech, but still extremely bias toward AMD.
As anybody ever thought Intel could be feeding these rumors to mislead people so that new release is a bigger surprise. They are taking a while to get it right. They have two threats AMD which I believe is the minor threat and ARM which is bigger threat.
Dell VP of Consumer Design: "That was a very, very secretive kind of pre-release. You will hear more about that in the next couple of quarters. I can probably safely say that it will be mid-year.”
Hopefully, this all means high volume by Q4, then.
That dell looks like possibly next generation of XPS 13 2in1 - interesting to see what CPU, maybe even Lakefield - very possibly could be faster than my XPS 13 2in1 with y processor or 10nm Y processor.
Sounds about right. Low-power cores/small this year (easy to make), large/low power cores datacentre 2020 (huuuge margins, genuine competition from AMD), desktop 2021. Smallest market, low margins.
Don't forget that Intel made a lot of their engineers redundant, and used a lot of the people they kept for the now-failed 5G effort. Intel don't have the R&D capacity they used to.
that's my question too. There's speculation that Intel's extremely long reliance on their 14nm process may give them an extra bite in the ass in the sense that 14nm parts are SO mature at this point that the new 10nm process won't be much of an improvement power and performance-wise, at least in its initial incarnation. That's just speculation though- and even if there's truth to it, fitting more transistors in a given space is still a boon for yields as you pointed out as well as benefiting extremely large many-core dies by cramming More Cores into less space
but for IPC you usually need more transistors. Adding more cores, for example, takes die size. AVX512 will take a huge part of next-gen Intel. (and Intel loves putting on new instructions so x86 cant is emulated faster on non-x86 parts like 2002 when G5 was released and the fastest windowsPC was a mac with emulation.
"Will Ice Lake increase IPC of Coffee Lake?" Yes. And it will provide AVX-512.
BUT that's not the interesting question. The interesting question is how MUCH will IPC increase. And that's where the Ice Lake supporters start getting extremely slippery. Every one I've spoken to who insists that this will be Intel's great comeback refuses to even estimate what the IPC increase will be. "Apple has managed an annual IPC increase of at least 15% since the A7. So will Ice Lake offer at least a 15% IPC increase?" ... crickets...
Yes, but if the OEM are made offers that they cannot refuse ? I mean, it would be a shame if an OEM were to be particularly affected by Intel's CPU shortage while working on new Ryzen based systems. These things can happen...
Do you really think OEM's like Dell and HP ( for example ) don't know what coming from Intel? They know and that is why they don't care about Ryzen systems - maybe some on desktop side and maybe some on mobile side.
But think of has OEM owner, why build a product for fans that would rather build there own system then purchase from OEM. Where the main purpose of fans is not for customer for fans to buy there systems but instead fans hoping to get OEM to make systems for other people so they hurt Intel. OEMs are smart and will not fall into this trap.
"As for 10 nm ramp in general, Intel is only talking about producing its relatively small Ice Lake-U processors in volumes this year, which is the company’s typical way of ramping up a new node."
No it's not. On 14nm and 22nm releases, they released a full range of desktop parts first. This strategy was typical as of 65nm and 45nm, but that was over 10 years ago.
I distinctly remember that the first 14nm products were Broadwell Y and Broadwell U in late 2014. Desktop parts didn't become available didn't become available until mid 2015, by which point they were nearly deprecated and superseded by Haswell.
yeah that used to be the case, afaik Broadwell was the first arch to change that with Intel's pivot to aggressively focusing on their extremely low-power (-Y and - U Core) and extremely high-power (Xeon) parts
You ever thought with new design in IceLake, that U Core provides the power of today higher watt CPU and even desktops and that 45watt notebooks are part of past.
Increasing your Throughput by 2x when your initial value is 1 and your original target in Pre 2015 were 10 meant nothing. It will still be a 5x difference
So as usual with the recent Intel, their 10nm words can not be trusted.
This is exactly what I thought when I read the statement: what's the baseline, and what's the target? 2x fuck-all is still fuck-all, and the way he phrased it implies that it's not even 2x.
What about PCIe 4.0? If Intel do not upgrade the PCIe controllers of their Sunny Cove / Ice Lake generation they might be stuck with PCIe 3.0 at least until late 2021 - early 2022. Willow Cove (Tiger Lake) is going to be an optimization of the same design, thus PCIe 4.0 might need to wait for Golden Cove (currently Alder Lake) if it is not added to Sunny Cove. I find that highly unlikely, however it is suspect that Intel have not talked about the PCIe generation and lanes of Ice Lake. Unless they have and I missed it.
With this news, I sense rumors that we will not see a 10nm desktop consumer chip till late 2021 till 2022 may be true. The fact that they are focusing on the ULV series in 2019 means the fab is still not ready to take on bigger chips. Even if they are ready for bigger chips, the yield is likely still significantly below their expectation. Intel's move to release Xeons on 10nm first is likely to try to fend their turf from AMD with their Rome chips.
"Furthermore, Intel believes that it will be able to ship more 10 nm parts than it originally anticipated." We already seen Intel play this game... they shipped 10nm CannonLake previously... Sure if they thought they could only ship 10K part in 2019 and now they can ship 50K that is a lot more, but not significant for the market.
Big details lacking in IceLake-SP Xeon is how many cores will it have and how different from the IceLake-U will it be? Will it just be a low power, slow speed part?
Intel might have raised 10nm processors volume because it can produce more but do computer manufacturers and customers benefit from performance improvement and better price over previous Gen processors ? If not than higher volume does not translates into higher adoptions/sale. 10nm processors has to offer more like better performance at lower power consumption increasing battery life. More integration on SOC of peripheral components to decrease overall cost.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
70 Comments
Back to Article
eva02langley - Thursday, April 25, 2019 - link
I don't understand...https://wccftech.com/intel-desktop-mobile-cpu-road...
ZeDestructor - Thursday, April 25, 2019 - link
One of them is lying, and one of them will have the SEC and shareholders asking some very hard questions if caught lying. Who to believe, eh?Alistair - Thursday, April 25, 2019 - link
No, there is no contradiction. 10nm Ice lake is for mobile only. Wccftech article said no desktop CPUs using 10nm for a long time, not mobile ones.Jorgp2 - Thursday, April 25, 2019 - link
Server CPUs usually come out after desktop CPUs.So we can expect to have them early next year.
Irata - Friday, April 26, 2019 - link
Maybe server is their priority due to Rome- on Desktop, Intel can probably rely more on inertia.psychobriggsy - Friday, April 26, 2019 - link
What do server CPUs have in common with mobile CPUs?Non-leading-edge clock speeds.
It looks like Intel has fixed the yield issue on 10nm, and can make larger dies now (Ice Lake U with 1 TFLOPS GPU, and future Xeons with higher core counts).
But they clearly have not got anywhere with the performance issue (whereby 14nm is still higher performance), which means no 10nm Desktop processors. Straight to 7nm I guess, in 2022?
Santoval - Friday, April 26, 2019 - link
This "leak" appears to be a fabrication. The tables are pretty good but not good enough to be genuine. Besides if Intel wait until... 2022(!) to upgrade the full range of their CPUs to 10nm they are completely toast.HStewart - Thursday, May 2, 2019 - link
I believe an article in AnandTech any day over an article in WCCFtech. I would agree leak in WCCFtech appears to be fabrication because it is on WCCFtech.HStewart - Thursday, May 2, 2019 - link
What if Intel has figured out way to use same chips for desktop as in mobile. Have you ever thought of that.WCCFtech is not a reliable source of technical information - they never provide there source. But they are fun to watch because they jump at news ( real or not ) quickly
Adonisds - Thursday, April 25, 2019 - link
If that's true, then shouldn't Intel be receiving some very hard questions from the SEC already?name99 - Thursday, April 25, 2019 - link
When has the SEC EVER held Intel to account.Compare
https://www.anandtech.com/show/7309/intel-14nm-pro...
Sept 2013 --- Broadwell is on track, ready for the end of the year
and
https://www.extremetech.com/computing/168799-intel...
Oct 2013 --- Oopsie, sorry we're delaying it one quarter.
BTW when did Broadwell ACTUALLY ship?
Oct 2014...
Still waiting for the SEC investigation into that. (Or the acknowledgement by Intel fans that there is a long term pattern of lying that didn't just start with 10nm.)
nevcairiel - Thursday, April 25, 2019 - link
Those "leaks" are based on enterprise timelines, which is slower to adopt new desktop CPUs, so even if those are true and still up to date, it may not necessarily say much about actual 10nm desktop parts being released in 2020.name99 - Thursday, April 25, 2019 - link
Whom to believe?Well, WTF does "our velocity is increasing" mean?
“On the [10 nm] process technology front, our teams executed well in Q1 and our velocity is increasing,”
I think I'd vote for the liar being the guy who chooses to say word salad...
flgt - Thursday, April 25, 2019 - link
I know it’s management speak but it’s positive management speak. The roadmaps were so conservative that he really didn’t need to say those things. He could have said we’re shipping a few parts and moved on. I take it as some recent good news.Spunjji - Friday, April 26, 2019 - link
There's no such thing as negative management speak, though - the entire point is to say nothing concrete while sounding positive. Your response is effectively to say "yeah but I want to feel good about this so I made up a reason". That's fine, but it's also completely illogical :)Rudde - Sunday, April 28, 2019 - link
He explained that their volume going through their factory has doubled in the recent four months. That is a 'velocity' increase.Bp_968 - Thursday, April 25, 2019 - link
The WCC article was rumor, this is a release by the company. This is much much better news in general. I am no more interested in seeing Intel nosedive than I am in seeing AMD nosedive. What is happening now is almost perfect, AMD is gobbling up a sizable chunk of traditionally intel product segments (server, enthusiast, etc) and hopefully gaining R&D budget to spend on GPUs to pressure Nvidia. The same goes for Intel. I want to see them succeed with their GPU and pressure Nvidia. Right now Intel is being forced to offer competitive products at more competitive prices. Nvidia now needs that same pressure. Right now in almost all market segments Nvidias only competitor is Nvidia. Thats bad for innovation and bad for customers.Dragonstongue - Thursday, April 25, 2019 - link
Radeons have only need to "optimize" just like Nv has done since the 600 series, AMD "more or less" have not been ABLE to do such in all honesty, now they are able, if something like an RX680 with optimizations became eventually the GTX 1080, titans and such, AMD did Vega and Radeon 7, but they are NOT "cut to the wick" to get them pure game grunt, Nv did exactly that and the "cost" was for some of the more advanced things that not everyone wants/needs to do anymore, such as "mining"if Nv could "tame a hot beast, such as RX 480/580/680 than AMD given the same flexibility can very very easily do this same thing, they have lots of experience and many generations to back it up, 4870 "base example, became the 7870 by far one of the best gpu released for the price, power, temps etc.
Anyways, this is a very very rare occurrence for AMD, last time this happened was Athlon days, but, they could not keep up with demand THEN (even if Intel did nothing) unlike now where they can push tends of millions product per quarter when "fully ramped up", seems w/e problems Intel was.is having AMD has been able to dodge such since Ryzen release, and with GPU wise, they can see what they must face, Nv did a slight stumble, if AMD does not do such a stumble with Navi, then likely will be a repeat of 4870 vs 280 days, 280 for more grunt but that much hotter etc or 4870 for near the same play but less cost and power etc.
it can happen, it did for other companies and these days AMD has a massive amount of raw talent with tons of manufacturing space unlike a decade or so back ^.^
drunkenmaster - Thursday, April 25, 2019 - link
Intel have been lying a ridiculous amount about their nodes through the past, well 6 years ago now is when 14nm was supposed to be a year away and on track, then every quarter it got pushed back another quarter, then it finally launched a year late. Throughout that delay they said 10nm had zero issues, was on track and the delays at 14nm don't effect 10nm at all. Then throughout the 4 years since 14nm launched and the 3 or so years since the originally planned 10nm launch they've lied about how ready it is. They were lying right up to the quarter before they launched that 10nm chip that ended up in two products in tiny volume then officially delayed the node an extra 18 months. Yields don't change overnight, they had to know pretty much latest mid 2017 that there was absolutely no chance for desktop or server chips in 2018 but they lied up to and including with the launch.Intel is not a better source of news in general, they've proven themselves to have effectively no credibility on news on their nodes at all. Until what they start saying is matched by actual events in the real world over a period of time such that they re-establish their credibility, then realise that these are messages to keep the stock prices going because that is their only real goal.
They can get around SEC lawsuits by both keeping the share prices high as they have done through the last 3 years, and by fudging the truth. Hey at that time we totally thought it was on track, but 3 months out from launch when we delayed by another quarter, then again, then again, that was new information.
I personally do want Intel to nosedive, they need less power, we need more AMD marketshare so they have a future and Intel have so much money and influence right now even with Zen being very good AMD's future is still precarious. We could do with a few bad years to level the playing field and let both companies survive long term without either one completely dominating the market and being able to use undue influence to prevent the other succeeding.
name99 - Thursday, April 25, 2019 - link
"'Throughout that delay they said 10nm had zero issues, was on track and the delays at 14nm don't effect 10nm at all."On the positive side, 7nm is totally on track and the delays at 10nm don't affect 7nm at all :-)
HStewart - Thursday, May 2, 2019 - link
AMD fans would rather believe Intel rumors on WCCFtech than real information direct from the source - Intel reaching goals of IceLake / Sunny Cove means that AMD must work to provide another solution - it not just Intel has gone to 10nm - that is small part of IceLake, Sunny cove is major enhancement.saratoga4 - Thursday, April 25, 2019 - link
Understand what? That leak says limited Icelake mobile this year. Intel says limited Icelake mobile this year. Both are probably correct.peevee - Friday, April 26, 2019 - link
Where do you see Ice Lake in this picture:https://cdn.wccftech.com/wp-content/uploads/2019/0...
bobhumplick - Thursday, April 25, 2019 - link
could be old info. plus arent fiscal quarters different than normal quarters? plus the wccftech stuff is rumored. as others say here enterprise timelines are differentksec - Friday, April 26, 2019 - link
Please, this is Anantech. Sources, News and discussions from WCC does not belong here. They are very much on a whole different level and spectrum.eva02langley - Friday, April 26, 2019 - link
Invalidating the article off the bat because it is from elsewhere is not discrediting it by any means.They were right with the RVII when NOBODY was.
On 10nm, I don't believe a single words from the mouth of Intel.
Valantar - Friday, April 26, 2019 - link
Considering that roadmap places Whiskey Lake U - which has been out for quite some time - in Q2 2019, it seems safe to assume that the roadmap isn't accurate.HStewart - Friday, April 26, 2019 - link
Wccftech is probably the worst place for technical forecast and they never provide there actual source. The comment are the worst of all - mostly people showing off their rigs and in politics.I would look else where for your information, but it can be fun to see what they state.
Rudde - Sunday, April 28, 2019 - link
If you want to read into leaks. Below is a more reasonable timeline.https://www.extremetech.com/computing/290298-intel...
HStewart - Thursday, May 2, 2019 - link
I would disagree with above, they are better than WCCFtech, but still extremely bias toward AMD.As anybody ever thought Intel could be feeding these rumors to mislead people so that new release is a bigger surprise. They are taking a while to get it right. They have two threats AMD which I believe is the minor threat and ARM which is bigger threat.
ikjadoon - Thursday, April 25, 2019 - link
Ice Lake, however, is paper launching in late Q2 (i.e., June 2019 = Computex), as Lenovo and Dell leaked out.https://www.trustedreviews.com/news/new-dell-xps-i...
https://www.reddit.com/r/thinkpad/comments/astmup/...
Dell VP of Consumer Design: "That was a very, very secretive kind of pre-release. You will hear more about that in the next couple of quarters. I can probably safely say that it will be mid-year.”
Hopefully, this all means high volume by Q4, then.
peevee - Friday, April 26, 2019 - link
Still "limited" on Intel's roadmap:https://cdn.wccftech.com/wp-content/uploads/2019/0...
Means they are still losing money on each CPU and don't want to lose more.
HStewart - Friday, April 26, 2019 - link
limited means "Limited realistic information"After all, this is wccftech link - I was fooled about there information before and never again.
HStewart - Thursday, May 2, 2019 - link
That dell looks like possibly next generation of XPS 13 2in1 - interesting to see what CPU, maybe even Lakefield - very possibly could be faster than my XPS 13 2in1 with y processor or 10nm Y processor.MDD1963 - Thursday, April 25, 2019 - link
I'm hearing (Youtube Hotnews/UFD Tech) there will be no desktop 10nm until late ...2021? That is simply pathetic...Ian Cutress - Thursday, April 25, 2019 - link
Ask if they second sourced their information.Jorgp2 - Thursday, April 25, 2019 - link
Their first and second sources are themselves.Irata - Friday, April 26, 2019 - link
How about a Bunny Suit bet with Charlie as you two do not seem to share the same outlook on Intel 10nm ?Alternatively, a Borat Mankini bet...well, no, let's stick with the Bunny Suit in everyone's interest :)
tipoo - Sunday, April 28, 2019 - link
Times like these I wish I could upvoteMeteor2 - Sunday, April 28, 2019 - link
Sounds about right. Low-power cores/small this year (easy to make), large/low power cores datacentre 2020 (huuuge margins, genuine competition from AMD), desktop 2021. Smallest market, low margins.Don't forget that Intel made a lot of their engineers redundant, and used a lot of the people they kept for the now-failed 5G effort. Intel don't have the R&D capacity they used to.
Hulk - Thursday, April 25, 2019 - link
Will Ice Lake increase IPC of Coffee Lake?Will power be reduced at the same frequency as a comparably clocked Coffee Lake part?
Or is this the point of 10nm to fit more parts on a wafer and increase profit?
KateH - Thursday, April 25, 2019 - link
that's my question too. There's speculation that Intel's extremely long reliance on their 14nm process may give them an extra bite in the ass in the sense that 14nm parts are SO mature at this point that the new 10nm process won't be much of an improvement power and performance-wise, at least in its initial incarnation.That's just speculation though- and even if there's truth to it, fitting more transistors in a given space is still a boon for yields as you pointed out as well as benefiting extremely large many-core dies by cramming More Cores into less space
saratoga4 - Thursday, April 25, 2019 - link
>Will Ice Lake increase IPC of Coffee Lake?Yes.
>Will power be reduced at the same frequency as a comparably clocked Coffee Lake part?
Yes.
>Or is this the point of 10nm to fit more parts on a wafer and increase profit?
Yes.
Meteor2 - Sunday, April 28, 2019 - link
We won't know the answers until review units arrive, but for the reasons KateH gave the likely answers are no.nevcairiel - Thursday, April 25, 2019 - link
The beauty of a new process node is that it can basically achieve both reduced power and smaller dies, covering 2 of yours already.And for IPC, you should check out the tech preview of Ice Lake/Sunny Cove that was on AT a few months ago.
shompa - Friday, April 26, 2019 - link
but for IPC you usually need more transistors. Adding more cores, for example, takes die size. AVX512 will take a huge part of next-gen Intel. (and Intel loves putting on new instructions so x86 cant is emulated faster on non-x86 parts like 2002 when G5 was released and the fastest windowsPC was a mac with emulation.name99 - Thursday, April 25, 2019 - link
"Will Ice Lake increase IPC of Coffee Lake?"Yes. And it will provide AVX-512.
BUT that's not the interesting question. The interesting question is how MUCH will IPC increase. And that's where the Ice Lake supporters start getting extremely slippery. Every one I've spoken to who insists that this will be Intel's great comeback refuses to even estimate what the IPC increase will be.
"Apple has managed an annual IPC increase of at least 15% since the A7. So will Ice Lake offer at least a 15% IPC increase?" ... crickets...
nandnandnand - Thursday, April 25, 2019 - link
Ice Lake supporters? All I've heard of is Zen 2 supporters. Alternatively, Zen 3.bug77 - Thursday, April 25, 2019 - link
Hm, we're talking 2020 here and no desktop part? Zen2 will have a field day.Irata - Friday, April 26, 2019 - link
If OEM decide to offer good systems with it, then yes.Oh, and also actively market them and not hide them on their own web sites.
bug77 - Friday, April 26, 2019 - link
I believe it's AMD's job to get the OEM interested and involved. And they're way better at this then they were back in AthlonXP/64 days.Irata - Friday, April 26, 2019 - link
Yes, but if the OEM are made offers that they cannot refuse ? I mean, it would be a shame if an OEM were to be particularly affected by Intel's CPU shortage while working on new Ryzen based systems. These things can happen...bug77 - Saturday, April 27, 2019 - link
As I said, that game is between AMD and OEMs ;)HStewart - Thursday, May 2, 2019 - link
Do you really think OEM's like Dell and HP ( for example ) don't know what coming from Intel? They know and that is why they don't care about Ryzen systems - maybe some on desktop side and maybe some on mobile side.But think of has OEM owner, why build a product for fans that would rather build there own system then purchase from OEM. Where the main purpose of fans is not for customer for fans to buy there systems but instead fans hoping to get OEM to make systems for other people so they hurt Intel. OEMs are smart and will not fall into this trap.
HStewart - Thursday, May 2, 2019 - link
Field on extreme minority of market is not a big deal.dew111 - Thursday, April 25, 2019 - link
"As for 10 nm ramp in general, Intel is only talking about producing its relatively small Ice Lake-U processors in volumes this year, which is the company’s typical way of ramping up a new node."No it's not. On 14nm and 22nm releases, they released a full range of desktop parts first. This strategy was typical as of 65nm and 45nm, but that was over 10 years ago.
dew111 - Thursday, April 25, 2019 - link
Nevermind, I am wrong.psychobriggsy - Friday, April 26, 2019 - link
It was usually only 6 months from U/Y series chip (basically a chip to let Intel crow that it has reached Xnm) to full desktop chips however.Desktop 10nm here is not happening.
10nm is for mobile and Xeon - neither need ~5GHz capable CPUs. That indicates where the problem is.
aryonoco - Thursday, April 25, 2019 - link
I distinctly remember that the first 14nm products were Broadwell Y and Broadwell U in late 2014. Desktop parts didn't become available didn't become available until mid 2015, by which point they were nearly deprecated and superseded by Haswell.Klimax - Friday, April 26, 2019 - link
You mean by Skylake. Broadwell was after Haswell.KateH - Thursday, April 25, 2019 - link
yeah that used to be the case, afaik Broadwell was the first arch to change that with Intel's pivot to aggressively focusing on their extremely low-power (-Y and - U Core) and extremely high-power (Xeon) partsHStewart - Thursday, May 2, 2019 - link
You ever thought with new design in IceLake, that U Core provides the power of today higher watt CPU and even desktops and that 45watt notebooks are part of past.ksec - Friday, April 26, 2019 - link
Increasing your Throughput by 2x when your initial value is 1 and your original target in Pre 2015 were 10 meant nothing. It will still be a 5x differenceSo as usual with the recent Intel, their 10nm words can not be trusted.
Spunjji - Friday, April 26, 2019 - link
This is exactly what I thought when I read the statement: what's the baseline, and what's the target? 2x fuck-all is still fuck-all, and the way he phrased it implies that it's not even 2x.Santoval - Friday, April 26, 2019 - link
What about PCIe 4.0? If Intel do not upgrade the PCIe controllers of their Sunny Cove / Ice Lake generation they might be stuck with PCIe 3.0 at least until late 2021 - early 2022. Willow Cove (Tiger Lake) is going to be an optimization of the same design, thus PCIe 4.0 might need to wait for Golden Cove (currently Alder Lake) if it is not added to Sunny Cove. I find that highly unlikely, however it is suspect that Intel have not talked about the PCIe generation and lanes of Ice Lake. Unless they have and I missed it.YoloPascual - Sunday, April 28, 2019 - link
I still remember that day when intel told us that we will have 10nm cpus by 2015..ajp_anton - Tuesday, April 30, 2019 - link
"Intel believes that it will be able to ship more 10 nm parts than it originally anticipated."So how many did they originally plan to ship 10nm parts in 2016?
watzupken - Wednesday, May 1, 2019 - link
With this news, I sense rumors that we will not see a 10nm desktop consumer chip till late 2021 till 2022 may be true. The fact that they are focusing on the ULV series in 2019 means the fab is still not ready to take on bigger chips. Even if they are ready for bigger chips, the yield is likely still significantly below their expectation. Intel's move to release Xeons on 10nm first is likely to try to fend their turf from AMD with their Rome chips.Freeb!rd - Saturday, May 4, 2019 - link
"Furthermore, Intel believes that it will be able to ship more 10 nm parts than it originally anticipated." We already seen Intel play this game... they shipped 10nm CannonLake previously... Sure if they thought they could only ship 10K part in 2019 and now they can ship 50K that is a lot more, but not significant for the market.Big details lacking in IceLake-SP Xeon is how many cores will it have and how different from the IceLake-U will it be? Will it just be a low power, slow speed part?
Rash1 - Tuesday, May 7, 2019 - link
Intel might have raised 10nm processors volume because it can produce more but do computer manufacturers and customers benefit from performance improvement and better price over previous Gen processors ? If not than higher volume does not translates into higher adoptions/sale.10nm processors has to offer more like better performance at lower power consumption increasing battery life. More integration on SOC of peripheral components to decrease overall cost.