The question is how much further can they go on binning and improving their 14nm process. Maybe they can deactivate part of the cores to get one to boost to 5 Ghz while staying reasonably within the power envelope, but I doubt this will suffice.
One core at idle speeds consumes very little power on my 6700HQ, under 0.15W. So the problem is the core that goes to 5Ghz, not the other ones. I think they can extract a bit more from it, but nothing spectacular, just enough to stay afloat. Another thing that we can learn from this very long period on 14nm is that throw away a bit too quickly very good processes that can be improved in time. Just compare how 5775C behaved and its limitations compared to what Intel has today.
That's already the case with the clocks of Whiskey, Amber and Comet Lake vs those of Ice Lake-U/Y. Intel canned the desktop (both -S and -H) CPUs of Ice Lake because their yields are horrible at desktop class clocks, and apparently will also can the Tiger Lake ones, replacing them with Rocket Lake-S/H. Rocket Lake will have Willow Cores and an Xe iGPU but will be fabbed at 14nm+++++ It will have a much higher clock than what Intel can reach with *any* 10nm variant (10nm, 10nm+ *and* 10nm++) with one important trade-off : a low power efficient and thus a very high TDP, much higher than the equivalent Ryzen CPU. That's the cost Intel will pay for screwing up fundamentally their 10nm node.
I think Intel has missed an opportunity by keeping the systems with eDRAM so selective. Think of how fast a 9900KS would be if it has 256 MB of eDRAM in the same package.
In fact, Intel hasn't been keen on increasing normal L2 and L3 cache sizes on their consumer chips. Granted that increases die size and thus costs which the consumer market margins are sensitive too. However, with Intel still struggling to get 10 nm out the door in high volume, that would have been a compromise to improve IPC while still on 14 nm.
I'm not certain, but I *think* the issue there was that the additional cost per processor didn't really map well to the performance boost from the eDRAM. In other words, performance per dollar did not benefit enough.
I still wonder what they could have got from 32nm with more tweaking - that had a particularly short life compared to 22nm and 14nm, yet the CPUs built on it could clock extremely well. They'd obviously lose out in terms of cache size and complexity compared to more recent CPUs, though, so perhaps it would just have been a wash.
32nm was a breakthrough for Intel though. Remember they were still on tick tock back then and Sandy Bridge retired 32nm production with a home run.
Early 22nm parts based on Ivy Bridge were also, in my experience, not as reliable, either. Obviously the process matured by the time Haswell hit, arguably Intel’s most successful architecture of the decade, but I’ve pulled so many failed 3rd gen i3 and i5 CPU’s from systems it almost seems there was a problem somewhere - most of them had malfunctioning GPU’s to the point if the IGFX drivers were installed the system would just BSOD.
Are we just giving them ideas now on how to scam their customers? Because if an 8 core processor deactivates cores and becomes a single core from 15 years ago, just so it can push 100-200MHz more, that in my opinion is a scam.
Has AMD got one that out performs Intel in laptops yet? Last I read on this site was that the Intel laptop trounced the AMD one in the same system. So, obviously Intel is building/opitimizing CPUs that relate to real world performances for consumers vs. arbitrary benchmarks as you imply.
Hopefully the performance matches what they presented during the keynote, because if it delivers, the 4800H would be a really impressive jump in performance for Ryzen mobile.
Hmm...so Intel is announcing new 5 Ghz mobile CPU (on 14 nm) without too many specifics, just that they will be available "soon".
"To be honest, if we’re so close to launching the product, I’m surprised as to why Intel isn’t disclosing."
Could this perhaps be that this is something similar to Intel demoing their 28 core 5 Ghz CPU prior to Ryzen's announcement ?
I would not be surprised if AMD were to announce their new Ryzen mobile CPU line up that also includes 8C/16T models and this is Intel's attempt at saying "no reason to get those, we'll have something available shortly that will be much faster."
The actual news is that they still can't manufacture any 45W CPUs using their latest core designs and the 10nm process node, so they're pushing Comet Lake - which is re-warmed Coffee Lake - which was bulked-up Kaby Lake - which was re-warmed Skylake... :D
Their "45W" i9 CPUs are already nothing of the sort; this 5Ghz nonsense will just make the gulf between their promises and reality even wider. It's a shame because their mobile CPUs are still objectively decent, and they'd likely be received better if you didn't have to read a product review just to find out whether there's a meaningful performance difference between an i5, i7 or i9 version of the same notebook.
And yet..."AMD’s Picasso platform, featuring its Zen+ cores and coupled with a Vega iGPU, has been a tremendous improvement for AMD. But Intel’s Ice Lake platform runs circles around it." https://www.anandtech.com/show/15213/the-microsoft... So, what Gen would you say AMD is on then???
Of course it did, it was a comparison between AMD's old Zen+ compared to Intel's high end Core i7-1065G7 Ice Lake cpu. Coffee Lake & Zen+ had comparable IPC, and both Ice Lake & Zen 2 both had similar IPC increases over their predecessors. If what AMD presented yesterday holds true, Ice Lake's glory will be short lived.
”To be honest, if we’re so close to launching the product, I’m surprised as to why Intel isn’t disclosing.”
Discrepancy between what they would like to announce for this generation to have any relevance, and what’s practical from a yield/volume perspective?
You’ve got to hand it to Intel, they’ve done magic with what they’ve got, being stuck on 14nm for so long, but they really got to be at a point now where they’ve pretty much pulled all their tricks out of their sleeve. LPDDR4 support is probably the most exiting thing coming this gen, and that’s saying something.
If Intel's are BS then what does that make AMD's that aren't even close? https://www.anandtech.com/show/15213/the-microsoft... "AMD’s Picasso platform, featuring its Zen+ cores and coupled with a Vega iGPU, has been a tremendous improvement for AMD. But Intel’s Ice Lake platform runs circles around it."
Intel's mobile offering, today, is noticeably superior to AMD's: just review the Surface 3 comparison. It's less than, say, 3 years ago, but still significant. The 4*** gen should fill the gap. I am hoping that it would bring a similar situation as on desktops (where AMD wins hands down today): should be good for consumers to have finally some competition on the laptops. I have some A10 laptops that I would like to replace: looking forward to the comparison.
Ice Lake's victory looks like it will be short lived if the 4000 series is as good as AMD claimed. I'm looking forward to seeing what kind of performance the 4800H is able to achieve in well-engineered laptop.
It is very clear, from their marketing they want to dominate the PR space with Intel still has so many great product in the pipeline, you might as well wait to buy from Intel rather than AMD.
I hope the Lisa Su's CES talk will reveal more information on Zen 3. At the point in time Intel does not deserve to enjoy any of its current success.
Yeah i totally agree. After years of doing the bare minimum there have been tons of Intel 'future' articles on here recently. Its basically a poor attempt to stay relevant and keep in people's minds.
Today article for you, "AMD’s Picasso platform, featuring its Zen+ cores and coupled with a Vega iGPU, has been a tremendous improvement for AMD. But Intel’s Ice Lake platform runs circles around it." https://www.anandtech.com/show/15213/the-microsoft...
LPDDR4X is currently being sold in 3733 MHz configurations with 10th gen processors, which is much faster then most laptop DDR4 setups. Why not pair it with a 45 watt CPU, you might, you know, want good battery life when not hammering the system.
LPDDRxx is only meaningfully reducing standby time (i.e., in sleep mode), where a few miliwatts matter. I don't think miliwatts are being optimized in a 45 W CPU system.
I thought only CML-U was getting LPDDR4 support, not H/S. CML-H should get the 400 Series ICP-H chipset, and I would expect it in the 16” MBP eventually, but Apple always uses Broadcom based 3x3:3 Wi-Fi solutions, never Intel.
I also thought we were looking at a Jan launch for CML-S 10+2 and Apr launch for CML-H 10+2 / U 6+2 LP4x.
Thermal Velocity Boost is a complete wank on the mobile parts it was first introduced in, designed to assist peak marketing in representing the processor's max turbo as 200MHz higher than it really is. Laptop manufacturers generally deprioritise fan noise at idle so idle CPU temperatures in the 40-50C range are extremely common, so how long do you think they stay under the 50C threshhold in which TVB adds its 200MHz...
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.
Facebook
Twitter
RSS
57 Comments
Back to Article
Chaitanya - Monday, January 6, 2020 - link
Another year of suffering from Blue Giant.nandnandnand - Monday, January 6, 2020 - link
Today is AMD's day, and 2020 is AMD's year.Chaitanya - Monday, January 6, 2020 - link
I really hope AMD can make meaningful inroads into HPC/Server/DC market which will really put Intel under pressure.prisonerX - Monday, January 6, 2020 - link
It already has.eSyr - Monday, January 6, 2020 - link
Why? IBM is doing relatively well.tipoo - Monday, January 6, 2020 - link
Not really, a slow decline buoyed by buybacks.yannigr2 - Monday, January 6, 2020 - link
5GHz with the 45W part. Like the 95W TDP 9900K that can go up to 210W based on Intel's internal document. Or the 125W 10900K that can go up to 250W.Irata - Monday, January 6, 2020 - link
The question is how much further can they go on binning and improving their 14nm process.Maybe they can deactivate part of the cores to get one to boost to 5 Ghz while staying reasonably within the power envelope, but I doubt this will suffice.
yeeeeman - Monday, January 6, 2020 - link
One core at idle speeds consumes very little power on my 6700HQ, under 0.15W. So the problem is the core that goes to 5Ghz, not the other ones. I think they can extract a bit more from it, but nothing spectacular, just enough to stay afloat.Another thing that we can learn from this very long period on 14nm is that throw away a bit too quickly very good processes that can be improved in time. Just compare how 5775C behaved and its limitations compared to what Intel has today.
Irata - Monday, January 6, 2020 - link
I think it's pretty amazing what Intel got out of their 14nm process.The danger here is that a very highly optimized mature process may make the following one(s) that are not yet optimized look bad in comparison.
Korguz - Monday, January 6, 2020 - link
um... too late....Santoval - Monday, January 6, 2020 - link
That's already the case with the clocks of Whiskey, Amber and Comet Lake vs those of Ice Lake-U/Y. Intel canned the desktop (both -S and -H) CPUs of Ice Lake because their yields are horrible at desktop class clocks, and apparently will also can the Tiger Lake ones, replacing them with Rocket Lake-S/H.Rocket Lake will have Willow Cores and an Xe iGPU but will be fabbed at 14nm+++++ It will have a much higher clock than what Intel can reach with *any* 10nm variant (10nm, 10nm+ *and* 10nm++) with one important trade-off : a low power efficient and thus a very high TDP, much higher than the equivalent Ryzen CPU. That's the cost Intel will pay for screwing up fundamentally their 10nm node.
Santoval - Monday, January 6, 2020 - link
edit : "..a low power *efficiency*..."Kevin G - Monday, January 6, 2020 - link
I think Intel has missed an opportunity by keeping the systems with eDRAM so selective. Think of how fast a 9900KS would be if it has 256 MB of eDRAM in the same package.In fact, Intel hasn't been keen on increasing normal L2 and L3 cache sizes on their consumer chips. Granted that increases die size and thus costs which the consumer market margins are sensitive too. However, with Intel still struggling to get 10 nm out the door in high volume, that would have been a compromise to improve IPC while still on 14 nm.
Spunjji - Monday, January 6, 2020 - link
I'm not certain, but I *think* the issue there was that the additional cost per processor didn't really map well to the performance boost from the eDRAM. In other words, performance per dollar did not benefit enough.Spunjji - Monday, January 6, 2020 - link
I still wonder what they could have got from 32nm with more tweaking - that had a particularly short life compared to 22nm and 14nm, yet the CPUs built on it could clock extremely well. They'd obviously lose out in terms of cache size and complexity compared to more recent CPUs, though, so perhaps it would just have been a wash.Samus - Friday, January 10, 2020 - link
32nm was a breakthrough for Intel though. Remember they were still on tick tock back then and Sandy Bridge retired 32nm production with a home run.Early 22nm parts based on Ivy Bridge were also, in my experience, not as reliable, either. Obviously the process matured by the time Haswell hit, arguably Intel’s most successful architecture of the decade, but I’ve pulled so many failed 3rd gen i3 and i5 CPU’s from systems it almost seems there was a problem somewhere - most of them had malfunctioning GPU’s to the point if the IGFX drivers were installed the system would just BSOD.
yannigr2 - Monday, January 6, 2020 - link
Are we just giving them ideas now on how to scam their customers? Because if an 8 core processor deactivates cores and becomes a single core from 15 years ago, just so it can push 100-200MHz more, that in my opinion is a scam.ironargonaut - Monday, January 6, 2020 - link
Has AMD got one that out performs Intel in laptops yet? Last I read on this site was that the Intel laptop trounced the AMD one in the same system. So, obviously Intel is building/opitimizing CPUs that relate to real world performances for consumers vs. arbitrary benchmarks as you imply.Trikkiedikkie - Tuesday, January 7, 2020 - link
The new ones are coming soon.Asus has announced a new line today, with one processsor exclusively for them
sarafino - Tuesday, January 7, 2020 - link
Hopefully the performance matches what they presented during the keynote, because if it delivers, the 4800H would be a really impressive jump in performance for Ryzen mobile.Irata - Monday, January 6, 2020 - link
Hmm...so Intel is announcing new 5 Ghz mobile CPU (on 14 nm) without too many specifics, just that they will be available "soon"."To be honest, if we’re so close to launching the product, I’m surprised as to why Intel isn’t disclosing."
Could this perhaps be that this is something similar to Intel demoing their 28 core 5 Ghz CPU prior to Ryzen's announcement ?
I would not be surprised if AMD were to announce their new Ryzen mobile CPU line up that also includes 8C/16T models and this is Intel's attempt at saying "no reason to get those, we'll have something available shortly that will be much faster."
Irata - Monday, January 6, 2020 - link
*prior to Threadripper 2's presentationSpunjji - Monday, January 6, 2020 - link
You're pretty much on the money, there.The actual news is that they still can't manufacture any 45W CPUs using their latest core designs and the 10nm process node, so they're pushing Comet Lake - which is re-warmed Coffee Lake - which was bulked-up Kaby Lake - which was re-warmed Skylake... :D
Their "45W" i9 CPUs are already nothing of the sort; this 5Ghz nonsense will just make the gulf between their promises and reality even wider. It's a shame because their mobile CPUs are still objectively decent, and they'd likely be received better if you didn't have to read a product review just to find out whether there's a meaningful performance difference between an i5, i7 or i9 version of the same notebook.
tamalero - Monday, January 6, 2020 - link
I was wondering about this.How long will intel BS with "demos" that are nothing similar to the finished product?
Irata - Monday, January 6, 2020 - link
As long as it works.69369369 - Monday, January 6, 2020 - link
10th gen? More like 8.5 gen at best.Spunjji - Monday, January 6, 2020 - link
6.5th Gen, even! Although we could be generous and say 8.5th if they finally admit that there were 2 generations of Core before "1st Gen" Core. :Dironargonaut - Monday, January 6, 2020 - link
And yet..."AMD’s Picasso platform, featuring its Zen+ cores and coupled with a Vega iGPU, has been a tremendous improvement for AMD. But Intel’s Ice Lake platform runs circles around it."https://www.anandtech.com/show/15213/the-microsoft...
So, what Gen would you say AMD is on then???
Lord of the Bored - Tuesday, January 7, 2020 - link
AMD is Gen X.sarafino - Tuesday, January 7, 2020 - link
Of course it did, it was a comparison between AMD's old Zen+ compared to Intel's high end Core i7-1065G7 Ice Lake cpu. Coffee Lake & Zen+ had comparable IPC, and both Ice Lake & Zen 2 both had similar IPC increases over their predecessors. If what AMD presented yesterday holds true, Ice Lake's glory will be short lived.prisonerX - Monday, January 6, 2020 - link
It's 10th gen 14nm, aka 14++++++++++.MarcusMo - Monday, January 6, 2020 - link
”To be honest, if we’re so close to launching the product, I’m surprised as to why Intel isn’t disclosing.”Discrepancy between what they would like to announce for this generation to have any relevance, and what’s practical from a yield/volume perspective?
You’ve got to hand it to Intel, they’ve done magic with what they’ve got, being stuck on 14nm for so long, but they really got to be at a point now where they’ve pretty much pulled all their tricks out of their sleeve. LPDDR4 support is probably the most exiting thing coming this gen, and that’s saying something.
prisonerX - Monday, January 6, 2020 - link
"Magic" as in marketing sleight of hand. Intel products are about 80% bullshit these days.ironargonaut - Monday, January 6, 2020 - link
If Intel's are BS then what does that make AMD's that aren't even close?https://www.anandtech.com/show/15213/the-microsoft...
"AMD’s Picasso platform, featuring its Zen+ cores and coupled with a Vega iGPU, has been a tremendous improvement for AMD. But Intel’s Ice Lake platform runs circles around it."
sarafino - Tuesday, January 7, 2020 - link
Do you just keep copy and pasting the same rant over and over?rahvin - Monday, January 6, 2020 - link
Because they've been promising 10nm chips were going to ship soon since 2016! I'll believe it when I see it.sarafino - Tuesday, January 7, 2020 - link
They have pushed 14nm much further than I would have expected. Quite refined.yankeeDDL - Monday, January 6, 2020 - link
Intel's mobile offering, today, is noticeably superior to AMD's: just review the Surface 3 comparison. It's less than, say, 3 years ago, but still significant.The 4*** gen should fill the gap. I am hoping that it would bring a similar situation as on desktops (where AMD wins hands down today): should be good for consumers to have finally some competition on the laptops. I have some A10 laptops that I would like to replace: looking forward to the comparison.
sarafino - Tuesday, January 7, 2020 - link
Ice Lake's victory looks like it will be short lived if the 4000 series is as good as AMD claimed. I'm looking forward to seeing what kind of performance the 4800H is able to achieve in well-engineered laptop.ksec - Monday, January 6, 2020 - link
>Intel also briefed us......It is very clear, from their marketing they want to dominate the PR space with Intel still has so many great product in the pipeline, you might as well wait to buy from Intel rather than AMD.
I hope the Lisa Su's CES talk will reveal more information on Zen 3. At the point in time Intel does not deserve to enjoy any of its current success.
DannyH246 - Monday, January 6, 2020 - link
Yeah i totally agree. After years of doing the bare minimum there have been tons of Intel 'future' articles on here recently. Its basically a poor attempt to stay relevant and keep in people's minds.ironargonaut - Monday, January 6, 2020 - link
Today article for you,"AMD’s Picasso platform, featuring its Zen+ cores and coupled with a Vega iGPU, has been a tremendous improvement for AMD. But Intel’s Ice Lake platform runs circles around it."
https://www.anandtech.com/show/15213/the-microsoft...
scineram - Wednesday, January 8, 2020 - link
A year old article, liar.Haawser - Monday, January 6, 2020 - link
So is this 45W 'Base TDP' or 45W actual max TDP ? Because they're not even close to being the same thing.Ian Cutress - Monday, January 6, 2020 - link
Intel always quotes the TDP as the sustained power draw. It goes without saying these days. Turbo power has the name 'PL2'vFunct - Monday, January 6, 2020 - link
These should have LPDDR4 & WiFi6 support, right?I'd expect these in MacBook Pro 16".
e36Jeff - Monday, January 6, 2020 - link
LPDDR4 is all about low power, you probably wouldn't bother with pairing it to a 45W CPU.TheinsanegamerN - Monday, January 6, 2020 - link
LPDDR4X is currently being sold in 3733 MHz configurations with 10th gen processors, which is much faster then most laptop DDR4 setups. Why not pair it with a 45 watt CPU, you might, you know, want good battery life when not hammering the system.Spunjji - Monday, January 6, 2020 - link
It has higher latency, and the high RAM clock-speeds don't make much difference to systems without an iGPU doing heavy lifting.It might make some sense if these processors had a proper 10th gen GPU, but they don't.
ikjadoon - Monday, January 6, 2020 - link
LPDDRxx is only meaningfully reducing standby time (i.e., in sleep mode), where a few miliwatts matter. I don't think miliwatts are being optimized in a 45 W CPU system.repoman27 - Monday, January 6, 2020 - link
I thought only CML-U was getting LPDDR4 support, not H/S. CML-H should get the 400 Series ICP-H chipset, and I would expect it in the 16” MBP eventually, but Apple always uses Broadcom based 3x3:3 Wi-Fi solutions, never Intel.I also thought we were looking at a Jan launch for CML-S 10+2 and Apr launch for CML-H 10+2 / U 6+2 LP4x.
Retycint - Monday, January 6, 2020 - link
Wouldn't count on it; even the 10th Gen U processors (Comet Lake) doesn't support LPDDR4, and that is a platform aimed at low-power usage.Silma - Monday, January 6, 2020 - link
So fake 10th Gen 45W mobile CPUs.Spunjji - Monday, January 6, 2020 - link
6th Gen ++++bennyg - Tuesday, January 7, 2020 - link
Skylake++ with a 4+2+2+2 design on 14nm++Process Architecture Optimisation Optimisation Optimisation Optimisation Optimisation Optimisation Optimisation Optimisation Optimisation Optimisation Optimisation Optimisation
bennyg - Tuesday, January 7, 2020 - link
Thermal Velocity Boost is a complete wank on the mobile parts it was first introduced in, designed to assist peak marketing in representing the processor's max turbo as 200MHz higher than it really is. Laptop manufacturers generally deprioritise fan noise at idle so idle CPU temperatures in the 40-50C range are extremely common, so how long do you think they stay under the 50C threshhold in which TVB adds its 200MHz...It was