Intel Ice Lake (10th gen/10nm) - desktop processors not until 2020?

grambo

[H]ard|Gawd
Joined
Apr 10, 2011
Messages
1,231
I know it's mostly rumours and speculation at this point, but I think Intel themselves have made some statements that Ice Lake desktop/enthusiast parts (i.e. 6C/12T+) will not be out until mid to late 2020 as they are focusing on mobile first. Is that accurate based on what we know today)? Debating when to upgrade my rig (see below), and if there's no chance of desktop Ice Lake for another year then I will make a decision based on Ryzen 3800X/3900X vs. 9700K/9900K and may as well upgrade in the next few months once independent benchmarks for Zen 2 are released.

I'm still running a 3770K rig I built in 2012, recently upgraded to a 1080 Ti from a 970 to go with my new 1440p 144Hz display. I know I'm CPU bottlenecked which is why I am interested in upgrading. I'm a huge CD Projekt Red fan and my main goal is to be able to play Cyberpunk 2077 maxed out, will likely sell my 1080 Ti and get a 2080 Ti/Super so I can enjoy the ray tracing in that game. But again, if Ice Lake isn't coming until after Cyberpunk is released then I do not want to wait it out.
 
If you do not have DDR4 at this point, you might want to hold off and just get the best 3770k overclock until Icelake as it might be DDR5 if it is not being released until 2020.

Otherwise, just get a 3700x/9700k, pick up a descent X470/Z370 mb with some cheap DDR4 and call it a day.
 
The mobile parts will be out in the next month or two- they've been released as have major laptop models using them.

The desktop parts will come later, probably on the abovementioned timeline.

For something like Cyberpunk, maxing out single-core speeds at the expense of cores doesn't make much sense. Grab the fastest-clocking Ryzen 2 you can get.
 
Thanks guys, you are confirming my thinking. I wasn't aware of Icelake and potential DDR5, but with DDR4 prices being what I would consider affordable ($200CAD for 32GB DDR4-3200 Corsair Vengeance) I'm not fussed. I had seen mobile parts trickling out but it seems like desktop is really far off.

3900X/9900K is massive overkill I know, and I should go with 3700/3800X or 9700K but I upgrade so infrequently that I might spend a few hundred extra and just go balls out.

Doing some research to determine if there's any reason to upgrade my case which I love (Corsiar 550D), PSU (Seasonic X-760 - awesome but 7 years old now) etc. Will definitely want NVMe SSD and my hard drives are old enough I should grab a new one for long-term storage of media files.
 
To be clear, I have no idea if Icelake will be DDR5, but it just seems that DDR4 will be on it's way out by that time.
 
The first bench results were not that impressive. Clocks down on 8th and 9th gen I know why they chose 6th gen to base it off. In reality it will barely be worthwhile if you are currently on 8/9th gen
 
Man I am so sick of the 9700k recommendations. The 8700k is a better chip for the average consumer who uses windows and games.
 
The first bench results were not that impressive. Clocks down on 8th and 9th gen I know why they chose 6th gen to base it off. In reality it will barely be worthwhile if you are currently on 8/9th gen

Well, the first benchmark results were in ultrabooks with hard TDP limits, so lower clockspeeds at the same performance (what Notebookcheck reported) means longer battery life, lower heat, and lower noise.

Which is what drives the thin and light market.

For desktop, when Intel gets around to it, that means more performance at the same power draw.
 
Well, the first benchmark results were in ultrabooks with hard TDP limits, so lower clockspeeds at the same performance (what Notebookcheck reported) means longer battery life, lower heat, and lower noise.

Which is what drives the thin and light market.

For desktop, when Intel gets around to it, that means more performance at the same power draw.

I thought I came across some analysis that attempted to extrapolate Ice Lake single core performance from those 15w parts up to what desktop might look like clock for clock but cannot find it now (I think they compared to 8th/9th gen 15w parts clock for clock on some benchmarks and there was some decent IPC gains). All speculation at this point though.
 
I thought I came across some analysis that attempted to extrapolate Ice Lake single core performance from those 15w parts up to what desktop might look like clock for clock but cannot find it now (I think they compared to 8th/9th gen 15w parts clock for clock on some benchmarks and there was some decent IPC gains). All speculation at this point though.

I would hesitate to make any more out of the benchmarks than what's been presented. That's true for all benchmarks, but there's a large degree of uncertainty with these given that patch levels for security issues are not reported with specificity.

This adds a whole new level of balancing attack surface with utility.
 
If you are not running VMs it is pretty much a moot point.

I would have said that a year ago- but so many developers including MS themselves are using virtualization instructions to better protect users, and these approaches are being compromised too.
 
I would have said that a year ago- but so many developers including MS themselves are using virtualization instructions to better protect users, and these approaches are being compromised too.

Please show me an example of where or when this has happened without VM involvement
 
Please show me an example of where or when this has happened without VM involvement

I'm not interested or speaking to 'has happened', rather to where software is going and in parallel, how these vulnerabilities might affect future software.

Virtualization and containerization of applications and services is an ongoing trend. Anything that affects hypervisors increases the attack surface and decreases real and perceived security.

So with respect to the topic and the question, while a year ago I would have likely recommended an 8700K, today I'd recommend a 9700K. Or if max FPS are not needed, a 2700 / 2700X or Ryzen 2 when available.
 
So with respect to the topic and the question, while a year ago I would have likely recommended an 8700K, today I'd recommend a 9700K. Or if max FPS are not needed, a 2700 / 2700X or Ryzen 2 when available.

I don’t disagree with your rationale.. However I truly think the risks are minimal for the average user. It’s like the risk that someone is likely to break into your house, if they really want to do it, they will find a way.

People are more likely to try to hack a business machine than a gamer’s machine these days.
 
I don’t disagree with your rationale.. However I truly think the risks are minimal for the average user. It’s like the risk that someone is likely to break into your house, if they really want to do it, they will find a way.

I don't disagree today. But I also don't purchase and make recommendations based solely on what's happening today either. Going forward, I see stuff like Docker and Windows Sandbox among others exploding in the consumer market. If I could open a Firefox / Chrome / Edge container at will to have a truly isolated browser window without the hassle of a VM? Sweet.

Unless it's all for naught because someone figured out how to string a few attacks together to break out of the hypervisor from some ad that made it through adblock.

People are more likely to try to hack a business machine than a gamer’s machine these days.

I'd say that depends. Attackers are more likely to use tuned attacks against businesses, but consumer machines are always good for botnet expansion as well as credential harvesting. They're also usually less secured, if only due to lacking institutional defense in depth as well as industrial IPS / IDS resources.

You're not wrong, but it's also not a clear-cut comparison.

In terms of 'more vulnerable' and 'less vulerable', as opposed to 'vulnerable' and 'invulnerable' (an impossibility), choosing 'less vulnerable' with nearly all other decision points being moot hypotheticals is pretty easy.
 
I know it's mostly rumours and speculation at this point, but I think Intel themselves have made some statements that Ice Lake desktop/enthusiast parts (i.e. 6C/12T+) will not be out until mid to late 2020 as they are focusing on mobile first. Is that accurate based on what we know today)? Debating when to upgrade my rig (see below), and if there's no chance of desktop Ice Lake for another year then I will make a decision based on Ryzen 3800X/3900X vs. 9700K/9900K and may as well upgrade in the next few months once independent benchmarks for Zen 2 are released.

I'm still running a 3770K rig I built in 2012, recently upgraded to a 1080 Ti from a 970 to go with my new 1440p 144Hz display. I know I'm CPU bottlenecked which is why I am interested in upgrading. I'm a huge CD Projekt Red fan and my main goal is to be able to play Cyberpunk 2077 maxed out, will likely sell my 1080 Ti and get a 2080 Ti/Super so I can enjoy the ray tracing in that game. But again, if Ice Lake isn't coming until after Cyberpunk is released then I do not want to wait it out.
What makes you think you will be CPU limited on DXR enabled game? :)
 
It will be 2045 before 10nm is available.

They will still be on 14+++^7

In the meantime AMD will be using quantum entangled super computing on a new substrate they mined from the asteroid belt using drones powered by Ryzen 17 chips.

Intel will be still re-engineering skylake cores for a few more IPC.
 
What makes you think you will be CPU limited on DXR enabled game? :)

Based on my own testing I am CPU limited today, even in single player games (Mass Effect Andromeda, Deus Ex Mankind Divided) and especially MP like Battlefield V. I plan to upgrade from the 1080 Ti I just got to an RTX card as well when 2077 comes out.
 
I know it's mostly rumours and speculation at this point, but I think Intel themselves have made some statements that Ice Lake desktop/enthusiast parts (i.e. 6C/12T+) will not be out until mid to late 2020 as they are focusing on mobile first. Is that accurate based on what we know today)? Debating when to upgrade my rig (see below), and if there's no chance of desktop Ice Lake for another year then I will make a decision based on Ryzen 3800X/3900X vs. 9700K/9900K and may as well upgrade in the next few months once independent benchmarks for Zen 2 are released.

I'm still running a 3770K rig I built in 2012, recently upgraded to a 1080 Ti from a 970 to go with my new 1440p 144Hz display. I know I'm CPU bottlenecked which is why I am interested in upgrading. I'm a huge CD Projekt Red fan and my main goal is to be able to play Cyberpunk 2077 maxed out, will likely sell my 1080 Ti and get a 2080 Ti/Super so I can enjoy the ray tracing in that game. But again, if Ice Lake isn't coming until after Cyberpunk is released then I do not want to wait it out.

Actually rumors for 10nm desktop are 2021 or 2022.
Leaked Intel roadmap shows its 10nm desktop CPUs won't arrive until 2022

Those are rumors. But Intel has a problem that 14nm is so good,since it easily clocks to 5GHz. 5GHz is looking hard to do for 10nm(7nm). So they are better off sticking to 14nm for desktop, until some combo of IPC and 10nm improvements can decisively beat 14nm parts.

Intel is probably in for a rough go of things during this transition.
 
Actually rumors for 10nm desktop are 2021 or 2022.

Those are rumors. But Intel has a problem that 14nm is so good,since it easily clocks to 5GHz. 5GHz is looking hard to do for 10nm(7nm). So they are better off sticking to 14nm for desktop, until some combo of IPC and 10nm improvements can decisively beat 14nm parts.

Intel is probably in for a rough go of things during this transition.


Rubbish

If intel can get 18% more performance then that's nearly 1ghz less they need to get out of their process tech.
 
Rubbish

If intel can get 18% more performance then that's nearly 1ghz less they need to get out of their process tech.

They don't need to just tie Coffee Lake, they have to beat it, and nearly everywhere, but in particular at gaming...

They also need to transition fabs from 14nm to 10nm, and while they transition, 10nm makes more sense to go first into Mobile, then Servers, and finally into desktop, in order of the importance of lower power usage 10nm should bring to the table.
 
Eh they don’t need to beat it. They need to beat it at the clock speed Zen2 runs
 
They don't need to just tie Coffee Lake, they have to beat it, and nearly everywhere, but in particular at gaming...

Something to consider: just meeting the performance with the new process could be enough. Not because it would be a selling point, but because they could transition more fabs over. You see this regularly with semiconductors, and for Intel, it's the only way 'forward'.
 
Something to consider: just meeting the performance with the new process could be enough. Not because it would be a selling point, but because they could transition more fabs over. You see this regularly with semiconductors, and for Intel, it's the only way 'forward'.

But the transition is going to go Mobile, then Servers. Both run lower clock speeds and require lower power than desktop. So transitioning a large part of that will take significant time.

By the time they get to desktop, they will probably be on 10nm+ (higher clocks) and have tweaked cores a bit more so it represents a delta over Coffee Lake.

This looks like a very rough year for Intel leaked roadmap is to be believed. It looks like 14nm 10 Core Comet Lake is the next stop for Intels top end desktop consumer chip. But it doesn't look like it will ship until early 2020, and if it is just another Sky Lake type core, who is really going to care? So 2020 will be a rough year as well.

Maybe there will be a late 2020 reveal of a 10nm desktop part, but maybe there won't...

I have to hand it to AMD, they are executing well, while Intel is still stumbling.
 
We had to endure some silly rumors this week, but it looks like there could be truth to some of it:
https://www.techpowerup.com/257283/...ce-lake-s-not-before-q3-2020-roadmap-suggests

Original Hong Kong source in the link.
So it looks to be 2020 for Comet Lake.

A new socket seems insane but if it is compatible with Icelake, why not.
Are they doing this for DDR5??
This just begs the question, if they are going to be compatible, why isn't Comet Lake just Ice Lake architecture running at 14nm instead of Skylake 5.0?

10 core at 125w TDP looks accurate but that was expected. Most likely a low base clock to meet that spec.

PCIe 3.0 looks correct as well, which is kinda lame.
 
10 core at 125w TDP looks accurate but that was expected. Most likely a low base clock to meet that spec.

Which is fine; we already know that high sustained boost clocks need good cooling, and that if you don't have good cooling, you're not going to get the highest clockspeeds. The same would be true if Zen 2 could hit >4.5GHz on all cores.

PCIe 3.0 looks correct as well, which is kinda lame.

It is to a degree, but if you really need more connectivity than is provided by Z390, you're really pushing into HEDT territory. Ryzen 3000 on X570 has more available lanes but the bandwidth still isn't available to saturate all of them.

Ice Lake architecture running at 14nm

We can ask this question all along Intel's product stack- if I hazard to guess, it would be due to the feature geometry not aligning. Ice Lake has likely been taped out for years awaiting Intel's 10nm process to become available, and taking that architecture and stuffing it into 14nm would likely not be economical.
 
We can ask this question all along Intel's product stack- if I hazard to guess, it would be due to the feature geometry not aligning. Ice Lake has likely been taped out for years awaiting Intel's 10nm process to become available, and taking that architecture and stuffing it into 14nm would likely not be economical.

Well that still seems like a minor feat compared to using the same platform when going from a 14nm monolithic chip to a 7nm multi chiplet design.

Weather Icelake and Cometlake or just Icelake are new architectures, they just need to stop with the 'Lake' to separate the new design classes!
 
Well that still seems like a minor feat compared to using the same platform when going from a 14nm monolithic chip to a 7nm multi chiplet design.

Well, AMD was already more or less 'there'. And Intel moving a 10nm arch to 14nm would be going backwards- we can probably assume that Ice Lake cores have more transisters than Skylake, at the very least, and may depend on cache architectures that are untenable on 14nm.
 
Nobody should care about clock speed, everyone should care about perf / $, outright performance and efficiency.
 
It's looking like Ice Lake is going to be another 10nm turd, just like Cannon Lake. So much so that the Ice Lake (10nm Sunny Cove) and Comet Lake (14nm Skylake) will both be considered 10th gen on laptops:
https://www.anandtech.com/show/1478...met-lakey-10th-gen-core-for-low-power-laptops

An 18% IPC increase just doesn't make up for 50% more cores. Comet Lake does provide 6 cores now, though we do not see a clock change on the 4 core parts from Whiskey Lake.

So what does this mean? Ice Lake will doubtfully come to desktop for good reason. It is also looking like we will see 2 architectures on the same platform, just like the laptops. However, it will most likely be Comet Lake and Rocket Lake (14nm Sunnycove). CML will most likely get the cores while RKL? will get the IPC. Frequency will probably be the same, as Cannon Lake did no better here than Ice Lake.

Just to be clear, this is not a rip on Sunny Cove - which looks VERY promising. It's a rip on Intel 10nm.

My guess would be CML for desktop mid FY20 and RKL late FY20. God knows when Tiger Lake comes out, which will most likely be the first 10nm desktop part.
 
6C on -U parts, bold
those claimed all-core clocks seem... ambitious... I agree w/ Dr. Cutress re: PL2's. Bespoke Surface Pro potholders, the new "hot" accessory /s
 
6C on -U parts, bold
those claimed all-core clocks seem... ambitious... I agree w/ Dr. Cutress re: PL2's. Bespoke Surface Pro potholders, the new "hot" accessory /s

AMD needs to hurry up with the 7nm APU, because it's parts trying to compete in this space are still 4c, and probably use more power.
 
AMD needs to hurry up with the 7nm APU, because it's parts trying to compete in this space are still 4c, and probably use more power.
lower CPU performance (and fewer cores, now); much, much higher GPU performance. You're right that power is an issue, same 15W PL1 TDP but AFAIK the Intel chips have lower idle power and better lightly-threaded efficiency. Really though the lack of strong OEM wins for AMD needs to change before they can make any headway into mobile, even if they manage a great performance success with mobile Zen 2. Alas, I'd be surprised if we see 7nm APUs this year...
 
I have an i5-4570 and GTX 1050 Ti. I'm thinking of possibly going with the Ryzen 3 3200G (+13% faster) when Microsoft ends support for Win7 in Jan.2020. I could then replace it with something else maybe a year later.
Or, maybe I'll use an SSD in my current rig with a new OS waiting for Comet Lake...

(I'm hoping by then there will be a new Microsoft OS with a similar desktop to Win7.)
 
I have an i5-4570 and GTX 1050 Ti. I'm thinking of possibly going with the Ryzen 3 3200G (+13% faster) when Microsoft ends support for Win7 in Jan.2020. I could then replace it with something else maybe a year later.)

Re-using the GPU in the new system, or keeping it with the old one? Because if you have a GPU, I would spend a little more ($140 at newegg) and get a Ryzen 2600, 6c/12t instead of the 3200g 4c/4t.
 
Back
Top