Is Intel likely to release 10nm desktop CPUs in 2021?

[g67575]

Panos - thanks for your reply, I wonder if your reply could be shortened to 2 separate responses, such as, 'I don't know the answer' regarding 10nm and release in 2021 and 'a move to 10nm is problematic for Intel because of potential performance issues related to changing the CPU die size' regarding reasons for delays to 10nm Intel CPUs. If that is so, couldn't a die size shrink potentially be avoided by improving the 10nm process to 10nm++ / whatever they want to call it?

On the other hand, they may simply be running into limits related to the laws of physics, as they approach 7nm/5nm, something that was speculated on years ago when these new process shrinks were hypothesized.

Then, you mention development of 7nm Nvidia GPUs and how you think Nvidia will not be able to shrink the die without removing parts of the chip's design. I suppose what I wonder about this is - Are CPU designs as large, complex and compartmentalised as modern GPUs? And if not, is a direct comparison between GPUs and CPUs informative? Also, isn't the graphics chip included with most CPUs a completely optional feature of chip designs?

 

[Panos]

Panos - thanks for your reply, I wonder if your reply could be shortened to 2 separate responses, such as, 'I don't know the answer' regarding 10nm and release in 2021 and 'a move to 10nm is problematic for Intel because of potential performance issues related to changing the CPU die size' regarding reasons for delays to 10nm Intel CPUs. If that is so, couldn't a die size shrink potentially be avoided by improving the 10nm process to 10nm++ / whatever they want to call it?

On the other hand, they may simply be running into limits related to the laws of physics, as they approach 7nm/5nm, something that was speculated on years ago when these new process shrinks were hypothesized.

Then, you mention development of 7nm Nvidia GPUs and how you think Nvidia will not be able to shrink the die without removing parts of the chip's design. I suppose what I wonder about this is - Are CPU designs as large, complex and compartmentalised as modern GPUs? And if not, is a direct comparison between GPUs and CPUs informative? Also, isn't the graphics chip included with most CPUs a completely optional feature of chip designs?

On the last question, is down to how big dies the manufacturers can make. There are physical limitations on the manufacturing process that apply both to CPUs & GPUs. Nothing to do with the inner parts complexity but the actual size of the chip.

Could it be 2021 when we see 10nm desktop CPUs from Intel? The answer is maybe if Intel manages to

• increase the IPC by a wooping 30%+ or get the speeds to over 5.5Ghz
• makes profitable yields on something bigger than quad core. Intel just managed to succesffuly make quad core with functional IGP in January 2019. Already 4 years late. Before Jan 2019, the only 10nm they could sell were dual & quad cores with the IGP fused, because of the low yields and high failure rate of the 10nm process.

Also the 10nm products Intel can make atm is a good indication of the current status.
https://ark.intel.com/content/www/u...ter.html?productType=873&3_Lithography-Max=10

Why there are no hexa or octa cores there? Without them, Intel going to lose the laptop market in 2020. We didn't even had an announcement for potential products later this year, to delay the AMD Ryzen 4000 laptop series.

 

[Dave2150]

10nm LGA 1700 Intel CPU's, with DDR5 and PCI-E v5 will likely crush anything AMD offers, as this will be their first CPU designed by Keller also. That's where I'm upgrading next, currently on a [email protected].

 

[jigger]

Do you guys think that 10nm desktop CPUs in 2021 makes sense, based on the schedule for 10nm++ in 2021? Or, could this just be Intel blowing smoke to their investors?

Another possibility is Xenon / server CPUs only, if production volumes are low, with 7nm desktop CPUs in 2022 or later. Which seems more likely?

Intel don’t just need a manufacturing base, they need to produce functional silicon and new chip designs. The desktop segments are kind of irrelevant considering the depth of the hole Intel have jumped into and the height of the walls AMD have built around them.

The time for 10nm to make any significant impact for Intels desktop parts has passed and bigger issues have been allowed to develop.

Producing garbage on 10nm is just more holes for AMD to build walls around. I think if Intel do push 10nm so late, then it’s a sign of just how much trouble the company is in. My hunch is Intels plans for 10nm was never much more than 4c8t chips at 4.5Ghz~ and now engineers need 10nm to scale to 64c 128t with a IPC jump of 25% and God knows how many silicon security fixes.

 

[Journey]

10nm LGA 1700 Intel CPU's, with DDR5 and PCI-E v5 will likely crush anything AMD offers, as this will be their first CPU designed by Keller also. That's where I'm upgrading next, currently on a [email protected].

Isn't waiting until 2022 and more likely 2023 a bit of a wait to be stuck on a quad core? Also I do realize that it will only put it at 8 years old, but technically it's not much better than a 2600k which was from 2011, so it'll be like using a good CPU from 2011 in 2022/3. Madness.

 

[Grim5]

10nm LGA 1700 Intel CPU's, with DDR5 and PCI-E v5 will likely crush anything AMD offers, as this will be their first CPU designed by Keller also. That's where I'm upgrading next, currently on a [email protected].

Let's assume everything goes perfect for Intel from here on out.

In 2022 they launch their fixed 10nm Desktop CPU's - 30% higher IPC than Coffee Lake and same core count as AMD.

All AMD has to do between now and then, is increase their clock speeds and get another 20-25% more IPC and that's enough to fight off Intel 10nm.

AMD is expecting 17% IPC and 100mhz extra clock this year and probably something similiar when it moves to 5nm.

I fail to see, how Intel will "crush anything AMD offers" - the only way that's possible is if AMD fails to hit their target every year for the next 2 years.

 

[Perfect_Chaos]

Let's assume everything goes perfect for Intel from here on out.

In 2022 they launch their fixed 10nm Desktop CPU's - 30% higher IPC than Coffee Lake and same core count as AMD.

All AMD has to do between now and then, is increase their clock speeds and get another 20-25% more IPC and that's enough to fight off Intel 10nm.

AMD is expecting 17% IPC and 100mhz extra clock this year and probably something similiar when it moves to 5nm.

I fail to see, how Intel will "crush anything AMD offers" - the only way that's possible is if AMD fails to hit their target every year for the next 2 years.

Yep, only recently CPUs became interesting again IMO. Won't be hanging around waiting myself, if AMD bring those rumoured nice gains with Zen 3 i will be replacing my 3700x ASAP.

DDR5 might not even be out til late 2021/2022, even Hardware Unboxed said similar. I personally will likely skip first round of DDR5 anyway give 6-12 months for faster speeds to come about, depends on how good early DDR5 is though.

 

[Panos]

10nm LGA 1700 Intel CPU's, with DDR5 and PCI-E v5 will likely crush anything AMD offers, as this will be their first CPU designed by Keller also. That's where I'm upgrading next, currently on a [email protected].

The day Intel has on sale something better than a quad core at 10mn let me know.

 

[LePhuronn]

10nm LGA 1700 Intel CPU's, with DDR5 and PCI-E v5 will likely crush anything AMD offers, as this will be their first CPU designed by Keller also. That's where I'm upgrading next, currently on a [email protected].

I thought this was a troll post from a new account, but then I saw your post count. So all I'll say is "lol OK, have you not been paying attention?".

 

[subbytna]

10nm LGA 1700 Intel CPU's, with DDR5 and PCI-E v5 will likely crush anything AMD offers, as this will be their first CPU designed by Keller also. That's where I'm upgrading next, currently on a [email protected].

Wish OCUK had a raised eyebrow smiley so I could fill the screen with them in my reply to this comment.

 

[Dave2150]

Isn't waiting until 2022 and more likely 2023 a bit of a wait to be stuck on a quad core? Also I do realize that it will only put it at 8 years old, but technically it's not much better than a 2600k which was from 2011, so it'll be like using a good CPU from 2011 in 2022/3. Madness.

I use my PC for gaming. I game at 4k, where I'm GPU limited. If I'd upgrade my CPU I'd see no difference.

 

[LePhuronn]

I use my PC for gaming. I game at 4k, where I'm GPU limited. If I'd upgrade my CPU I'd see no difference.

Which makes your suggestion that Intel's 10nm vapourware will crush AMD to be even more eyebrow-raising.

 

[Dave2150]

Which makes your suggestion that Intel's 10nm vapourware will crush AMD to be even more eyebrow-raising.

Go ahead and play with your eyebrows mate. I'll be gaming @ 4k powered by a mighty 6700k in the meanwhile

I've had my [email protected] for 4.5 years. I have 16GB DDR 3000, a 1TB 970 evo NVME drive, and it drives my Radeon 7 well at 4k, as well as any other CPU on the market. I'm surprised that game development/GPU development hasn't yet mandated a CPU upgrade, though least my setup will last longer.

I imagine it will be another 1-3 years before Intel's DDR5 LGA1700 (next gen) platform will launch, that seems like the perfect time for me to refresh my PC. If AMD is faster I'll get that, though I highly doubt they'll beat LGA1700 CPU's for gaming.

Have to remember, AMD can't beat Intel's 4.5 year architecture, 5 year old 14nm process with their latest Zen 3000 series from a purely gaming point of view (9900k overclocked still wins by a few %, while using lots more power), so I'd wager they have no hope of beating Intel's Keller designed LGA1700 DDR5 monster. Doesn't really matter if it's on 14, 10 or 7nm, all that matters is performance.

 

[Grim5]

I'm surprised that game development/GPU development hasn't yet mandated a CPU upgrade, though least my setup will last longer.

Because consoles, you know the thing that most gamers in the world play AAA games on, have a CPU weaker than your 6700k. But don't worry, within 10 months developers will be able to target 16 threads of Zen 2 for games and then your 6700k will suffer a quick death.

 

[LePhuronn]

Eyebrow raising intensifies, Dave...

Not sure what your 6700K adds to your point: it's a 14nm CPU that's going to be equivalent to a 10 series i3. And when you're GPU constrained, as you point out, the CPU doesn't mean a lot, to the point that the 3600X gets you the same frame rate, and often better lows. And as a fellow owner of the "mighty" 6700K, I can say hand on heart you're talking tripe. It's not mighty, it's inferior and just going to get ground into the dirt very soon.

You also say AMD can't beat Intel's Skylake arch in gaming, but that's just not true: Threadripper 3000 does (amazingly). Ergo, given the right circumstances, Zen 2 can beat Skylake re-re-refresh at gaming. Whatever the next Desperation Lake is called is still just Skylake again, and Zen 3 is going to batter it in gaming. Now granted the only person who can beat Jim Keller is Jim Keller, so it's possible his Intel arch will take on AMD, but that's going to be Zen 5 and nobody knows what that's going to be yet.

You keep saying you "highly doubt" AMD will beat LGA1700, but you literally have nothing to support that stance, just as I have nothing to support Zen 5 will win. It's baseless and really sounding fanboi-ish.

 

[jigger]

Go ahead and play with your eyebrows mate. I'll be gaming @ 4k powered by a mighty 6700k in the meanwhile

I've had my [email protected] for 4.5 years. I have 16GB DDR 3000, a 1TB 970 evo NVME drive, and it drives my Radeon 7 well at 4k, as well as any other CPU on the market. I'm surprised that game development/GPU development hasn't yet mandated a CPU upgrade, though least my setup will last longer.

I imagine it will be another 1-3 years before Intel's DDR5 LGA1700 (next gen) platform will launch, that seems like the perfect time for me to refresh my PC. If AMD is faster I'll get that, though I highly doubt they'll beat LGA1700 CPU's for gaming.

Have to remember, AMD can't beat Intel's 4.5 year architecture, 5 year old 14nm process with their latest Zen 3000 series from a purely gaming point of view (9900k overclocked still wins by a few %, while using lots more power), so I'd wager they have no hope of beating Intel's Keller designed LGA1700 DDR5 monster. Doesn't really matter if it's on 14, 10 or 7nm, all that matters is performance.

A 3700X gives a 9900K a hard time. The 3900X and 3950X are in another league.

 

[Vince]

Have to remember, AMD can't beat Intel's 4.5 year architecture, 5 year old 14nm process with their latest Zen 3000 series from a purely gaming point of view (9900k overclocked still wins by a few %, while using lots more power), so I'd wager they have no hope of beating Intel's Keller designed LGA1700 DDR5 monster. Doesn't really matter if it's on 14, 10 or 7nm, all that matters is performance.

Id wager that one of the reasons it's as fast as it is, is because Intel heavily favoured performance over security during design, a decision that has hurt and is continuing to hurt Intel in the DC. Do you really think that with their current trajectory they can manage, even with the help of Jim, a chip that is faster, secure, scalable, with massively improved IPC on a node that will deliver reduced clock speeds? One of the reasons they are where they are is because they don't know how to go significantly faster and even if they do they haven't got the node to do it on.

Even when they do get the node they will have to deliver in a big way, the spotlight is on Intel right now, they need to deliver an entirely new arch because what they have right now is at the limit, a properly new ground up design is something that intel haven't really done in how long? I mean core/everything after it is broadly based on pentium 3 from 1999.

I'm not saying they will fail but 10nm ramp up was scheduled for 2015. Does that sound like a company that is smashing it to you? It's not the first time they have been caught with their pants down so it will be interesting to see how and what they deliver. I personally hope you are right and they deliver a monster but I think they are going to struggle, right now their fabric is an issue as it is inferior to AMD's infinity fabric and they don't really have anything to rival what IF is capable of. They have EMIB and Mesh which are nice elegant on die solution which is nice and all but all very specific, as far as I can see they have nothing to rival the "go anywhere" protocol that is Infinity Fabric. They need better "mesh" technology to even stand a chance.

 

[humbug]

10nm LGA 1700 Intel CPU's, with DDR5 and PCI-E v5 will likely crush anything AMD offers, as this will be their first CPU designed by Keller also. That's where I'm upgrading next, currently on a [email protected].

This is the sort of language being thrown around back in 2016/17. "oh Intel are this big fantastical company who will just pull a new CPU out of their arse in months and crush Ryzen"

That sort of crap, it didn't happen, instead Intel refreshed the same CPU's over and over again... have you been sitting around for the past 3 years waiting for a new hope opportunity to rave about how Intel will crush AMD?

You should wait until its actually happened because Intel right now are desperate for attention, desperate to continue looking like the company you think they are, in that all they have done in the last 3 years is scream and shout "look at me, i'm still the best"

 

[EsaT]

10nm LGA 1700 Intel CPU's, with DDR5 and PCI-E v5 will likely crush anything AMD offers, as this will be their first CPU designed by Keller also.

Designing new CPU takes many years.
Anything designed by Keller from the start would be couple years away.

And assuming he's even working on desktop/x86 CPUs, I wouldn't wonder if lots of things under work when he joined went back to first phase drawing board to fix security.
But most likely he was hired to lead Intel's AI/GPGPU development.
That's growing market where Intel hasn't done that great.

He certainly can't help with Intel's manufacturing node SNAFU.
Making cores beefier while increasing core counts would require lot more transistors/bigger chips when manufacturing capacity already has problems.


As for PCIe v5...
Well, Intel couldn't make PCIe v4 work for their this year's Skylake rebranding.
https://www.tomshardware.com/news/i...lans-then-nixes-pcie-40-support-on-comet-lake

And with already PCIe v4 having strict signal integrity requirements, I don't expect wide use of PCIe v5 on desktop soon.
Higher speed signaling will certainly demand even more and even now trace length is limiting without signal re-timers.

 

[humbug]

BTW @Dave2150 Your your silly little quad core is probably no better than a Ryzen 1600 here.....