Intel Expands 9th Gen Core Desktop and Mobile Processor Families

[HwGeek]

Intel CPU 2018-2021 Roadmap Leaks Out – Up To 10 Core Comet Lake-S Desktop CPUs in 2020, 14nm Rocket Lake-S in 2021, No 10nm LGA Parts Till 2022
https://wccftech.com/intel-desktop-mobile-cpu-roadmap-leak-14nm-comet-lake-10nm-ice-lake-tiger-lake/

We all know that AMD can go upto 16 cores, and we also know how hard was for Intel to make Core i9 9990XE 14C 5.0Ghz, this can give you an Idea that there is no more Clock speed bump to save Intel with New Refresh.

 

[Darmok N Jalad]

Intel CPU 2018-2021 Roadmap Leaks Out – Up To 10 Core Comet Lake-S Desktop CPUs in 2020, 14nm Rocket Lake-S in 2021, No 10nm LGA Parts Till 2022
https://wccftech.com/intel-desktop-mobile-cpu-roadmap-leak-14nm-comet-lake-10nm-ice-lake-tiger-lake/

We all know that AMD can go upto 16 cores, and we also know how hard was for Intel to make Core i9 9990XE 14C 5.0Ghz, this can give you an Idea that there is no more Clock speed bump to save Intel with New Refresh.

If true, then Intel is going to have to make architectural adjustments to get more IPC. I believe their 14nm is actually pretty dense compared to competitors’ 14nm, but I suspect we are starting to see a combination of node and architecture limits. Much like the first two versions of Ryzen seem architecturally limited to current clocks. Zen 2 will likely be revised for better clockspeed headroom.

 

[notb]

Intel CPU 2018-2021 Roadmap Leaks Out – Up To 10 Core Comet Lake-S Desktop CPUs in 2020, 14nm Rocket Lake-S in 2021, No 10nm LGA Parts Till 2022
https://wccftech.com/intel-desktop-mobile-cpu-roadmap-leak-14nm-comet-lake-10nm-ice-lake-tiger-lake/

We all know that AMD can go upto 16 cores, and we also know how hard was for Intel to make Core i9 9990XE 14C 5.0Ghz, this can give you an Idea that there is no more Clock speed bump to save Intel with New Refresh.

There's a slide missing. ;-)
"Intel client commercial" means CPUs put into OEM desktops.
DIY desktop processors for gamers/enthusiasts have a separate roadmap. If needed, Intel should be able to provide 10nm CPUs for enthusiast market (it's tiny).

Personally I think it's fake, but it actually looks fine.
Intel won't be able to counter 16-core Ryzen with this, but that's a marketing excercise. Everything else is within range.
Ice Lake-U in May is a fantastic information. It'll be interesting to see how limited 10nm supply is.

 

[DarthJedi]

There's a slide missing. ;-)
"Intel client commercial" means CPUs put into OEM desktops.
DIY desktop processors for gamers/enthusiasts have a separate roadmap. If needed, Intel should be able to provide 10nm CPUs for enthusiast market (it's tiny).

Personally I think it's fake, but it actually looks fine.
Intel won't be able to counter 16-core Ryzen with this, but that's a marketing excercise. Everything else is within range.
Ice Lake-U in May is a fantastic information. It'll be interesting to see how limited 10nm supply is.

What 10nm? Intel is dead if this is true, and it can't be. They've announced Cove for mid 2019. If Zen 2 catches up per-core performance, Intel will suffer as long as they fail to deliver something 10% faster. Core count also. If Zen 2 desktop gets 16 cores, at the per-core level of 9900K 5GHz, Intel won't sell HEDT either, let alone 9990XE.

 

[notb]

What 10nm? Intel is dead if this is true, and it can't be.

Intel has been dead so many times over the last 2 years that I don't notice any more.

If Zen 2 catches up per-core performance

Yeah, "if".

Intel will suffer as long as they fail to deliver something 10% faster.

On the contrary. It's AMD who has to show something 10% faster (or better: 50%).
Intel has the brand, the features, the ULV market, the ecosystem and - obviously - partnership contracts.

If Zen 2 desktop gets 16 cores, at the per-core level of 9900K 5GHz, Intel won't sell HEDT either, let alone 9990XE.

If Zen 2 gets 16 cores at per-core level of 9900K, it'll need a separate 8-pin on the SoC, because AM4 will melt from the current needed.

Lets say they'll manage to keep the 2700X single-core performance and that 7nm DUV is 30% more efficient.
2700X all-core boost draws around 170W. 170*2*0.7 = 238. You know... AMD once had a CPU with TDP like that. It didn't end well for them.

As for HEDT: no one buys it anyway. Who cares?

 

[TheGuruStud]

Intel has been dead so many times over the last 2 years that I don't notice any more.

Yeah, "if".

On the contrary. It's AMD who has to show something 10% faster (or better: 50%).
Intel has the brand, the features, the ULV market, the ecosystem and - obviously - partnership contracts.

If Zen 2 gets 16 cores at per-core level of 9900K, it'll need a separate 8-pin on the SoC, because AM4 will melt from the current needed.

Lets say they'll manage to keep the 2700X single-core performance and that 7nm DUV is 30% more efficient.
2700X all-core boost draws around 170W. 170*2*0.7 = 238. You know... AMD once had a CPU with TDP like that. It didn't end well for them.

As for HEDT: no one buys it anyway. Who cares?

Yeah, sure, 170W. A crappy 1700 OCed to 4.0 ghz at 1.395V uses that much maxed out (mostly stress testing). You're really confusing AMD with intel, here lol.
A 2700X system from the wall only pulls about 190 lulz.

 

[Deleted member 158293]

If intel can't bring out any new competitive parts, then they'll bring out new non-competitive parts... if anything just to keep selling something new.

 

[notb]

Yeah, sure, 170W. A crappy 1700 OCed to 4.0 ghz at 1.395V uses that much maxed out. You're really confusing AMD with intel, here lol

What can I say? Ryzen XFR is very good. Didn't you praise it earlier?
https://www.techpowerup.com/reviews/AMD/Ryzen_7_2700/17.html
https://www.techpowerup.com/reviews/AMD/Ryzen_7_2700X/18.html
After OC both CPUs draw roughly the same (with similar performance as well). But 2700X gets awfully close on its own. And that's the power draw we should look at, since it corresponds to performance that Ryzen is praised for.
Looking at the specs and idle figures, I'd say the rest of the PC draws around 40W. This leaves 170W for the Ryzen.
If you don't believe TPU reviews, here's another one:
https://www.guru3d.com/articles-pages/amd-ryzen-7-2700x-review,7.html

Oh my. Are you one of the people that believe Ryzens stay under the TDP and only Intel CPUs boost beyond? Poor you...

 

[DarthJedi]

Intel has been dead so many times over the last 2 years that I don't notice any more.

Yeah, "if".

On the contrary. It's AMD who has to show something 10% faster (or better: 50%).
Intel has the brand, the features, the ULV market, the ecosystem and - obviously - partnership contracts.

If Zen 2 gets 16 cores at per-core level of 9900K, it'll need a separate 8-pin on the SoC, because AM4 will melt from the current needed.

Lets say they'll manage to keep the 2700X single-core performance and that 7nm DUV is 30% more efficient.
2700X all-core boost draws around 170W. 170*2*0.7 = 238. You know... AMD once had a CPU with TDP like that. It didn't end well for them.

As for HEDT: no one buys it anyway. Who cares?

AMD needs to match, that's it. I buy loads of machines every generation. This year I'm on hold. I'm waiting for Zen 2 to either go with 9990XE or similar from AMD. If AMD can reach 5GHz 9900K performance and drop 16 cores, I won't think twice, I'll build 2 such machines. If AMD manages to get to the level of 4.7GHz (Hz for Hz AMD and Intel are draw now, they just can't catch the clock ATM) that'll also be reason to go AMD. Anything less than 4.7 and I'm staying on Intel.
Keep in mind workstation and server cost for Intel is huge. Rome is already better per socket.

Not sure where you draw your data from, but powers you mention are a joke. 7nm TSMC gives 50% less power requirements. So, even if AM4 was limited to 300W, they could double the core count power-wise.

 

[TheGuruStud]

What can I say? Ryzen XFR is very good. Didn't you praise it earlier?
https://www.techpowerup.com/reviews/AMD/Ryzen_7_2700/17.html
https://www.techpowerup.com/reviews/AMD/Ryzen_7_2700X/18.html
After OC both CPUs draw roughly the same (with similar performance as well). But 2700X gets awfully close on its own. And that's the power draw we should look at, since it corresponds to performance that Ryzen is praised for.
Looking at the specs and idle figures, I'd say the rest of the PC draws around 40W. This leaves 170W for the Ryzen.
If you don't believe TPU reviews, here's another one:
https://www.guru3d.com/articles-pages/amd-ryzen-7-2700x-review,7.html

Oh my. Are you one of the people that believe Ryzens stay under the TDP and only Intel CPUs boost beyond? Poor you...

Do you have some sort of reality distortion field when looking at websites, so you can't read numbers? Here, let me simplify it for you:

Ok, ABSOLUTE worst case scenario (blender/cinebench), and we'll even use the guru3d link. Look at CPU+SoC power in wprime (101). Look at the increase from wprime to cinebench in the graph above (33). Add the difference. 101+33= what? SURE AS SHIT AIN'T 170! Typical loads are gonna be like 105-115.

Bye, ban yourself from using the internet.

 

[medi01]

Intel is dead if this is true

Nvidia sells cards that are half speed or less for the same price as 570, so, hold your horses.

AMD might gaining (so needed) market share is all that can come out of it.
Note how Ryzen gains are largely driven by Intel chip shortages.

 

[DarthJedi]

Nvidia sells cards that are half speed or less for the same price as 570, so, hold your horses.

AMD might gaining (so needed) market share is all that can come out of it.
Note how Ryzen gains are largely driven by Intel chip shortages.

Ok, let's be clear: not that the company is dead itself , but they've lost the battle for the 2019-2020. All until they make 10nm reality.
Of course, if AMD fails to catch up, nothing changes. Intel will slowly keep loosing server market.

 

[notb]

Do you have some sort of reality distortion field when looking at websites, so you can't read numbers? Here, let me simplify it for you:

Ok, ABSOLUTE worst case scenario (blender/cinebench), and we'll even use the guru3d link. Look at CPU+SoC power in wprime (101). Look at the increase from wprime to cinebench in the graph above (33). Add the difference. 101+33= what? SURE AS SHIT AIN'T 170! Typical loads are gonna be like 105-115.

Bye, ban yourself from using the internet.

So people believe in HWiNFO figures? Is this the reason of all this power draw misunderstanding?

Anyway, in guru3d test 2700X system pulled 199W in Cinebench, getting 1828 points.
8700K system pulled 150W getting 1402 points.

Now, you tell me, based on HWiNFO readouts, that the rest of the system draws around 55W under load.
That means 8700K itself used 95W (its TDP! ding ding ding!).
This also means 2700X itself used 144W (not exactly its TDP).

Moreover,
1802/144 = 12.7 [pts/W]
1402/95 = 14.8 [pts/W]
which makes the i7 quite a bit more efficient.

Do you agree with all that?

 

[Darmok N Jalad]

I don’t see anything remotely close to 199W for a 2700X on the chart above. Even a 12c/24t TR can’t get there. Intel’s “95w” 9900K does look a little funny though.

https://www.anandtech.com/show/13400/intel-9th-gen-core-i9-9900k-i7-9700k-i5-9600k-review/21

And no, Intel will not be dead anytime soon. They still have big volume and the supply chain advantages, not to mention long-standing relationships with OEMs.

 

[ShurikN]

Well on one hand its a huge issue if they get stuck on 14 for 2 more years, AMD wll undoubtedly get a massive market share in all 3 major segments.
On the other side it's wccf, and I don't remember if their leaks have ever come remotely true.

 

[TheLostSwede]

Now, now ladies, please try to be civilised to one another, no need for this bickering and name calling.

 

[efikkan]

There's a slide missing. ;-)

"Intel client commercial" means CPUs put into OEM desktops.

DIY desktop processors for gamers/enthusiasts have a separate roadmap. If needed, Intel should be able to provide 10nm CPUs for enthusiast market (it's tiny).

It depends on what you mean by tiny. Yes, OEM volumes are much larger, but the OEM builds are mostly low-end CPUs with low margins, so in terms of profits the "enthusiast" and workstation markets are quite important.

It has been rumored for a while that Intel will be prioritizing server and ultra low voltage chips on 10nm, and if it's accurate that 10nm never will be widely deployed, then I'm fine with it as long as HEDT is included. Low-end and lower mid-range CPUs don't really need 10nm yet, so if the 10nm production volume will remain limited, it's better to focus on those chips that need the new node.

What remains unclear to me is what changes Comet Lake (and "Rocket Lake") will bring. Will this be another tweaked Skylake, or is it a backported Ice Lake/Sunny Cove?

 

[notb]

It depends on what you mean by tiny. Yes, OEM volumes are much larger, but the OEM builds are mostly low-end CPUs with low margins, so in terms of profits the "enthusiast" and workstation markets are quite important.

Every company will fight for a 1% revenue growth. That's what shareholders expect.
But this 1% has to make sense.
At the moment enthusiast segment likely is profitable - even if small compared to OEM.
If AMD forces the prices down, Intel won't be interested anymore. They won't fight for the profit margin that AMD has right now.
Intel makes a lot of stuff. They can easily enter new markets. They'll find a better investment.

It has been rumored for a while that Intel will be prioritizing server and ultra low voltage chips on 10nm, and if it's accurate that 10nm never will be widely deployed, then I'm fine with it as long as HEDT is included. Low-end and lower mid-range CPUs don't really need 10nm yet, so if the 10nm production volume will remain limited, it's better to focus on those chips that need the new node.

We really don't know what "limited" means - other than the fact that Intel simply can't replace all their products with 10nm successors.
It could turn out that Intel 10nm will outsell AMD 7nm anyway.

If AMD's market share grows, they'll be faced with the same problem.

 

[efikkan]

We really don't know what "limited" means - other than the fact that Intel simply can't replace all their products with 10nm successors.

It could turn out that Intel 10nm will outsell AMD 7nm anyway.

If AMD's market share grows, they'll be faced with the same problem.

Excellent point.
Also, Intel is pretty deeply committed to ship Ice Lake-SP, which is already postponed. Canceling it would have huge financial consequences for Intel, so making it even if it means eating up most of the 10nm capacity is probably the best idea, and the best use of "limited" 10nm production volume.

 

[voltage]

H is for Hot
K is for much Klock
F is for freaking avoid the ...ucker

Bottom line, not much changed, this is just 8th gen all over again with more cut down parts in the stack.

I'll see myself out

yawn...

 

[Nihilus]

Sort of curious to what overclock that 9350kf could get.

 

[jmgbjr]

That 9980HK will boost to 5GHz for about 0.5 of a second and then throttle.

.5 seconds before your battery dies