10 nm Intel desktop CPUs discussion (performance vs Zen 2/3)

[LUK]

With Intel 10 nm desktop CPUs reportedly arriving early next year (https://wccftech.com/intel-10nm-desktop-cpus-arriving-early-2020/), what is your take on how their performance will stack up against the latest and greatest from AMD? The SKUs will probably not be high-end/enthusiast yet, but still, how do you think the team blue's parts will perform against the same grade/similarly priced AMD counterparts? (Ice Lake mobile has about 18 % IPC improvement over the previous gen.)

 

I think that as soon as intel has something to offer in all the importanat price brackets, they will retake the lead firmly. I'm inclined to that opinion because they already have the lead in gaming, clock speeds and some productivity instances. As soon as they figure out how to pack more cores on a chip efficiently, that will be it.

[Skiiwee29]

pointless to speculate as we dont have performance numbers for intels 10nm desktop products yet... where as we do with Ryzen. 

[Fasauceome]

1 minute ago, LUK said:

As soon as they figure out how to pack more cores on a chip efficiently, that will be it.

they have ring bus and mesh bus, what more do they need?

 

the lack of cores on Intel chips is by choice.

[Zando_]

"What do you think will happen after Intel drops 10 nm architecture in desktop CPUs?"

I think we'll have actual performance numbers. We can speculate as much as we want or not, but fact is nobody knows or has the numbers. 

[LUK]

9 minutes ago, Skiiwee29 said:

pointless to speculate as we dont have performance numbers for intels 10nm desktop products yet... where as we do with Ryzen. 

true, but we do have it for mobile for both of them. How do you think it will translate to desktop?

 

[GoldenLag]

5 minutes ago, Skiiwee29 said:

pointless to speculate as we dont have performance numbers for intels 10nm desktop products yet...

best guess we have is 10nm mobile, but that is thermally constrained parts and not very representative, not to mention early silicon. 

 

also its very likely that by "desktop" they mean NUC style devices for Desktop. 

 

8 minutes ago, LUK said:

As soon as they figure out how to pack more cores on a chip efficiently,

the issues isnt packing cores. the issue is packing cores and yielding it. Intel`s mature 28 core Server CPU is allready a pain to yield. 

 

[Jurrunio]

10 minutes ago, LUK said:

(Ice Lake mobile has about 18 % IPC improvement over the previous gen.)

but no increase in efficiency. It's not really faster than 14nm Comet Lake and with increased thermal density, unlikely to beat Coffee Lake Refresh in a max OC battle (if immaturity of new process doesn't stop it before that)

[vanished]

The real question is not what will 10 nm look like but how long will it take them?  At this rate it could easily still be a year or two away, and I'd take any claims from Intel about timeline with a huge mound of salt at best.  Just look back over the last 4 years at how many times they said they'd have it by year X and then missed it.  It's also reasonable to expect them to make claims that stretch the truth just to stop people (shareholders) from panicking and realizing how bad they really have it.  They've been doing that for quite a while already, sadly to seemingly decent success.

 

As for how it will do, I've heard that due to 14nm+++++ being so well optimized that 10 nm isn't going to even offer any meaningful improvement anymore, and may even incur a clockspeed regression.  We're certainly seeing this on the mobile side so far.  The only thing Intel has going for them currently is a tiny single-threaded lead.  If they lose that to Zen 3, which they almost certainly will if improvements aren't made, they will have nothing.  That means they're stuck pushing 8 cores harder, because if they go for 10 and are forced to drop clockspeeds, that's exactly what's going to happen.  Either way though they fail to make any meaningful progress at catching up.  AMD has more multi-threaded, more efficiency, less cost, more memory bandwidth, more PCIe bandwidth, fewer security issues, and basically everything else going for them right now.

 

Intel's only good market is in the mobile space because AMD's offerings there are a generation behind and only target the mid-low end.  It's going to be many years before Intel can hope to compete in the high-end desktop and server space again, so their best chance is to keep pushing for gaming supremacy while they work on a broader solution (their own ryzen, or sandy bridge 2.0) to the rest of their issues that will let them actually compete with and beat AMD.

[Mister Woof]

13 minutes ago, Ryan_Vickers said:

The real question is not what will 10 nm look like but how long will it take them?  At this rate it could easily still be a year or two away, and I'd take any claims from Intel about timeline with a huge mound of salt at best.  Just look back over the last 4 years at how many times they said they'd have it by year X and then missed it.  It's also reasonable to expect them to make claims that stretch the truth just to stop people (shareholders) from panicking and realizing how bad they really have it.  They've been doing that for quite a while already, sadly to seemingly decent success.

 

As for how it will do, I've heard that due to 14nm+++++ being so well optimized that 10 nm isn't going to even offer any meaningful improvement anymore, and may even incur a clockspeed regression.  We're certainly seeing this on the mobile side so far.  The only thing Intel has going for them currently is a tiny single-threaded lead.  If they lose that to Zen 3, which they almost certainly will if improvements aren't made, they will have nothing.  That means they're stuck pushing 8 cores harder, because if they go for 10 and are forced to drop clockspeeds, that's exactly what's going to happen.  Either way though they fail to make any meaningful progress at catching up.  AMD has more multi-threaded, more efficiency, less cost, more memory bandwidth, more PCIe bandwidth, fewer security issues, and basically everything else going for them right now.

 

Intel's only good market is in the mobile space because AMD's offerings there are a generation behind and only target the mid-low end.  It's going to be many years before Intel can hope to compete in the high-end desktop and server space again, so their best chance is to keep pushing for gaming supremacy while they work on a broader solution (their own ryzen, or sandy bridge 2.0) to the rest of their issues that will let them actually compete with and beat AMD.

I get the feeling Intel is shifting (further) into what it seems they feel to be the most profitable segment - mobile computing. At least consumer end, anyway.

 

Desktop powerhouses are something we as enthusiasts love, but like video games, mobile is the place to put your money. Most of the consumer demographic doesn't care about powerful desktop PCs. They'll probably work to get a powerful SoC design and go from there.

 

IMO, the days of big, consumer owned and powerful PCs are numbered, sadly.

[vanished]

1 minute ago, Plutosaurus said:

I get the feeling Intel is shifting (further) into what it seems they feel to be the most profitable segment - mobile computing. At least consumer end, anyway.

 

Desktop powerhouses are something we as enthusiasts love, but like video games, mobile is the place to put your money. Most of the consumer demographic doesn't care about powerful desktop PCs. They'll probably work to get a powerful SoC design and go from there.

 

IMO, the days of big powerful PCs are numbered, sadly.

It's certainly a smart move for them at the moment.  Whether that's the long term focus I couldn't say, but the needs and restrictions of the mobile environment lines up nicely to take advantage of what they can do well and hide what they can't.  Not perfectly mind you, but it's the best option imo

[Harry Voyager]

From what we've seen from Ice Lake, the Sunny Cove architecture is very good, but the 10nm process is letting it down. 

 

My expectation is we'll see something like the early Athlon vs Pentium 4 days pan out, with Intel getting to parity in 2020, and 2021 being a game of "who drops the ball first".

 

Of course, all of the is dependent on Intel getting good clock speeds out of their Cove chips. 

 

I find myself wondering how the Sunny Cove architecture handles chiplettes? Part of how AMD tamed 7nm is to split off parts of the package into 14nm parts. 

[redbread123]

Smh, I keep hearing "5 nm this" "14 nm that" "HAhahhHAha 14Nm+++++++++" Let me tell you, the manufacturing process by TSMC and intel's fabs are measured differently. If 7 nm is so superior to intel's "14nm" then why is the performance speeds down to the cores? It's not performance based race, it's price. Intel fabrication is expensive because they do it themselves while AMD and other companies like nvidia, etc. use TSMC. The reason AMD is cheaper is because they use chiplets which have higher yields. This post isn't meant to fanboy intel but it's to not completely shutdown intel as an option just because they are on a meaningless number of nm. 9700k and 3700x are around same price, but the only difference is 16 threads which is only really useful during rendering,etc. Amd has been doing good lately and they are a competitor now, but it's stupid to hear people basing the performance of the word Nanometer.

[Mister Woof]

1 hour ago, redbread123 said:

Smh, I keep hearing "5 nm this" "14 nm that" "HAhahhHAha 14Nm+++++++++" Let me tell you, the manufacturing process by TSMC and intel's fabs are measured differently. If 7 nm is so superior to intel's "14nm" then why is the performance speeds down to the cores? It's not performance based race, it's price. Intel fabrication is expensive because they do it themselves while AMD and other companies like nvidia, etc. use TSMC. The reason AMD is cheaper is because they use chiplets which have higher yields. This post isn't meant to fanboy intel but it's to not completely shutdown intel as an option just because they are on a meaningless number of nm. 9700k and 3700x are around same price, but the only difference is 16 threads which is only really useful during rendering,etc. Amd has been doing good lately and they are a competitor now, but it's stupid to hear people basing the performance of the word Nanometer.

I'm an Intel "fanboy" and I think it's stupid intel segments hyperthreading/smt period, and even worse that it's absent from the flagship i7 sku.

 

It's an efficiency feature and should be standard in 2019, especially on the i7.

 

The 9700k was a mistake. Defective 8/16s should have just been segmented down into i5s, but Intel wanted to compromise it's lineup's integrity to save a few bucks, and pulled an Nvidia with the i9.

 

And if rumors are to be believed on the 10th generation, Intel agrees.