Intel Core i9-10900K 10-core Processor and Z490 Chipset Arrive April 2020

[Chrispy_]

Three questions (well okay, four because Q2 is a two-parter):

1. Does this finally have enough hardware mitigations for specultaive-execution attacks? Without them, all these hyperthreading improvements are meaningless because the vulnerability patches hurt performance and for a secure system, HT needs to be disabled in its entirety.
2. Why do we need a new socket? It's still using DDR4, not DDR5 (that's AMD's excuse for changing to socket AM5) and it's still using PCIe 3.0. Is Intel unable to physically fit more than 8 cores on a package at 14nm?
3. Is 10C the upper limit or are there hints that Intel have 12C and 16C models in the pipeline for later in 2020?

 

[The Egg]

Should have kept it on z390, at least people with existing boards would have bought it. But now, nobody cares.

Yeap. Folks such as myself might have considered them, were they a drop-in on current boards. Forget it now.

 

[Patriot]

price it <$300 and I might be interested, otherwise fuck it.

 

[notb]

1. Does this finally have enough hardware mitigations for specultaive-execution attacks? Without them, all these hyperthreading improvements are meaningless because the vulnerability patches hurt performance and for a secure system, HT needs to be disabled in its entirety.

It'll have more than earlier generations. Hard to say at the moment.
Disabling HT does NOT make a system secure. CPU has other security flaws that haven't been found or revealed yet.

2. Why do we need a new socket? It's still using DDR4, not DDR5 (that's AMD's excuse for changing to socket AM5) and it's still using PCIe 3.0. Is Intel unable to physically fit more than 8 cores on a package at 14nm?
3. Is 10C the upper limit or are there hints that Intel have 12C and 16C models in the pipeline for later in 2020?

They're adding pins. This could be because of added PCIe, higher IGP transfers or other features unknown at this point.
Socket/package size is the same as in LGA1151 and LGA1150.

As for socket size being a limiting factor for core count...
A 4-core SB die was 24% larger than an 8-core CL (below, from wikichip).
Without GPU, 12 cores should fit in SB's 216mm2.
And, of course, both cores' and cache's shape/size can change, so only Intel and partners know the answer at this point.

 

[ppn]

We can get 16 core, but it would sacrifice the iGPU, it can't fit, outside the limits. and it will draw 500 watts in prime 95 at 5.00Ghz, 9900K draws 250 watts.

For now, only gets 10 core. next year 12, every year is +2 then +2.

 

[torsoreaper]

Still on DMI 3.0, a 5 year old technology... Great job Intel, way to push the envelope.

 

[GoldenX]

Still on DMI 3.0, a 5 year old technology... Great job Intel, way to push the envelope.

Run, the blue squad is coming for you.

 

[Steevo]

I figured they would glue some dies onto the NB for morez corez, but that would be different than their strategy of "lets make a new socket, and the fanbois will rejoice as they pay" or whatever it is they are doing now since they have no new process and are also starting production of older chips.

 

[Nater]

Still on a 6700K at work. Boss has green-lit a new build for the end of January. I can't see waiting around for this. Probably go 9900KS or maybe 3900X. Lean Intel because SolidWorks wants clockspeed/IPC, not cores.

Not seeing any reason to wait till April to spend even MORE capital on 2 more cores I don't need.

 

[chodaboy19]

How many PCIe lanes?

 

[Chrispy_]

View attachment 139016

We can get 16 core, but it would sacrifice the iGPU, it can't fit, outside the limits. and it will draw 500 watts in prime 95 at 5.00Ghz, 9900K draws 250 watts.

For now, only gets 10 core. next year 12, every year is +2 then +2.

10nm would give them enough space
also, LOL; 10nm.

 

[efikkan]

Why do we need a new socket? It's still using DDR4, not DDR5 (that's AMD's excuse for changing to socket AM5) and it's still using PCIe 3.0. Is Intel unable to physically fit more than 8 cores on a package at 14nm?

Mostly to redesign it to be more suitable for higher power draw.
I don't think it will be much larger.

Is 10C the upper limit or are there hints that Intel have 12C and 16C models in the pipeline for later in 2020?

Physically, Intel could easily have fitted 16 cores on this socket (in a 4x4 mesh), but it's not really practical nor is it really needed. On 14nm, their i9-9900K is already throttling unless the power limit is removed.

While it's good to have more than 4 cores on the mainstream platform, we really don't need 12-16 cores for non-server software (except for special use cases). Most applications will scale much better on a 30% faster CPU vs. a CPU with 30% more cores. Intel need to focus on bringing their new architectures to the market rather than winning "the core race".

 

[medi01]

to match iPad

Come on, which freaking ipad... Did you want to say "a tablet"?

 

[R0H1T]

They're adding pins. This could be because of added PCIe, higher IGP transfers or other features unknown at this point.
Socket/package size is the same as in LGA1151 and LGA1150.

You know that's BS, a large number of OC records on 8700k were on previous gen Zxxx boards.

For what exactly, they're still limited by DMI 3.0 IIRC.

Nope.

The only "unknown" feature they're adding is CNVi & I question it's much touted utility on desktops.

 

[Tomorrow]

Disappointed with Intel recently but still rest assured it will still and easily best any ryzen 4xxx cpu's launched next year easily and with a bigger margin than this year as far as gaming is concerned with no misleading advertisement on the product.

Ladies and genteman. We have a Nostradamus here. I seriously hope this is a deliberate troll post and you're actually not that stupid.

 

[notb]

We can get 16 core, but it would sacrifice the iGPU, it can't fit, outside the limits. and it will draw 500 watts in prime 95 at 5.00Ghz, 9900K draws 250 watts.

It would not draw double of 9900K in the same way 3900X doesn't draw 250W at full load. It would be limited and utilize clever load allocation.

Also, why would anyone run 9900K at const 5GHz? It makes absolutely no sense. And that's exactly why people think Intel CPUs draw so much power.

 

[scouserpcgamer]

I am hoping that PCIE 4.0 same as AMD boards are supported, I think it will be massive mistake if the new motherboards only support PCIE 3.0 for SSD’s

 

[Darmok N Jalad]

It'll have more than earlier generations. Hard to say at the moment.
Disabling HT does NOT make a system secure. CPU has other security flaws that haven't been found or revealed yet.

They're adding pins. This could be because of added PCIe, higher IGP transfers or other features unknown at this point.
Socket/package size is the same as in LGA1151 and LGA1150.

As for socket size being a limiting factor for core count...
A 4-core SB die was 24% larger than an 8-core CL (below, from wikichip).
Without GPU, 12 cores should fit in SB's 216mm2.
And, of course, both cores' and cache's shape/size can change, so only Intel and partners know the answer at this point.

View attachment 139015

Except Intel probably wants to produce today’s CPUs at less than Sandy Bridge sizes, as production costs and overhead costs increase over time. Not only that, Intel can’t price these like they want to, so the smaller the node, the better.

 

[candle_86]

Hmm i3, looks like i7 from 2 years ago, this should tank the used market entirely

 

[EarthDog]

While it's good to have more than 4 cores on the mainstream platform, we really don't need 12-16 cores for non-server software (except for special use cases). Most applications will scale much better on a 30% faster CPU vs. a CPU with 30% more cores. Intel need to focus on bringing their new architectures to the market rather than winning "the core race".

Intel doesn't seem to be focused on a core race. They aren't stuffing 18c/36t CPUs down the mainstream pipe when 95% of consumers are good with half that (and will be for the next few years). I think their next gen will go up to 10c/20t on the mainstream platform. Who knows moving forward.

Blame AMD for starting a core race in the first place. They are the ones who brought this HCC crap to the mainstream segment.

 

[candle_86]

So does this mean Celeron can get an upgrade to 2c4t finally if Pentium is 4/4 and i3 is 4/8

 

[notb]

So does this mean Celeron can get an upgrade to 2c4t finally if Pentium is 4/4 and i3 is 4/8

Celerons serve their purpose in cheap, low power applications (general servers, NAS, HTPC, NUC etc).
There's very little need for more multi-thread performance in this segment... and even less will to sacrifice single thread potential.

It's hard to say how Intel will segment their next generation into traditional product lines. They'll have to stretch something. But it's extremely unlikely that 2-core, non-HT CPUs won't be included.

 

[Darmok N Jalad]

Celerons serve their purpose in cheap, low power applications (general servers, NAS, HTPC, NUC etc).
There's very little need for more multi-thread performance in this segment... and even less will to sacrifice single thread potential.

It's hard to say how Intel will segment their next generation into traditional product lines. They'll have to stretch something. But it's extremely unlikely that 2-core, non-HT CPUs won't be included.

I dunno, once they drop so far, the Atom-based architecture is able to step in. I think a 2C/2T Core-based architecture is finally being eclipsed by Gemini Lake in most tasks. I had an Apollo Lake quad core that could handle a 4K stream. It’s a low bar, but we’re talking about really small dies and super cheap prices.

 

[Prima.Vera]

14+(+)nm, PCI-E 3.0, 16xPCI-E lines, new chipset.....

 

[Zach_01]

Oh come on people... give Intel a slack... Dont have enough "in the head" already struggling to spit a modern working chip/platform out? ...need this kicking while its down? yes!

Come on, which freaking ipad... Did you want to say "a tablet"?

No he said and ment iPad... If you never own one then you dont know how these "tablets" work and behave.