18-core Core i9

[Gfted1]

Intel unveils monster 18-core Core i9: 'First teraflop-speed' consumer PC chip.

[Bartimaeus]

Probably only on one of their useless Extreme sockets, though...and $2000 on top of it...

Edited May 30, 2017 by Bartimaeus

[Gfted1]

Yeah its crazy expensive. Costs more than my entire current PC.

[Malcador]

Great for servers. Lord knows we could use one of those at work

[Sarex]

Probably only on one of their useless Extreme sockets, though...and $2000 on top of it...

 

Why useless?

[Zoraptor]

X99 had big initial problems, and still has a very high relative failure rate for its motherboards. They're also expensive.

 

Great for servers. Lord knows we could use one of those at work

 

No ECC support though, at this stage at least. Also no soldered heat dispersal, though hopefully that won't matter so much for lower clocks.

[Sarex]

X99 had big initial problems, and still has a very high relative failure rate for its motherboards. They're also expensive.

 

No ECC support though, at this stage at least. Also no soldered heat dispersal, though hopefully that won't matter so much for lower clocks.

 

Huh guess I got lucky with my rig, although I did get the rampage v board.

 

Also as far as I know all E variants are soldered to their lid.

[Zoraptor]

The i9s (so kabylake-e and skylake-e) aren't soldered, some of the Engineering Samples have already been delidded- yes, despite the previous enthusiast chips all having solder. I guess they could still change it since it seems an odd choice to make to save a little cash, and since the 7700k's heat issues have gained some traction. They do recommend watercooling too, so there clearly are some concerns about thermals.

 

The failure rate on x99 is just relatively high, not absolutely high- about 5x that of a z170/ z97, iirc. But you'd still be unlucky to get an actual failure.

[Keyrock]

Switching from solder to the toothpaste they use for their consumer-grade chips for a $2000 chip?  For shame, Intel.

[Sarex]

I can't believe it's for saving cost though, I mean how much can they really save on it?

[injurai]

When 18 core chips are standard and consumer friendly, we may finally see that revolution in AI that has been a decade over due. Still probably a decade out. Maybe more.

 

Enthusiasts never truly win in the end when nobody builds to their specs.

[Zoraptor]

The unconfirmed price for the 16/32 AMD offering is ~$850, which is highly competitive to say the least and roughly half the 16/32 skylake-e price. That pricing seems fairly likely to be accurate though, as AMD have dropped prices on the r7 8 cores to get some separation (a week after I bought a 1700, no surprises there).

[mkreku]

https://www.youtube.com/watch?v=TWFzWRoVNnE

 

I kind of agree with Linus here. Intel are panicking.

[Zoraptor]

And he didn't even mention the solder issue either.

 

The i9/x299 situation makes a lot more sense if Intel planned the whole thing under the assumption there would be no external competition and that their chips would be competing with each other rather than another company's products. The variable PCI lane numbers, dual/quad channel RAM depending on processor and no ECC make sense if you see the competition as being your own lower range chips and need reasons to sell the expensive ones. But if the competition suddenly becomes external you end up with a decidedly odd looking and inconsistent line up with gimped features relative to the competition instead.

 

There's definitely some panic there when the flagship chip won't be available at launch (and may not even make 2017).

[injurai]

Let's not forget that the current processors are still part of a family that launched when Intel was still on the tick-tock cycle. Personally I'm waiting to take a sniff of hot silicon off the die until Intel kicks off a proper process-architecture-optimization cycle. So basically Tigerlake at the earliest which is 2019.

 

Intel got caught with their pants down, because AMD while still lagging is killing it in the price per power arena. Big data centers prefer to scale linearly and throw cheaper power efficient blades at their problems. AMD has locked in enthusiast consumer gaming devices, and is on the path to not only continue winning those contracts. But will encroach on Intels annual margins by selling to enthusiast "whales."

 

At this point I think Intel is in hail-mary mode. Basically they got ahead enough that they thought they'd try to revolutionize themselves before AMD caught up, all while sitting pretty for their investors. Only thing is Intel may have cut too many costs and bleed talent into the competition. Plenty of young graduates with fresh research hearing horror stories of Intel being swepted up by AMD. Nevermind AMD unilaterally integrated with it's GPU department. A far more promising company to work at than just Intel or just NVidia.

[teknoman2]

When 18 core chips are standard and consumer friendly, we may finally see that revolution in AI that has been a decade over due. Still probably a decade out. Maybe more.

 

Enthusiasts never truly win in the end when nobody builds to their specs.

software companies ALWAYS make stuff based on what the surveys show as being the majority in the hardware specs. if 80% of the consumers have a R5 1500x / i5 7500 and RX 480 / GTX 1060 that's the target hardware for the software makers, even if stuff like R7 2800X / i7 8900k and RX 695X / GTX 1190Ti are out there for enthusiasts.

[Terminator]

Look there exist rich people who own Ferrari car. A chip like that is mainly for rich people or someone that really need super power for major server for some reason. There are also some people that are not rich, but really want new technology and want to own it despite it being expensive.

 

Really most game run perfectly fine for consumers already on Intel I5 with 4 cores. Intel I7 might be slightly advantage over I5 in some rare games. That said if you go down from Intel I5 to say Intel I3 with 2 cores then you go down really much in gaming performance in many modern games.

Edited June 9, 2017 by Terminator

[teknoman2]

Look there exist rich people who own Ferrari car. A chip like that is mainly for rich people or someone that really need super power for major server for some reason. There are also some people that are not rich, but really want new technology and want to own it despite it being expensive.

 

Really most game run perfectly fine for consumers already on Intel I5 with 4 cores. Intel I7 might be slightly advantage over I5 in some rare games. That said if you go down from Intel I5 to say Intel I3 with 2 cores then you go down really much in gaming performance in many modern games.

however, games are starting to scale better with cores if you look at the relative results between i5 and r5 on older and newer games. the newer the game, the smaller the gap in performance between the 2 CPUs

[Zoraptor]

At this point I think Intel is in hail-mary mode. Basically they got ahead enough that they thought they'd try to revolutionize themselves before AMD caught up, all while sitting pretty for their investors. Only thing is Intel may have cut too many costs and bleed talent into the competition. Plenty of young graduates with fresh research hearing horror stories of Intel being swepted up by AMD. Nevermind AMD unilaterally integrated with it's GPU department. A far more promising company to work at than just Intel or just NVidia.

 

Intel's R&D budget is higher than AMD's entire turnover- it's certainly sufficient (or should be) to stay well ahead. They also have a host of research agreements with universities and the like.

 

Their big mistake has been focusing on going after ARM's low power business which so far has been an utter failure. It was an understandable mistake though, that's the one market segment they have no real stake in so an obvious target for expansion. Also AMD somewhat sandbagged Ryzen's expected performance increase at 40% above Bulldozer when it was closer to 60% increase in reality, and that transforms it from competitive under certain circumstances to highly competitive under almost all circumstances. AMD's cluster based architecture is also exactly the sort of thing that a monolithic company like Intel would have difficulty coming up with when in a dominant position as there would be no incentive to try new approaches.

[LadyCrimson]

"18 cores" sounds good in theory (although from that posted video I guess it may not be as cool as it sounds) but I can't see myself ever wanting to spend that much on a CPU for a home PC, ever. Even if it also baked and served me cookies.

[Bartimaeus]

What, really? Do you know how much time baking and serving cookies can take? Definitely worth the investment.

[SonicMage117]

I would say not worth having for gaming. Nothing will take advantage of it or use it properly (optimally at least) for the some years to come.

 

There are different perfomance values for single core, dual-core, quad-core... meaning that you won't benefit from 18-cores just yet realistically. Games are just now getting optimized for quad core cpu's and next is octo-core cpu's.

 

Workloads on the other hand and production.... now that's a different story. You might actually get close to reverse engineering and emulating PS3 to run games at 15fps with that kind of power!

Edited June 21, 2017 by SonicMage117

[teknoman2]

interesting stuff.

[Keyrock]

Interesting that Bill Gates talks about Intel's despicable strong arm tactics in his book. I wonder if he also talks about how Microsoft used many similar tactics to gain and hold on to market dominance?

[Azdeus]

Yes, that'd be interesting to know.

 

I liked the video, and I am actually happy that I've been running AMD processors since I had a K6/2 450