Intel Shows Off 28 Core, 56 Thread Core-X HEDT Processor For Enthusiasts, In Market Q4 ’18

[coercitiv]

It probably does not run at 5GHz all core stock, and they overclocked it using watercooling to show off.

AMD presentation is tomorrow. Wonder if they're overclocking as we speak

It's days like this that showcase just how boring the CPU scene was before Mar 2017.

 

[Hitman928]

It probably does not run at 5GHz all core stock, and they overclocked it using watercooling to show off. Kind of like in the clockspeed war days they showed us overclocked Pentium 4 chips.

Yeah, I would be completely shocked if this thing was 5 GHz all boost clock out of the box. Rumors are that they used a custom loop with chilled water for the system. Kind of hard to pack that in a cpu box or have that cooling requirement listed on the side.

Yeah maybe they only show maximum boost clock, but that still freaking awsome

Going off the benchmark score, it seems to be running all cores at 5 GHz, but again, pretty sure that is a substantial overclock to get there.

 

[Bouowmx]

Will LCC and HCC HEDT move to socket P, or will socket R get the next generation?

 

[EXCellR8]

Looking forward to AMD's update on 2nd Gen Ripper chips. There's obviously no way these Intel HEDT processors are going to be priced to compete with TR4. That is some impressive hardware though, minus all of those tiny fans. If Lian-Li chassis weren't so darn expensive I'd probably only buy them exclusively. I like subtle and sleek.

 

[LTC8K6]

Well, Intel uses base frequency to calculate TDP, but the chip will not run at base frequency very often.

If all the cores run at 5.0 on chilled water cooling, can we extrapolate what they'd do with the best air cooling?

How much is the water cooling likely to give you?

Also, could we guess what it would do on LN2?

 

[wahdangun]

Yeah, I would be completely shocked if this thing was 5 GHz all boost clock out of the box. Rumors are that they used a custom loop with chilled water for the system. Kind of hard to pack that in a cpu box or have that cooling requirement listed on the side.



Going off the benchmark score, it seems to be running all cores at 5 GHz, but again, pretty sure that is a substantial overclock to get there.

What I mean is, maybe intel will introduce their own version of speculative boost clock, like AMD cpu and nvdia gpu. So as long as the cooling/power is sufficient they will boost to the maximum regardless of TDP.

 

[jpiniero]

If the 5912 score is the stock frequency, the FCT is 3.9 or 4. This at 300W.

 

[Bouowmx]

Point of reference: Xeon Platinum 8180 @ 3340.8 MHz: 5010 cb

 

[epsilon84]

Why all this talk about whether it can hit 5.0GHz? If you are getting this CPU it would be for the core count, not because the architecture (without thermal limits) is capable of 5GHz+.

If you are going to the trouble of overclocking such a chip, I would think 4.7 - 4.8GHz would be a much more realistic goal, based on my experience overclocking the 8700K. Pretty much every 100MHz over 4.7/4.8 requires an extra 0.05V which adds about 10% power per 100MHz, it's not really worth it for a 28 core CPU.

 

[WingZero30]

With the introduction of this 28 core cpu does this mean its the end of the line for x299 and there won't be any further refreshes/releases on the current 2066 platform ?

 

[PliotronX]

Why all this talk about whether it can hit 5.0GHz? If you are getting this CPU it would be for the core count, not because the architecture (without thermal limits) is capable of 5GHz+.

If you are going to the trouble of overclocking such a chip, I would think 4.7 - 4.8GHz would be a much more realistic goal, based on my experience overclocking the 8700K. Pretty much every 100MHz over 4.7/4.8 requires an extra 0.05V which adds about 10% power per 100MHz, it's not really worth it for a 28 core CPU.

Something about having a cake and eating too

 

[jpiniero]

With the introduction of this 28 core cpu does this mean its the end of the line for x299 and there won't be any further refreshes/releases on the current 2066 platform ?

Cascade Lake-X is still coming. I guess it's still unclear if Intel is going to refresh the entire Skylake-X lineup some more with more cores/tdp and then release Cascade Lake-X sometime in 2019 or just release Cascade Lake-X with 18 cores max alongside the 3647 Skylake-X.

 

[JoeRambo]

This is going to be one sweet CPU, probably the swan the song of monolithic CPUs. One die, without Numa nonsense or multi ring penalties, ~38MB of rather uniform L3 cache, 5Ghz clocks, six channel DDR4 with 150GB/s bw. One can dream about it

 

[Hitman928]

What I mean is, maybe intel will introduce their own version of speculative boost clock, like AMD cpu and nvdia gpu. So as long as the cooling/power is sufficient they will boost to the maximum regardless of TDP.

Gotcha. I don't know about Nvidia, but AMD still has an upper limit in frequency at default settings and has an upper power limit (130 W for Ryzenn 2xxx series if I remember correctly). The upper power limit can be turned off, at least on some boards, but is on by default AMD settings.

 

[LTC8K6]

What I mean is, maybe intel will introduce their own version of speculative boost clock, like AMD cpu and nvdia gpu. So as long as the cooling/power is sufficient they will boost to the maximum regardless of TDP.

Intel has a introduced a couple new TB levels recently. TB 3.0 Max, then adding a second favored core to 3.0 Max, and Thermal Velocity Boost to the newest mobile chips.

 

[wahdangun]

Gotcha. I don't know about Nvidia, but AMD still has an upper limit in frequency at default settings and has an upper power limit (130 W for Ryzenn 2xxx series if I remember correctly). The upper power limit can be turned off, at least on some boards, but is on by default AMD settings.

Yeah forgot about 130 tsp limit, but it's can be disabled, AMD called it boost overdrive.

Intel has a introduced a couple new TB levels recently. TB 3.0 Max, then adding a second favored core to 3.0 Max, and Thermal Velocity Boost to the newest mobile chips.

But that's still predetermined boost, quite different implementation.

 

[moonbogg]

It seems this is one of their bread and butter server platforms? This is cool and everything, but the core counts are entering the area where it only matters for servers and maybe a few individuals with some really unique work loads. These people probably already have a multi socket xeon system or something like that anyway, right? If you are the kind of person who needs 28 cores, you probably already have access to something like that. I am feeling like this is just a wee wee swinging contest between Intel and AMD at this point, which is fine of course. 28 and 32 core CPU's for consumers though...my eyes honestly started to roll at 16 and 18 cores.
Hopefully this means 8/16 chips will be seen as cheap entry level crap and the best of them will soon go for about $300. I'll take my overkill in 8/16 flavor.

 

[Jan Olšan]

https://www.tomshardware.com/news/intel-cpu-28-core-5-ghz,37201.html

For instance, we tested the 16-core Core i9-7960X and could only reach up to 4.3GHz before the processor was overwhelmed by the intense thermal output and 400W of power consumption. In fact, we've pushed past 350W of power consumption during some of our overclocked tests with the smaller 10-core Core i9-7900X.

EDIT: According to this image we sourced from Engadget's compressed keynote video on YouTube, Intel apparently was running some sort of closed-loop cooling that required insulating material around the tubing. This could be a multi-stage phase cooler (sub-zero cooling), or possibly a more mundane water chiller, under the table. We also spot more shielding over the long rectangular waterblock and what appears to be six sticks of RAM flanking the processor on each side. That would imply this platform is based on the (until now) enterprise-class LGA3647 socket. We'll dig for more details and update as necessary. This means the processor could be a variant of the $8,700 Xeon Platinum Scalable processor we reviewed here, albeit with an unlocked multiplier. (That doesn't bode well for pricing).

https://twitter.com/witeken/status/1004081461513814022

It's 2.7GHz base, but yeah, 14(++) is making Intel $_$.

I suspect the chip might not be as impressive as many people think (or better said, expectations are probably inflated) - if base clock is just 2.7 GHz, then that isn't much higher than Xeon Platinum 8180 - a year and something after its introduction - and that's with server SKUs being binned conservatively with some headroom, while HEDT less so.
Sure, overclock will make it much faster, but if you run some actual task like encoding that takes longer than Cinebench, the power consumption and cooling is bound to be a challenge. I would not be surprised if practical OC was way way lower than 5.0 GHz on this. Probably quite sub-4GHz. Add to that a high price of the CPU and cooling needed, and we're probably looking at a close to purely halo/benchmark part.

It could be of use to some people that would otherwise be able to shell for highend Xeons, I guess. Maybe it should be seen as something like Titan V? Price could probably be similar too, $3000-3500. I think Nvidia and Intel might actually be watching each other's steps in the pricing-creep action in highend hardware and are half checking what can the other company get away with it, half one-upping it afterwards.

 

[WingZero30]

Cascade Lake-X is still coming. I guess it's still unclear if Intel is going to refresh the entire Skylake-X lineup some more with more cores/tdp and then release Cascade Lake-X sometime in 2019 or just release Cascade Lake-X with 18 cores max alongside the 3647 Skylake-X.

Cascade Lake-X is 3647 as shown today. Unless intel splits the Cascade Lake-X into both 2066 & 3647 then I don't think Skylake-X on 2066 would get any further refreshes or successor on the X299 platform.

 

[Timmah!]

Honestly, did not see this coming, i mean, not this year already. IMO this means there will be 24C-32C TR2 this year...or at least Intel believes there will be.

 

[IntelUser2000]

I think Nvidia and Intel might actually be watching each other's steps in the pricing-creep action in highend hardware and are half checking what can the other company get away with it, half one-upping it afterwards.

You know how Intel and Nvidia both said things like "they outpace the market" and "record revenue", despite IDC and Gartner reporting eternal PC market declines?

Well, this is how they do it. They use marketing teams to upsell you the higher end chips, which increases ASP. Remember some guys complaining how Nvidia took their x70 die but call it x80? That's probably true too. That's a hidden price creep. Ultrabooks? 2-in-1s? Surface Pros? They cost like two thousand dollars but marketed as the best thing since sliced bread. And they make it really fancy too, like a fashion accessory. Fashion items sell for a illogical price.

Computers used to go down in price for decades. I remember a Dell PC flyer ad that had the high end system up for $7K! No one would buy such a system nowadays. There's probably greed mixed in as well, but part of it is to counter the decline in PC volume. You make the same money whether you sell 350 million units for $100 or 250 million units for $140.

Cascade Lake-X is 3647 as shown today.

Where have you seen Cascade Lake? Are links out there that show pictures? Or are you saying the 28-core, 5GHz water-cooled demo setup is Cascade Lake?

 

[Zucker2k]

https://www.tomshardware.com/news/intel-cpu-28-core-5-ghz,37201.html





https://twitter.com/witeken/status/1004081461513814022



I suspect the chip might not be as impressive as many people think (or better said, expectations are probably inflated) - if base clock is just 2.7 GHz, then that isn't much higher than Xeon Platinum 8180 - a year and something after its introduction - and that's with server SKUs being binned conservatively with some headroom, while HEDT less so.
Sure, overclock will make it much faster, but if you run some actual task like encoding that takes longer than Cinebench, the power consumption and cooling is bound to be a challenge. I would not be surprised if practical OC was way way lower than 5.0 GHz on this. Probably quite sub-4GHz. Add to that a high price of the CPU and cooling needed, and we're probably looking at a close to purely halo/benchmark part.

It could be of use to some people that would otherwise be able to shell for highend Xeons, I guess. Maybe it should be seen as something like Titan V? Price could probably be similar too, $3000-3500. I think Nvidia and Intel might actually be watching each other's steps in the pricing-creep action in highend hardware and are half checking what can the other company get away with it, half one-upping it afterwards.

I'd be surprised if that chip runs at anything near 2.7GHz with the proper cooling at stock. Intel catches a lot of flack, unfairly, from people using high overclock, stress-testing (AVX) numbers to justify their skewed arguments about power consumption, but what other manufacturer even comes close to the fmax/ipc numbers which Intel has continually laid down over the last 8-year stretch, TIM and all? The other x86 chipmaker of note falls far short of these numbers with just 4 cores. This is an engineering feat, both in the silicon process and architectural design. Core and the 14++ process have just consolidated their legendary status in my book.

 

[jpiniero]

Cascade Lake-X is 3647 as shown today. Unless intel splits the Cascade Lake-X into both 2066 & 3647 then I don't think Skylake-X on 2066 would get any further refreshes or successor on the X299 platform.

The 28 core at least is almost certainly Skylake-SP XCC die and not Cascade Lake.

 

[TheGiant]

So the future of PC computing is more of the moarz coarz a oced above the point of balance of power/absolute performance

WTB personal nuclear reactor (balcony 1,5x7m)

I am waiting for the benchmark setup that shows us- if you don't have 16cores, your standard OS +bloatware will slow you down and your PC is unusable-recommended 24 cores....but yes, I can imagine that this is the strategy to increasy dying PC market....

 

[tamz_msc]

I'd be surprised if that chip runs at anything near 2.7GHz with the proper cooling at stock. Intel catches a lot of flack, unfairly, from people using high overclock, stress-testing (AVX) numbers to justify their skewed arguments about power consumption, but what other manufacturer even comes close to the fmax/ipc numbers which Intel has continually laid down over the last 8-year stretch, TIM and all? The other x86 chipmaker of note falls far short of these numbers with just 4 cores. This is an engineering feat, both in the silicon process and architectural design. Core and the 14++ process have just consolidated their legendary status in my book.

You don't need to use stress-test apps to justify high power draw while utilizing AVX on these CPUs. Just run a couple of simultaneous x265 transcodes and you'll get similar power draw as any other stress-testing application. Far more realistic use-case and will quickly identify stable overclocks within a few minutes instead of hours.

There is a reason why they used Cinebench and exotic cooling. (Probably) Hiring a pro overclocker behind the scenes to keep tabs while the benchmark is running does not make this a feat of engineering.