Intel Skylake / Kaby Lake

[formulav8]

That remains to be seen. No official confirmation either way yet.

Wow, quick reply, thanks.

Still stinks Intel won't say anything. No reason for them not to support it though. Same ole 14nm based arch and everything overall it seems. If they don't it only points to 1 thing I guess. I hope for many peoples sake they will update the firmware and not do the typical artificial stuff.

 

[VirtualLarry]

If they don't it only points to 1 thing I guess. I hope for many peoples sake they will update the firmware and not do the typical artificial stuff.

I hope that they do segregate the chipsets, if nothing else, to illustrate Intel's greed. And then I hope ASRock and Biostar (among others) figure out a way to hack CFL to work with 100- and 200-series chipset boards anyways, against Intel's wishes.

 

[TheF34RChannel]

Wow, quick reply, thanks.

Still stinks Intel won't say anything. No reason for them not to support it though. Same ole 14nm based arch and everything overall it seems. If they don't it only points to 1 thing I guess. I hope for many peoples sake they will update the firmware and not do the typical artificial stuff.

I hope they'll make it work and save me money (I've got a very good board now) so I'll only have to chuck in the new CPU and not bother with a fresh Windows installation etc., but it's Intel = I have little hope.

 

[DrMrLordX]

yes

That chart doesn't show clockspeeds at all. There is actually software out there that runs slower with HT enabled right off the bat without any thermal throttling (depending on the iteration of HT), and you've done nothing to show that SpecAPC Render is losing performance due to HT from reductions in clockspeed.

Try again?

 

[beginner99]

I have two m.2 drives in Raid 0 as my boot drive. The performance is less than awesome. I have two Asus m.2 PCIe adapter cards. Can I install them into two pcie slots and still use them in raid and be bootable?

Maybe. The feature is called VROC (virtual raid over CPU) because these lanes go directly to the CPU and previous raid versions form intel are implemented in the chipset. VROC only allows RAID0 for free else you need to pay intel for the dongle ($100) to enable RAID1 or 5 on PCIe lanes from cpu.

But the main issue is it will only work with intel drives, even raid 0.

EDIT:

If you haven't bought Skylake-X yet I would wait for threadripper and see if it offers such a feature albeit I doubt it.

 

[2blzd]

They did.

I am 99% sure this was the intended pricing lineup for SKX/KBX:

7900X: $1699
7820X: $1099
6C12T with full PCIe: $649
7800X: $449
7740X: $380
7640X $280

Ya, I was aware of that too..But it's still not enough.

I was hoping for Broadwell-EP price drops since technically Ryzen launched before Sky-X.

I should be able to by a 6950x for $899.

 

[wildhorse2k]

I just reran the test with flush caching turned off. When I switched ports, it turned it on by default (thought it was a new drive).

Is this when running M.2 through PCH? Could you remove the other M.2 drivers and retest?

As others noted 4K performance is still not ok. More interesting is that with Kabylake-X its ok.

 

[wildhorse2k]

They are being conservative to protect the value of the more expensive 14C/16C/18C models. There's a 12C Xeon Gold at 165W TDP with 3.2 GHz base / 3.9 GHz all-core Turbo. I still expect Core i9-7920X to have an all-core Turbo equal/above 3.4 GHz, so it will no doubt be faster than Core i9-7900X in MT tasks.

It seems that these are the likely candidates for the 12/14/16/18C SKUs:
12C: Xeon Gold 5118 2.3/3.2GHz 105W
14C: Xeon Gold 6132 2.6/3.7GHz 140W
16C: Xeon Gold 6142 2.6/3.7GHz 150W
18C: Xeon Gold 6150 2.7/3.7GHz 165W

AnandTech has already provided the Turbo modes - none of these parts exceed 3.4GHz on all core non-AVX workloads.

I think its a good idea to use low base clocks, as that allows Intel to use higher few core turbo clocks on these CPUs. Notice that the highest turbo on these Xeons is 3.7Ghz. That is quite low. I would prefer 2.2Ghz base clock on 18C and 4.3Ghz top turbo, with 3.4Ghz all core turbo.

All core turbo is quite high on 16C/18C, about the same as competition.

 

[frozentundra123456]

I hope that they do segregate the chipsets, if nothing else, to illustrate Intel's greed. And then I hope ASRock and Biostar (among others) figure out a way to hack CFL to work with 100- and 200-series chipset boards anyways, against Intel's wishes.

So you would rather see intel look bad than have a positive feature for the consumer? Really Larry? I assume you are being at least partially sarcastic, but not the kind of post I expect from you.

 

[VirtualLarry]

So you would rather see intel look bad than have a positive feature for the consumer? Really Larry? I assume you are being at least partially sarcastic, but not the kind of post I expect from you.

No, I am vehemently opposed to the SEVERELY ANTI-CONSUMER segmentation and "overclocking tax" that Intel has done.

I want them to double-down on their stupidity (like they already have, with the Skylake-X TIM issue), so even everyday joe consumer can see what a rotten end of the stick that they've been getting from Intel.

Then maybe consumers will "wake up", and AMD's desktop market share will double or triple by Christmas.
(Not too hard, they didn't really have that much market share before Ryzen's introduction, except with their die-hard fans.)

I guess I'm just hoping for enough Consumer push-back to Intel, that they'll see the error of their greedy ways.

 

[TheF34RChannel]

What's the daily tidbit @Sweepr? I'm a full blown addict now I usually come home, go outside, have a smoke and a beer and read your daily teaser, and the world is A-okay

 

[Phynaz]

So the major performance deficit seems to be in sequential operations, with a moderate deficit in 4k write. Everything seems about the same.

Hmm wonder if this is a mesh vs ring DMI interface thingy? Or am I just reaching here?

Caution, sample size of 1.

 

[Edrick]

Is this when running M.2 through PCH? Could you remove the other M.2 drivers and retest?

As others noted 4K performance is still not ok. More interesting is that with Kabylake-X its ok.

That is now running through the CPU PCIe lanes. Running through the PCH was much worse (see previous post).

I think the motherboard may be to blame more so than the CPU. Kabylake-X vs Skylake-X should not impact PCH bandwidth. But motherboard design/BIOS could.

 

[Edrick]

So the major performance deficit seems to be in sequential operations, with a moderate deficit in 4k write. Everything seems about the same.

Hmm wonder if this is a mesh vs ring DMI interface thingy? Or am I just reaching here?

Keep in mind that this maybe a motherboard/BIOS issue still. AdamK47 has a different motherboard than me. I think that is more likely the case than the CPU itself.

 

[formulav8]

against Intel's wishes.

Ha, Asrock has definitely done some things Intel didn't like. Like non-k ocing and non Z chipset ocing. Asrock would flaunt it while others like MSI was much more discreet.

 

[formulav8]

like they already have, with the Skylake-X TIM issue

Intel is going to change the TIM or soldering the chips? A DC refresh or something I guess?

 

[wildhorse2k]

Intel is going to change the TIM or soldering the chips? A DC refresh or something I guess?

After looking at delidded 7740X and 7900X it looks like a possibility for 12-18C. The reason is 7980XE die very likely won't fit on the stacked interposer. I speculated this is what may have prevented Intel from soldering 7900X. I posted it few pages back.

 

[Sweepr]

Coffee Lake tidbit of the day: Core i7-8700K offers close to 50% throughput performance increase over its predecessor (based on Turbo).

 

[dullard]

Coffee Lake tidbit of the day: Core i7-8700K offers close to 50% throughput performance increase over its predecessor (based on Turbo).

I assume you mean in multi-threaded applications?

 

[scannall]

Coffee Lake tidbit of the day: Core i7-8700K offers close to 50% throughput performance increase over its predecessor (based on Turbo).

With %50 more cores I would certainly hope so.

 

[Zucker2k]

With %50 more cores I would certainly hope so.

If you take multicore scaling penalty into consideration, this probably means identical turbo frequencies as the 7700k.

 

[Shivansps]

No, I am vehemently opposed to the SEVERELY ANTI-CONSUMER segmentation and "overclocking tax" that Intel has done.

I want them to double-down on their stupidity (like they already have, with the Skylake-X TIM issue), so even everyday joe consumer can see what a rotten end of the stick that they've been getting from Intel.

Then maybe consumers will "wake up", and AMD's desktop market share will double or triple by Christmas.
(Not too hard, they didn't really have that much market share before Ryzen's introduction, except with their die-hard fans.)

I guess I'm just hoping for enough Consumer push-back to Intel, that they'll see the error of their greedy ways.

Wake up to what? AMD has been doing so many bad things over the years that they gained way too much bad fame, and they cant recover as fast as Intel does. They had been competitive for 5 months now with a super agressive strategy (the only option they had), its going to take time.

 

[IntelUser2000]

According to Intel, 14nm++ (Coffee Lake) will use 52% less power than the 14nm process (Skylake). See page 4, top graph:
https://newsroom.intel.com/newsroom...es/11/2017/03/14-nm-technology-fact-sheet.pdf

So, a very rough estimate would be to take any 4-core Skylake, add 2 more cores (+ 50% power) and then use 14nm++ (- 50% power) and you are still right at about the same TDP. That is a pretty close match to the results that have leaked so far:

It's very rough dullard.

52% less power equals 0.48x. Adding 2 more cores to 4 is 1.5x. 1.5 x 0.48 = 0.72. So you see such simplistic calculations do not work.

You also cannot relate amount of cores and clock speeds. 10% more cores does not mean 10% difference in TDP, because other factors come into play.

-Non-core takes power
-Binning
-Market positioning. Server chips have lower Turbo than SKL-X chips. Simply because consumer workloads REQUIRE high clocks

Really, I will say the same thing I always do. Wait for the release. You'll never accurately guess what they will release. And that's for everyone.

 

[frozentundra123456]

Wake up to what? AMD has been doing so many bad things over the years that they gained way too much bad fame, and they cant recover as fast as Intel does. They had been competitive for 5 months now with a super agressive strategy (the only option they had), its going to take time.

Yea, without all those "anti-consumer" tactics, we would have had to live with much inferior cpu performance for the past ten plus years.

 

[IntelUser2000]

Wake up to what? AMD has been doing so many bad things over the years that they gained way too much bad fame, and they cant recover as fast as Intel does. They had been competitive for 5 months now with a super agressive strategy (the only option they had), its going to take time.

NOTHING happens instantly, nothing. These things are usually related to loss/gain in trust. What happens when you lose trust with someone? It'll take YEARS if possible to gain it back.

But if Intel keeps faltering AMD can come back. Just like Intel will not be at its most desperate position for a while, AMD won't see big success for a few years. But it will if it continues.

AMD had the largest marketshare(nearly half) with the original Athlon? Nope, it was with Athlon 64. It's then very interesting they stopped making better CPUs after that.

For Intel I think they have the biggest challenge ahead of them. It's dwarfs the problem they had with AMD. Everyone is against them. Even internally.