Intel Skylake / Kaby Lake

[TheF34RChannel]

DigiTimes reaffirms Coffee Lake-S in late August, this is their third article mentioning the earlier than expect launch:

http://www.digitimes.com/news/a20170518PD204.html

Good! All I need now is more CFL-S specifications to decide which way I'm leaning towards until both Coffee Lake-S and Skylake-X are out and properly compared to each other...

 

[krumme]

Skylake-SP is already dominating SiSoftware's Multi-Media ranks.

Rumor says Xeon Platinum 8180 scores 4200+ in CB R15, ahead of AMD's Epyc 32C/64T (4000+). This is really impressive considering how well Zen performs in CB R15, Ryzen 7 1800X actually beats stock Core i7-6900X here, even if it's slower overall.

You link to Hardware.fr that show the 1800x is 5.4% slower than than the 6900. Pretty good perf for less than half the cost at 95w tdp.
It thats not dominating the Intel solution i dont know what is. So ryzen with it lower tdp have a headstart because of tdp for the higher core solutions especially on hedt side.
A 16c threadripper will probably score 3100 or so in cb or 2x 1800x. It will 100% dominate Intel 12c here for all except heavy avx2 loads.
For the 32c variant its remains to be seen it it can get to eg 4200. It depends on the tdp. And there is tons of variables we dont know yet.

 

[blue11]

Meh, they also have a Quad Xeon Gold 6152 score now too, which gets 8041. Kind of interesting that Xeon Gold 6xxx will allow 4P (?)

Intel relaxed(?) their segmentation strategy for multi-socket with the SP series.

Platinum - 8S
Gold 6000 - 4S with crossbard
Gold 5000 - 4S with ring
Bronze/Silver - 2S only

 

[Sweepr]

A 16c threadripper will probably score 3100 or so in cb. It will 100% dominate Intel 12c here for all except heavy avx2 loads.

Even AMD recognizes a higher-clocked monolithic 12C/24T Core i9-7920X will provide better performance overall across a variety of AVX & non-AVX workloads, except for corner cases. That's why they are rumored to price Threadripper competitively against 10C/20T Core i9-7900X instead. I also know a lot of people with Haswell-E/Broadwell-E systems that don't want to downgrade/sidegrade on performance per core, so 4.5 GHz Turbo Skylake-X is the only way to go.

 

[LTC8K6]

Good! All I need now is more CFL-S specifications to decide which way I'm leaning towards until both Coffee Lake-S and Skylake-X are out and properly compared to each other...

CL-S is going to be released 2 months or more after SL-X, though.

Direct bench comparisons are going to have to wait awhile after SL-X is released.

I would expect the higher clocks and new cache structure of SL-X to keep the HEDT chips out in front of the DT chips this time around.

I don't expect the DT chips to have the upper hand in gaming anymore.

 

[TheF34RChannel]

CL-S is going to be released 2 months or more after SL-X, though.

Direct bench comparisons are going to have to wait awhile after SL-X is released.

I would expect the higher clocks and new cache structure of SL-X to keep the HEDT chips out in front of the DT chips this time around.

I don't expect the DT chips to have the upper hand in gaming anymore.

Yeah I know unfortunately it'll be a while after. I can wait (well, forcing myself to in all honesty). To your last paragraph; it depends on how high CFL-S clocks I think. However the cache will be much better on SKL-X, which, coupled with so far sweet looking clocks it just may be true. Affordability can be messy into play as well, as they simply cannot continue these hideous prices imo.

 

[pcslookout]

So CoffeeLake will be on LGA1151 socket!

 

[scannall]

Even AMD recognizes a higher-clocked monolithic 12C/24T Core i9-7920X will provide better performance overall across a variety of AVX & non-AVX workloads, except for corner cases. That's why they are rumored to price Threadripper competitively against 10C/20T Core i9-7900X instead. I also know a lot of people with Haswell-E/Broadwell-E systems that don't want to downgrade/sidegrade on performance per core, so 4.5 GHz Turbo Skylake-X is the only way to go.

Platform cost does matter. If for example it's $700 cheaper to go with 16 cores of Threadripper vs 12 cores of Intel, that will matter. $700 spent in other areas of your build will make a much bigger difference. A second 1080ti? More, bigger and faster storage? I know there are some who just throw money at a project, cost no object. But most have a specific objective, and budget to meet that objective when they go shopping around.

 

[Jan Olšan]

DigiTimes reaffirms Coffee Lake-S in late August, this is their third article mentioning the earlier than expect launch:



http://www.digitimes.com/news/a20170518PD204.html

Just something from my hw news experience: these articles usually don't mean "reaffirmation". They just refer to teh particular subject as a piece of information that is "known" from before, but the actual "ground" behind it doesn't change, it is still based only on the original article it was first reported in.

Basically, once you report on something like that, you can refer to it in your next news items (unless it gets disproven by a new conflicting report), but this is just repeating, not "new information". Repeating something doesn't reaffirm it. When analysing this stuff you should always split the news text into the parts that report what is supposed to be the actual new, added information ("scoop"), and what is just background taken from the "general pool of current knowledge".

People often make this mistake - particularly when they see reposts and reports of sites B, C, D, E that draw from previous publications by site A (or indirectly through other reports by F, G, H...), and don't realize that it is just repeating of the same instance of the information. The kernel of new information naturally can't grow or get "stronger" via such reposts, because the reposters don't know the things the original source knew (although good deduction can guess things the original poster failed to guess from his info). Often such reposts can appear quite late, which makes it easier to get fooled, but of course you have to keep sharp about that.

Anyway, that was just a methodical bit, the actual bit of Coffee Lake launch in August is probably very very likely to be true. It was posted by BenchLife.info after all IIRC, which is a great source?

 

[wildhorse2k]

Platform cost does matter. If for example it's $700 cheaper to go with 16 cores of Threadripper vs 12 cores of Intel, that will matter. $700 spent in other areas of your build will make a much bigger difference. A second 1080ti? More, bigger and faster storage? I know there are some who just throw money at a project, cost no object. But most have a specific objective, and budget to meet that objective when they go shopping around.

Not to all. Platform cost mostly matters to AMD buyers as they cannot afford Intel. I do not have a budget limit but that doesn't mean I will be buying Skylake-X 12C or Epyc 32C as I could not use all the CPU power and they will be very poor at few thread loads. I will go with Skylake-X 10C only because it will be a good all-around CPU for desktop and superior to AMD per core. Even if I save some money with AMD, I will not be buying more as I do not need more and there is no clear budget that has to be spent.

 

[UncleCrusty]

Given the Intel process slides posted a while back showed 14++ being the same improvement over 14+ as 14+ was over 14... I wonder if we'll be seeing 5.6 GHz overclocks possible on CFL-S. Doubt it, but looking forward to finding out this August.

 

[formulav8]

5

Platform cost mostly matters to AMD buyers as they cannot afford Intel.

You can't be serious with that remark.

 

[Topweasel]

Not to all. Platform cost mostly matters to AMD buyers as they cannot afford Intel. I do not have a budget limit but that doesn't mean I will be buying Skylake-X 12C or Epyc 32C as I could not use all the CPU power and they will be very poor at few thread loads. I will go with Skylake-X 10C only because it will be a good all-around CPU for desktop and superior to AMD per core. Even if I save some money with AMD, I will not be buying more as I do not need more and there is no clear budget that has to be spent.

Then why does anyone get an i3, or an i5, why isn't every CPU they sell a 6900K? That's a pretty inflamatory statement. I wanted a eight core CPU, I was actually gearing up for an upgrade of my 3930k to a 6900k when I got my Ryzen. Why did I get the Ryzen, I got the 8 cores I needed for 1/3 the price, spent more on NVME drives and saved the rest for Video card once Vega is reviewed (will get either a 1080 or Vega depending on results). Why pay more for the same or in some cases less? When you can get more of other stuff.

 

[tamz_msc]

Even AMD recognizes a higher-clocked monolithic 12C/24T Core i9-7920X will provide better performance overall across a variety of AVX & non-AVX workloads, except for corner cases. That's why they are rumored to price Threadripper competitively against 10C/20T Core i9-7900X instead. I also know a lot of people with Haswell-E/Broadwell-E systems that don't want to downgrade/sidegrade on performance per core, so 4.5 GHz Turbo Skylake-X is the only way to go.

Where did AMD compare its 12C Threadripper with the 7920X, like it did with Ryzen 8C ES and 6900K back in August 2016?

 

[LTC8K6]

Then why does anyone get an i3, or an i5, why isn't every CPU they sell a 6900K? That's a pretty inflamatory statement. I wanted a eight core CPU, I was actually gearing up for an upgrade of my 3930k to a 6900k when I got my Ryzen. Why did I get the Ryzen, I got the 8 cores I needed for 1/3 the price, spent more on NVME drives and saved the rest for Video card once Vega is reviewed (will get either a 1080 or Vega depending on results). Why pay more for the same or in some cases less? When you can get more of other stuff.

I think the point was that people who have the money generally buy the fastest parts. They generally don't bargain shop.
Also people who want bragging rights and benchmarks generally don't bargain shop. They willingly pay more for that 10% extra performance.
There are a few exceptions of course.

People in the Porsche/BMW/Audi and up brackets generally don't even look into the Ford and Chevy lots.

For a short time now, RyZen has captured a bit more of the market, with Intel caught between chips.

We will have to wait and see how long that lasts.

 

[LTC8K6]

Where did AMD compare its 12C Threadripper with the 7920X, like it did with Ryzen 8C ES and 6900K back in August 2016?

Priced, not compared. There haven't been any 7920X performance leaks as far as I can tell.

 

[coercitiv]

I think the point was that people who have the money generally buy the fastest parts. They generally don't bargain shop.

But that's not a point, that's a truism (not yours, his), and it doesn't even apply in the context it was mentioned (which was that platform cost does not matter for HEDT).

If we agree Intel's HEDT is a platform for the rich, then we can safely agree it's utterly irrelevant in the grand scheme of things, since the niche is so narrow (not the rich, but the high core count perf enthusiast rich). However, if we choose to include a wider spectrum of performance enthusiasts then dismissing platform cost is a rookie mistake, since people with the means to buy into HEDT are not necessarily big spenders. Either way you put it price matters, since it affects TAM.

Bonus thinking material: how many times has Kaby Lake X been referred on this thread as some kind of affordable entry point into Intel's HEDT? What's that, a pocket friendly CPU for the wealthy?! It's clear that even Intel knows HEDT needs to address a wider audience, if not migrating oc enthusiasts then at least by grabbing a few more of them from the mainstream line.

For a short time now, RyZen has captured a bit more of the market, with Intel caught between chips.

The revolutionary development with Zen is not the current offering, but rather the solid foundation AMD developed to build CPUs with a competitive price/perf/power ratio. That is not going away any time soon, even if it's not a guarantee AMD will ever hold the performance crown.

 

[LTC8K6]

But that's not a point, that's a truism (not yours, his), and it doesn't even apply in the context it was mentioned (which was that platform cost does not matter for HEDT).

If we agree Intel's HEDT is a platform for the rich, then we can safely agree it's utterly irrelevant in the grand scheme of things, since the niche is so narrow (not the rich, but the high core count perf enthusiast rich). However, if we choose to include a wider spectrum of performance enthusiasts then dismissing platform cost is a rookie mistake, since people with the means to buy into HEDT are not necessarily big spenders. Either way you put it price matters, since it affects TAM.

Bonus thinking material: how many times has Kaby Lake X been referred on this thread as some kind of affordable entry point into Intel's HEDT? What's that, a pocket friendly CPU for the wealthy?! It's clear that even Intel knows HEDT needs to address a wider audience, if not migrating oc enthusiasts then at least by grabbing a few more of them from the mainstream line.


The revolutionary development with Zen is not the current offering, but rather the solid foundation AMD developed to build CPUs with a competitive price/perf/power ratio. That is not going away any time soon, even if it's not a guarantee AMD will ever hold the performance crown.

Kaby Lake X as an entry level HEDT chip was merely put forth as an attempt to explain it's existence. As was the idea that it might hit 5.0. Almost all posters were mystified by it and thought it was likely DOA. I still don't see it selling at all.

Well, I'm not rich and I tend to buy the best because over the years I have just decided that buying less than what I really want is ultimately disappointing and costs me more money in the long run.
I have the money for reasons other than being rich. You can have money for toys without being rich.

Zen is not solid yet, imo.
AMD can still screw it up or screw up Zen+.
Intel can still crush it, also.
The future is very much variable with CPUs.
The desktop market could collapse, too.

The idea that Intel won't fire back strong is just as silly as the idea that AMD wouldn't come back strong with a new CPU.
In both cases it is/was just a matter of time.

 

[Eug]

I think the point was that people who have the money generally buy the fastest parts. They generally don't bargain shop.

I only occasionally buy the fastest parts, because I don't really care about having the fastest speed. I care a lot more about feature set of the machines I buy, in the context of my needs.

Cuz well, that makes a heluvalot more sense than needlessly overpaying just for bragging rights.

 

[frozentundra123456]

I only occasionally buy the fastest parts, because I don't really care about having the fastest speed. I care a lot more about feature set of the machines I buy, in the context of my needs.

Cuz well, that makes a heluvalot more sense than needlessly overpaying just for bragging rights.

Both cases can apply. And price does matter to a lot of buyers. I will say however, that if one wants to spend the money to have the "best", computer equipment is a relatively inexpensive way to do that.

 

[tamz_msc]

Priced, not compared. There haven't been any 7920X performance leaks as far as I can tell.

All we know is that according to AMD's own words, it will be 'disruptive', which presumably means that pricing will be more attractive compared to 7920X. That's about all we know about Threadripper, and as for Skylake-X, we only know the clock speeds and cache sizes.

AMD does not recognize anything about the performance of the 7920X, but from what we know about Skylake it is almost certain that it will be faster than a 12C Threadripper.

 

[Eug]

Both cases can apply. And price does matter to a lot of buyers. I will say however, that if one wants to spend the money to have the "best", computer equipment is a relatively inexpensive way to do that.

That's just it. "Best" is highly variable. CPU speed is only one of many features of a computer. Indeed, the most interesting CPU of late for me is actually Y series, despite being amongst Intel's slowest offerings. My next computer purchase will likely be either Core i5 Y or even Core m3.

For my next desktop, as long as the CPU is half decent I'm fine with it. Everything else about the computer factors heavily into the decision of course.

 

[LTC8K6]

I only occasionally buy the fastest parts, because I don't really care about having the fastest speed. I care a lot more about feature set of the machines I buy, in the context of my needs.

Cuz well, that makes a heluvalot more sense than needlessly overpaying just for bragging rights.

Well, no one said hobbies made any sense...
Nor is buying the fastest parts necessarily "needlessly overpaying".
Why does everything have to seem like a pissing contest?

 

[Eug]

Well, no one said hobbies made any sense...
Nor is buying the fastest parts necessarily "needlessly overpaying".
Why does everything have to seem like a pissing contest?

You're the one suggesting people don't buy Intel's fastest chips because they can't afford them, whereas many of us have pointed out that CPU speed is just one of many factors to consider. If your contention made any sense, the 1% would only ever buy the fastest CPUs but obviously that is not the case. Most people just don't care as much about raw CPU speed as you seem to.

 

[coercitiv]

Well, I'm not rich and I tend to buy the best because over the years I have just decided that buying less than what I really want is ultimately disappointing and costs me more money in the long run.

So do you agree with me that there is a sizeable customer base that affords and desires HEDT but is also price sensitive?