Who's buying Skylake-X? (You may now change your vote)

[jj109]

No Intel thread can exist on the planet where the AMD army doesn't come stomping in. I wonder if it's the other way around too in AMD threads...somehow I doubt it. It's becoming boring and I'll definitely refrain from engaging in any such futile discussions. Apologies for derailing folks.

I'll close my post by stating that I wish AMD was on par with Intel and that there would be one/two more equal players in the field.

I just lurk here mostly to catch up on the latest rumors, but it seems that way yes. I found the last guy's rant about "having grown up and stopped chasing performance" really funny. After getting a decent paying job, a couple extra hundred bucks every 3-4 years for a hobby barely registers on my radar. The efficiency claims are bonkers, too. Skylake X uses more power, but it's also faster.

6% increase in task energy in exchange for 32% lower time is a fair trade off, I think.

Regarding the original topic, I'm looking to upgrade my overclocked 4770K with a 7800X, pending some data about clocks. OC3D got it to 5GHz@1.3V, but I need to get a gauge on how much of a unicorn that sample ended up being. Coffee Lake is a back up plan. The non-soldered IHS puts a damper on the enthusiasm, for sure.

Edrick, please report back when you get the system up and running

 

[Ajay]

They were able to run cinebench at 4.8 but only passed stress testing at 4.4.

Stress testing just means running AVX code. There is an AVX offset in the BIOS for a reason.

 

[ozzy702]

Why not go with a Ryzen 8 core if you feel that way?

If I'm going to upgrade I want to go big which means either Threadripper or Skylake-X.

 

[beginner99]

Changed to "no". I will sure wait for coffeelake and TR and then make a decision. But my guess is I will go for coffeelake as my rig is mostly either standard usage or gaming and TR/Skylake-x would more be for fun/ e-pen.

 

[DrMrLordX]

No Intel thread can exist on the planet where the AMD army doesn't come stomping in. I wonder if it's the other way around too in AMD threads...somehow I doubt it. It's becoming boring and I'll definitely refrain from engaging in any such futile discussions. Apologies for derailing folks.

I'll close my post by stating that I wish AMD was on par with Intel and that there would be one/two more equal players in the field.

As much as I'd like to bite on some of the above statements, I'm going to refrain since this is . . . a Skylake-X thread, not an Intel vs. AMD thread. The point being is that there are plenty of opportunities for those discussions in other threads.

I would certainly consider myself to be a member of the "AMD Army" since I swore off Intel products in 1997 for my own reasons. No sense in going on about that now.

Regardless, Intel and AMD (and in their own funky way VIA) are all that's left of the desktop world. I'll show interest here without bringing up the other company since I spend about 95% of my time discussing AMD anyway. And really, Skylake-X IS an interesting topic even for those who are already owners of x99. The performance is there, but the power usage . . . egads. I was fully expecting more power from Skylake-X without that much more power draw as compared to Broadwell-E. Something is wrong, I just can't put my finger on exactly why this is happening. I expected more perf/watt than what we're seeing from Skylake-X.

6% increase in task energy in exchange for 32% lower time is a fair trade off, I think.

Erm well maybe, maybe not. The problem here is that task energy already takes time factor into account. In order to calculate task energy, you take avg power draw in watts, multiply that by how many seconds it takes to complete the task, and produce the total number of joules to complete the task.

You want that number to be as low as possible no matter what your time constraints unless you're rendering with a one-off machine independent of any other hardware.

If you're setting up a render farm that (for whatever reason) has to use x86 CPUs, then you'd be basically insane to pick Skylake-X over Broadwell-E. Though really you'd be looking at Xeons. The fact remains that both CPUs feature dice derived from server products - Xeons. Why would you take a Skylake-based Xeon over a Broadwell one if these kinds of task efficiency numbers are reflected by Xeons? Hell why wouldn't you be using Broadwell-D for that matter? Talk about efficient.

Of course that's just rendering but certainly you can see the point?

6% higher task energy for 32% faster completion really isn't that good. In order for Skylake-X to be a clear advancement from Broadwell-E, it should have been lower task energy with some non-zero improvement in completion time.

 

[tamz_msc]

I was fully expecting more power from Skylake-X without that much more power draw as compared to Broadwell-E. Something is wrong, I just can't put my finger on exactly why this is happening. I expected more perf/watt than what we're seeing from Skylake-X.

I wouldn't be surprised if 3647 Skylake-SP Xeons don't have the high power consumption and the temperature issues that these 2066 Skylake-X processors have owing to lower clocks and larger package size.

Intel is crazy if it thinks it can get away with a 473 sq.mm 18C chip in package that previously housed a ~250 sq.mm 10C chip, and clocking it higher than the server equivalent, which seems to be the general expectation.

 

[jj109]

If you're setting up a render farm that (for whatever reason) has to use x86 CPUs, then you'd be basically insane to pick Skylake-X over Broadwell-E. Though really you'd be looking at Xeons. The fact remains that both CPUs feature dice derived from server products - Xeons. Why would you take a Skylake-based Xeon over a Broadwell one if these kinds of task efficiency numbers are reflected by Xeons? Hell why wouldn't you be using Broadwell-D for that matter? Talk about efficient.

Of course that's just rendering but certainly you can see the point?

6% higher task energy for 32% faster completion really isn't that good. In order for Skylake-X to be a clear advancement from Broadwell-E, it should have been lower task energy with some non-zero improvement in completion time.

If you had an effectively infinite number of tasks and time to complete has little value, sure. But at that point you wouldn't be using a hot clocked Skylake X.

 

[tamz_msc]

Xeon Phis get upto 50% price-cut, on the review day of Skylake-X.

 

[coercitiv]

6% higher task energy for 32% faster completion really isn't that good. In order for Skylake-X to be a clear advancement from Broadwell-E, it should have been lower task energy with some non-zero improvement in completion time.

We currently lack the complete picture of SKL-X, that is measurements based on firmware that actually enforces Intel policies with respect to both Turbo and TDP and also a clock for clock comparison between BDW-E and SKL-X.

There's a lot we can blame Intel for doing wrong with this launch, but expecting efficiency improvements from a performance oriented platform is a bit much, especially in the context of Intel taking HEDT SKUs from the cozy area of 3Ghz+ where efficiency still thrives to the far more challenging area of 4Ghz+, where things get nasty really quick.

Going back to your example, 6% higher task energy for 32% faster completion can easily be turned into a clear win on both task energy and task completion by lowering clocks by 10%. Power usage will drop like a rock, performance will still be above last gen. But why wish for that in HEDT?

 

[lolfail9001]

Intel is crazy if it thinks it can get away with a 473 sq.mm 18C chip in package that previously housed a ~250 sq.mm 10C chip, and clocking it higher than the server equivalent, which seems to be the general expectation.

Intel already got away with 2679v4, 470ish (or whatever BDW 24c die was) sq.mm. 200W TDP CPU in this package, so Intel are hardly crazy.

Now, Intel may be crazy if they think they will get away with bad assembly process they look to repeat once again on these.

Why would you take a Skylake-based Xeon over a Broadwell one if these kinds of task efficiency numbers are reflected by Xeons?

Why would they be reflected? I did not see any 4Ghz Xeons in Skylake-SP line-up, did you?

 

[Shmee]

Probably passing on this. Seems like a downgrade/sidegrade compared to my OCd 5930k as far as 6 core CPUs are concerned. Not worth the change.

 

[tamz_msc]

Intel already got away with 2679v4, 470ish (or whatever BDW 24c die was) sq.mm. 200W TDP CPU in this package, so Intel are hardly crazy.

Largest Broadwell-E Xeon die is around 450 sq. mm, and that was for 16+ cores clocked at 2.6GHz or below. Their rated TDP is for base clock operation in an "Intel-specified" workload, who knows what that means. So it's no surprise that they can claim to 'satisfy' their TDP target with these requirements.

The 22C 2699A v4 at 2.4GHz base has the same 145W TDP as the 16C 2697A v4 at 2.6GHz base. So it is obvious that latter turbos higher than the former, unless Intel lies about TDP that is. Given what we know about Intel's 14nm process post-Broadwell, it'd be very surprising if we don't have the same issues with power consumption and high temperatures plaguing the high core count parts when they launch in October, especially now when Intel seems to leading this round of the gigahertz war.

 

[coercitiv]

I wonder if it's the other way around too in AMD threads...somehow I doubt it. It's becoming boring and I'll definitely refrain from engaging in any such futile discussions.

It's the same in AMD threads. There are always a couple of agitators making a lot of noise and unfounded claims, no matter the brand. One must simply conserve energy with these people, make good use of the Ignore function. Most of them will be gone in 1-2 months anyway.

I'll close my post by stating that I wish AMD was on par with Intel and that there would be one/two more equal players in the field.

It is on par, in the sense that depending on one's requirements they may have some compelling alternatives. One simply has to acknowledge their performance/throughput/price requirements and go with the platform that makes most sense. And no, as of right now that is not a clear cut in favor of one brand or the other.

What we get instead in these forums is people trying to skew other people's requirements to better fit their agenda. That's where the real friction starts, when some of us start viewing their ideal PC as the ideal PC for everyone.

If you want a better picture about the x86 field, you have to deal with the noise and use your own energy and judgement, or at least identify people with decent rep & experience on the forums and weigh in their input as well. There's too much ego & money involved in this business to get the truth for free.

/end of rant

 

[R0H1T]

In other news ~
Debian Project Warns: Turn off Hyperthreading with Skylake and Kaby Lake

This warning advisory is relevant for users of systems with the Intel
processors code-named "Skylake" and "Kaby Lake". These are: the 6th and
7th generation Intel Core processors (desktop, embedded, mobile and
HEDT), their related server processors (such as Xeon v5 and Xeon v6), as
well as select Intel Pentium processor models.

TL;DR: unfixed Skylake and Kaby Lake processors could, in some
situations, dangerously misbehave when hyper-threading is enabled.
Disable hyper-threading immediately in BIOS/UEFI to work around the
problem. Read this advisory for instructions about an Intel-provided
fix.

SO, WHAT IS THIS ALL ABOUT?
---------------------------
This advisory is about a processor/microcode defect recently identified
on Intel Skylake and Intel Kaby Lake processors with hyper-threading
enabled. This defect can, when triggered, cause unpredictable system
behavior: it could cause spurious errors, such as application and system
misbehavior, data corruption, and data loss.
It was brought to the attention of the Debian project that this defect
is known to directly affect some Debian stable users (refer to the end
of this advisory for details), thus this advisory.
Please note that the defect can potentially affect any operating system
(it is not restricted to Debian, and it is not restricted to Linux-based
systems). It can be either avoided (by disabling hyper-threading), or
fixed (by updating the processor microcode).
Due to the difficult detection of potentially affected software, and the
unpredictable nature of the defect, all users of the affected Intel
processors are strongly urged to take action as recommended by this
advisory.
WHAT SHOULD I DO IF I DO HAVE SUCH PROCESSORS?
----------------------------------------------
Kaby Lake:
Users of systems with Intel Kaby Lake processors should immediately
*disable* hyper-threading in the BIOS/UEFI configuration. Please
consult your computer/motherboard's manual for instructions, or maybe
contact your system vendor's support line.
The Kaby Lake microcode updates that fix this issue are currently only
available to system vendors, so you will need a BIOS/UEFI update to get
it. Contact your system vendor: if you are lucky, such a BIOS/UEFI
update might already be available, or undergoing beta testing.
You want your system vendor to provide a BIOS/UEFI update that fixes
"Intel processor errata KBL095, KBW095 or the similar one for my Kaby
Lake processor".
We strongly recommend that you should not re-enable hyper-threading
until you install a BIOS/UEFI update with this fix.

Skylake:
Users of systems with Intel Skylake processors may have two choices:
1. If your processor model (listed in /proc/cpuinfo) is 78 or 94, and
the stepping is 3, install the non-free "intel-microcode" package
with base version 3.20170511.1, and reboot the system. THIS IS
THE RECOMMENDED SOLUTION FOR THESE SYSTEMS, AS IT FIXES OTHER
PROCESSOR ISSUES AS WELL.
Run this command in a command line shell (e.g. xterm) to know the
model numbers and steppings of your processor. All processors must
be either model 78 or 94, and stepping 3, for the intel-microcode fix
to work:
grep -E 'model|stepping' /proc/cpuinfo | sort -u
If you get any lines with a model number that is neither 78 or 94, or
the stepping is not 3, you will have to disable hyper-threading as
described on choice 2, below.
Refer to the section "INSTALLING THE MICROCODE UPDATES FROM NON-FREE"
for instructions on how to install the intel-microcode package.
2. For other processor models, disable hyper-threading in BIOS/UEFI
configuration. Please consult your computer/motherboard's manual for
instructions on how to do this. Contact your system vendor for a
BIOS/UEFI update that fixes "Intel erratum SKW144, SKL150, SKX150,
SKZ7, or the similar one for my Skylake processor".
NOTE: If you did not have the intel-microcode package installed on your
Skylake system before, it is best if you check for (and install) any
BIOS/UEFI updates *first*. Read the wiki page mentioned below.​

Debian developers have issued an advisory to disable hyper-threading with Intel and Skylake- Kaby Lake generation processors. The CPUs can cause problems under certain conditions, in fact on all operating systems.

They advise you to turn it off in the BIOS, and thus forfeit the extra performance you gain. It seems there is s defect for the CPUs in question that can cause compiler- and application crashes, unexpected behavior of programs and incorrect output software: More info here.​

 

[tamz_msc]

In other news ~
Debian Project Warns: Turn off Hyperthreading with Skylake and Kaby Lake

Debian developers have issued an advisory to disable hyper-threading with Intel and Skylake- Kaby Lake generation processors. The CPUs can cause problems under certain conditions, in fact on all operating systems.

They advise you to turn it off in the BIOS, and thus forfeit the extra performance you gain. It seems there is s defect for the CPUs in question that can cause compiler- and application crashes, unexpected behavior of programs and incorrect output software: More info here.​

There's already a thread about this, no need to post this information again here.

 

[R0H1T]

There's already a thread about this, no need to post this information again here.

Didn't see it, anyway posting it here will make it a little more visible.

 

[krumme]

Stress testing just means running AVX code. There is an AVX offset in the BIOS for a reason.

Ahh. Its like amd bios switch in the 290 gpu. They cant have high perf and low noise at the same time. You have to choose.
The meaning of this cpu is very much connected to the avx perf. You dont gimp on it. Then you are better of with either a cl for gaming or tr for productivity.
It is 4.4 tops.
All this water cooling is crazy for this kind of perf imo. And then on a cpu with this tim issue. Stupid. Like a racecar with 14 inch tires. Makes no sense.

 

[TheGiant]

Ahh. Its like amd bios switch in the 290 gpu. They cant have high perf and low noise at the same time. You have to choose.
The meaning of this cpu is very much connected to the avx perf. You dont gimp on it. Then you are better of with either a cl for gaming or tr for productivity.
It is 4.4 tops.
All this water cooling is crazy for this kind of perf imo. And then on a cpu with this tim issue. Stupid. Like a racecar with 14 inch tires. Makes no sense.

Exactly,
I don't know who is the target audience for this CPU. Gamer ?- No it is for 95% certain coffee lake 6C will have better OC and higher gaming performance while not demanding water cooling. And
Profi Multithreaded performance is very nice with SKL-X but also the power and the price will be definitely higher than TR

 

[CHADBOGA]

If I'm going to upgrade I want to go big which means either Threadripper or Skylake-X.

Personally I think you should just accept that for someone with a system like the one you have, now is not the time to upgrade, maybe in 2019 or later.

Considering how much CPU advancement has slowed down in the last 5+ years, having to wait for 3 or so years to upgrade, seems like the new normal.

 

[jmelgaard]

Probably - Yes...

I am still running on an old Core i970, so some would say it's about time to upgrade to something... But actually haven't had a need to upgrade had I not hit the memory barrier hard a good while back. (Sure, there are plenty of performance gains to have in newer CPU's, but it has never been my CPU that made me feel constrained)

I was going for an X99 system at first, but then everything began to get a little interesting with the rumors around Ryzen and then the release. So I wanted to wait out and see Intel's play and in turn I might end up having to wait long enough for AMD to play their next card (Threadripper) as well before a final decision (I won't be able to get a Core i9 right away due to availability here)...

But this doesn't come without mixed feelings... I am going for the memory support (amount) and PCIE, and the Core i9-7900X was certainly not my first choice, but because Intel cut down the PCIE lanes on their Core i7-7800X and Core i7-7820X that is going to be the one I am aiming for... Then there is this whole talk about the thermals and TDP which has me raising my eyebrows... (Not the slightest interested in OC, so I should be fine, but I am use to running my i970 just under a heatsink, no fans except the 180mm case fans)...

Could really use some reviews that wen't more into power draw, head and cooler options at stock frequencies, but granted... That perspective is probably rare in the HEDT segment >.<...

 

[ManyThreads]

The way I see it, and the way that it has been hyped, is that if you are a potential Skylake-X customer, then you either: 1) Care about AVX512, in which case, you should buy the i9 variants exclusively, or 2) You don't care about AVX512, in which case, you should invest in Ryzen or ThreadRipper, and save some cash.

I find it hard to see a middle ground, unless you just prefer to buy Intel for emotional reasons.

If I may offer a third scenario, which is why I am buying. Heavy photo editing. Photoshop combines very single-threaded and very multi-threaded processes (especially in some plug-ins I use), so if I can get an 8C with clock speeds and IPC roughly equivalent to the best quad core (say, a 7700K), that gives me the best of both worlds. I expect to see some pretty significant time savings.

 

[evrae]

I've ordered a 7800x to upgrade from my i5-4590. A hexacore with powerful gaming performance is exactly what I was waiting for. Ryzen was the only other alternative and I just didn't want that much FPS loss.

 

[DrMrLordX]

If you had an effectively infinite number of tasks and time to complete has little value, sure. But at that point you wouldn't be using a hot clocked Skylake X.

It isn't about an infinite number of tasks and an infinite amount of time. The benchmark in question was blender. If you're going to do rendering - which is a task that can be split up along nodes in a cluster if you do it right - then it's not like you're restricted to one machine, even as a hobbyist. Yeah you get extra system overhead per machine but there are ways to compensate for that. Anyone looking to render as a hobby with HEDT hardware would have to think long and hard about using a "hot clocked" Skylake-X system (and why are we even using that term here?) vs. a pair of . . . well let's just say underclocked/undervolted 6800k systems (set them to 3.0 GHz to mimic the 6950X since that's where the clocks should be for any sustained MT workload, such as a large render job).

Two 6800ks should be able to complete the job in ~193 seconds assuming the clustering software works properly and the job is broken up seamlessly. No guarantees that it will. Task efficiency might go down a little from extra system overhead. But you'll get it done faster than you would with the 7900X, and you'd do it more efficiently.

Now do you see the problem? Low efficiency is trouble even if you're a "high end" user, unless you have some very specific need to complete the task within a single node/machine.

We currently lack the complete picture of SKL-X, that is measurements based on firmware that actually enforces Intel policies with respect to both Turbo and TDP and also a clock for clock comparison between BDW-E and SKL-X.

I would like to see what Intel can do with UEFI/firmware updates to improve the situation. The reported current draw of 200 amps @ 3.8 GHz with 1v vcore is disturbing.

There's a lot we can blame Intel for doing wrong with this launch, but expecting efficiency improvements from a performance oriented platform is a bit much, especially in the context of Intel taking HEDT SKUs from the cozy area of 3Ghz+ where efficiency still thrives to the far more challenging area of 4Ghz+, where things get nasty really quick.

We're supposed to see efficiency improvements. We were told 14nm+ and 14nm++ would give this to us. Besides, that's been the general arc of x86 CPU development for years, at least until recently; hell even Broadwell-E managed it. We all bashed BDW-E for its low clockspeed limits, but let's face it, @ 4 GHz, core per core, BDW-E beats HSW-E in performance while using as much or less power. Skylake-X, on the other hand, sometimes LOSES performance vs. BDW-E at the same clock, and it certainly doesn't use LESS power.

If we downclock BDW-E and SKL-X to 3 GHz, will the picture change at all? I would be interested to know.

Going back to your example, 6% higher task energy for 32% faster completion can easily be turned into a clear win on both task energy and task completion by lowering clocks by 10%. Power usage will drop like a rock, performance will still be above last gen.

Again, do we know that for sure? And would the drop in power usage match up with the power drop we'd see on x99?

But why wish for that in HEDT?

See above for one example. Sometimes, burning more power for little gain gives you less, especially when there is an alternative, such as the last generation of HEDT CPUs.

Overclockers in general aren't supposed to care, but remember, we're talking about a stock CPU here. OC is just going to make it worse to the point that heat may be unmanageable.

Why would they be reflected? I did not see any 4Ghz Xeons in Skylake-SP line-up, did you?

Why wouldn't it? We have no concrete explanations thus far for why the current draw is so much higher on Skylake-X. I don't know that lower clockspeeds will solve the problem. Intel has only been selling Skylake Xeons to select customers. We won't see broader reviews of the products until July. Maybe we'll find out then.

 

[coercitiv]

If we downclock BDW-E and SKL-X to 3 GHz, will the picture change at all? I would be interested to know.

That's what I'd like to see as well, a fixed clock comparison at 3Ghz and maybe 4Ghz between the two.

The reported current draw of 200 amps @ 3.8 GHz with 1v vcore is disturbing.

I did not know about that, although it wouldn't surprise me at this point, not with the numbers reviewers are gathering in the so called "stock" configurations current motherboards have.

 

[LTC8K6]

Probably still need to wait for BIOS/Microcode/Windows changes and updates to settle down.