Intel Skylake / Kaby Lake

[tential]

Well the Corsair H110i is supposed to be pretty good, but the reason I bring it up is because it allows you much more room around the memory slots and PCIe slots. Big tower coolers extend over them sometimes.

How about the newly released NH-C14S?

http://forums.anandtech.com/showthread.php?t=2443210

I compiled some benches from it there but it looks to be VERY close to the DH15 in performance. While being VERY VERY quiet. It also gives you very very good space around the components and has some other cooling benefits that I'm probably not qualified to talk about but are within the reviews I linked.

I mean, the issue is the watercooler is louder and not benefit over the DH15 right? So I just don't think the H110i is worth it with the NH-C14S coming so close to the DH15 (and most of these are in single fan testing you can add a second fan) and with how quiet it is and how it leaves you full room for your components (unless I'm missing something).

Take a look though, see what you think. I'm no cooling expert, I have an intel stock cooler on my 4770k, I'm JUST updating it now and the NH-C14S was my premium option.

I think I'll go with the Thermal Right True Spirit 140 for $50 since it seems to be as quiet as it gets without spending a pretty penny. The DH15 just looks scary to me lol....

I like the THermal Right as it will blow air directly out of my case, but would like to hear more about the benefits of a C cooler like the NH-C14S as I just don't know much about CPU cooling.

 

[Madpacket]

Intel Skylake is like that super pretty girl who everyone hyped up to be the girl of their dreams but when you actually spent some time with her getting ready to dump my current girlfriend , she turns out to be barely any better than my girlfriend. So instead of dumping my current girlfriend for her, i choose to stick with my current girlfriend. I will dump my current girlfriend for a new boyfriend next year (AMD ZEN)

Insulting Skylake purchasers while insinuating going AMD is going gay. I know I shouldn't laugh but that was funny.

 

[witeken]

While the supply chain for chipsets and MBs is different and also has different lead times, I also think the likely culprit is Intel 14nm. Intel really delivered in terms of costs, I really doubt that either Samsung or TSMC could deliver Skylake at 4Ghz with 120mm^2. OTOH it seems that Intel pushed itself a little too hard in terms of how hard they pressed the current technology.

It's been almost a year after the first 14nm process and only today we are starting to see 14nm chips to appear in meaningful quantities, and even now we are hearing about issues in the supply chain. I can't recall the last time a 1-year old node at Intel got availability issues, and given that 10nm was also pushed back, it rather more probable that something more fundamental at Intel R&D pipeline is causing the problems than this snafu being a one off issue.

I completely disagree. Broadwell has been in large volumes on the market since January and only higher (Core m, -U, Xeon D, Xeon Phi will be in production in H2, -M, now Skylake-S since end of Q2 with more SKUs and -U coming).

The culprit is the shrinking pc market. Intel thought the pc market would be about flat, so they overestimated sales. This was a big blow for them in Q1, and they took some measures with factory utilisation er cetera. I don't really recall the details, but you can find them in Q1 and 2 conference call, but as I understood, in Q2 the inventory went down less than they expected. Now they've added more 14nm capacity and those will produce Skylake. However, because Skylake is new and has added features, initially it is higher cost than Broadwell. I refer back to the Investor Meeting: http://forums.anandtech.com/showthread.php?t=2409131.

So probably yields still aren't on par with the superb 22nm yields, but I'd say it is probably healthy. It shouldn't be exclusively yields as the cause.

They would have had the same problems without the little bit extra density aggression cause the patterning would have been the same.

 

[Fjodor2001]

it's pointless to bring up a 3W TDP increase or whatever.

32->22nm: SB->IB was down 95->77W, so a 22 W decrease.
22->14 nm: Haswell->Skylake was up 84->91/95W, so a 7-11 W increase.

In total the difference is 22 + 7 or 22 + 11 W, i.e. 29 to 33 W.

That's quite a big difference actually. And also quite interesting, since I suppose it tells us a bit about the properties of 32 vs 22 vs 14 nm.

Whether or not you care about the TDP on desktop CPUs is another story...

 

[mrmt]

The culprit is the shrinking pc market. Intel thought the pc market would be about flat, so they overestimated sales. This was a big blow for them in Q1, and they took some measures with factory utilisation er cetera. I don't really recall the details, but you can find them in Q1 and 2 conference call, but as I understood, in Q2 the inventory went down less than they expected. Now they've added more 14nm capacity and those will produce Skylake. However, because Skylake is new and has added features, initially it is higher cost than Broadwell. I refer back to the Investor Meeting: http://forums.anandtech.com/showthread.php?t=2409131.

Not sure what to make up of your post, Witeken:

- You are saying that Intel missed their sales forecast, and this should have pulled, not pushed 14nm schedule if the node was healthy enough for prime time. Simply put in the stressful market conditions of 2015 a healthy 14nm node would bring a welcome cost reduction in COGS, a perfect measure on this market condition. The fact that they pushed 14nm back means that 14nm cannot compete with 22nm after an entire year of HVM.

- This comparison with the 22nm yields curve is a smoke screen. 22nm might have been a good exception to Intel yields curve, but what if we compare 14nm with 45nm or even 32nm, the other nodes developed under tick/tock, would we find a shallow ramp up curve like 14nm? We don't. Usually after one year the majority of the product stack was mainly comprised of the new node parts, while this will take about 18 months on 14nm.

 

[cmdrdredd]

How about the newly released NH-C14S?

http://forums.anandtech.com/showthread.php?t=2443210

I compiled some benches from it there but it looks to be VERY close to the DH15 in performance. While being VERY VERY quiet. It also gives you very very good space around the components and has some other cooling benefits that I'm probably not qualified to talk about but are within the reviews I linked.

I mean, the issue is the watercooler is louder and not benefit over the DH15 right? So I just don't think the H110i is worth it with the NH-C14S coming so close to the DH15 (and most of these are in single fan testing you can add a second fan) and with how quiet it is and how it leaves you full room for your components (unless I'm missing something).

Take a look though, see what you think. I'm no cooling expert, I have an intel stock cooler on my 4770k, I'm JUST updating it now and the NH-C14S was my premium option.

I think I'll go with the Thermal Right True Spirit 140 for $50 since it seems to be as quiet as it gets without spending a pretty penny. The DH15 just looks scary to me lol....

I like the THermal Right as it will blow air directly out of my case, but would like to hear more about the benefits of a C cooler like the NH-C14S as I just don't know much about CPU cooling.

The difference would only be a few degrees which is probably not enough for me to say one sucks and the other is great. That's really the problem. I suppose for air cooling it depends on how much space you are willing to sacrifice for that extra bit of performance as the fin towers get huge as the performance increases.

I bought the NH-D14 as it was the top end air cooler at the time and the AIO water solutions didn't perform much better if any and were fairly new so there were some problems. Now things are so close with all the top end products. There are only a few standouts anymore and the NH-C14S seems to perform about on par with the NH-D14 that I have which is really good. Some water coolers are better than the D15 but as you can see in the chart in the TweakTown article you linked to the difference between their top cooling solution and the NH-C14S was about 6.83degrees which isn't a huge difference when both of them keep the temps under 75c in an overclocked situation on a 4770k. Remember though that this product pushes the hot air up into the case and you need to have adequate flow to get it out to keep the rest of the system temps in check. Something like a Thermalright tower or a Noctua D14 or D15 can be oriented to blow the air toward the back of the case for an exhaust fan to pull out.

I'm going to use my D14 for as long as I can. There have only been minor advances in air cooling IMO. The benefit of Noctua products is they will give you a mounting bracket for a new motherboard socket for free if you prove you own the cooler and the motherboard with the updated socket.

 

[witeken]

If you look at Stacy Smith's slide 28, he forecast Broadwell to go under Ivy Bridge (which is the right comparison) and slightly under Haswell at the same point in HVM if his forecast is right, this quarter is when Moore’s Law kicks in for 14nm, but those will be in stores in Q4.

 

[DrMrLordX]

I bought the NH-D14 as it was the top end air cooler at the time and the AIO water solutions didn't perform much better if any and were fairly new so there were some problems. Now things are so close with all the top end products.

I have been using the NH-D14 since 2009, and it has yet to fail me. Granted, I am not running a 9590 or a 5960x . . . for something like that, I'd probably need an NH-D15, an AiO cooler, or a custom loop.

Though if I ever do want to step up the cooling performance, I'm sorely tempted by that EK AiO that comes with a 360mm rad for $240. Expensive, yes, but it's the only AiO I know of with that big of a rad. That will make it the top performer among all the AiOs unless the block is a major letdown and/or the fans in use are clearly inferior. Unless someone's got a deep 240/280mm rad that makes better use of fans with high static pressure, in which case that might edge out the EK AiO, but whatever.

Bottom line is, for the stuff I run, the NH-D14 is major overkill, which is exactly what I want.

 

[phillyman36]

I had a Corsair H110 on my z97. For z170 I decided to go air. Had a fear of coming home and finding leaked fluids on my components. The Ek 240 AIO does look nice but im sticking with the Noctua NH-U14s.

 

[CakeMonster]

Thanks for the input about coolers, I will need to look this up with regards to size, weight and mounting mechanism.

I found out that I currently have the Cooler Master Hyper 412S. It's rather big, but it doesn't sit well at all. I think I've had it for 2 or 3 builds, and after each install, I've still been able to wiggle it a little. I'm either not doing it right, or its too heavy for its own good, or the fastening mechanism is not up to what it should be. It might just be this units fault, but I want something that will fit and stay in place and is a no brainer to install.

 

[alecmg]

Spcr.com had such great things to say about mounting mechanism and performance of Scythe Ninja 4 that i bought it for my skylake build
Cant comment yet, still sits in the box. But i have Ninja 3 on my Sandy and it served me well. No wiggle of course

 

[Fjodor2001]

I noticed Intel has released an official CPU cooler for Skylake (since it doesn't come with that included as before).

http://www.kitguru.net/components/c...-cooler-to-date-ts15a-for-lga1151-processors/

It's called Thermal Solution BXTS15A:

Apparently they charge $41.5 for it! Isn't that quite a lot for a mediocre cooler?

I always tossed mine away and replaced it with a third party cooler anyway, so I personally don't mind it not being included anymore. But I guess it will add to the cost for the OEMs who build computers, compared to previous desktop CPUs where it was included.

After all $41.5 is quite a lot relative to the price of the CPU itself, especially if getting a lower end SKU for around $200. That's in fact 41.5/200 => ~20% price increase for the total CPU + cooler cost!

Nice trick by Intel and it should give them quite a lot of extra profit margin.

Also good for the third party CPU cooler companies, since I bet they will sell a lot more units now that Intel's cooler is no longer included for free.

 

[Absolute0]

RE: NH-D15 vs. NH-C14S:

There is such a thing as cooling "capacity." And though the #'s on the smaller, quieter C14S seem nearly the same as the D15, I would wager (heavily) that the D15 has a significantly greater cooling capacity.

I was astonished to see such good #s coming from the C14S; with one less fan and about half the metal of the D14/D15, how can it be so close? They must both be cooling pretty light loads. Indeed, if you look at Tweaktown's OC'd profile, they are using a 4770K @ 4.5 Ghz, 1.14 V.
That is not a very aggressive overclock, light on voltage, and not dumping a huge amount of watts.
Their testing parameters do not allow for the bigger D15, or the significantly more poweful water cooling, to show their enhanced cooling capacity.

When it comes to which cooler to get, it should apply to the thermal load you're going to cool. A modest overclock on 4 cores? The quiet C14S looks perfect.
What can be said of the AIO water coolers and the NH-D14/15 is they will give you an increased capacity to dump more voltage for a more extreme overclock, or (probably more relevant) for a 6 core, or 8 core in the future. So to me, much more future proof.

Personally I own an NH-D14 and would buy NH-D15 if I got something new. I like the assurance of the extra metal.

I had a Corsair H110 on my z97. For z170 I decided to go air. Had a fear of coming home and finding leaked fluids on my components. The Ek 240 AIO does look nice but im sticking with the Noctua NH-U14s.

This fear is normal but I've been watercooling since 2005 - no leaks . I use hose clamps and leave nothing loose

 

[cmdrdredd]

That's crazy. For around $50 you can get a 3rd party cooling system with a much better mounting system (the plastic push pins are terrible).

 

[mrmt]

If you look at Stacy Smith's slide 28, he forecast Broadwell to go under Ivy Bridge (which is the right comparison) and slightly under Haswell at the same point in HVM if his forecast is right, this quarter is when Moore’s Law kicks in for 14nm, but those will be in stores in Q4.

And why do you think that the graphics invalidates anything I'm saying here? Look at IVB curve, IVB was launched in Q212, so the 22nm curve starts 1 quarter prior to commercial launch of the first product on the node.

But look at the 14nm curve. It starts in the quarter Broadwell U was launched, not one quarter before, and Broadwell U wasn't the first 14nm product, that honor befalls to Core M in Q314, so the correspondent point in the curve would be Q214, not Q115, but even in Q115, 3 quarters after the launch of the first commercial product the economics of 14nm are much worse than IVB at the same point.

You could say that Core M is a completely different product and shouldn't count for this comparison, but even then we should be talking about Q414, not Q115 as the start point for the BDW curve, and even then the economics on 14nm would still be much worse than 22nm.

 

[witeken]

Moore’s Law is getting hard. That's why 10nm is also delayed by a year. Tick Tock is closer to 3 years nowadays. I blame ASML.

 

[mrmt]

Moore’s Law is getting hard. That's why 10nm is also delayed by a year. Tick Tock is closer to 3 years nowadays. I blame ASML.

ASML produces the equipment that Intel uses for manufacturing processors, and things like lack of reliable EUV certainly had an impact on Intel plan of record, but the other part of the equation, the "recipe" Intel uses to make its chips, is Intel's responsibility. Did Intel push too hard on the targets they were trying to reach with 14nm and 10nm?

 

[witeken]

ASML produces the equipment that Intel uses for manufacturing processors, and things like lack of reliable EUV certainly had an impact on Intel plan of record, but the other part of the equation, the "recipe" Intel uses to make its chips, is Intel's responsibility. Did Intel push too hard on the targets they were trying to reach with 14nm and 10nm?

When was the last time before 14nm something like this happened? Bill Holt was talking that I'd something that can happen and has happened before, but now suddenly 2 times in a row. Intel has a great track record, they aren't amateurs. It's continually getting harder, and lack of EUV makes it worse.

Edit: TSMC and Samsung have the same problems.

 

[ShintaiDK]

I noticed Intel has released an official CPU cooler for Skylake (since it doesn't come with that included as before).

http://www.kitguru.net/components/c...-cooler-to-date-ts15a-for-lga1151-processors/

It's called Thermal Solution BXTS15A:

Apparently they charge $41.5 for it! Isn't that quite a lot for a mediocre cooler?

I always tossed mine away and replaced it with a third party cooler anyway, so I personally don't mind it not being included anymore. But I guess it will add to the cost for the OEMs who build computers, compared to previous desktop CPUs where it was included.

After all $41.5 is quite a lot relative to the price of the CPU itself, especially if getting a lower end SKU for around $200. That's in fact 41.5/200 => ~20% price increase for the total CPU + cooler cost!

Nice trick by Intel and it should give them quite a lot of extra profit margin.

Also good for the third party CPU cooler companies, since I bet they will sell a lot more units now that Intel's cooler is no longer included for free.

They dont charge 41$.

It cost around 20euro in europe. And its a 130W cooler. Vast majority of people with K CPUs never used the shipped cooler. It just went directly to the landfills/recycling yards.

 

[Fjodor2001]

They dont charge 41$.

It cost around 20euro in europe. And its a 130W cooler. Vast majority of people with K CPUs never used the shipped cooler. It just went directly to the landfills/recycling yards.

According to the article the list price is $41.5. What you can get it for in different stores in different countries is another story. I did a quick check and most stores in Europe charged €30-40 (depends on VAT in each country too), so quite close to the official $41.5.

Anyway, point is that Intel effectively has raised the official price of Skylake by $41.5. Quite a lot compared to the price of the CPU itself, especially for the lower end SKUs.

 

[Arachnotronic]

ASML produces the equipment that Intel uses for manufacturing processors, and things like lack of reliable EUV certainly had an impact on Intel plan of record, but the other part of the equation, the "recipe" Intel uses to make its chips, is Intel's responsibility. Did Intel push too hard on the targets they were trying to reach with 14nm and 10nm?

That certainly seems like it. Until 14nm, Intel had generally been more relaxed with respect to shrinking minimum metal pitches relative to competitors. At 14nm, Intel went from being behind on a node-to-node comparison to ahead in that same comparison.

 

[TeknoBug]

I noticed Intel has released an official CPU cooler for Skylake (since it doesn't come with that included as before).

http://www.kitguru.net/components/c...-cooler-to-date-ts15a-for-lga1151-processors/

It's called Thermal Solution BXTS15A:

Apparently they charge $41.5 for it! Isn't that quite a lot for a mediocre cooler?

I always tossed mine away and replaced it with a third party cooler anyway, so I personally don't mind it not being included anymore. But I guess it will add to the cost for the OEMs who build computers, compared to previous desktop CPUs where it was included.

After all $41.5 is quite a lot relative to the price of the CPU itself, especially if getting a lower end SKU for around $200. That's in fact 41.5/200 => ~20% price increase for the total CPU + cooler cost!

Nice trick by Intel and it should give them quite a lot of extra profit margin.

Also good for the third party CPU cooler companies, since I bet they will sell a lot more units now that Intel's cooler is no longer included for free.

ha a $16 Hyper TX3, $23 TX4 or $26 212 would be better off.

 

[Sweepr]

Some impressive numbers from Phoronix's Core i5 6600K Linux preview:

H.264 Video Encoding
Core i5 6600K (Skylake 4C/4T 3.5-3.9GHz): 171.28 FPS
Core i5 4670 (Haswell 4C/4T 3.4-3.8GHz): 136.59 FPS
Core i7 4770K (Haswell 4C/8T 3.5-3.9GHz): 182.23 FPS
FX8350 (Vishera octo-core 4.0-4.2GHz): 155.39 FPS
FX8350 OCed to 4.6GHz: 173.93 FPS

25.3% faster than Core i5 4670, almost as fast as Core i7 Haswell in this particular multi-threaded benchmark, also basically matching a 4.6GHz Vishera.

Also Ian Cutress got his Core i7 6700K to 4.9GHz:
http://valid.x86.fr/ln6rms

 

[mrmt]

That certainly seems like it. Until 14nm, Intel had generally been more relaxed with respect to shrinking minimum metal pitches relative to competitors. At 14nm, Intel went from being behind on a node-to-node comparison to ahead in that same comparison.

It seems that Intel is gunning high for the mobile market, regardless of getting at it as a IDM or foundry partner, but at the same time they are keen to keep an edge over electric parameters over their competitors. That kind of absolute leadership is really hard to pursue within development framework.

 

[ShintaiDK]

According to the article the list price is $41.5. What you can get it for in different stores in different countries is another story. I did a quick check and most stores in Europe charged €30-40 (depends on VAT in each country too), so quite close to the official $41.5.

Anyway, point is that Intel effectively has raised the official price of Skylake by $41.5. Quite a lot compared to the price of the CPU itself, especially for the lower end SKUs.

Its not the list price. You didnt read the article...again.