Skylake second thoughts; how hot is Haswell-E?

Cooling Processors quietly

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
Ken5d
Posts: 130
Joined: Sat Aug 21, 2004 3:06 pm
Location: USA

Skylake second thoughts; how hot is Haswell-E?

Post by Ken5d » Wed Dec 02, 2015 5:46 pm

Well, while waiting for the supply of i7-6700K chips to catch up with the demand, I've begun to have second thoughts about Skylake.

My main use is for 4K video editing, and I know that the 6-core Haswell-E i7-5820K chip is likely to outperform the Skylake i7-6700K for some of my things I do. There's the added advantages that the 5820K is readily available, and the motherboards are far more mature and therefore (hopefully) less buggy.

I would likely overclock the 5820K to around 4GHz (hopefully at stock voltage). But I assume it would run hotter than a moderately overclocked 6700K (slight voltage increase, hopefully, to get to around 4.5GHz).

Question: Can the Scythe Kotetsu that I've already purchased do a decent job of cooling the hotter 5820K? Or should I give up and get a Ninja 4? (Other relevant components for that build would be a Fractal Design Define R5, a couple of HDs, a Seasonic 660 Platinum, and a GTX 960.)

Ken5d
Posts: 130
Joined: Sat Aug 21, 2004 3:06 pm
Location: USA

Re: Skylake second thoughts; how hot is Haswell-E?

Post by Ken5d » Fri Dec 04, 2015 10:38 am

No one wants to risk sharing their opinion?

Well, I'll share mine: Why not just build it with the Kotetsu and then replace it if it proves inadequate for the job?

A bit of a hassle, and it means I couldn't return the Kotetsu for a refund, but that's one way of finding out for sure!

Though I'm still not sure if I want to go with the 5820K or not. I wish I could just make up my mind and stop being so wishy washy. :lol:

Abula
Posts: 3662
Joined: Tue Nov 02, 2004 12:22 pm
Location: Guatemala

Re: Skylake second thoughts; how hot is Haswell-E?

Post by Abula » Fri Dec 04, 2015 4:45 pm

These are not CPUs that silent seekers peruse, not that they are bad, just that most users can live fine with quads and even dual cores, Haswell E should put a little more heat out, and thus you need to deal with it, that said, i don't think anyone should fear the outcome, specially someone that edits a lot, honestly imo the 5820K is one of the best offering by intel price/performance wise, specially since don't seem mainstream chipsets will get 6 core anytime soon. If the next couple of gens of intel continue to give us marginal upgrades on the mainstream segment, ill probably move toward E edition, would love to get 8 core, my eyes are on skylake E, but its going to take a while, my guess is 2017.

The most important thing for you to decide is the programs you will use, see how much improvement will a 5820K will net over 6700K on stock clocks both, and overclocked, and see if you can deal with the heat either will produce, then you can resort into how you going to cool them to achieve a quiet setup. Contact the program companies that develop the software you produce, try to search for reviews on the 6700K and 5820K, check benchmarks into similar programs that you use for editing and decide there whats best for you.

A weakness that X99 has is no iGPU, but at the end it could end up better, if your editing programs also take advantage of the dedicated GPU, all this things you need to check with the company that develops the software you use.

A big weakness of Z170/6700K is that you are going with the best CPU already, while with X99 you are entering with the lowest offering, you still have room to upgrade toward a 8 core, and probably with the release of Broadwell E a 10 core, so you have room to keep on growing, and memory wise the same, you get more banks/slots with X99, in some cases double, so its a platform that you can invest toward the future.

Research, research and research before you commit to either platform, and good luck =P
Last edited by Abula on Fri Dec 04, 2015 4:52 pm, edited 2 times in total.

CA_Steve
Moderator
Posts: 7650
Joined: Thu Oct 06, 2005 4:36 am
Location: St. Louis, MO

Re: Skylake second thoughts; how hot is Haswell-E?

Post by CA_Steve » Fri Dec 04, 2015 4:45 pm

Here's a swag.

The Kotetsu was tested with a 130W TDP i7-965. The i7-5820K has a 140W TDP at 3.3GHz base speed. Assuming you could OC it at stock speed to 4GHz, then it raises the stress power to 4/3.3*140 = ~170W. Everything else being equal (say they have similar die size and IHS configuration which leads to similar thermal transfer), and let's swag it to say the thermal response is linear between 130W and 170W, you'll see ~30% higher temp rise over ambient than what's in the review. So, call it 55C rise instead of 42C at 12dB, 50C rise instead of 38C at 16dB. If your room ambient is 25C, the CPU under stress load will be at 80C for 12dB and 75C at 16dB.

Real loads seldom approach Prime95 kinds of loads, so it could be doable.

Ken5d
Posts: 130
Joined: Sat Aug 21, 2004 3:06 pm
Location: USA

Re: Skylake second thoughts; how hot is Haswell-E?

Post by Ken5d » Fri Dec 04, 2015 5:36 pm

Thank you, Steve, that's exactly what I was looking for (but I didn't know how to do it myself). And I think it validates my best guess stated above (why not give a try and the replace it if it turns out to be inadequate for the job).

Abula, I think that's great advice, thanks! I have been trying to do exactly that, but it isn't easy. My main editing program is After Effects CC, which I know benefits from both multiprocessing and GPU acceleration for some things (e.g., rendering). What I don't know, and haven't been able to really even guess at, is how much difference the 2 extra cores on the 5820K would actually translate to. For example, just now on my lowly, old i7 920, I rendered a 30-second 4K clip that took just under 12 minutes to complete. But watching my resources, CPU usage rarely went over 50%, and usually stayed around 40%, during the render. Why? I'm not sure; it might be memory bound (I have 8GB of my 12GB allocated to AE). But, anyway, I'm hoping a 6700K would improve that by at least 50%, to say 6 minutes. How much more could I possibly expect of a 5820K? I've read some general advice saying perhaps a 20% improvement over the 6700K for multi-threaded applications. So down to under 5 minutes. But how important is another minute plus savings, compared to dropping the needed time in half? That's what I ask myself, anyway. :D

I can buy a dedicated GPU regardless of which processor I go with. Getting a 6700K gives me the luxury of waiting for a Pascal-based card, which perhaps might come along by the time I start to realize that 6 minutes for a 30-second clip is still a LONG time to wait. :)

I did read about the Broadwell-E 10-core coming out, but it's likely to be priced at $1k or more. As this is mostly a hobby, as much as I might love that, I can't honestly justify it (tempting, though, to be sure!). So I probably shouldn't base my purchase on it, though keeping it in play couldn't hurt, either.

Anyway, I knew this was going to get OT because the cooler is just part of my conundrum. But I think Steve answered that question very well, so I will try not to drag out my indecisiveness in this thread.

I'm still leaning towards Skylake because in my heart I am a silent PC lover, and the lower heat and energy usage of Skylake makes my heart happy.

baii
Posts: 69
Joined: Fri Nov 20, 2015 3:32 pm

Re: Skylake second thoughts; how hot is Haswell-E?

Post by baii » Fri Dec 04, 2015 6:16 pm

The actual power consumption for the 5820k seem to be lowered than its tdp, though power consumption does not scale linearly with added vcore, so you are at the mercy of silicon lottery when ocing. At 3.5ghz, every 100mhz give you 2-3% performance increase, but when the voltage scaling go off, the temp went up much more than that.
If you carefully monitor your work load, you can pick a meaning oc that may not survive sff prime 95, but stable under your typical load(with some spare headroom).

Ken5d
Posts: 130
Joined: Sat Aug 21, 2004 3:06 pm
Location: USA

Re: Skylake second thoughts; how hot is Haswell-E?

Post by Ken5d » Fri Dec 04, 2015 8:51 pm

baii wrote:At 3.5ghz, every 100mhz give you 2-3% performance increase, but when the voltage scaling go off, the temp went up much more than that.
2-3% of theoretical performance increase.

I run my i7 920 at about a 30% OC (3.48GHz instead of 2.66 at stock). I just repeated the render I mentioned above, it took 13:29 minutes (compared to 11:45 at OC speed). That means the 30% OC amounted to only a 13% improvement for this particular task. I knew it wouldn't be 30%, but I was hoping it was close to 20%.

Which illustrates how CPU clock speed is just one part of performance (which I'm sure we all know, anyway).

Post Reply