PDA

View Full Version : Kaby Lake i7-7700K power efficiency...



Raja@ASUS
01-18-2017, 04:55 PM
KBL is very power efficient. Small FFT load of the latest Prime with AVX, using ~1.35Vcore and 5GHz CPU freq with DDR4-4000 on the memory...




61939




Sub 150W. SKL pulls more current than this at 4.7GHz.

lufiron
01-18-2017, 05:21 PM
Which motherboard are you using to obtain these results? *Also what are your thermals when under full current load?

Raja@ASUS
01-18-2017, 05:35 PM
Which motherboard are you using to obtain these results? *Also what are your thermals when under full current load?

Not that it matters much, but, Apex.

Circa 80C, with 19C water inlet temp..

bn880
01-19-2017, 05:52 PM
Did you also test the SKL in the same Apex board? :) If not... please try it to compare.

Raja@ASUS
01-19-2017, 07:19 PM
Old figures show around 150W at 4.7GHz for SKL at lower Vcore with the same load. Even an optimistic favor in VRM efficiency variance isn't going to make up for that. 300MHz down and around the same current. My delidded SKL needs around 1.50Vcore for 5GHz, and there, it would be closer to 200W. KBL has the edge "fo'sho'";) Knowing that, I'm not inclined to waste my time with it any further. Never been the type to waste time on things when there's no mystery. KBL stays in the socket for good.

61961

The main thing to take away here is that you dont need much of a VRM solution to overclock these CPUs.

Please feel free to share your own compares here, though. Nothing stopping you if there's a mystery you'd like to solve. :)

bn880
01-19-2017, 08:15 PM
Okay just saying, we don't know the actual % as it's a different board. But you are probably right.
I don't have a Skylake to compare with unfortunately :)

Edit: if you are saying the Skylake was IN THE SAME motherboard than your figures are valid. Otherwise you are comparing the CPU +GPU combination, Skylake Z170 vs Kabylake + Z270 (You never specified which mobo was used for the Skylake)

Raja@ASUS
01-19-2017, 08:39 PM
Okay just saying, we don't know the actual % as it's a different board. But you are probably right.
I don't have a Skylake to compare with unfortunately :)

Edit: if you are saying the Skylake was IN THE SAME motherboard than your figures are valid. Otherwise you are comparing the CPU +GPU combination, Skylake Z170 vs Kabylake + Z270 (You never specified which mobo was used for the Skylake)

I isolated the EPS 12V line which doesn't supply ANY GPU or "chipset" power. ;)

VRM efficiency at these types of load is typically north of 85%. Even if you account very generously for a 10% swing, SKL will lose. Like I said, no mystery for me, but if there is for you, feel free to get your mitts on a SKL CPU and solve to put your own mind at rest. I await your results. :)

lufiron
01-19-2017, 09:12 PM
Nothing stopping you if there's a mystery you'd like to solve. :)

There is if a certain specific motherboard still isn't available :) *

Get them to release the "ape" from the cage Raja! *

Raja@ASUS
01-19-2017, 09:16 PM
There is if a certain specific motherboard still isn't available :) *

Get them to release the "ape" from the cage Raja! *


Give it a week or two and it will be on sale.

bn880
01-19-2017, 10:19 PM
I isolated the EPS 12V line which doesn't supply ANY GPU or "chipset" power. ;)

VRM efficiency at these types of load is typically north of 85%. Even if you account very generously for a 10% swing, SKL will lose. Like I said, no mystery for me, but if there is for you, feel free to get your mitts on a SKL CPU and solve to put your own mind at rest. I await your results. :)

Hah, yeah I mean CPU + Mobo, total typo. I think it's no mystery that the Kabylake is more efficient, but by how much is still a mystery. (I'm not getting a Skylake now, I don't get paid by Intel or Asus ;)

But the point is in scientific experiments, you should compare apples to apples wherever possible. This is not about anybody getting personal or ego's, it's just reality.

Cheers.

Raja@ASUS
01-19-2017, 10:28 PM
Hah, yeah I mean CPU + Mobo, total typo. I think it's no mystery that the Kabylake is more efficient, but by how much is still a mystery. (I'm not getting a Skylake now, I don't get paid by Intel or Asus ;)

But the point is in scientific experiments, you should compare apples to apples wherever possible. This is not about anybody getting personal or ego's, it's just reality.

Cheers.

Interesting back-peddle, I mean typo, er I mean motherboard, no, I mean GPU, or maybe Z170/Z270. ;)

Yeah, even the "motherboard" isn't really part of it; it's just the VRM efficiency. Motherboard implies "a lot" of power circuits. The rail distribution is the same across boards (only the main power rails for CPU power connected to EPS12V). Now, if I was measuring stuff like most other people do (wall power etc), more variables would be in the equation, and I'd be inclined to go further. See no need to go that far here, though, even if it doesn't jive for you.


I'm usually very scientific, just in this case, I dont mind putting my name against this without going that far - seen enough to know and that's enough for my peace of mind. However, if it's disturbing yours, well, you know what you can do about it. I always encourage people to learn for themselves.:)

Arne Saknussemm
01-21-2017, 04:50 PM
I like when you post squiggly yellow lines Raja...always interesting...there are too few

:)

Raja@ASUS
01-23-2017, 12:10 PM
Thanks, Arne.

FlanK3r
01-23-2017, 09:42 PM
How is it possible?? Technically, the architecture of SKylake and Kabylake is the same in CPU part...same 14nm process, same P-states etc. Maybe it can be based also at leakage of chip (remember this only from 32nm Vishera chips, there were 50W diffrence with same settings in 20 tested pieces)?

Raja@ASUS
01-24-2017, 08:12 AM
How is it possible?? Technically, the architecture of SKylake and Kabylake is the same in CPU part...same 14nm process, same P-states etc. Maybe it can be based also at leakage of chip (remember this only from 32nm Vishera chips, there were 50W diffrence with same settings in 20 tested pieces)?


It's called process enhancements which lead to lower leakage at equivalent voltage and frequency. If there were no advances, Kaby Lake wouldnt clock any higher than Skylake, when, in fact, it does. The base TDP for both architectures is the same, yet KBL has a higher boost clock. That tells you right there it's more efficient. It's worth having a think about what you're posting about before you post, Michal. Or, at least bring your own measurements to the table before making these posts (citing AMD's process issues doesn't count). ;)

FlanK3r
01-24-2017, 08:46 AM
You right man, sometimes Im too much quickly in writing than in thinking :) Thank you.