i7-7700k multiplier locked?


#21

Oh damn that’s a sexy cooler. Kinda makin’ me feel ashamed of my CRYORIG H7. . .


#22

Well, I got the Z270-A mobo installed. I guess it only takes 30 minutes, but still, a pain. Will test some more later.


#23

OK! I had the CPU fan connected to the alt CPU fan connector so that was a hiccup, but it’s up and running aaannnndddd…

Now it works fine. So whatever. Money solves all problems.


#24

Your chip can’t make it to 5Ghz? Or have you not really OC’d it yet?


#25

Shouldn’t you stop at 42 for obvious reasons?


#26

I have, this is all I need, the mild 10% is ok for me.


#27

That’s a 4.5% overclock, as the 7700K boosts to 4.5Ghz stock.

5.0Ghz is a 10% overclock.


#28

I couldn’t get it over 4.4 Ghz on the Z170 board. I did see 4.5 Ghz reported for single core loads in a different CPU monitor though. Weird that they reported differently.


#29

At stock, it will only boost to 4.5Ghz with a single core under load. You should be able to hit 5Ghz on your Z270 with little difficulty.

I mean, I have my 6700K at 4.6Ghz no problem. Kaby Lake doesn’t offer a sizable IPC improvement so literally the only reason to buy one is that it clocks higher.


#30

That and 10-bit HEVC decoding support. See http://jell.yfish.us/ and try playing back any 4k HEVC 10-bit clips on Skylake. It’s brutal

Kaby Lake is about 3% faster at the same speed, which isn’t much.


#31

Depends on which tests you use, last I saw at HardOCP they were identical to within the margin of error. But even 3% is far less than we’ve become accustomed to from Intel.


#32

I actually have a perfect apples-to-apples for this, because my old i7-6700k was overclocked to 4.4Ghz (from 4.2) and the new i7-7700k was stuck at 4.4 Ghz.

When I ran browserbench.org speedtest I got 193 vs. 183 – same browser, same mobo, same OS, same max clock speed. Literally only the CPU changed. So 5% based on JavaScript benchmarks. I understand Kaby has better speedshift which would help it do a tad better on bursty loads which represent most typical web browsing

At 4.7 Ghz I get 209 on http://browserbench.org/Speedometer/ now – that is a 12% improvement over my 4.4 Ghz i7-6700k.

(Reading that chart a little closer, it takes Skylake 35ms to reach max clock, whereas Kaby does it in 15ms.)


#33

I went ahead and increased the multiplier to 48x, just for you @stusser. Seems stable initially on quick Prime95 run.

Since my old i7-6700k was “only” overclocked to 4.4 Ghz my e-peen does not require anything higher than 4.8 Ghz with the i7-7700k, that suffices for me. I don’t need to reach the mystical 5.0 Ghz barrier to get to sleep at night.

edit: ooo I spoke too soon. I eventually get errors in Prime95 at 4.8 Ghz. Oh well.


#34

You will need to increase the voltage to get higher than the automatic motherboard overclock. Basically clock as high as you can go without going over 1.35v. As long as you don’t go over 75C or so under full load you’re good to go. I betcha 1.35v will get you to stable 5Ghz.

Those are conservative numbers, too. I like to play it safe when overclocking-- I take it as high as possible with borderline crazy voltages, then back it off until the voltage is somewhat close to stock. It’s impossible to destroy a CPU these days, if you run them outside limits they just shut it down. You can even startup without a cooler at all and you won’t damage it.


#35

It is already going over 80c at the current levels. You’d need water cooling to do better than this high end air cooler.

(Note that I am allergic to super noisy high speed fans though.)


#36

Ahh. Guess you aren’t going any higher then. Surprised it’s running so hot with high-end air. My NH-D15 does much better, at 1.31v I can do a stable 4.6Ghz and temps top out at 76C. Fans are audible during benchmarks, of course, but not otherwise.

I can get it up to 4.8Ghz but have to put voltage at 1.42v to get it stable and temps come dang close to boiling water. That’s Skylake, of course, not Kaby.


#37

Well, Noctua does make the best cpu air coolers…


#38

You mean 76c while running 8 MPrime (Prime 95) torture test “max power consumption” threads? That’s how I test.

edit: I just ordered a new Noctua 120mm fan as my old one was 3 pin instead of 4 pin and I figured it should be replaced. But instead of tossing my old fan – it’s from 2011, and was ordered at the same time as the Venomous cooler I still use – I cleaned it up and installed it on the front using the rubber pads and metal clips I (surprisingly) kept around. Now I have push-pull 120mm Noctua fans versus the single pull fan I had before. I still get eventual failures in Prime95 though. I guess the best of 2011 ain’t enough six years later… I’ll upgrade to the NH-D15 like you when the next upgrade cycle rolls around!

Interestingly, it is always the ‘same’ core that errors out on Prime95.


#39

Hmm looking a little more, the Ninja 4 outperforms the DH-15, in both single and dual fan configurations – and by a fair bit:

http://www.silentpcreview.com/article1462-page6.html

I’ll look into the Ninja 4.


#40

I use LinX not Prime95 as some people have said Prime95 isn’t safe on modern CPUs. I suspect that’s bullshit, but LinX brings my cores up to 100% and pumping out the heat just fine.

Your chip isn’t stable at higher clocks because it needs more voltage, that’s all. But voltage=heat and your temps are already >80C, so you really can’t get higher unless you improve your cooling.

Note that silentpcreview article is comparing against the NH-D15 with one stock fan removed. I use both fans in push/pull oriented towards the back of my case, where an exhaust fan gets the heat quickly out of my case.