Intel Kaby Lake Core i7-7700K Overclocking Preview @ [H]

mmm
Getting a 7600k today or tomorrow (I'm not gonna spring 150€ extra just for ht, I'm not doing anything that could benefit from heavy multithreading)

Wanted to delid the thing and replace the TIM

But

My shiny new Kraken x62 is taking a week longer to be available
So I'll run it stock on my Formula VIII with an air cooler and see where OC'ing takes me

Kinda before and after :D
 
Looking good! Someone got 7Ghz using the older Z170 oc-formula and LN2..
Not sure if Zen will have the same luck considering the AMD past history.
 
lukart said:
Looking good! Someone got 7Ghz using the older Z170 oc-formula and LN2..
Not sure if Zen will have the same luck considering the AMD past history.
Click to expand...
The wall on good water looks to be 5.1GHz. And from what I am told, those CPUs are few and far between.
 
What, no assurances of 5ghz overclocks on air (else you're doing it wrong noob!) like a certain other release? Whatever happened to that guy anyway?
 
I'm guessing you are still under NDA for the motherboard? Or are you using a Z170 board?

What I'm wondering is regarding the memory overclock is that due to the newer motherboards or actually Kaby Lake. From what I can tell with all the z270 leaks from Asus specifically advertise higher memory support compared to equivalent z170 boards. And the Asus Z270 press release today specifically mentions that the new motherboards all use a newer design to facilitate higher memory speeds.

Third-generation ASUS T-Topology takes DDR4 memory overclocking to new heights — up to 4133MHz or even beyond[2]. Refined manufacturing ensures ultra-consistent quality and minimal crosstalk between custom trace paths, so users can enjoy the thrill of overclocking DDR4 with time-aligned signaling for superior stability. And ASUS motherboards are tested for compatibility with industry-leading 790+[3] DDR4 memory modules — for great flexibility and reassurance.
Click to expand...
 
finally we are back up to what sb would hit years ago. kyle do you have any guess to what % will hit 5 on good water?
 
were you guys able to get it close enough to throttle at these stable settings?
 
limitedaccess said:
I'm guessing you are still under NDA for the motherboard? Or are you using a Z170 board?

What I'm wondering is regarding the memory overclock is that due to the newer motherboards or actually Kaby Lake. From what I can tell with all the z270 leaks from Asus specifically advertise higher memory support compared to equivalent z170 boards. And the Asus Z270 press release today specifically mentions that the new motherboards all use a newer design to facilitate higher memory speeds.
Click to expand...
All Z270 boards used.

tunatime said:
finally we are back up to what sb would hit years ago. kyle do you have any guess to what % will hit 5 on good water?
Click to expand...
No clue. I am not feeling as good about it this week as I was last.

Trimlock said:
were you guys able to get it close enough to throttle at these stable settings?
Click to expand...
I never saw any throttling of any kind in ALL of my OC testing. When they failed they just BSODed.
 
Something AMD shill, screwing Intel out of their magic with 5.x++ O/C. And sticking it to the giant of the computer industry. etc.

Keep up the good work.
 
What sort of temps were you getting at those speeds? What did you have as "good water"? Do you think by the time retail comes around there might be board changes/BIOS updates that allow for better overall OCs?
 
BlueFireIce said:
What sort of temps were you getting at those speeds? What did you have as "good water"? Do you think by the time retail comes around there might be board changes/BIOS updates that allow for better overall OCs?
Click to expand...
Pic is there so you can see the load temps yourself.
Written in the article.
No.
 
Last edited:
Dan_D said:
Speaking for myself, I only see temps like that under Prime95 or encoding tasks. I do not see that while gaming or in general usage. I'm using a Koolance Exos 2.5 with a CPU-370 waterblock.
Click to expand...

For myself, it depends on what radiator was used and what fans and what speed, as I like silent, which often means running pumps and fans much slower than most, and as such I tend to get much higher temps even while gaming. Being a triple, however, even with slower fans, bigger and faster probably would not do much, would probably need a delid to see real drops.
 
BlueFireIce said:
Had to check on my phone, seems work connection was blocking the images? Who knows. I only saw "triple radiator" listed, nothing about what kind of rad or if that was 120x3 or 140x3 etc, along with fans, as 82C doesn't seem to good. My Swiftech H220-X only topped 78C hitting 5GHz on my 6700k and that was at much higher voltage than being reported for this CPU.
Click to expand...
http://koolance.com/ex2-1055-exos-2.5-liquid-cooling-system-aluminum and new 370 block. Running pump and fans at highest settings.

We do not shoot for silent when doing overclocking reviews.

Yes, 6700K vCores are going to look very different from 7700K vCores.
 
There weren't that many chips in the Sandy Bridge days that could do 5GHz. I think I had three out of ten I personally tested that could. Water was required to do it in all cases. The rest all did 4.8GHz easily. I don't think I ever saw a Sandy Bridge CPU that could only do 4.7GHz or less. I know they exist but I never saw them. So far Kaby Lake is shaping up to be about the same. We'll see though as time goes on and we get more chips on the bench.
 
The only real advantage it seems to have is running faster memory speeds.

Problem is that the improvements have been so incremental over the last few years that you really can't get excited by it.

I think that's why everyone is so excited by Ryzen, we know it won't be as good, but dammed if it is something different.
 
Nimisys said:
The only real advantage it seems to have is running faster memory speeds.

Problem is that the improvements have been so incremental over the last few years that you really can't get excited by it.

I think that's why everyone is so excited by Ryzen, we know it won't be as good, but dammed if it is something different.
Click to expand...

I think Kaby Lake is a bit better than that. Yeah it's incremental but the memory controller is better and we finally get clock speed parity with Sandy Bridge. Finally those IPC improvements don't feel like taking one step forward and two steps back.
 
So I'll have to delid my chip to hit the same speeds without fancy custom watercooling? Challenge accepted!

Thanks for the preview guys!
 
Dan_D said:
I think Kaby Lake is a bit better than that. Yeah it's incremental but the memory controller is better and we finally get clock speed parity with Sandy Bridge. Finally those IPC improvements don't feel like taking one step forward and two steps back.
Click to expand...

Since Sandy Bridge there have been roughly 5% increases in aggregate average IPC, but also roughly 5% reductions in max OC clockspeed so it was a wash except for improvements in special instruction sets and motherboard features like m2 slots.

Performance per watt is pointless for single socket home use/ prosumer once you start overclocking. I've never really understood why people who would buy chips like a 7700K, overclock it to the limit, and then complain about wattage or thermal loads.
 
Teenyman45 said:
Since Sandy Bridge there have been roughly 5% increases in aggregate average IPC, but also roughly 5% reductions in max OC clockspeed so it was a wash except for improvements in special instruction sets and motherboard features like m2 slots.

Performance per watt is pointless for single socket home use/ prosumer once you start overclocking. I've never really understood why people who would buy chips like a 7700K, overclock it to the limit, and then complain about wattage or thermal loads.
Click to expand...

I never understood why people would want to bother OC a 7700K anyway besides obvious e-peen issues.
 
Dan_D said:
I think Kaby Lake is a bit better than that. Yeah it's incremental but the memory controller is better and we finally get clock speed parity with Sandy Bridge. Finally those IPC improvements don't feel like taking one step forward and two steps back.
Click to expand...

I asked this earlier but is the higher memory speeds actually due to Kaby Lake memory controller improvements or z270 board design improvements? Or how much is attributable to each factor? As I believe the mobo makers are advertising higher memory speeds due to improved design.

This is of course largely academic to some extent but I'm wondering what memory speed results would with a 6700k in a z270 board vs a 7700k in a z170 board? This information be somewhat relevant to those looking to pair Kaby Lake CPus with z170 boards or even the reverse (although a less likely scenario).

On a more practical personal note I'd be interested in a gaming review with memory speeds clocked high targeting games such as Fallout 4 and similar scenarios which might be highly affected. Can we play FO4 without the dips in certain sections?
 
limitedaccess said:
I asked this earlier but is the higher memory speeds actually due to Kaby Lake memory controller improvements or z270 board design improvements? Or how much is attributable to each factor? As I believe the mobo makers are advertising higher memory speeds due to improved design.

This is of course largely academic to some extent but I'm wondering what memory speed results would with a 6700k in a z270 board vs a 7700k in a z170 board? This information be somewhat relevant to those looking to pair Kaby Lake CPus with z170 boards or even the reverse (although a less likely scenario).

On a more practical personal note I'd be interested in a gaming review with memory speeds clocked high targeting games such as Fallout 4 and similar scenarios which might be highly affected. Can we play FO4 without the dips in certain sections?
Click to expand...

To some degree, there is an improvement in the motherboards. Last generation there were several motherboards that had fewer PCB layers and lacked the capability of carrying the higher frequencies over the memory bus. Additionally, not all motherboards are equal when it comes to memory power implementation. That said, those things can still happen this generation but we are told we shouldn't see that this time around. At least not with the earlier crop of motherboards hitting the market. We were also told that the memory controller on Kaby Lake is flat out better and so far our testing seems conclusive that this is indeed the case. Kaby Lake doesn't seem to lose stability when pushed to the edge of an overclock with the memory clocks being way up there. Its still possible to find CPUs that may clock higher clock speed wise but do less well on the memory front. So far the two I've tested can do 3600MHz easily at 4.8GHz. I have some DDR4 3866MHz memory as well but there are some issues with it and Kaby Lake. It doesn't detect properly and is flaky at those speeds with those CPUs. I couldn't even push that memory to DDR4 3866MHz on all Z170 or X99 motherboards regardless of the CPU used.

So indeed, there are multiple factors that effect memory clocks but I think its safe to say that its one area where Kaby Lake is much improved over even Skylake. Its an even larger improvement over Haswell-E and Broadwell-E.

I do have a 6700K on hand. I didn't have time to do it with this review but I was curious at how it behaves on the newer Z270 motherboards.
 
lolfail9001 said:
Well, how about the proper [H]ard question: there were reports that delid improved OC potential as well as temps. Will you consider doing it to one of CPUs you have or shall purchase?
Click to expand...
Send me one I can afford to destroy and I will fully document the entire procedure on video.
 
limitedaccess said:
I asked this earlier but is the higher memory speeds actually due to Kaby Lake memory controller improvements or z270 board design improvements? Or how much is attributable to each factor? As I believe the mobo makers are advertising higher memory speeds due to improved design.
Click to expand...
My guess is that it is nearly 100% due to the CPU's IMC. That said, let me throw one on right now and see. :)
 
limitedaccess said:
I asked this earlier but is the higher memory speeds actually due to Kaby Lake memory controller improvements or z270 board design improvements? Or how much is attributable to each factor? As I believe the mobo makers are advertising higher memory speeds due to improved design.
Click to expand...
OK....I only have one 6700K here left to test with at this time, so this is a narrow look. This is on an MSI Z270 Gaming M7. I just took the 7700K off this board that was running a rock solid 5GHz/3600MHz with 32GB of Corsair RAM using the XMP profile settings.

With the 6700K set to full stock optimized "stock" settings with just the XMP profile enabled, the board will not even POST. I tried with two sticks and four sticks, same no-POST.

Hardware pr0n for fun. :)

Z270XPr0N.jpg
 
Now this is extremely interesting. Given these CPUs have an iGPU, I uninstalled the TITAN X card I had on this system. Now "logic" tells me this should not impact the IMC positively at all, but I figured I would give it a shot. I have this 6700K up and running at stock 4.2GHz but with the Corsair RAM at 3600MHz with the XMP profile. Running a HandBrake encode to see if it will work then maybe some RealBench. Still very strange. Keep in mind this is all "from the hip" testing, which is not normally how we would do things we were writing an article about. We like to see something work more than once before drawing conclusions.

6700Kat3600Desktop.png

Z270Pr0n2.jpg
 
HeadRusch said:
A lot of us are still running 2600's....I'd love to see a clock-for-clock comparison between that 5 year old part and this latest-and-greatest, see that 4 fps improvement..........but I'm guessing HardOCP has no need to keep platforms that old alive for testing....still, Kaby Lake seems to be more of what we already have....
Click to expand...

I don't know about Kyle but I don't have anything quite that old. I do have a Core i5 3570 I pulled from my girlfriend's machine this last year or so and the motherboard that goes with it. That's as far back as I've got without going to something even older than Sandy Bridge. I've still got my Skulltrail and Gulftown CPUs and motherboards floating around.
 
HeadRusch said:
A lot of us are still running 2600's....I'd love to see a clock-for-clock comparison between that 5 year old part and this latest-and-greatest, see that 4 fps improvement..........but I'm guessing HardOCP has no need to keep platforms that old alive for testing....still, Kaby Lake seems to be more of what we already have....
Click to expand...
I just looked and could not find one here. I very likely doled those out to editors for testing platforms back in the day, and quite frankly, those likely never came back to me and if they did, I likely used those to build machines for others.
 
It looks like a nice cpu to run all day at 4.7 or 4.8 which isn't shabby one little bit.
Plus looks like some very fast RAM will be paired up.

It may be time to get off the 4770Ks I am using.
 
Did you use the "AVX offset" to increase the chances of a better overclock, as described here:
http://www.anandtech.com/show/10968...-review-the-new-stock-performance-champion/11 ?
There’s also a new feature worth mentioning before we get into the meat: AVX Offset. We go into this more in our bigger overclocking piece, but the crux is that AVX instructions are power hungry and hurt stability when overclocked. The new Kaby Lake processors come with BIOS options to implement an offset for these instructions in the form of a negative multiplier. As a result, a user can stick on a high main overclock with a reduced AVX frequency for when the odd instruction comes along that would have previously caused the system to crash.
Click to expand...
 
EODetroit said:
Did you use the "AVX offset" to increase the chances of a better overclock, as described here:
http://www.anandtech.com/show/10968...-review-the-new-stock-performance-champion/11 ?
Click to expand...
No we did not. There is this option in many of the motherboards we have. But you are trading off AVX performance for a 100MHz of so. Now what is AVX used for? Most notably Prime95 uses it and is why we have generally dropped it from our stability tests because it so heavily uses AVX that it does not rep real world usage.

But would you start turning off instruction sets for clocks? For benchmarking and stability we are going to that instruction set in-tact as set up by how Intel wants the CPU to work.

Surely you can do what you want. But when it comes to reviewing hardware, how much do we hack into the way the hardware works just to get a little better clock, and where do you stop at doing that?
 
You must log in or register to reply here.
Back
Top