Overclocking Intel’s Core i7-7700K: Kaby Lake Hits The Desktop!

[Xajel]

Im going to wait for cannonlake. For now I'll stick with my good old Ivy Bridge i5 3570k.

I don't see a reason to upgrade either, still happily running along with my i5-3570K.

Yeah, My i5 3570K is good for now.. lets see how Zen will do and how these two camps will fight while we stick with the good old i5-3570k...

My next upgrade will be mostly a laptop anyway...

 

[cats_Paw]

So... still within the ballpark of 3770k, 4770k, 6700k.
If zen is not a massive improvement I will probably get something like a 5820k just to avoid win10.

 

[Olle P]

The results are kind of disappointing for a 200mhz clock bump...

A 5% boost clock increase is expected to result in <5% overall performance increase.
So what's disappointing?

... The thermal figures are surprising, I would have thought that they wouldn't have done that bad.

The drop in idle power consumption is nice!
The increase in maximum power consumption is puzzling, but it might be an effect of better HT (and thus core) utility?

There are quite a few pieces of content out there showing a skylake i5 vs a 2500k. I don't remember specifics, but it was a pretty handy increase.

I recently went that road myself, upgrading from i5-2500K with 1,600 MHz DDR3 to i5-6600K with 2,666 MHz DDR4. (The only reason I did it was that I erroneously thought my motherboard was broken. Performance was sufficient for my needs.)
At stock speeds the power consumption is roughly the same while performance increase is 20-25%. Far from "pretty handy" in my book, especially considered that there's a five years and four generations difference between them. Given that they're about the same price Moore's Law stipulate an expected eightfold performance increase in that time period!

Possible reasons I can think of for buying a Kaby Lake i7-7700K:

• ■ You want to stay at the bleeding edge of CPUs, and have the money to do so.■ You have an i3 or i5 CPU on a Z170 motherboard and want a substantial performance increase.■ You have an older generation i7 on an older motherboard and need the extra features of a new motherboard without sacrificing CPU performance.

 

[Jonathanese]

Soooo.... From what I gather, if you take a 6700K, overclock it by 200MHz, replace the thermal paste with some off-brand junk or peanut butter, and get a slightly less efficient power supply, you end up with the 7700K?

And this is an upgrade?

I've been looking for something to upgrade my 3570K 4.6GHz from. 4 generations later and there still isn't enough of a difference to upgrade to a new i7.

 

[Isokolon]

Well this is disappointing
I expected Skylake+
No Big Step Up but a little

But all we've got is a factory overclocked Skylake it seems. oc the 6700k 200MHz and you'll end up with the same heat and power draw.

Well let's hope 10nm can deliver what it implies so far.

 

[Crashman]

Soooo.... From what I gather, if you take a 6700K, overclock it by 200MHz, replace the thermal paste with some off-brand junk or peanut butter, and get a slightly less efficient power supply, you end up with the 7700K?

And this is an upgrade?

I've been looking for something to upgrade my 3570K 4.6GHz from. 4 generations later and there still isn't enough of a difference to upgrade to a new i7.

Or, you could gather that the firmware wasn't quite ready yet, since there was a 40W deviation between hard-cleared settings and soft-cleared settings. There is the 200MHz additional overclocking headroom, and there are guys who aren't so worried about heat that they'll run eight threads of Prime95 with AVX enabled, who are reaching 5GHz at higher voltage levels. Yeh, we should probably ignore those 5GHz guys for purposes of this article since we're actually looking for test consistency

 

[xFeaRDom]

Not sure why a lot of people are complaining, I don't get what the point of upgrading from the current generation 6600K or 6700K to the Kaby Lake, its purely a waste of money. And the way I see it, as because they haven't been fully released yet, they're just basically bug testing, fixing things such as the power usage, aka the firmware.

And for those complaining that they may as well wait for the 10nm process, it can't be too major to you, just get a 6700K and overclock it if performance is that important.

The Kaby Lake generation, the way I see it, is purely there for people who didn't think the Skylake process was worth it at the time, as their systems might be quite old and they're ready for an upgrade, and they can then upgrade to the Kaby Lake generation.

Although this upcoming generation isn't going to blow anybody away, because its merely not as possible as it was 10, 15 years ago or so, when increasing clock speed, core speeds and extra technologies to processors, now until a new conductor or processes are found, the clock speeds and performance gains are going to barely move anywhere, the only thing they can really do, is improve current technologies, and decreate power usage and making things more efficient, whereas if you look at something like AMD, who used to be classed as the king, or close to being there, as they're aimed at the lower end more than the high-end intels, if they pull the Zen generation and it lives up to the hype they're making it, then they could make a large jump towards Intel, rather than Intel taking a walk in the park and not having to do much to stay there.

As someone who is coming up to wanting a nice upgrade, the availability of overclocking, then I would look at this being an upgrade, or the 6600K if the price drops enough, and there's probably going to be a dump on Ebay or other websites for people who just want the bragging rights of 'I got a new Kaby Lake overclocked to bla bla bla', so I could easily just pick up one of them, and as long as it runs at stock then I've not really lost anything.

 

[theyeti87]

I have been looking to Kaby Lake to replace my old i7-950 system. Perhaps I'll wait for the next cycle...

 

[Crashman]

I have been looking to Kaby Lake to replace my old i7-950 system. Perhaps I'll wait for the next cycle...

Would you like Tom's Hardware to test it on another motherboard? I don't think we'll know anything with certainty regarding its heat and power figures until Z270 comes out, but I'm sure more info could be useful prior to Z270. These processors are Z170 compatible, after all.

 

[sarinaide]

It shouldn't really make any difference as the Z170 can support the Kabylake parts. While it won't give a true indication on performance and features, the article was not about that, for a person with said board this may be alarming. Maybe they can try find whatever it is that is causing the issue before deducing anything out of it, maybe it is just this board, maybe it is all Z170's either way just dumping it on Z270 seems wrong as the end user expects it to work on a Z170

 

[hixbot]

The only thing good abount this release is perhaps skylake will drop in price slightly. Very disappointing indeed. AMD need to step up fast because consumers are getting gouged. Really hope AMD can push Intel into releasing 8 cores to the mainstream.

 

[CaptainTom]

I'm sorry but 4.8 GHz is nothing special lol.

We have been able to reach that since Sandy Bridge (Yes even with IB - HW). Hell my 6700K hit 4.8GHz with ease.

This thing is nothing but a rebadged 6700K.

 

[CaptainTom]

Tick -> Tock -> Tock should really just be Tick -> Tock -> Tweak

It's not even really Tick > Tock > Tweak


It's more like Tick > Tock > Milk.

 

[turkey3_scratch]

Tick - Ety - Tock

 

[JackNaylorPE]

A 10% performance increase for a speed bump from 4.6 to 4.8 is a pretty good return and one of the larger increases we have seen in years. A Reason to upgrade ti the next generation ? We haven't had that since Sandy Bridge.

Would love to know what PSU was used ... or better said what efficiency point it was at, so one could get an indication of actual CPU draw. And since heat increases exponentially with higher voltages and voltage increase also increase greater than linear with each extra multiplier increase, the heat numbers aren't a surprise either.

So while it's an improvement better some of the recent generations, it's still not a reason to break out the credit cards and expect to be impressed. Also worth mentioning that we are comparing an early stepping CPU with one that is quite mature. Unless the Kaby was hand picked, we can expect some slight improvements along the way.

What is somewhat disconcerting ... or not depending on PoV is that even tho KL has broken into double digit % point increases, the last GFX incremental generational improvement brought us > a 50% improvement. This has changed the way we look at system building as for example, it takes the shine off SLI for example ... We are not seeing the scaling (70% average in TPU Game Test Suite, 95-100% in most demanding games) we have seen in the past. There are several reasons for this but one of them is some games are being CPU limited such that outside 4k, it no longer the performance improvement doesn't justify the cost.

Would be an interesting article charting / comparing generation to generation performance increases over time (say past 20 years or this millenium") for the components (CPU / Memory / GFX / Storage / Networking) that go into our systems

 

[RedJaron]

It shouldn't really make any difference as the Z170 can support the Kabylake parts.

Considering the power delivery is managed by the board, and thus by the firmware, a different board with better tuned firmware could make all the difference in regular use heat and efficiency on Kaby.

Maybe they can try find whatever it is that is causing the issue before deducing anything out of it, maybe it is just this board, maybe it is all Z170's either way just dumping it on Z270 seems wrong as the end user expects it to work on a Z170

It could be all that, and one way to find out is to isolate variables. That would include using other boards / firmwares.

I'm sorry but 4.8 GHz is nothing special lol. We have been able to reach that since Sandy Bridge.

Um, you do realize that a Kaby Lake at 4.8 GHz would be considerably faster than a Sandy Bridge at 4.8 GHz, right? More important, while some individual chips in the last few CPU generations have been able to reach 4.8 GHz, Sandy Bridge is basically the last time such clocks could be regularly expected, and achieved, at safe voltage/temps for 24/7 use. Notice this preview is only using an NH-U12S cooler, not an exotic liquid loop. Outside a few Golden Samples, most Haswell chips topped out in the 4.5 - 4.65 range on a mid-size air cooler.

 

[TJ Hooker]

A 10% performance increase for a speed bump from 4.6 to 4.8 is a pretty good return and one of the larger increases we have seen in years. A Reason to upgrade ti the next generation ? We haven't had that since Sandy Bridge.

Where does the 7700k at 4.8 outperform the 6700k at 4.6 by 10%?

 

[Compuser10165]

When the leaked benchmarks appeared a while ago I did a comparison with my own i7 6700K clocked at 4.5 ghz and I ended up with ~7% performance difference in both single an multi-threaded performance. This seems to be in line with that result more or less.

 

[lp231]

Your Skylake isn't going to stop working when these CPUs comes out, and Intel isn't forcing people to upgrade. What's with all the moaning about how bad the performance is? It's been like that since Sandy Bridge, a reason why some have not seen the point to upgrade. Now your looking to upgrade because of performance, don't bother. Upgrading for new features then Kaby Lake will have current latest features.

 

[JackNaylorPE]

Where does the 7700k at 4.8 outperform the 6700k at 4.6 by 10%?

I'm responding to the statement in 1st post .. so if ya want the answer, that would be the person to ask.

I'm sorry but 4.8 GHz is nothing special lol.

We have been able to reach that since Sandy Bridge (Yes even with IB - HW). Hell my 6700K hit 4.8GHz with ease.

I think that's a bit misleading.... unless you don't define something that 95 % of CPUs couldn't do as "something special". And let's not forget, we are talking 4.8 at 1.3 volts. ... not 4.8 at 1.45v.

We got a lot of 4.8 results on SB ... even on 1st build on that platform ... however it was @ 1.41 volts and core temps averaged 75C. 5.0 Hz was doable for between 1 and 2% without going too crazy. The honeymoon was over after that.

Yes, after that was a bit messy, Ivy was much worse in that results were all over the place ... some could get to 5.0 Ghz but they were much hotter than SB, requiring something "better than your typical air cooler. Most who got good multipliers used water . K even came out with the "Naked Ivy" CPU mount as delidding suddenly became something for even the casual enthusiast.

Haswell was not much better with just under 5% reaching 4.8. So yes with under 5% capable of reaching 4.8, I have to call that pretty special.

Devils Canyon did better ... the overclock.net spreadsheet has about 18% making 4.8 Ghz, many with voltages above 1.5, 90C and even a few with LN2...quite a number used water. Keep in mind that these are "self reported" often w/o validation and when they do get it, how long did it remain stable ? More importantly, how many web site reviews can ya find where MoBo review got OC better than 4.6 ? Bet if we look, we find more who didn't than ones that did.

For Skylake, things improved a bit, jumping to 22% getting 4.8 or better...slightly better than 1 in 5. Only a third of those were able to do it under 1.40 vcore setting tho. And again, these results are self reported.

So I would have to say that managing 4.8 GHz in a published review is something pretty special in that it is quite rate. Managing 4.8 Ghz at 1.3 volts is something that is very rare.

Here's what Guru3D Wrote for each generation:

6700k: Reached 4900 @ 1.45 volts

4790k: "Many of you will likely end up with a maximum OC of 4600 maybe 4700 MHz max we think."

4770k: "we had a hard time getting over 4600 MHz stable"

3770k: Guru3D made 4900 ... but at 1.4 volts.

2600k: Guru 3D managed 4300 at 1.3 volts

 

[CaptainTom]

Honest question for everyone here: WTF happened to Tom's Hardware?


The Haswell review had a "Yawn" in the title for only having a 10% IPC increase over IB, and yet now a near ZERO increase merits a glowing review?

When did TH become a puff-piece website that does nothing but shill marketing? I used to recommend this place to all of my new-to-PC friends, but now I avoid it like the plague. Nothing but ads and misinformation.

This is the website that in-depth analyzed micro-stuttering and proved that the R9 295X2 was fine with a 650w PSU.


Now TH considers a $400 1070 t be "Good for the money" over a $260 R9 Fury?!?!?!?! WTF?! This is pathetic!

 

[LeeRains]

Why would anyone want to buy a 7th gen Intel CPU when the last 5 generations have been nothing but sad little bumps in performance, just barely good enough to milk you, the customer?

 

[turkey3_scratch]

Cause the last gen cost the same amount.

 

[Crashman]

Honest question for everyone here: WTF happened to Tom's Hardware?


The Haswell review had a "Yawn" in the title for only having a 10% IPC increase over IB, and yet now a near ZERO increase merits a glowing review?

When did TH become a puff-piece website that does nothing but shill marketing? I used to recommend this place to all of my new-to-PC friends, but now I avoid it like the plague. Nothing but ads and misinformation.

This is the website that in-depth analyzed micro-stuttering and proved that the R9 295X2 was fine with a 650w PSU.


Now TH considers a $400 1070 t be "Good for the money" over a $260 R9 Fury?!?!?!?! WTF?! This is pathetic!

What happened TO YOU? Did you NOT want this data published? The article doesn't tell people to go out and buy a new 7700K to replace their 6700K unless that 6700K is burned, in which case the 7700K is a better chip to buy because it overclocks better. You can expect Intel to maintain its price structure, and anyone who thinks that the 6700K will have a drastic price drop hasn't learned anything from Intel's history yet.

 

[rantoc]

Mote like night time Tic->Toc->ZZZzzz.... With those performance "gains" its not even worth launching beside for money-milking purposes.

Intel is a very responsible party in the PC sales figure with lackluster "advances" like this, how much ipc/clock performance increases have we seen over the latest say 3 rounds? Perhaps they should just stop to waste all the milked pc cpu money on their repeated failed mobile attempts and instead spend some of that on pc cpu research and development. The later part have been absent for quite some while....