i7-7700K first benchmarks: 40% up in single threaded, 20% up in multithreaded (UPDATED)

[qubit]

And that's over Skylake. Over my old Sandy Bridge, the improvement is going to be very large indeed. This finally sounds like the CPU generation that will be worthy of me upgrading from SB and should improve framerates in CPU limited games.

If Intel are making such a big performance improvement, then perhaps AMD's Zen really is an effective competitor? The reviews will finally tell us.

http://wccftech.com/intel-core-i7-7700k-benchmarks

EDIT Important update from wccftech. Thanks @silentbogo for spotting it.

Update [10/4/2016]: Some readers have pointed out something concerning about the benchmark in question. We were using this verified result of the Core i7 6700k as a baseline to compare the new results against. As you can see, it shows a single core score of 4300 and a multi core score of 16756, which gave the percentage figures of 40% and 20% respectively. Unfortunately, however, there is a lot of inconsistency in the scores reported by this benchmark and not all results are alike. The biggest issue is that the clock speed readings appear to be read completely wrong. I was able to find Core i7 6700k multi core scores as low as 11000 and as high as 22000 at seemingly the same clocks. This means that analysing the performance gains depicted in this particular leak gets very difficult without injecting significant subjective opinion, which is why I have decided to edit out the percentage gains from the headline. I apologise for the err on my part in not conducting the necessary due diligence.

 

[Kanan]

Wccftech and only 1 benchmark. Sorry but I call that irrelevant. It's most likely a 5% ipc gain like every other "tick" architecture before it (ivy bridge, Broadwell). You can call me out if im wrong. But I'm not. 40% single threaded hahahaha good joke. Just a reminder: skylake is a new architecture and kabylake just a refinement means small gains like always. Wccftech click bait at its best.

 

[CounterZeus]

previous leak was like 2% improvement (IPC)? 40% is a joke.

 

[FYFI13]

Either fake or CPU was running at very high frequency. There's no way they could have achieved 40% performance improvement on the same tech process and same architecture. I could bet that actual performance increase will be well bellow 5%. And here I'm being very optimistic...

 

[Easo]

Suuure. Way too fantastic increase, since this is just a die shrink...

 

[evernessince]

Yeah, don't drink the kool-aid.

 

[Kanan]

Suuure. Way too fantastic increase, since this is just a die shrink...

Not even, the die shrink (tick) happened with Broadwell. Kabylake is even less than what's Broadwell worth basically. Evolutionary spoken, a tick without shrinking because they stay at 14nm.

 

[newtekie1]

They already said they are abandoning tick/tock. There likely won't be a die shrink for 1150, and Kabylake is almost identical to Skylake CPU wise, the improvement was almost entirely on the GPU side.

 

[FordGT90Concept]

They probably used one instruction that was added to Kabylake over Skylake. In general computing, no way man. I want to see the equivalent in Geekbench on 6700K to see where the improvements are coming from, specifically. Oh, hey, I have a 6700K...

 

[Vario]

I guess we will see. I am really happy with the lasting performance Ivy has offered but we need to see some real IPC gains. Same problem with Videocards. Seems like storage (SSD) is the current frontier.

 

[FordGT90Concept]

Here we go (I made sure to get 4.0.0 instead of 4.0.1 on the website):
https://browser.geekbench.com/v4/cpu/621108

Comparison:
https://browser.primatelabs.com/v4/cpu/compare/621108?baseline=583064

I scored 700 points higher than WCCF claimed...
Single-Core Score: 5033
Multi-Core Score: 15068
...they must be smoking some good shit.


SGEMM is way up (almost double). Not entirely sure what that even is...

I'm thinking most of the performance gain could be from using faster memory (note how Geekbench fails to read the memory speed on both). A lot of the tests mirror the difference in memory performance (21% gain). Add in the 200 MHz difference and...whatever instruction is boosting SGEMM and I think you'd end up with 0-10% improvement in the cores itself.

If Intel are making such a big performance improvement, then perhaps AMD's Zen really is an effective competitor? The reviews will finally tell us.

If memory serves, Intel is planning on launching a 6-core processor on the consumer LGA. I highly doubt Intel would do that if they didn't feel Zen was a potential threat.

 

[Melvis]

Isnt it meant to turbo to like 4.5GHz or something?

 

[slozomby]

i'd like to know what math makes a score of 6131 40% higher than a score of 5612 ( stock 6700k) on geekbench.

maybe its a typo and it was supposed to be 4%.

 

[biffzinker]

Isnt it meant to turbo to like 4.5GHz or something?

It's rumoured turbo clock speed is 4.5 GHz with the base clock set at 4.2 GHz.

Intel’s K Series Kaby Lake CPUs:
So talking about the details, first up, we have two “K” series models. These chips have an unlocked design that indicates that they are built for overclocking. The flagship is the Core i7-7700K features 4 cores, 8 threads and 95W TDP. This chip has a base clock of 4.2 GHz and boost clock of 4.5 GHz while featuring 8 MB of L3 cache. The Core i5-7600K is a quad core chip without multi-threading support that packs 6 MB of L3 cache and a 95W TDP. The chip features a base clock of 3.8 GHz and would turbo beyond 4 GHz.

http://wccftech.com/intel-kaby-lake-desktop-lineup-leak/

 

[D007]

Even if it was 20%.. I'd still be like.. Bleh..lol..

 

[Vayra86]

WCCFTech and Kaby Lake has already been benched at 2-4% IPC improvement.

So yes, this is definitely BS and clickbait

https://tweakers.net/reviews/4909/kaby-lake-de-i7-7500u-getest-in-de-asus-zenbook-3.html

For some realistic numbers on the low power parts.

 

[BiggieShady]

40% higher single threaded performance than skylake can be only achieved by 33.33 % higher clocks ... and isn't score 4300 too low for skylake?

I scored 700 points higher than WCCF claimed...
Single-Core Score: 5033
Multi-Core Score: 15068
...they must be smoking some good shit.

interesting ... and they scored higher in multi threaded test. Must be that they have disabled turbo on 6700k and enabled it on 7700k. Problem solved.

 

[Bill_Bright]

Come on! Why would WCCFTech fake this? That makes no sense. Nor does it make sense to accuse WCCFTech of click bating, BS, or whatever. They are just reporting what they saw with "leaked" results with a pre-production processor.

I think it is a very honest report. It clearly states it is one benchmark. Everyone knows (including WCCFTech) that there is no single benchmark that shows all the pluses and minuses of every CPU in every scenario. WCCFTech is not trying pretend this one does. In fact, they raise their own doubts by saying,

if Geekbench is anything to go by...

It is prudent to note here that due caution is advised in a few aspects of the matter. 1) It is never ever wise to take a single benchmark as being indicative of real life performance of a processor and 2) we will have to wait for a clock to clock analysis to actually figure out just how much better the Kaby Lake architecture is from Sky Lake.

Other reports with other benchmarks include this which shows some nice performance gains too, and this and others.

I say let's reserve judgement on the processor (like WCCFTech is doing) and these sites until the final production processor is RTM and solid platforms are out that are designed to take advantage of and can illustrate the advantages of this processor.

This is just a preliminary report. No need to start bashing what are little more than rumors with a pre-production, leaked product.

 

[RejZoR]

And yet it's still a 4/8 configuration. C'mon Intel, we've had that back in 2008...

 

[Alduin]

Skylake and Kaby Lake have the same IPC.
Kaby Lake features new graphics architecture to improve performance in 3D graphics.
Kaby lake features higher clock rates.
The difference between kaby lake and skylake is the new graphics architecture

 

[FordGT90Concept]

40% higher single threaded performance than skylake can be only achieved by 33.33 % higher clocks ... and isn't score 4300 too low for skylake?

interesting ... and they scored higher in multi threaded test. Must be that they have disabled turbo on 6700k and enabled it on 7700k. Problem solved.

My turbo never really engages. If they changed it in 7700K so turbo works when the core CPU is at full load when there is thermal overhead to allow it, then yeah, 500 MHz difference. And seriously, why not? My CPU never hits 60C yet it never turbos. Kind of wasted potential.

 

[64K]

I would take most of what comes out of WCCFtech with a grain of salt imo

 

[Bill_Bright]

I would take most of what comes out of WCCFtech with a grain of salt imo

Again I will point out that even WCCFtech raised their own doubts about those finding - that's a sign of integrity.

 

[RejZoR]

My turbo never really engages. If they changed it in 7700K so turbo works when the core CPU is at full load when there is thermal overhead to allow it, then yeah, 500 MHz difference. And seriously, why not? My CPU never hits 60C yet it never turbos. Kind of wasted potential.

Turbo used to kick in only when just 2 cores were utilized. As soon as more of them kicked in, it dropped to base clock. Which is kinda lame. You could set the multiplier to achieve boost clock on all cores without any change in voltage. This way it can boost on all cores to a designated clock.
There were some changes how Turbo works, but I don't think they've changed this.

 

[Kanan]

Don't forget that boost is also limited by tdp not only temperatures. Intel states that max turbo is only on by default settings when not all cores are in use.