Intel i3 7350K (4.2Ghz) GeekBench Benchmarks Leaked Allegedly Outperforms an i5 6400 & 4670K CPUs

[VerticalDiscussions]

32 minutes ago, Energycore said:

-snip-

Yes it is

[Energycore]

1 hour ago, VerticalDiscussions said:

Yes it is

I won't sugarcoat this for you

[Space Reptile]

[Wolther]

3 hours ago, Sauron said:

No, quite simply geekbench is a very bad benchmarking tool, especially for x86.

Oh, bummer. Why is the price increase so big then? Am I missing something, cause overclocking can be done on skylake i3s too

[Sauron]

6 minutes ago, Wolther said:

Oh, bummer. Why is the price increase so big then? Am I missing something, cause overclocking can be done on skylake i3s too

skylake i3s don't have an unlocked multiplier, this does. It makes oc much easier and you can get better frequencies usually. however in my opinion it's not worth the money over an i5

[Jumper118]

got 18K with my i5 6500 so.............................. this is poo lol pretty sure i get nearly 10K with an i3 6320

[MrDynamicMan]

"Outperforms" by as little as 12 points. Eh, its on the same tier.

[Swordtiny]

but you know that you dont need to upgrade just because a new generation of processor came out.
3rd & 4th gen of i5 and i7is still good enough except you want to do like extreme editing or stuff like that.

[Bananasplit_00]

i find it annoying that people are taking this as the "first overclockable I3" when its not, all first gen Core I series CPUs can be overclocked, all I3s, all I5s and all the I7s, there were no CPUs you couldent overclock and people seem to forget that

[Jumper118]

3 hours ago, Bananasplit_00 said:

i find it annoying that people are taking this as the "first overclockable I3" when its not, all first gen Core I series CPUs can be overclocked, all I3s, all I5s and all the I7s, there were no CPUs you couldent overclock and people seem to forget that

cos they are all noobs thet cantt remember before hasweasel

[Phate.exe]

1 hour ago, Jumper118 said:

cos they are all noobs thet cantt remember before hasweasel

Not as bad as the noobs that don't remember what Intel was like before Conroe.

[PlayStation 2]

On 11/24/2016 at 9:45 PM, RagnarokDel said:

An I3 better then a 4 years old I5 chip? Shocking! I'm pretty sure my 4670k was better then my I7 920 I had previously...

I'm fairly sure the 4670K just hit 2 years.

[PlayStation 2]

5 hours ago, Bananasplit_00 said:

i find it annoying that people are taking this as the "first overclockable I3" when its not, all first gen Core I series CPUs can be overclocked, all I3s, all I5s and all the I7s, there were no CPUs you couldent overclock and people seem to forget that

In terms of truly unlocked multiplier, your point is kinda moot.

In terms of overclocking in general, then your point stands.

[DarkBlade2117]

3 minutes ago, Dan Castellaneta said:

In terms of truly unlocked multiplier, your point is kinda moot.

In terms of overclocking in general, then your point stands.

can confirm http://imgur.com/a/Nmaqd

 

[DarkBlade2117]

On 11/24/2016 at 2:27 PM, ONOTech said:

It's not too surprising honestly. I tested it myself, and a non HT dual-core @ 4 GHz beats a non HT quad core @ 2 GHz in both multi-threaded and single-threaded applications (though results may vary a bit), assuming they use identical CPU architectures.

 

Now, add HT to the dual-core and give it the atypical generational 5% IPC increase, and voila, you have these results. It's why Intel has locked their i3s for so long - so people will buy the more expensive i5 variants. An unlocked i3 with good memory speed is a wolf in sheep's clothing.

As proved in multiple Digital Foundry videos, The i3 can tussle with a locked i5, yeah I could of Overclocked an i5-6500 also but I think it would of been hard to get a merely 1.3GHz overclock out of it compared to the .75Ghz I pulled out of my i3. Which I also don't get tall this BS on how i3's are shit, because I can recording ( Though I prefer using a hardware encoder for recording over a software encoder ), I can stream, I can edit, I can use Photoshop. All pretty optimal. I do however plan on getting a 6700k and 8 more GBs of RAM after I add in my money and hopefully have my parents split the cost with me

[Jumper118]

28 minutes ago, Phate.exe said:

Not as bad as the noobs that don't remember what Intel was like before Conroe.

idk they are pretty bad. these new gen noobs eat nvidia pr and intel pr pie for breakfast, lunch and dinner 

[Bananasplit_00]

23 minutes ago, Dan Castellaneta said:

In terms of truly unlocked multiplier, your point is kinda moot.

In terms of overclocking in general, then your point stands.

my point is that people take this as the first ever overclockable I3 just because its a K skew, you can overclock I3s from Skylake if you can find a motherboard that hasent had the update aplied yet to lock the multipliter and as i said first gen I3s you could overclock aswell

[PlayStation 2]

1 minute ago, Bananasplit_00 said:

my point is that people take this as the first ever overclockable I3 just because its a K skew, you can overclock I3s from Skylake if you can find a motherboard that hasent had the update aplied yet to lock the multipliter and as i said first gen I3s you could overclock aswell

So my latter statement.

Fair enough.

[MageTank]

Don't know if it's been mentioned yet (only skimmed the thread) but this shows nothing concrete. Geekbench only reports DMI strings, NOT the actual clock rate of the hardware. You do not know what the CPU is actually clocked at, nor do you know what the memory is clocked at. Memory speed has a huge impact on the score, so the difference between the Haswell (DDR3) i5 and Kaby (DDR4) i3 are expected somewhat in that regard.

 

Take this i5 6400 result for an example: https://browser.primatelabs.com/v4/cpu/1076817 . You are looking at a 10% difference in single core with that result, and a 60% difference in multicore. My point is, do not trust DMI strings as if they represent the absolute fact. They can be quite deceptive. Kaby Lake does not bring a single 1% IPC boost (over skylake), that you can be certain of. 

[Drak3]

On ‎11‎/‎25‎/‎2016 at 5:48 AM, Sauron said:

Too bad geekbench is garbage, as clearly shown by the fact the multithreaded score for the i3 is double that of single core (which makes sense since it's a dual core) but in the case of the i5 it's less than 3 times the single core score despite the i5 having 4 cores. Also, the 4670k has significantly higher single core results compared to the 6400, but is only barely faster in the multithreaded test - what gives?

It's possible (possible, not likely) that the TDP was given some margin to accomodate for high overclocking.

Performance scaling across cores isn't linear, and performance increases aren't consistent across all tasks. That's why we use multiple benchmarks, both real world and synthetic.

 

Also, the 6400 isn't all that good. It's a 2.7GHz CPU, and being compared to a 3.4GHz CPU that isn't exactly far behind in terms of IPC. Even if Skylake has an 11% IPC boost over Broadwell (which is 3% better than Haswell at best), it doesn't make up for the clock difference.

There's also the i7-4970K, which either matches the 6700K in most benchmarks, or is right behind it (and usually within the margin of error). While generalizing an IPC boost to an entire generation has some accuracy, there can be outliers that go against that.

 

Also, Intel's TDP is their estimated heat output (in Watts), under a "real world load" (basically, a similar scenario to an Aida64 stress test). Without any indication of what the new dual core might use, WCCF has to go off of assumptions as to what it could be, and they have to be based off of existing trends.

[MageTank]

5 minutes ago, Drak3 said:

Performance scaling across cores isn't linear, and performance increases aren't consistent across all tasks. That's why we use multiple benchmarks, both real world and synthetic.

 

Also, the 6400 isn't all that good. It's a 2.7GHz CPU, and being compared to a 3.4GHz CPU that isn't exactly far behind in terms of IPC. Even if Skylake has an 11% IPC boost over Broadwell (which is 3% better than Haswell at best), it doesn't make up for the clock difference.

There's also the i7-4970K, which either matches the 6700K in most benchmarks, or is right behind it (and usually within the margin of error). While generalizing an IPC boost to an entire generation has some accuracy, there can be outliers that go against that.

 

Also, Intel's TDP is their estimated heat output (in Watts), under a "real world load" (basically, a similar scenario to an Aida64 stress test). Without any indication of what the new dual core might use, WCCF has to go off of assumptions as to what it could be, and they have to be based off of existing trends.

What? Skylake's IPC boost over Haswell is 5% at best. The biggest boost Skylake brought to the table in terms of bench scores was DDR4. The rest of the tests showed roughly 5% over Haswell and 2-3% over Broadwell (depending on task).

 

http://www.anandtech.com/show/9483/intel-skylake-review-6700k-6600k-ddr4-ddr3-ipc-6th-generation/9

Quote

This graph shows that:

Sandy Bridge to Ivy Bridge: Average ~5.8% Up
Ivy Bridge to Haswell: Average ~11.2% Up
Haswell to Broadwell: Average ~3.3% Up
Broadwell to Skylake (DDR3): Average ~2.4% Up
Broadwell to Skylake (DDR4): Average ~2.7% Up

Oh dear. Typically with an architecture update we see a bigger increase in performance than 2.7% IPC.  Looking at matters purely from this perspective, Skylake does not come out well. These results suggest that Skylake is merely another minor upgrade in the performance metrics, and that a clock for clock result compared to Broadwell is not favorable. However, consider that very few people actually invested in Broadwell. If anything, Haswell was the last major mainstream processor generation that people actually purchased, which means that:

Haswell to Skylake (DDR3): Average ~5.7% Up.

The biggest problem with comparing Skylake to Haswell is the fact that most people seem to do so when using DDR4. Once you factor in the results when using comparable DDR3, Skylakes lead diminishes. There are plenty of CPU benchmarks that can be manipulated with faster ram. I've done so many times, even with the Geekbench benchmark discussed in this very thread. For a fair comparison to be made, one must remove as many variables as possible. 

[Drak3]

7 hours ago, MageTank said:

Skylake's IPC boost over Haswell is 5% at best

 

7 hours ago, MageTank said:

The rest of the tests showed roughly 5% over Haswell and 2-3% over Broadwell (depending on task).

Looking at the conclusion graph, which summarizes the gains saw in each test, Skylake ranged from 5% gain over Sandy Bridge, to 11% gain from Broadwell.

POV Ray pins the 6700K as being 9% better than the 4770K. In the single thread Cinebench test, Skylake saw a 9.5% improvement over the i7-5775C. FastStone saw between 4 to 6% increase from Broadwell to Skylake.

Anandtech's conclusions are sporadic, and show no trend. Their method is faulty for actual testing on IPC (two of their tests also rely on iGPU), and they should go back to the drawing board, or give up on finding IPC themselves.

 

I also never claimed that Skylake did have an 11% gain, I went with the trend the previous two 'tock' generations set up: roughly 11-15% increases, as per:

 

https://us.hardware.info/reviews/6215/2/intel-five-generation-ipc-test-broadwell-haswell-ivy-bridge-sandy-bridge-and-nehalem-results

[Biggerboot]

I may just get it just to say "I have a 5ghz i3 CPU".  That's if these things are hitting 5ghz with ease.  Maybe make it my last sub-i7 processor before finally building a Skylake-X workstation or something.  Taking benchmarks out of the equation, that means my single thread performance should be about 30% better than it is now with a i3 6100 on stock.  CS:GO players, you will love it.

[MageTank]

4 hours ago, Drak3 said:

 

Looking at the conclusion graph, which summarizes the gains saw in each test, Skylake ranged from 5% gain over Sandy Bridge, to 11% gain from Broadwell.

POV Ray pins the 6700K as being 9% better than the 4770K. In the single thread Cinebench test, Skylake saw a 9.5% improvement over the i7-5775C. FastStone saw between 4 to 6% increase from Broadwell to Skylake.

Anandtech's conclusions are sporadic, and show no trend. Their method is faulty for actual testing on IPC (two of their tests also rely on iGPU), and they should go back to the drawing board, or give up on finding IPC themselves.

 

I also never claimed that Skylake did have an 11% gain, I went with the trend the previous two 'tock' generations set up: roughly 11-15% increases, as per:

 

https://us.hardware.info/reviews/6215/2/intel-five-generation-ipc-test-broadwell-haswell-ivy-bridge-sandy-bridge-and-nehalem-results

Well, I happen to own both a 4790k and 6700k, and I've tested this myself (Though, I cannot use myself as a valid source of information as I am not considered a third party in this discussion). After testing each CPU using the exact same tests, and even used comparable memory speeds (and memory speeds that completely dwarfed the fastest DDR3, both in raw bandwidth and latency) the results showed roughly a 5% difference on average in favor of Skylake, with the biggest outlier being heavy compression, where DDR4 3600 C14 ran away from Haswell's 2400 C11.

 

Even using the specific workload (Blender) from our "Resident Intel Expert" Patrick, I came to a very similar conclusion when comparing Skylake to Haswell. Looking at other third party sources that tested DDR3 on Skylake, the results also showed roughly 5% on average: http://www.hardware.fr/articles/940-6/cpu-sandy-bridge-vs-ivy-bridge-vs-haswell-vs-skylake-4-g.html

 

If you are to believe the LTT video review of the 6700k, it actually failed to show a 5% improvement in most of their tests (other than Attila, which went from 12fps to 19fps, roughly a 60% gain). 

If you have any specific tests you would like me to run to test IPC with my systems, i'll gladly do so. Sadly, I no longer have a DDR3 Skylake board to test with, as I sold my ASUS board a few months back. That being said, I should be able to underclock my DDR4 to a similar speed/latency setup.

[RagnarokDel]

On 28/11/2016 at 9:23 AM, Dan Castellaneta said:

I'm fairly sure the 4670K just hit 2 years.

I7 920 Q4 2008

I5 4670k Q2 2013

Kaby Lake Q1 2017