First Intel Kaby Lake benchmarks leaked!

AthlonXP1800 · 1 May 2016 at 16:37

Here are first Kaby Lake i7 7500U Sandra benchmarks I found from HWBattle.

http://www.hwbattle.com/bbs/board.php?bo_table=news&wr_id=20033

i7 7500U Kaby Lake benchmarks:

http://ranker.sisoftware.net/show_d...ab96a781e9d4e1c7bf82b096f396ab9bbdcef3cb&l=en

i7 6500U Sky Lake benchmarks:

http://ranker.sisoftware.net/show_d...aa97a680e8d5e0c6be83b294f194a999bfccf1c0&l=en

Look like Kaby Lake have a nice boost in performance compared to Sky Lake.

Possible i7 7700K Kaby Lake benchmarks?:

http://ranker.sisoftware.net/show_s...e1dceccaa29faf89f1ccfcdabfdae7d7f182bf87&l=en

DragonQ · 1 May 2016 at 22:31

6% per clock in CPU arithmetic, pretty surprising jump. More than the change from Broadwell to Sky Lake...makes you think Sky Lake was rushed out the door and Kaby Lake is what they wanted to create from the start?

Retron · 2 May 2016 at 05:16

Hmm, from those links:

6500U @2.5GHz maths score: 46.49 GOps
7500U @2.9GHz maths score: 55.39 GOps

2.5*1.16=2.9, therefore the 7500U is clocked 16% faster. If you increase the 6500U result by 16% you get 53.92 GOPs.

55.39 is 2.7% more than 53.92, therefore the actual increase in performance is more like 3%.

DragonQ · 2 May 2016 at 11:46

You are correct, my late night maths was poo.

eddyr · 2 May 2016 at 12:09

Seems like the kind of increase that could easily be due upping the ram speed at stock spec, say kabylake running 2666mhz ram as standard stock configuration instead of skylake's 2133.

AllBodies · 2 May 2016 at 12:43

Well, really looking like the last hope is Zen being competent, to kick Intel up the arse.

pastymuncher · 2 May 2016 at 13:12

What did we expect though. Intel has given us very little performance increase over the past few generations and have been improving igpu performance and efficiency instead. If Zen is any good you can bet the days of AMD cpu's being much cheaper than Intel's offerings are over and I expect them to be similarly priced to Intel's cpu's. Just look at the launch prices for the Nano/Fury gpu range. I think the days of AMD being the budget brand are likley to be over.

Telecaster · 2 May 2016 at 13:41

I'm not really sure what sort of CPU performance people are expecting, what do most desktop users do that a current CPU won't demolish? How CPU limited are current games?

I get the impression there isn't a massive demand for huge leaps in CPU performance which is why we aren't getting it. Law of diminishing returns and all that.

Cooper · 2 May 2016 at 15:00

what do most desktop users do that a current CPU won't demolish? How CPU limited are current games?

Mame, PCSX2, Dolphin and just about every other CPU limited emulator out there says hello!!!

McGraw · 2 May 2016 at 15:33

Telecaster said:
I'm not really sure what sort of CPU performance people are expecting, what do most desktop users do that a current CPU won't demolish? How CPU limited are current games?

I get the impression there isn't a massive demand for huge leaps in CPU performance which is why we aren't getting it. Law of diminishing returns and all that.

FSX, Prepar3D, Arma 3...there are loads of games/sims that could use more horsepower.

Sem · 2 May 2016 at 16:11

McGraw said:
FSX, Prepar3D, Arma 3...there are loads of games/sims that could use more horsepower.

more horsepower or better optimised?

AthlonXP1800 · 2 May 2016 at 17:59

Retron said:
Hmm, from those links:

6500U @2.5GHz maths score: 46.49 GOps
7500U @2.9GHz maths score: 55.39 GOps

2.5*1.16=2.9, therefore the 7500U is clocked 16% faster. If you increase the 6500U result by 16% you get 53.92 GOPs.

55.39 is 2.7% more than 53.92, therefore the actual increase in performance is more like 3%.

Interesting maths.

I noticed there are Kaby Lake m7-7Y75 benchmark:

http://ranker.sisoftware.net/show_d...94a998bed6ebdef880bd8fa9cca994a482f1ccf4&l=en

Sky Lake m7-6Y75 benchmark:

http://ranker.sisoftware.net/show_d...95a899bfd7eadff981bc8dabceab96a680f3ceff&l=en

m7-6Y75 @3.02GHz maths score: 22.57 GOPS
m7-7Y75 @1.61GHz maths score: 34.59 GOPS

Kaby Lake Core M is really very interesting one, it pushed 53% more GOPS at just 1.61GHz nearly half speed. :eek:

If m7-7Y75 increased clock speed by 87% to 3.02GHz you get 64.68 GOPS. :eek:

The actual increase in Kaby Lake Core M performance is 86.99%! :eek:

Look like Kaby Lake is Zen killer.

Poor AMD. :eek:

joeyjojo · 2 May 2016 at 18:20

AthlonXP1800 said:
m7-6Y75 @3.02GHz maths score: 22.57 GOPS
m7-7Y75 @1.61GHz maths score: 34.59 GOPS

Kaby Lake Core M is really very interesting one, it pushed 53% more GOPS at just 1.61GHz nearly half speed. If m7-7Y75 increased clock speed by 87% to 3.02GHz you get 64.68 GOPS.

There must be an explanation for that. Either the core count is wrong (4?), the speed is wrong, the score is wrong, or it's got AVX-512 (to explain a doubling of peak performance vs. AVX2). What else could do it?

AthlonXP1800 · 2 May 2016 at 18:48

joeyjojo said:
There must be an explanation for that. Either the core count is wrong (4?), the speed is wrong, the score is wrong, or it's got AVX-512 (to explain a doubling of peak performance vs. AVX2). What else could do it?

I doubled checked it, all looked correct both used 4T. It don't used AVX-512 as I found out it only use for Xeon.

https://en.wikipedia.org/wiki/Kaby_Lake

CAT-THE-FIFTH · 2 May 2016 at 20:01

joeyjojo said:
There must be an explanation for that. Either the core count is wrong (4?), the speed is wrong, the score is wrong, or it's got AVX-512 (to explain a doubling of peak performance vs. AVX2). What else could do it?

Probably reporting the actual clockspeeds incorrectly??

The m7-6Y75 has a base clockspeed of 1.2GHZ with a max Turbo of 3.1GHZ:
http://ark.intel.com/products/88199/Intel-Core-m7-6Y75-Processor-4M-Cache-up-to-3_10-GHz

The m7-7Y75 probably has a 1.6GHZ base clockspeed.

Caracus2k · 2 May 2016 at 21:40

Probably best to wait for some more solid benches... Still anyone expecting miracles out of a chip based on the same (if slightly tweaked) design on the same 14nm process may be disappointed. Some models may get a reasonable uplift if they have the large l4 cache that we first saw on some broadwell processors but otherwise expect more of largely the same....

AiiR · 2 May 2016 at 22:38

Pretty small increases, is this Intel squeezing the very last bit of IPC performance out of silicon?

Telecaster · 2 May 2016 at 23:09

McGraw said:
FSX, Prepar3D, Arma 3...there are loads of games/sims that could use more horsepower.

But you have to ask yourself, what is the % of the market that plays those games that also has the disposable income to keep buying the latest CPU with a mobo change etc.?

It can't be a profitable business surely, just constantly designing a new line of CPUs for such a small market?

Look how many "enthusiast gamers" on sites like this that are still on 1366/1155 systems.

rossiboy001 · 3 May 2016 at 07:27

Telecaster said:
But you have to ask yourself, what is the % of the market that plays those games that also has the disposable income to keep buying the latest CPU with a mobo change etc.?

It can't be a profitable business surely, just constantly designing a new line of CPUs for such a small market?

Look how many "enthusiast gamers" on sites like this that are still on 1366/1155 systems.

Im still on 1155, Intel have not given anything worth upgrading to. Not because there is not demand on my part. However i seem to have upgraded by GPU in this time every time a new top card is released that offers improvement. I would apply the same to CPU"s given a reasonable upgrade choice.

=XDC=FluphyBunny · 3 May 2016 at 11:35

I was just expecting the difference between 4770k > 4790k given its basically filling the gap till Cannonlake.

Expecting Zen to be catching up to maybe IvyBridge clock for clock.