Alleged Kaby Lake CPU shows its face in SiSoft Sandra database

Back in March, we reported on the demise of Intel’s long-held “tick-tock” product development strategy. For years, Intel has predictably released a line of CPUs with a new manufacturing process one year, and a line on the same process with a new architecture the next. This cadence is changing with the current 14-nm process node. Under the tick-tock schedule, we would have seen products made on a new process node—10-nm—this year. Instead, we are looking at an optimization of the 14-nm process, called "Kaby Lake." Now we have purported benchmarks of a Kaby Lake processor, courtesy of an anonymous posting in the SiSoftware Sandra database.

The leaked benchmark shows a four-core, eight-thread processor with 3.6GHz base and 4.2GHz Turbo clock speeds, along with 8MB of L3 cache. That base clock puts it about in the middle of the i7-6700 (3.4GHz) and i7-6700K (4.0GHz), and the 4.2GHz Turbo speed aligns with the i7-6700K. Even if this benchmark does actually represent a Kaby Lake processor, it is likely an engineering sample, so these numbers should be taken with a grain of salt. That said, these frequencies and cache numbers do seem to align with either a slightly more conservatively-clocked "Core i7-7700K," or a "Core i7-7700" with a 200MHz bump to both its base and Turbo frequencies.

The GPU specs in SiSoft’s database show a 24-execution-unit, 1.15GHz graphics processor on this mystery chip. Those specs look a lot like the HD 530 GPU in the i7-6700 and i7-6700K, which supports the argument that this is an i7-7700 chip. Sadly, Sandra doesn’t report anything about TDP on the chip, so we are missing a crucial part of the puzzle. Since Kaby Lake is an extra optimization on the 14-nm process node, a drop in TDP could make this a more attractive package than a mere 200MHz clock speed bump. On the other hand, if this turns out to be a K-series part, we could see a return to past K-series TDPs, at the cost of a few hundred MHz of base clock speed.

Comments closed
    • DavidC1
    • 4 years ago

    It looks like their so-called “Process-Architecture-Optimization” is not at this phase or just a story to cover-up the real one. That being the “Optimization” part really sucks and is really a 1 year delay with a different name.

    5% CPU and 0% on iGPU would be absolute disappointing. Polaris/Pascal is going to hit you with 2x perf/watt gains. Hello? They are already barely competitive with perf/watt on their GPUs, despite being an integrated one. Iris Pro 580 is going to look like a HD 530 once Polaris/Pascal hits.

    I’ve seen the benchmark with Iris Pro 6200 that a 750Ti uses slightly less power while performing 50-70% better.

      • Ninjitsu
      • 4 years ago

      [quote<]I've seen the benchmark with Iris Pro 6200 that a 750Ti uses slightly less power while performing 50-70% better.[/quote<] Are you comparing system power, GPU to GPU power, or what?

      • Anonymous Coward
      • 4 years ago

      Intel ought to invest in GPU’s I think. They will be in some danger if AMD pairs a reasonable CPU core with superior graphics.

        • faramir
        • 4 years ago

        Raven Ridge please …

      • w76
      • 4 years ago

      Didn’t everyone just assume this would be a Sky Lake rebadge to tread water while Cannonlake gets delayed? I know I sure did. I just hoped they’d throw Iris Pro with that huge cache on some desktop parts as a form of apology and differentiation.

        • Ninjitsu
        • 4 years ago

        Something like this? [url<]http://www.anandtech.com/show/10281/intel-adds-crystal-well-skylake-processors-65w-edram[/url<]

          • w76
          • 4 years ago

          Yes, but those are embedded, I meant available in consumer packaging so we can pair it with the other components we please.

        • faramir
        • 4 years ago

        I expected “Devil’s Canyon” style of release, higher clocks, better binning. It appears I was correct in my expectations.

    • hasseb64
    • 4 years ago

    Intel is cracking, why even give this update a name? They are loosing my respect.

    • rudimentary_lathe
    • 4 years ago

    If Zen does bring the rumored 8 full cores to their mainstream consumer platform, I’ll be interested to see how Intel responds. So far all signs point to Kaby Lake having a maximum of four physical cores. Maybe Cannonlake will break with tradition?

    I don’t see why most people who already have a newish Intel CPU would be motivated to upgrade to an incrementally better CPU. The 2011 socket platform has some nice options, but they’re extremely expensive for your average consumer.

      • blastdoor
      • 4 years ago

      I would think Intel has a ton of options. One simple option would be to wait to see what Zen’s performance/watt really is and then go rebrand an older 8 core model and sell it at the same price or slightly lower, carefully marketing it only to the audience of people who will know that Zen even exists. For example, sell 8 core Haswell-E to people under some goofy brand like Pefromium and only advertise on websites like TR.

      • Archer
      • 4 years ago

      Actually, Intel has a fat core CPU in the waiting in the wings in case this does happen. It’s called Ricki Lake, and has 8 cores.

        • rudimentary_lathe
        • 4 years ago

        I *almost* Googled that. Nicely done.

        • faramir
        • 4 years ago

        I thought Ricki Lake was a big.LITTLE kind of design, two *extremely* large cores with two moderately sized companion cores?

        The very large cores can process a single thread at an amazing rate though so this is what Intel is gambling on I suppose, “reverse-hyperthreading” …

      • Unknown-Error
      • 4 years ago

      Zen will require 6 to 8 core with SMT to keep up with Sky-Lake 4 cores with SMT when it come to FPU heavy workloads. So Intel won’t need to respond. Difference is, Zen will be good at legacy work loads and even more importantly power efficiency will be much much better than the construction cores (excavator).

        • Anonymous Coward
        • 4 years ago

        If AMD launches a competent Zen and Intel can’t progress much, its just a matter of time before they close the gap with Intel. They would just need calmly refine their design, one generation following another.

      • Anonymous Coward
      • 4 years ago

      Intel’s challenge is to make something that [i<]isn't[/i<] incremental. Mature markets of all kinds are full of incremental product upgrades. They can stick on more cores or cache or whatever, it would cost more to make and perform worse, for most relevant performance metrics. Its remarkably honest of them to have stayed with 4 cores so long, probably they'll stop that once Zen arrives and they need to sell product based on whatever impressive-looking statistic it takes. But on the other hand, such a strategy could simply erode the same profit margin which Zen will (hopefully) be applying pressure to. We could enter an interesting new phase of desktop CPU marketing.

      • Ifalna
      • 4 years ago

      And what benefit would having more than 4 cores have in regards to consumer software?

        • Anonymous Coward
        • 4 years ago

        Greater sales at the expense of a competitor with fewer cores.

          • Ifalna
          • 4 years ago

          I’m talking about technical benefits.
          Because for most consumer software there is none.

            • w76
            • 4 years ago

            What’s the technical benefit for most drivers, realistically, for 6 cylinders or 8 cylinders vs 8? Because there is none, all vehicles go from A to B. Welcome to the real world.

            Asides from that, as much as I just don’t get it culturally, YouTube and Twitch become a bigger thing by the day, and while the sort of content I’m referring to is very different at each site, they have one thing in common: video. And video capture, editing and encoding/transcoding eats CPU cores up like candy. Photography as a hobby doesn’t seem to be a quick transient fad, either, and that can use additional cores as well.

            Also, Anandtech did a nice article showing how smartphone apps seem to make great use of 4 – 8 cores in some cases. Developers have enabled that there because most (Android) phones do have lots of cores. Your average Wintel cheap laptop is still probably dual core (even if it’s an “i7”). Chicken and egg situation.

            • Anonymous Coward
            • 4 years ago

            I have to say, when 95% of vehicles have 4-bangers (most of them diesel), a technical person comes to miss the sound of a V6 (diesel) or V8 (gas or diesel).

            • Ifalna
            • 4 years ago

            “Chicken and egg” describes it pretty well.

            If there would be mainstream software (besides video rendering) that actually benefits from more cores, people would want more (and pay a premium). But so far it seems to me that the added technical difficulties associated with more cores and coding for more cores aren’t worth it on a desktop.

            Why? Probably because a desktop has much more single threaded punch than your average smartphone.

            • w76
            • 4 years ago

            You’re half way there; it’s not worth the added technical difficulties (which translates to “added cost” to everyone in business) when your competition isn’t mounting a challenge. Those that would pay a premium for more cores will either pay up for Xeons, -E versions, or simply pay top dollar for the highest end consumer part you do offer, because there’s no alternative. No business impetus at all to increase product costs, beyond “just enough” to keep an upgrade cycle somewhat alive.

            I’ve said it dozens of times, some folks look at the tech industry through too much of an engineers eyes. It’s a business, too, asides from just an engineering issue.

    • DPete27
    • 4 years ago

    Do you think Intel/AMD/Nvidia would actually conduct a criminal investigation to find and locate someone in possession of a leaked engineering sample and interrogate them to uncover their supplier? (if the supplier and owner weren’t the same person)

      • chuckula
      • 4 years ago

      Probably not unless there was direct evidence that it was actual espionage to provide secret information to a competitor instead of just leaking vague benchmarks on a website.

      This is Intel we are talking about: A multi-billion dollar company that jealously guards its technology and is a ruthless competitor. That is to say, they are practically Mr. Rogers compared to some real control freaks like Apple.

      • Pitabred
      • 4 years ago

      They probably would. But it’s likely not leaked, they send engineering samples to manufacturing partners all the time so they can have motherboards and such ready when the chip is announced. So this is likely someone at a motherboard company running the engineering sample chip on an engineering sample motherboard to see if it works like they expect it to, and all the benchmark programs just tend to call home like this.

      • cygnus1
      • 4 years ago

      And you don’t think some engineer is stupid enough to accidentally let Sandra upload the data while benchmarking an engineering sample? That’s the default option I believe.

      edit: In other words, never ascribe to malice what can be explained by stupidity or ignorance.

    • chuckula
    • 4 years ago

    OK Robert, now I know you are trying really hard, but if you are going to bore us to death with Rebrand Lake stories, can we at least get a purportedly thrown in for good measure?

      • hexr
      • 4 years ago

      What, is “purported” in the first paragraph not close enough for you?

        • chuckula
        • 4 years ago

        Yeah, like I actually read anything beyond the headline.

          • cygnus1
          • 4 years ago

          you failed at the internets today, good sir

            • chuckula
            • 4 years ago

            Given the level of the “internets” wouldn’t failing entail carefully reading the article in full?

            • Anonymous Coward
            • 4 years ago

            That would have saved you time compared to having this discussion.

      • Meadows
      • 4 years ago

      Didn’t I just ask you quite recently about trying too hard? (Although in this case I’m not quite sure what it is that you were trying.)

Pin It on Pinterest

Share This