Forgot your password?
typodupeerror
Intel Hardware

Intel Bay Trail Brings New Architecture and Performance To Atom 68

Posted by Unknown Lamer
from the small-cpus-are-so-cute dept.
Vigile writes "Today at the Intel Developer Forum in San Francisco, the company officially released the Atom Z3000 series of SoCs (Bay Trail) based on the Silvermont architecture. Unlike previous Atom designs, the Z3000 and Silvermont is a completely re-architected product from the ground up and is no longer based on legacy processors. Changes include a move to an out-of-order x86 architecture with drastically improved single threaded performance but the removal of Intel's HyperThreading technology. Dual-core modules with 1MB of shared cache can be paired up to create a quad-core SoC that also includes upgraded graphics design. Intel is no longer depending on PowerVR for a GPU and has integrated a 4 EU (execution unit) Intel HD Graphics design that is very similar to the one used in Ivy Bridge. As a result, as tested at PC Perspective in both Windows 8.1 and Android 4.2.2, the Bay Trail part is as much as 4x faster in single threaded tasks and 3.5x faster in gaming and graphics. Power consumption remains nearly the same as it did with Clover Trail (Atom Z2760) but with improved power gating and support for Connected Standby, Intel's new Atom looks and feels completely different than any before it." MojoKid notes that Intel also announced an "open" SoC architecture (where open involves you giving Intel tons of money).
This discussion has been archived. No new comments can be posted.

Intel Bay Trail Brings New Architecture and Performance To Atom

Comments Filter:
  • by Anonymous Coward

    This chip family is Intel's first real answer to arm SoCs. I look forward to seeing devices that feature it. Supposedly it will enable sub-100 dollar windows 8 tablets. (Well, excluding the win8 license probably. MS- You have a problem when your OS costs 2x more than the hardware itself.) - I'd love to pick up a 99 dollar tablet and see what I can do with a linux distro. I'd also love to see some ultra-small low cost SoC based boards. (Atomberry pi anyone?)

    The x86 android port is supposed to be pretty damn

    • by hairyfeet (841228) <bassbeast1968&gmail,com> on Wednesday September 11, 2013 @01:51PM (#44821171) Journal

      But how does it compare to the AMD Bobcats and Jaguars? To me THAT is the question. I have built many a system around the Bobcat and its pretty damned nice, more than powerful enough to replace the aging P4s in office environments while taking less power under load for the entire system than your average P4 idles at, great for netbooks, even low power servers, and the Jaguars are powerful enough that they are the chip of the Xbox N and PS4 while supposedly using little more than the Bobcats.

      Every system with an Atom chip that has came into the shop...sigh...the words "mediocre" and "weak" instantly come to mind. AMD was able to boost their Bobcat by pairing it with a powerful GCN GPU and splitting the load, but lets face it Intel has never been great at GPU design so you ended up with a weak sauce CPU tied to a weak sauce GPU and that equals just painful to use.

      So while I'll be happy to give 'em a spin if any ever walk through my door unless someone can show me some real world tests, NOT benchmarks because as we have seen the "Intel Cripple Compiler" makes any benchmark that doesn't announce which compiler they use suspect at best, I'll have to tell my customers to avoid Atoms, if they want Intel the i3s are ULV and at least have good performance.

      • Re: (Score:3, Informative)

        by Anonymous Coward

        According to Intel this is a complete redesign that has more in common with haswell than it does with older atoms. It's interesting to note that all Intel chips, pre haswell, have had bolt-on powersaving tech. Haswell and later chips are designed from the ground up with power management in mind.

        The family has chips aimed at true mobile applications, like tablets and smartphones. It also has chips aimed at low end laptops and deskops, like the previous atom. I don't know how well they will compete with bobca

      • Now that Intel is focusing on beating ARM, they have less energy to focus on AMD, so now is AMD's chance to move back in.
        • by 0123456 (636235)

          Isn't Intel's R&D budget greater than AMD's entire revenue?

          I think they can afford to keep pushing high-end CPUs as well as low-end.

          • Yes, but they don't have infinite resources. It can determine questions like, once they build a new fab for 14nm chips, should it be used to manufacture mobile chips or server chips before another comes online?
        • by hairyfeet (841228)

          The sad part is the way forward for AMD is obvious but nobody there is listening. How did Intel beat AMD after years of having to bribe the OEMs to use the P4? By ditching netburst, going back to the proven P3 technology and focusing on ramping up performance of it instead of the P4.

          How can AMD beat Intel again? By dropping the turkey that is the Bulldozer "half core" design and going back to a solid proven performer...the K8, which powered the Athlon64 and 64x2 to such great heights. Use THAT as the basis

      • by timeOday (582209)

        unless someone can show me some real world tests, NOT benchmarks because as we have seen the "Intel Cripple Compiler" makes any benchmark that doesn't announce which compiler they use suspect at best

        RTA. The 4x improvement isn't an Intel claim, they are what PC Perspective observed "without intervention from Intel and without restrictions on allowed benchmarks." They ran SiSoft Sandra, CineBench, TrueCrypt, 3DMark, and a few more under Windows 8.

        This looks to me to be the part that would enable the Mi

        • by hairyfeet (841228)

          How about YOU read what I posted, where I clearly said you can't trust benchmarks unless the companies that MADE the benchmark state clear which compiler they use. if you want to know why simply Google "Intel Cripple Compiler" to see how simply changing the CPUID of a Centaur CPU (the only CPU that allows you to change CPUID) from "Centaur Hauls" to "Genuine Intel" cause the benchmarks to magically improve by over 30% simply by changing CPUID and nothing else, amazing huh?

          Its actually VERY simple, any and A

          • by Rockoon (1252108)
            Its not just the benchmark code either. Benchmarks often link to pre-compiled math libraries that themselves were compiled with ICC, so compiling the benchmark with Visual Studio or GCC doesnt help the situation any if you continue to link to those libraries.

            In addition, you need to trust the benchmarkers themselves, and most of these tech sites that so often do benchmarks have conflicts of interest (advertising money, free review hardware, etc..) With this in mind I trust PassMarks online benchmark data
  • by h4rr4r (612664) on Wednesday September 11, 2013 @12:55PM (#44820595)

    Hopefully, this will end any reliance on PowerVR by Intel. I dream that this is the beginning of the end for those bastards.

    • by PhrostyMcByte (589271) <phrosty@gmail.com> on Wednesday September 11, 2013 @01:07PM (#44820709) Homepage
      PowerVR has some of the most pathetic support for x86 and Windows I've ever seen, and it hasn't got any better. With my fanless Shuttle PC using an Atom N2800, I have a choice of either 32-bit Windows and glitchy graphics, 64-bit Windows and VGA output, or Linux with VGA output. It's pretty obvious why, of course... PowerVR's x86 market is so infinitesimally small compared to their ARM market, they probably hired some old printer driver developer to be the sole guy working on it stashed in a closet somewhere. It is really surprising that Intel ever decided to use them, without some sort of support contract built in.
      • by tlhIngan (30335)

        PowerVR has some of the most pathetic support for x86 and Windows I've ever seen, and it hasn't got any better. With my fanless Shuttle PC using an Atom N2800, I have a choice of either 32-bit Windows and glitchy graphics, 64-bit Windows and VGA output, or Linux with VGA output. It's pretty obvious why, of course... PowerVR's x86 market is so infinitesimally small compared to their ARM market, they probably hired some old printer driver developer to be the sole guy working on it stashed in a closet somewher

        • by h4rr4r (612664)

          At least GMA950 works. I would rather have that than anything PowerVR.

        • by turgid (580780)

          They're making a nice living doing GPU cores for ARMs,

          But for how long? The ARM folks are designing their own GPUs now.

          • by Narishma (822073)

            And the PowerVR folks just bought their own CPU (MIPS).

            • by h4rr4r (612664)

              Which competes directly with ARM and while it has some market is not exactly as big as it once was.

            • by turgid (580780)

              Yes, at great expense and with a shrinking market share. ARM has the smartphone market cornered. So the PowerVR are losing GPU market share and they've got no chance of competing against ARM in the embedded CPU market in phones...

              Maybe they have other plans?

              Then one day OpenRISC will come along and ARM will start to lose out. It's already started in some very small, currently not very significant niches.

        • Why are you running 64-bit Windows when all the Atom chipsets only support 2GB of RAM?

          This is not true. Cedar Trail (the current gen, out for about 2 years now) supports 4GB max, which is what I have in mine.

    • by h4rr4r (612664)

      To the idiot with modpoints, here is another comment you can mod down. Have fun, my karma can take it.

    • Just to add to what you're saying here, allow me to illustrate.

      I was doing some embedded FreeBSD work on an Atom board, and it had the PowerVR GPU. The VGA text console was all scrambled. This was in 2011, it's not like VGA is all that new.

      It turns out that the VGA spec itself says that data can be written to the frame buffer in half-byte words. Every VGA implementation that FreeBSD had ever encountered worked with writing VGA text console data to the frame buffer a byte at a time. Until 20[8-10] when P

  • by CajunArson (465943) on Wednesday September 11, 2013 @01:01PM (#44820655) Journal

    It turns out that Bay Trail has some very solid performance numbers and that the power consumption is very good too, but frankly, you can get similar results from high-end ARM SoCs.

    What you can't get, however, are 100% GLPd GPU drivers that are already in the mainline Linux kernel. THANK YOU INTEL and I hope this is a wakeup call to the ARM vendors that the days of crappy, unsupported binary blobs are hopefully coming to an end.

    • by h4rr4r (612664)

      Hopefully we finally see some x86 android devices.

      BTW your sig is stupid. If a smoker tells you not to smoke as it causes cancer do you ignore him too?

    • It turns out that Bay Trail has some very solid performance numbers and that the power consumption is very good too, but frankly, you can get similar results from high-end ARM SoCs.

      But will that still be true in 2014 when Intel drops to 14nm?

      • Who knows? Even last year the results that Bay Trail is showing would have been considered next to impossible for an x86 part. Intel is still an underdog in mobile devices like tablets/smartphones, but it is good to see some competition.

      • by h4rr4r (612664)

        Does it matter?
        If I can get even comparable power consumption sign me up. I am sick of this binary driver ARM bullshit.

        • Well said.

          I got one of the Samsung Android devices with an Intel chip on my desk the other day, and I was running native code, and didn't even notice it wasn't ARM until some obscure library wasn't opening correctly, and I tried to run objdump on it. So Intel seems to have the emulation capability, they should have no problem moving right in to the place where ARM is.

  • Given that Intel wants so badly to push into mobile and their biggest weakness has been the relatively high power consumption of previous Atoms compared to the incumbent Arm offerings, it seems odd that their big re-design improves performance but not power. Have they given up on the phone market?

    Meanwhile, Haswell improved power consumption greatly but gives meager improvements in performance over the previous generation of Core.

    So, instead of broadening to cover new markets, it looks like Intel's line i

  • I hate metro and the latest atoms force you to use Windows 8. The SOI is proprietary and linux and Windows 7 support are not existant.

    If Intel has drivers for this one I will withdrawl my compliant

  • Bay Trail and Windows 8.1, trailblazing a new generation of failures.

    • by 0123456 (636235)

      Not really. Bay Trail can also run Android and should run Linux (unlike the last generation of Atoms with the PowerVR GPU), so it gives Intel a decently performing mobile chip that isn't reliant on Windows.

  • The Z3770 is the fastest Silvermont at 2.4 Ghz, I think they should be really comparing with the fastest Jaguar which is A6 5200 (at 2.0 Ghz), not the A4 5000 (1.5 GHz).

    • by 0123456 (636235)

      The Z3770 is the fastest Silvermont at 2.4 Ghz, I think they should be really comparing with the fastest Jaguar which is A6 5200 (at 2.0 Ghz), not the A4 5000 (1.5 GHz).

      So they should be comparing a 4W CPU to a 25W CPU?

      • by edxwelch (600979)

        They already compare to the 4.5W Temash if you want something with equilivilant power consumption, but if you want a raw power comparison why choose the second fastest?

  • I misread the title as "Intel Bay Troll Brings New Architecture and Performance To Atom", which didn't make much sense.

  • by DarthVain (724186) on Thursday September 12, 2013 @11:50AM (#44831001)

    OK. I fail to the the point (market) of the Atom processor. Ditto for the AMD counterpart.

    The reason I say this is because of A) miniaturization and power efficiency gains in traditional processors, and B) ARM and Motorola.

    I just bought an i5 Haswell on an itx format. I could have got an i7 (or an i3 for that matter). They make laptops with all of those. Power usage is way down. If looking for "cheapness" past an i3, they still over Celerons more less. AMD likewise has some cheap lower powered chips.

    ARM and Motorola (A# and Snapdragon basically) own the phone/tablet market. Nothing Intel or AMD do to their lines is going to change that.

    For Atom and it's AMD counterpart fall somewhere in between the very cheap low end chip, and the ARM/Motorola chips. What are you making with these chips? Shitty netbook laptops? Sorry the processor is only part of the price of these things. Not to mention the death of the netbook due to the popularity of Tablets.

    Anyway I just do not see the point of this processor segment at all.

Our business in life is not to succeed but to continue to fail in high spirits. -- Robert Louis Stevenson

Working...