Comments Locked

53 Comments

Back to Article

  • jaydee - Thursday, January 29, 2015 - link

    I was hoping for a bit more here from Intel. Looking at the encoding, compressing, encrypting benchmarks, almost nothing was gained performance-wise, or in terms of power consumption.
  • Flunk - Thursday, January 29, 2015 - link

    Broadwell is an incremental update so it's not a big surprise.
  • gonchuki - Thursday, January 29, 2015 - link

    Doesn't even look incremental, more like a step revision of Haswell (just a node shrink and nothing else). It's the same processor Intel launched in 2013, but on a 14nm node.
  • Refuge - Thursday, January 29, 2015 - link

    Tick Tock Tick Tock.

    Although I've noticed their "Tick Tock" Has been slightly off cadence since their "Mobile First" strategy.
  • quophog - Friday, January 30, 2015 - link

    No surprise really. No competition at all at the performance end. All their effort is going in to the ultra low power & cost end where ARM based architectures currently dominate. Very sad for everyone (except intel share holders) that AMD couldn't keep up with Intel R&D and fab spending.

    It's a long way off anyway, but I really hope Intel fab advantage doesn't let them eventually monopolise mobile procs too..
  • Haravikk - Friday, January 30, 2015 - link

    Is that really such a bad thing? Making a smaller, much more power efficient processor without sacrificing any performance is still a pretty big accomplishment. While it might not make existing machines much more exiting, the really interesting part is what new things it makes possible.

    It's for this reason I hate this new Brix machine; it's the same horrific chassis as before, the with the same anaemic cooling. It should be possible for this to be entirely passively cooled, particularly if they want for a smaller, flatter, aluminium design.
  • eanazag - Friday, January 30, 2015 - link

    Intel has done well, but with a competitive AMD we get better prices and more features. ARM has kept Intel moving, but look at the server market. Prices run from $200+ - north of $2000 for 2P systems. And Intel has the middle filled out too. Intel is charging whatever the hell they want there.
  • eanazag - Friday, January 30, 2015 - link

    It does look incremental in some benches. In others it is on par. For this type of system I would be leaning towards the 4770R. I don't care for the gaming performance on the 5500U. It is fine at 1280x1024/1366x768. I'd rather have the option to take a step up. The 4770R can do decent in 1600x900/1600x1000. For $750+ I'd like to be doing better. Broadwell Iris Pro may be what I'm hoping for.

    Otherwise, I'm a 5500U with a mobile GPU built-in - pricing and features may see which way I lean. I'm not a HTPC user.
  • nathanddrews - Thursday, January 29, 2015 - link

    I don't think we should jump to judgment based upon multiple platforms with vastly different components. The closest competitor in this review is the 4500U, which the 5500U decimates in half the benchmarks, while leading it by at least a small margin in the rest. I'm looking forward to a more controlled, thorough review of Broadwell.

    I know that the gaming tests are set in stone and you try not to change them, but can we please drop the 1280x1024 test and replace it with 1280x720? It's 30% less taxing for these weak GPUs and is much more likely to be used by gamers - either on HDTVs, notebook displays (768p), and widescreen monitors. Nearly all 360/PS3 games are 720p, so many people are comfortable with that resolution and it scales to 1080p, 1440p, and 4K. It's just more realistic. No one games at 1280x1024 anymore unless they like pillarboxing and letterboxing at the same time. 720p and 1080p are really the only meaningful resolutions for these GPUs. You can save yourself some time, too...
  • ATC9001 - Thursday, January 29, 2015 - link

    +1 for 720p resolution. If possible atleast add this to existing HTPC benchmarking, and of course when the benchmark is revamped for it's next iteraion, drop 1280x1024 completely!
  • gonchuki - Thursday, January 29, 2015 - link

    Please define "decimates". In non-OpenCL and non-GPU bound tests, it's a 5-7% win at most, which can be easily explained by the 33% higher base clock of the CPU cores, plus the die shrink that allows for better thermals (more headroom for higher bins of turbo boost).

    All of the test results point to Broadwell having the exact same IPC as Haswell in all situations. If anything improved it can only be because of the new stepping that might have fixed some errata.
  • nathanddrews - Thursday, January 29, 2015 - link

    Agreed, the word "decimates" is a bit extreme - but I consider anything in the 10-20% range to be significantly better.
  • Refuge - Thursday, January 29, 2015 - link

    These days I agree, long gone are the days of Sandybridge... Tis a shame, they were fun.
  • Laststop311 - Friday, January 30, 2015 - link

    I'm glad some 1 else noticed this. In the benchmarks that strictly use only the cpu, broadwells haswell equivalent is barely and i mean barely any faster. Sure the gpu is a pretty decent improvement but who cares about intels integrated gpu's? Anyone that relies heavily on an integrated gpu is going to get an apu from amd. The only reason the gpu is so much better is its such a poor performing part to begin with, it's a lot easier to improve lower performing things than things that are already highly optimized like the cpu.

    This is bad news for people using desktops with discrete gpu's and were hoping broadwell would be a decent boost. In those situations the iGPU means nothing so big deal it got better. This also means broadwell-e is going to rly suck and be basically identical to haswell-e almost no reason to even bother designing broadwell-e chips since they dont even use iGPU there is no performance increase at all to talk about in those.

    The silver lining though is we get to save money another year. With intel having no pressure on them we get to save our money till there is a real performance boost. Basically anyone with an i7-920 or higher doesn't have to spend money on a pc upgrade till maybe skylake/skylake-e MAYBE, intel has put out underwhelming tocks lately as well. My x58 i7-980x system still has no cpu bottleneck. This allowed me to buy a 55" LG OLED tv as normally i was buying a new pc every 2-3 years before the core i7 series started then all the sudden performance upgrades became pathetic, my new pc fund built up and i found the oled tv for 3000 and figured why not i can easily go another couple years with the same pc. So thanks intel for making no progress i got a new oled tv.
  • BrokenCrayons - Friday, January 30, 2015 - link

    I care about iGPU benchmarks and the computer I use for gaming has an Intel HD3000 and probably will do so for at least another year or more before even thinking about an upgrade. Having dedicated graphics in my laptop seems pointless when I can just wait 5-7 years or so to play a game after it's fully patched and usually avaiable with all of it's DLC for very little cost plus runs well on something that doesn't need a higher end graphics processor. So yes, for serious gaming, iGPUs are fine if you manage expectations and play things your computer can easily handle.
  • purerice - Saturday, January 31, 2015 - link

    BrokenCrayons, agreed 100%!! I recently upgraded from Merom to Ivy Bridge myself.

    There are tons of games now selling for $5-$10 that wouldn't run on Merom when the games cost $40-$60. In addition to being patched and DLC'd, guides and walkthroughs exist to get through any of the "less awesome" parts. More money saved for real life and less frustration to interrupt gaming. Patience pays indeed.

    Seeing the ~20% boost over 4500U in Ice Storm and Cinebench Open GL was actually exciting, even if it represents performance below 90% of other Anandtech users' current levels.
  • DrMrLordX - Tuesday, February 3, 2015 - link

    Decimates means to destroy something by 10% of its whole. All things considered, I'd rather be decimated than . . . you know, devastated, or annihilated.
  • DanNeely - Thursday, January 29, 2015 - link

    While I agree that replacing 1280x1024 is past due; I disagree with picking 1280x720. Back when it was picked 1280x1024 was the most common resolution on low end monitors. Today the default low end resolution is 1366x768 (26.65% on steam); it's also the second most commonly used one (after 1080p).
  • Oxford Guy - Thursday, January 29, 2015 - link

    Agreed. It's pretty silly to "replace" a higher resolution with a lower one.
  • frozentundra123456 - Saturday, January 31, 2015 - link

    I would disagree. It is quite reasonable, because many laptops use 768p, as well as cheap TVs.
  • cobalt42 - Friday, January 30, 2015 - link

    I think you're actually kind of agreeing -- as implied by the OP, 1366x768 is closer to 1280x720 than it is to 1280x1024. (768p is 80% of 12x10, 720p is 70% of 12x10). Either way, 1280x1048 is rather too many pixels.
  • milkod2001 - Thursday, January 29, 2015 - link

    This seems more reasonable build compared to ASrock a few days back you posted review about. But Im still not sold on temperatures it reaches. 90c is way too much. Looks like even Broodwell could not help. Problem must be the case, it's just way too small.

    My suggestion to vendors would be to completely ignore Intel's NUC standard build entirely and put everything into bigger aluminium case Apple mini like.
  • milkod2001 - Thursday, January 29, 2015 - link

    in other words, clone Apple mini, make RAM + SSD upgradable and slap windows OS on that. Then we talk.
  • nwrigley - Thursday, January 29, 2015 - link

    I haven't seen anything that comes close to the Mac Mini on an engineering level. It's a real shame too, the Mac Mini has been out for 10 years and nothing has bested that design.

    I hope we start to see some more innovation. If you go a step above the Mac Mini in size/weight/performance you've probably doubled your size and weight (at least). I wish they would make parts for systems larger than a laptop, but smaller than a tower. I would gladly trade off some of the size and weight of the Mini for more performance.
  • Kalessian - Thursday, January 29, 2015 - link

    This review hit my browser like a sack of brix. Thx.
  • ericgl21 - Thursday, January 29, 2015 - link

    This i7-5500U is not impressive at all.
    Even now with 14nm, a 15W CPU (with iGPU) from Intel still doesn't cut it.
    Hopefully, Intel will release a proper quad-core CPU that would take on the i7-4770R in all respects, including a lower TDP.
  • BlueBlazer - Thursday, January 29, 2015 - link

    Getting close and often surpassing AMD mobile APU offerings that has higher TDPs. Source of comparisons http://www.anandtech.com/show/8119/amd-launches-mo... and http://www.anandtech.com/show/7106/amds-a105750m-r... Although differrent resolutions, 1280x1024 has more pixels than 1366x768. Furthermore not forgetting that this Core i7 5500u is just a 15W part!
  • Hulk - Thursday, January 29, 2015 - link

    Not that I don't appreciate this fantastic review but are we going to see a Broadwell review?

    Specifically:
    Broadwell, Haswell, Ivy, and Sandy with clocks normalized so we can see IPC improvements Work normalized so we can see a few generations of power efficiency improvements?
    And finally head-to-head for each iGPU (with games and compute benches)?
  • TheinsanegamerN - Thursday, January 29, 2015 - link

    Looks good. cant wait to pick up the hd 6000 varient. I may have missed it, but does the broadwell NUC allow for the TDP to be raised like the haswell ones did? that, combined with 2133 memory let hd 5000 stretch its legs and perform quite well. hd6000 would be even better.
  • kgh00007 - Thursday, January 29, 2015 - link

    Yeah I hope so! I have raised the TDP on my D54250WYK to 25W and also have 2133 RAM, makes a big difference in performance over 15W TDP and 1600 RAM in games!!
  • kgh00007 - Thursday, January 29, 2015 - link

    No mention of the i7-5500U turbo speeds?

    3GHz single core turbo and 2.9GHz dual core turbo, source notebookcheck.
  • voicequal - Saturday, February 7, 2015 - link

    These were provided in a previous article.
    http://anandtech.com/show/8814/intel-releases-broa...
  • Shadowmaster625 - Thursday, January 29, 2015 - link

    The only people who would pay that much money for so little performance already gave their money to apple.
  • gr8pcguy - Thursday, January 29, 2015 - link

    Doesn't it bother anyone that the box is printed with the statement "Supports 2.5" Hard Drivers"? Obviously Gigabyte needs to do a bit of QA on their package production line!
  • Refuge - Thursday, January 29, 2015 - link

    Ha! I noticed this to when I first clicked the link.

    Hopefully they fix the typo before full production for consumers. :P
  • skifiddle - Saturday, January 31, 2015 - link

    to?
  • Refuge - Thursday, January 29, 2015 - link

    The Brix, now supports 2.5" Hard "Drivers"

    Is this some kind of new, super strong drivers for our HDD's?

    And yes, this is me being sarcastic :P
  • tspacie - Thursday, January 29, 2015 - link

    The huge iGPU BIOS carve-out intrigues me. Does the iGPU not handle allocations in shared system memory (in 4K pages) ?
  • Pissedoffyouth - Thursday, January 29, 2015 - link

    Nah, neither does AMD APUs. You lose whatever you allocate.
  • rootheday3 - Friday, January 30, 2015 - link

    the BIOS carve out is not necessary. The bios reserves a little (~32mb) for some internal data structures used by the driver (graphics page tables, memory for content protection, auxilary buffer for display frame buffer compression). Otherwise, the driver gets allocations out of system memory from the OS to map into the graphics page tables.

    the DVMT pre-Allocated is a hold over from Windows XP driver model and is no longer meaningful since Vista. In fact, if the OS comes under memory pressure and starts asking the driver to stick data into that "dedicated" segment, the PCI aperture copy/swizzle process used to load the data is slower than if the driver simply used OS allocations. Likewise, on standby/hibernate the OS "pages out" all the data from dedicated segment to "normal" memory which is slow... whereas for normal system allocations on iGPU, the pages are just "there" - no extra copy required.

    The sole exception is a handful of games that are incorrectly coded to look for "Dedicated" graphics memory from OS API call and then make bogus decisions based on that (e.g. refuse to run=>PES, restrict available game resolution/settings =>SW:TOR, older Total War games, render incorrectly because they think they don't have room to load textures =>GTA IV). For this small set of games, having the BIOS option to preallocate memory (which the driver won't actually use) is a workaround to fool them into running correctly.
  • toshz - Friday, January 30, 2015 - link

    Does anyone know what's the difference between the AC7260 and AC7265 wireless adapters?
    I was looking into purchasing the AC7260 to update my ultrabook (currently using AC3160) and then I saw the AC7265. Couldn't find any difference between them on Intel's site.

    Thanks.
  • kevith - Friday, January 30, 2015 - link

    Nice review, but why have you stopped opening the cases? I would be interested in one of these, but changing/adding cooling is a must.
  • jrs77 - Friday, January 30, 2015 - link

    Allthough I very much like the idea of the smallest PC possible, the NUC or the BRIX (or anything else in this formfactor) is still too expensive compared to a more powerful and better customizable mITX-system.
    I can build a low-powered mITX-system for $600 (i5-4590T, H97 board, 16GB RAM, 256GB SSD, 20x20x8cm case incl 90W PSU), which leaves me with money for the Win8.1 license (the one you didn't include in your price there!). Such a system, has much more value and can be strapped to the back of your screen not using anymore space than the NUC or BRIX.
  • zodiacfml - Friday, January 30, 2015 - link

    I agree. That's what I thought of this review. The processor is just too pricey.
    Someone could buy a notebook with an i5-Haswell with a AMD/Nvidia GPU near that price.
  • piasabird - Friday, January 30, 2015 - link

    If most of what you plan on doing viewing video an i3 with 4 megs of cache will work just fine.
  • deathwombat - Friday, January 30, 2015 - link

    Does the top of the box actually say "Supports 2.5" Hard Drivers"? At least have someone proofread your packaging!
  • eanazag - Friday, January 30, 2015 - link

    I agree on the networking. 2x2 Wifi would be better and Intel NIC. I dislike the Realtek NICs. With an Intel NIC there are more options from the software side, like a tiny VM server.
  • vision33r - Saturday, January 31, 2015 - link

    Intel has no pressure to improve performance.
  • Mikad - Tuesday, February 3, 2015 - link

    This is just a small thing, but it has bothered me much lately: Many of the new articles have a quite bad "featured image". For example this one: Just a picture of the box with a bad lightning. The product itself is interesting but the picture is a turn off. IMHO it would be great if you could put more effort into these pictures.
  • Teknobug - Wednesday, February 4, 2015 - link

    Looks like gaming is out of the question, at least at 1920x1080. May as well go with an i3 or N2940 and play via Steam stream from a gaming PC.

    Otherwise everything else about this is awesome.
  • voicequal - Saturday, February 7, 2015 - link

    The frequency and temperature characteristics are the most interesting and uniquely Anandtech parts of these reviews. Please keep up the good work.

    I'm puzzled as to how a CPU with a base clock specification of 2.4 GHz can drop to 1.6 GHz during the CPU+GPU loading test? Doesn't that make the base clock effectively 1.6 GHz at the given TDP of 15W?

    It's disconcerting that the GPU can steal power from the CPU, and yet this doesn't show up anywhere on the spec sheet but is continually exposed by these reviews.
  • xchaotic - Monday, February 16, 2015 - link

    Actually for the most obvious scenarios I can think of - gaming and other GPU accelerated tasks, I think it's better to give GPU priority, but I think there should be a software option to control how this works, along with other power saving features like idle timers...
    I can think of scenarios where both the CPU and GPU are loaded and the CPU should be given priority...
  • BtotheT - Monday, March 30, 2015 - link

    I'm thinking the next big leap for Brix will be the Skylake H series processors(Q4'15-Q2'16). The Broadwell model here was a dual core, with non edram graphics(HD5500 series no less, Broadwell's low end) , you can't expect it to keep up with a quad core with edram the 4770R had.

Log in

Don't have an account? Sign up now