Ivy Bridge Vs Trinity benchmarks are in: Intel CPU wins, AMD GPU and battery win

Discussion in 'MacBook Air' started by Jobsian, May 15, 2012.

  1. macrumors 6502a

    Jobsian

    Joined:
    Jul 30, 2009
    #1
    There's been some talk about Apple's next mobile CPU choice for either MBA or a slimmer MBP. Rumors have pegged an Intel plus Nvidia for at least the 15" and up models. Nothing strongly rumored yet for MBA/13"MBP.

    Anandtech have now benchmarked these two processor classes. Unfortunately I don't think they were able to benchmark the only 2 Intel CPUs that seem viable (at present at least) for the next MBA, ie the Core i5 3427U and Core i7 3667U (both 17W TDP, 1.8GHz 2.8 turbo, 2.0GHz 3.2 turbo respectively), and I don't think AMD's ULV ones are there, but these results should give us a decent indication as to strengths/weaknesses in the class.

    As expected really, Intel's Ivy Bridge CPUs seem to outperform AMD's CPU but AMD's integrated GPU Trinity 7660g outdoes Intel's HD 4000. The really disappointing thing for me is the battery with Anandtech saying:

    "We’ve only looked at one Ivy Bridge laptop so far, but if the pattern holds than Ivy Bridge will generally be a moderate step back in battery life relative to Sandy Bridge, giving AMD an even larger lead in this area."


    [​IMG]


    [​IMG]

    [​IMG]

    Note that the HD 3000 corresponds to Sandy Bridge, the Ivy Bridge HD 4000 lagging at the bottom of battery tests.

    More details, graphs and source at Anandtech: http://www.anandtech.com/show/5831/amd-trinity-review-a10-4600m-a-new-hope
     
  2. thread starter macrumors 6502a

    Jobsian

    Joined:
    Jul 30, 2009
    #2
    Hellhammer: any more insight/thoughts?

    It's very disappointing to see Ivy Bridge lag on battery (except H.264). I thought the 22nm process was going to help, anyone know why we're seeing this?

    If Ivy Bridge battery is going to be worse than Sandy Bridge, that especially doesn't bode well for those like me looking to upgrade from C2D MBA, which has a better battery than SB MBA.
     
  3. macrumors 604

    theSeb

    Joined:
    Aug 10, 2010
    Location:
    Poole, England
    #3
    That ivy bridge CPU is not an ULV chip so it has no bearing on the MBA.
     
  4. macrumors 65816

    oneMadRssn

    Joined:
    Sep 8, 2011
    Location:
    Boston, MA
    #4
    I was under the impression the ULV ivy bridges were not coming out until mid summer or later.
     
  5. macrumors 6502a

    Joined:
    Nov 11, 2008
    Location:
    iLondon/iDurban
    #5
    That graphics comparison didn't quite make sense to me :eek:

    How did the HD4000 fare?
     
  6. macrumors 6502a

    JasonR

    Joined:
    Nov 11, 2008
    #6
    So does this mean the new Airs are at least a few months away?
     
  7. macrumors 604

    theSeb

    Joined:
    Aug 10, 2010
    Location:
    Poole, England
    #7
    It all depends on when the ULV IB chips are released by Intel.
     
  8. thread starter macrumors 6502a

    Jobsian

    Joined:
    Jul 30, 2009
    #8
    As I said the only ULV Ivy chips (Core i5 3427U and Core i7 3667U) weren't benched. But like Anandtech imply, there may be a bearing depending on whether the pattern holds.
     
  9. macrumors 603

    nuckinfutz

    Joined:
    Jul 3, 2002
    Location:
    Middle Earth
    #9
    I wouldn't mind having a Trinity based Mac mini
     
  10. macrumors regular

    Joined:
    Apr 30, 2012
    #10
    For a PC that needs long battery life and decent graphic performance like the MBA, AMD is the way to go.
    Who cares about raw CPU power? I mean, all the modern processors give enough horse power for common tasks and anyway I don't believe that people who have a MBA use for CPU intensive operations. I am doing heavy scientific image analysis and I barely see the difference between a 2008 C2D and a modern i5.

    BTW, I think that we will never see any AMD CPU on any Mac. AMD has a bad reputation (I don't know why, I have been using AMD on non-apple desktops since 2001 and never had a single problem), Apple is synonym to "perfect". I can't see any possible marriage between "bad" and "perfect" at the horizon.
     
  11. macrumors 603

    nuckinfutz

    Joined:
    Jul 3, 2002
    Location:
    Middle Earth
    #11
    There was a rumor about Apple looking at the Llano APU and wrote it off when AMD couldn't deliver the numbers they need.

    AMD needs to get their foundry game polished up. Intel is beating the breaks off them in the process node wars.
     
  12. macrumors 6502a

    Joined:
    Mar 24, 2011
    #12
  13. macrumors 68000

    KylePowers

    Joined:
    Mar 5, 2011
    #13
    Of course it's going to run hotter. You're packing X amount more Ivy Bridge transistors in the same (or smaller) space in which you packed Y amount of Sandy Bridge transistors. The heat density increases drastically. //I'm pretty sure at least

    The only way to make it run cooler, aside from using more efficient thermal paste and a good cooling system, is to decrease the voltage (thus ULV processors) which decreases the overall power consumption. But heat is seldom an issue nowadays in the sense that any modern day CPU has thermal monitors that once tripped will shut down the computer to prevent any damage.
     
  14. macrumors 6502a

    Joined:
    Feb 3, 2011
    #14
    Aren't the Ivy ULV chips due in June? I see Air updates coinciding with MtnLion in July/Aug.

    Intel's tri-gate architecture may not be the holy grail it was thought to be. It was supposed to lower leakage across the gate and thus increase battery life, but something tells me the new architecture and the smaller process wasn't fully baked before production, similar to the SB Arrandale release, which ran hot.

    Ivy is looking more like one to skip, based on these reports, and wait for Haswell or see if Apple switches to Trinity (still using 2010 C2D MBA).
     
  15. macrumors 603

    nuckinfutz

    Joined:
    Jul 3, 2002
    Location:
    Middle Earth
    #15
    I can see your point here. I'm a bit shocked that IB isn't quite as fast or power efficient as I had guessed.

    I wasn't expecting the world but I was expecting perhaps a bit more power savings along with an improved GPU. I got 50% of my desire.
     
  16. macrumors 604

    theSeb

    Joined:
    Aug 10, 2010
    Location:
    Poole, England
    #16
    Unfortunately that story completed missed the point of what is really going on. Ivy Bridge (at stock settings) runs cooler than Sandy Bridge. This has been proven and over again by sites like Anandtech and Tomshardware. If you overlock IB CPUs (not something that you will do on a Mac), then it does run hotter and requires some care when setting the overclock.

    http://www.anandtech.com/show/5763/undervolting-and-overclocking-on-ivy-bridge

    ----------

    You're going to base your entire buying decision on one graph from a preview of preview hardware?

    From the preview of that Asus Ivy Bridge a couple of weeks ago:

    At least read the article. :)
     
  17. macrumors 6502a

    Joined:
    Feb 3, 2011
    #17
    Quote:
    There are Sandy Bridge laptops that have better normalized idle and video battery life compared to the ASUS N56VM, but it does come out on top in the normalized Internet battery life results. Given the early nature of the hardware we’re not ready to draw any firm conclusions, but we’re pretty certain Ivy Bridge won’t be substantially worse than Sandy Bridge; it's just not substantially better either.

    ^^^ This! The more I read, the more Ivy looks like a wash. Without a more power efficient cpu, and adding Retina at the expense of higher power consumption, more heat, and bigger/heavier battery, the next Air is going the way of the iPad3. I just bought an iPad2 for these reasons.
     
  18. macrumors 601

    Mr. Retrofire

    Joined:
    Mar 2, 2010
    Location:
    www.emiliana.cl
    #18
    Regarding ASUS and the Intel HD 4000:
    ASUS installs probably a lot of crap on the HDD/SDD, which means higher energy consumption and has nothing to do with the HD 4000.

    The idle battery life (the computer does "nothing") of the ASUS is also one of the lowest:
    http://www.anandtech.com/show/5831/amd-trinity-review-a10-4600m-a-new-hope/8

    So the low battery life has nothing to do with the HD 4000. Windows and Mac OS X use the best power management features, which are available within the hardware. So i doubt it is a driver issue.
     
  19. macrumors 601

    Mr. Retrofire

    Joined:
    Mar 2, 2010
    Location:
    www.emiliana.cl
    #19
    The structures are smaller (22 nm) in Ivy Bridge processors, which means less surface, and less current, and therefore less heat.

    Heat ist still an issue today.
    1. It shortens the lifetime of many components.
    2. It disables the Turbo-Mode of modern processors.
     
  20. macrumors 601

    Mr. Retrofire

    Joined:
    Mar 2, 2010
    Location:
    www.emiliana.cl
    #20
    No, you should wait for the first test of real MBPs or MBAs. The test above shows only, that a certain OS with certain drivers and certain additional software (most likely unnecessary junk) on a certain hardware configuration requires X Wh. And what has the GPU to do with an internet benchmark (Anandtech)? Answer: Nothing.
     
  21. macrumors 603

    thekev

    Joined:
    Aug 5, 2010
    #21
    You're not looking at ULV chips, and you don't know if Apple is running in place on battery technology and power management. I'd wait to see a system you would actually consider purchasing before passing a final verdict.

    Rumors like this have come up before. Weren't they considering AMD over Intel in 2006 as well?

    Intel always talks things up. I may still jump on this just for an update, and it'll have whatever the latest gpu is, but I wouldn't be grabbing an Air.
     
  22. throAU, May 16, 2012
    Last edited: May 16, 2012

    macrumors 68030

    Joined:
    Feb 13, 2012
    Location:
    Perth, Western Australia
    #22
    even HD3000 is good enough for most people, anyone wanting serious GPU will buy discrete for the next few generations.

    on battery, the ivy bridge parts are quad core, the sandy bridge at the top of the battery life chart aren't.... they're only duals.

    the only quad core sandy bridge chip in there (the i7-2820QM) is not a lot better than ivy bridge - 5.48 vs 5.86. that's worse, yes - but given it is a different model of machine with different motherboard, screen, etc i wouldn't be overly concerned that "OMG ivy bridge sucks compared to SB!!!" just yet. You're talking 23 minutes difference in run time there.

    Not a lot to write home about, again especially considering they're in different manufacturer's machines - it could be something as innocuous as a wifi chip or bluetooth module causing the difference. or even simply a less effective power management plan. I know on the HP elitebooks we have here at work, installing the HP recommended power plan improves battery life by about 20% vs the inbuilt Windows 7 default profiles.



    The HD4000 was considered "100%"

    the graphs are of the trinity GPU, longer than 100% = better, shorter = worse than HD4000.


    So it was nearly 2x as fast in some tests, not as fast in others. Mostly it is within 25% either way, apart from a couple of outliers - civ5, starcraft2, portal 2... as shown by the top bar: Trinity = average of ~20% faster over 15 titles.

    Whoopie. I'll take the better CPU thanks (ivy bridge = 2x performance in pcmarks vs trinity)... and if HD4000 is inadequate, go for a discrete GPU. 20% faster than "meh" is still "meh".

    Anyway... that's my 2c. If CPU performance is not an issue for you and you're chasing half decent GPU (on a portable - with integrated GPU, why?) and battery, your opinion may differ.... but getting half the CPU performance for a 20% boost in GPU performance (which is still pretty average) is a bit of a big trade-off if you ask me.
     
  23. macrumors 68000

    KylePowers

    Joined:
    Mar 5, 2011
    #23
    Yes, I am aware, but that's a gross overgeneralization. By going to 22nm, Intel essentially gets to double (more or less) the amount of transistors in the same amount of silicon they were selling you last year, this is advantageous for them, this is advantageous for you. But die shrinking isn't without its issues.

    Previously, shrinking past the 60ish-nm (or was it 40ish-nm?) meant much more leakage current, which was why high-k dielectrics were created. Even so though, going down to 22nm required a radical different type of transistor - the "trigate", which is closely related to the finfet transistor. Yes, die shrinking mean less switching current, but leakage current will always increase as the gate oxide thickness decreases, causing an increase in capacitance (i.e. C = [k(epsilon)A]/t, where k is the dielectric and t is the oxide thickness)

    The trigate helped alleviate this to an acceptable [marketable] amount, but my initial statement still isn't false - you have a greater heat density, more heat per area over the entire chip (because you've essentially packed twice as many transistors). Yes, 22nm transistor requires less voltage, less power than a 32nm one, but considering the fact you have more of them, doesn't necessarily makes you have less heat or break even. It's obviously all very complicating. But nonetheless, we haven't seen anything about the ULV chips, so who knows what they'll actually be like.

    Your last two points are correct though, but I wouldn't say there's anything to worry about how long it'll last. For most consumers, it'll last longer than they'll want to keep the computer anyway. That doesn't make it okay though, of course.
     
  24. thread starter macrumors 6502a

    Jobsian

    Joined:
    Jul 30, 2009
    #24
    Thanks for the clarifying posts above. It would be great if they could run these benchmarks on completely standardised hardware and software.
     

Share This Page