Intel's TDP rating is NOT max power and SB heat

Discussion in 'MacBook Air' started by calvol, Jul 7, 2011.

  1. calvol, Jul 7, 2011
    Last edited: Jul 8, 2011

    calvol macrumors 6502a

    Joined:
    Feb 3, 2011
    #1
    There's a lot of misinformation on this board about TDP ratings and many are in denial about the potential heat issues with Sandy Bridge. What many don't know is: 1) Intel's TDP rating significantly underrates maximum power consumption, 2) TurboBoost duty cycle is NOT included in the TDP rating, and 3) TurboBoost is causing heating problems in the MBP, which has better cooling capacity than the Air.

    1) Misconception 1: TDP represents max power rating

    From the Intel Datasheets:
    “The TDP numbers are not indicative of the maximum power the processor can dissipate under worst case conditions.”


    Intel is listing TDP numbers that are significantly lower than the actual maximum power draw of their CPUs. They are then relying on the fact that most applications barely use the CPU, assuming that it will remain idle most of the time.
    http://www.silentpcreview.com/article169-page3.html


    2) Misconception 2: Turbo-boost over-clocking is captured within the heat profile of TDP rating.

    From Techreport on SB: The Turbo algorithm does something that may seem a little counterintuitive at first, allowing the CPU to ramp beyond its maximum TDP... taking advantage of the lag between when a relatively cool idle chip begins to warm up its environment.... the chip opportunistically pushes beyond its rated thermal peak by running at higher-than-usual frequencies within its Turbo Boost range. Once the surrounding system has warmed up or enough time has passed (the algorithm is complex, and Intel hasn't shared all of the details with us), the chip will drop back to operating within its TDP max... Turbo Boost algorithm incorporates not just the CPU cores but the IGP, as well; it can raise the operating frequency of the graphics processor when the CPU cores aren't at full utilization.

    3) Misconception 3: SB heat issues in 2011 MBP won't appear in 2011 MBA

    Since the Air ULV chips also use TurboBoost, expect the same thermal behavior as this poor MBP owner observes:

    "I just ran my "old" 13 MBP (2010 edition - 2.66 GHz) and my day-old 13 MBP (2011 edition - 2.7 GHz) with identical conditions, and performed the same test on both: launch the game "Portal" and have it just sit on the main intro screen (which shows 3D graphics slowly panning around a room, plus the game's start menu).

    On the 2010 MBP: CPU Temp: 149 F, Fan: 1999 rpm
    On the 2011 MBP: CPU Temp 192 F, Fan: 6200 rpm

    If I'm only running a simple application (like MS Word), the 2011 MBP actually runs a few degrees cooler than the 2010 system. However, as soon as the system is having to track a large number of open windows (whether those programs are graphically intensive or not), the CPU starts to heat up rather dramatically, and the fans ramp up to "jet turbine" mode.


    With the Air's limited cooling capacity by virtue of it's case design, I expect SB heat problems to be worse than the MBP because it has less ability to respond to instantaneous heat caused by TB, even though it will be using the ULV chips. Also, keep in mind that the Intel's solution to prevent heat problems involves "throttling" the CPU through it's power control unit, but ALL major PC OEMs are experiencing throttling issues with SB!
     
  2. fyrefly macrumors 6502a

    Joined:
    Jun 27, 2004
    #2
    Some really interesting points here.

    I have a few questions, but before I do, let me say that things have gotten kindof heated in these forums before, and I am absolutely not trying to blindly defend Intel or the SB chip - I just want to have an intelligent discussion about these points.

    So here we go:

    1. If Intel is routinely underestimating it's TDP, would it not have done the same thing on the C2D's? Therefore, if Apple put 17W i5/i7 chips into the 13" MBA then the TDP (w/o turbo boosting) would be lower than the C2D regardless, no?

    I can understand how (given the 13" MBP) there would be concerns about heat issues, but the 13" MBP (and 15/17" too for that matter) went up 10W in TDP (as reported by Intel). With no case re-design. If Apple uses 17W i5/i7s in the 13" MBA, then the TDP goes *down* from the C2D usage. Even if they use 25W i5/i7 processors, the TDP remains the same or slightly less than the current C2D TDP. And again, if Intel's reporting TDP erroneously on the i5/i7, would they not also have done it on the C2D?

    2. re: Turbo boosting - So what happens if Apple doesn't enable Turbo Boosting at all? Or throttles the processor's Turbo so that it does 1/2 or what it's supposed to to stay w/in a certain thermal envelope? Would that not be sufficient? They've certainly throttled both CPU and GPU's in the MBA before.

    That's not to say heat issues won't appear. They've practically been in every Air since it came out (I know as I've owned a RevA, RevB and RevD).

    3. Where does that leave the Samsung Series 9? It has at least one of the potential processors for the 11" or 13" MBA refresh... and Engadget noted: "We should also note that the rig stayed fairly cool during both corded and cordless use -- the keyboard got slightly warm, but only after quite a bit of use." Is Samsung finding some other way of masking the Sandy Bridge Heat issues here? Or did Engadget not test hard enough?

    4. When comparing the two machines the article you quoted mostly used a gaming benchmark, where we know the i5/i7 GPU will struggle. And the point about the multiple windows causing fans to spin up... well, I have a 13" 1.86Ghz 2010 MBA model with 4GB of RAM. If I open enough windows (or just the iTunes visualizer), it heats up and the fans spin at 6000RPM too. The battery life on the 2.13Ghz model I tried for a week was worse than the 1.86Ghz - noticeably so. So the 13" model is already not whisper quiet in those scenarios either.
     
  3. bp1000 macrumors 65816

    Joined:
    Jul 7, 2011
    #3
    This is an interesting set of points

    I was about to post a topic on this, of concern.

    As i understand it, turbo boost is only enabled if the CPU is under a certain temp and is disabled or throttled when it goes above a certain pre-defined temp.

    With the MBA case design and cooling giving it a less efficient heat exchange than say a chunky dell laptop surely that would mean the turbo boost would be a short and rare event in the MBA?

    I am naive on the point of TB, so i would like some advice. Also would like to know when turbo boost is enabled, perhaps when the chip is running at 100% for x amount of time does it up the clock speed to help?

    I know we will be buying blind unless we wait for reviews but that is boring and using the Samsung Series 9 as the first respondent pointed out as an idea, it doesnt seem to have significant heat issues.
     
  4. Duke15 macrumors 6502

    Duke15

    Joined:
    May 18, 2011
    Location:
    Canada
    #4
    From what I understand is that the MBPs are actually using higher TDP in there 2011 versions in relation to the 2010 ones, I could be wrong but I know i read that on one of the other threads.
     
  5. sporadicMotion macrumors 65816

    sporadicMotion

    Joined:
    Oct 18, 2008
    Location:
    Your girlfriends place
    #5
    We'll find out for sure in a few days. My SB MBP got pretty hot but it didn't ever malfunction or act odd. It was fast and ran very well.

    I felt the big issue with the new MBP's was just the insane fan volume! Mind bogglingly loud. Louder than both of the fans in my old 15" together at 6200 RPM each.

    You do make a great point though... at the very least we can probably expect loud MBA's.
     
  6. calvol thread starter macrumors 6502a

    Joined:
    Feb 3, 2011
    #6
    In the coming weeks, I am expecting a lot of threads like this from 2011 MBA owners:

    "I just bought this Macbook Pro 2gz 2011. I previously owned a MBP late 2008

    My initial impression what's been happening with this new system is unacceptable.

    - The fans spin up to a loud volume 6400rpm during some simple activities that dont seem to warrant it ie. starting a browser with a handful of restored tabs.
    - Running the simple "yes > /dev/null" in the shell immediately causes the fans to spin up.

    None of these things happen with my MBP 2008.

    Is this the same for all 2011 MBPs or is it possibly system specific?

    What's the likelihood of the next MBP not having these issues? I might wait a year and make due with my ol trusty 2008.

    Also running a simple video stream like you tube gets the fans up to 4500rpm.
    I could run multiple streams simultaneously on the 2008 without fans kicking in.
     
  7. sporadicMotion macrumors 65816

    sporadicMotion

    Joined:
    Oct 18, 2008
    Location:
    Your girlfriends place
    #7
    Yeah, but this is nothing new anyways. This goes back well over 6-7 years. It's just the noise that irritates people and gets their attention. The actually failure rate is very low.

    The old overclocker in me hates high temps, the mac user in me has only had one mac actually fail in 19 years (a macbook).

    If it does, I'm glad we got warranty. :D
     
  8. pil0tflame macrumors member

    Joined:
    Apr 19, 2011
    Location:
    London, Ontario
    #8
    Unfortunately, you've just added to it. ;) It's not my intention to be rude, but from what you've posted you appear to be misinformed on the very topics you attempt to clear up.

    1. TDP has nothing to do with CPU power draw.
    TDP is the maximum amount of heat (measured in watts) that a CPU may throw off. A measurement of maximum heat energy coming out of the CPU, not the electrical energy going into it. Source: http://en.wikipedia.org/wiki/Thermal_design_power

    The SilentPCReview page that you link (from June 2004, btw) is simply suggesting that when SPCR measured the heat output from Pentium 4's, they found the maximum output exceed that defined by the TDP rating Intel had assigned. This is in line with Intel's datasheet information at that time which explains that Intel's TDP numbers weren't calculated based on continuous full CPU loads, but instead near-continuous. (As an aside, some of the language in that SPCR article is poorly chosen and could cause a lot of confusion for those not familiar with the topics covered.)

    If Intel has continued to under-spec their TDP ratings, then this could be some cause for concern. As fyrefly points out though, if Intel has continued to use the same methods to arrive at their TDP numbers, then the Core 2 Duo found in the late 2010 MBA would also be under-spec'ed to a similar degree. Since those systems perform favourably with regards to heat and fan noise, I would believe similar TDP rated Sandy Bridge i5/i7 CPUs to produce similar results.

    2. Turbo Boost takes into account a number of factors, including current temperature.
    When the CPU is not using the maximum amount of power and current and the temperature is below a set maximum, Turbo Boost will temporally allow the CPU to operate at higher frequencies for a short duration if the OS is requesting the highest CPU performance state. During this Turbo Boost, if the CPU reaches it's maximum power, current or temperature rating, it will begin slowing down to remain within safe operating parameters. So, yes, Turbo Boost will allow the heat output of the CPU to exceed the rated TDP for a short duration until the temperature of the CPU hits the maximum safe limit at which point it slows down again until the temperature returns to safe limits. Source: http://www.intel.com/technology/turboboost/index.htm

    3. This is going to happen with Turbo Boost. It's a trade off of extra performance as the cost of extra heat and fan noise. If you're planning on gaming with the 2011 MBA or other prolonged and very CPU intensive tasks, then maybe the 2011 MBA isn't the ideal system for you. In addition to the switch from the Nvidia 320M to the Intel HD 3000 GPU, the 2011 MBA isn't going to be very appealing to the gamer or lengthy CPU intensive application user.

    Turbo Boost is ideal for situations where the CPU needs a short extra burst of speed to get a task done a bit quicker. If it's a prolonged task, then Turbo Boost is going to kick on and off depending on how well the CPU is being cooled and generally throw off a lot of heat in the process.

    In the future someone is likely going to create a program to let Mac users switch on and off Turbo Boost, but currently you cannot on the Mac. On the PC, you can disable Turbo Boost in the BIOS, so the Sandy Bridge CPUs do have the setting exposed in some manner.

    Maybe the 2011 MBA's will have Turbo Boost disabled (or set less aggressive) because of their more limited cooling abilities.
     
  9. jackyyeow macrumors 6502

    Joined:
    Jul 5, 2011
    #9
    Looking at the list of mobile i5/i7 sandy bridge, TDP is the same at 17w from 2537m to 2677m, 1.4GHz to 1.8GHz respectively.

    I don't quite believe one can expect the heat will be the same across the clock speed, or I'm missing something.:confused:
     
  10. KillerTree macrumors regular

    Joined:
    Jul 27, 2008
    #10
    Samsung Series 9 doesn't have heating problems does it?
     
  11. jackyyeow, Jul 12, 2011
    Last edited: Jul 13, 2011

    jackyyeow macrumors 6502

    Joined:
    Jul 5, 2011
    #11
    Apparently, yes (I mean it stay cool generally), with the 1.4GHz CPU i5 2537m.
     
  12. Davidkoh macrumors 65816

    Joined:
    Aug 2, 2008
    #12
    This is something that worries me and I might actually buy the old one today before the new one is released due to this.
     
  13. lukekarts macrumors regular

    Joined:
    Mar 16, 2009
    #13
    It's an interesting one actually, as I picked this up on the Alienware forums over at Notebook review - when I bought my M11x for my gaming needs.

    Seems the SB runs far hotter than the C2D's, especially the ULV versions. People don't report failures very often, but I suppose early into product life cycles, failures are unlikely. But laptops are running a lot hotter with the new architecture, meaning a lot more fan noise, and also a lot more throttling. It's of concern, particularly as the Air's will always struggle with cooling.
     
  14. Davidkoh macrumors 65816

    Joined:
    Aug 2, 2008
    #14
    Yeah, I don't care about failures that much, I only keep my computers for max 12 months anyhow. But I hate heat and fan noise. I wonder if I should get the 2010 model now while I am still able.
     
  15. Hellhammer Moderator

    Hellhammer

    Staff Member

    Joined:
    Dec 10, 2008
    Location:
    Finland
    #15
    Nobody has ever claimed that TDP is the maximum power draw under the worst conditions. TDP is merely an average maximum power draw that commercial apps can use and it's for OEMs to design cooling systems (i.e. their cooling system should be able to dissipate 35W if the CPU has 35W TDP). Some heavier benchmark tools like FurMark can easily make the CPU to run at its maximum power draw, which is called Maximum power dissipation.

    However, the CPU will only exceed the TDP (or any point) if the thermals allow that. It won't trigger Turbo if the core temperature is already 95°C. If the temps are too high, then idle cycles will be added to cool the CPU down to avoid overheating and possible damages.
     
  16. thunng8, Jul 16, 2011
    Last edited: Jul 16, 2011

    thunng8 macrumors 6502a

    Joined:
    Feb 8, 2006
    #16
    Seems turbo boost is extremely aggressive in the new Sandy Bridge processors and will exceed TDP by a large margin.

    From anandtech:

    http://www.anandtech.com/show/4205/the-macbook-pro-review-13-and-15-inch-2011-brings-sandy-bridge/14

    Sandy Bridge quad core (45W TDP) in the 2011 15" MBP uses almost 40W
    (93 vs 54.5W) more than an Arrandale Dual Core (35W) in the 2010 15" MBP when running a CPU intensive benchmark.

    I've used the 2011 MBP Sandy Bridge models, although they are much faster, they are also give out much more heat and are a lot noisier .. much more than the 10W difference in TDP would suggest.
     
  17. seepel macrumors 6502

    seepel

    Joined:
    Dec 22, 2009
    #17
    Looking at those numbers it seems the heat is mainly a product of the jump to quad core plus the discrete graphics. Look at the 13", it barely gets warmer at all so it seems like a better comparison to what will happen with the Air.
     
  18. thunng8 macrumors 6502a

    Joined:
    Feb 8, 2006
    #18
    The 15" 2010 model also has discrete graphics. Anyway, the quoted power numbers was for a CPU intensive benchmark, so the discrete graphics in both system would not have been heavily used.
     

Share This Page