11 Replies Latest reply on Mar 16, 2012 9:10 AM by Sunshineguy

    Turbo Boost and predicting TDP of cpu

    Evilsizer

      I struck me the other day when setting up my i3-330M in a Jetway mITX board, as i was watching HWmontior.  it was showing the TDP of the cpu, not sure how it was getting or predicting the TDP of the cpu.  if this program is accurate at getting/predicting TDP, would it not be possble to keep the tubo boost constantly enabled on a say a i7-640M by disabling HT.  i guess the next question would just how much does HT add or effect the cpu TDP in the first place.  as i would think that without HT and turbo boost enabled it would still be around 35watt TDP or around there.

       

      the next part would be, is it possible to modify the intel turbo boost driver to pass the OS command to enble it.  that way the cpu would be constanly at 3.46ghz as it should be or would be near/under the 35watt TPD.  this is going on the fact the Jetway board can only support a max of 35watt cpu.

       

      thoughts?

        • 1. Re: Turbo Boost and predicting TDP of cpu
          mechbob

          Disable speed step.

          • 2. Re: Turbo Boost and predicting TDP of cpu
            Evilsizer

            yea im not talking about speed step, this is something completely different.

            • 3. Re: Turbo Boost and predicting TDP of cpu

              I have'nt heard or read a driver for turbo boost except in this thread. Can you please clarify? Thanks.

              • 4. Re: Turbo Boost and predicting TDP of cpu
                Doc_SilverCreek

                I got real deep into Turbo Boost in this thread http://communities.intel.com/message/115876#115876

                 

                Yes & No

                 

                If you are under the max TDP (which is really CPU voltage times current) , Imon current (which is a sensor on the CPU that recieves feaeback from the mother boards VRM telling the CPU how much current it is drawing) and temperature and it is Turbo Boost is enabled in your BIOS (C-states),

                Turbo boost will engage,

                 

                So, better cooling, less work (either in workload or by disabling CPU features like HT or even full cores) will keep in in turbo boost longer or forever.

                The question comes down to can the system process more information with all the cores and HT enabled running at related speed, or will it process more info with a core disable and running at turbo boost speed?

                The answer is application specific.

                 

                I have a older game I enjoy that is older single thread and uses only one core in my Dual Xeon workstation.

                It very rarely comes out of turbo boost anyway while playing unless I am running other things at the same time.. If I disable cores in boot, it never comes out of turbo boost.

                But if I did this while running my CAD rendering, I could be waiting all night to get any work done. 

                 

                Driver -- No.

                Turbo boost is a feature of the processor and is enable \ disable via processor C-states, but there is no driver.

                Some Turbo-boost monitors use a driver, but that is just keying in the monitor and has nothing to do with the Turbo boost itself.

                 

                You might do a application to enable \ disable HT or cores within an OS, but it would take some interesting programming to make sure the thread \  core is not active before you disable it. Easier to do in BIOS set-up so the OS never gets involved. 

                • 5. Re: Turbo Boost and predicting TDP of cpu
                  Evilsizer

                  i am just going what i have found in windows 7 and what i have read [url=http://download.intel.com/design/processor/datashts/322812.pdf] here section 3.4.1].  where it states "Intel Turbo Boost Technology processor frequencies are only active if the operating system is requesting the P0 state", which sounds like to me the OS needs to pass the P0 state to enable turbo boost.

                   

                  here is a screen shot of device manager on my i3-330m mITX setup.

                  http://img.photobucket.com/albums/v733/Evilsizer/i3330mTBdriver.jpg

                  then for 25watts that is with HT enabled right now, at a little under 1v it is getting 25amps or a little over.  i guess in this case using turbo boost wont matter seeing as my WU's for rosetta run from anywhere for 9k to 10k secs or 2.5hrs to 2.8hrs.  though i was tryting to figure out a way to get tubo boost to work if i was playing a game, that way i could get the higher cpu speed for the GPU.

                   

                  thanks for the link to that other thread, i shall check it out.

                  • 6. Re: Turbo Boost and predicting TDP of cpu
                    parsec

                    3dfx,  Your's is an interesting situation, a mobile CPU in a desktop PC environment, if I'm reading your post correctly.  If so, you have much more potential of keeping your CPU operating at the Turbo maximum (or close to it) frequency than it would be working in a laptop.  As you know from the replies in this thread, and studying the document on your CPU, accomplishing that mainly if not completely depends on keeping the TDP within it's working area, or less than it's maximum.  As the tech document for your CPU states, the graphics processor in your CPU will ramp up to it's higher frequency when it is under load and the overall TDP has not exceeded the maximum.

                     

                    While I initially thought it would be much easier cooling your CPU in a desktop board and case, when I checked the socket type, BGA1288/PGA988/G1, I realized it is likely difficult to obtain a CPU cooler for this platform.  That turned out to be true, I found a few which reminded me of early AMD CPU coolers, with small fans and minimal metal in the heat sink itself.  I'm sure you have better knowledge about that item than I do, and I can imagine a high air-flow case helping cool this CPU very well.

                     

                    In Turbo 2.0, used in Sandy Bridge processors, there are several limits defined, including time limits, for how long Turbo boost is active, or the circumstances in which it is stopped or limited.  Those limits can be overridden, or set to "infinite" in a mother boards BIOS, but not directly by the user, they seem to be behind the scenes, although I have seen the settings displayed by a hardware monitoring and info program.  I am less familiar with these things in standard Turbo boost, and they may not exist at all in standard Turbo.  The point here is it is possible that your mother board may be a limiting factor in attempting to achieve "endless Turbo", or may even enable it.

                     

                    In my experience and given what I have read, disabling Hyper-Threading in the BIOS will give you back about 5C under load, which will be significant with the low TDP you are working with.  I have never seen TDP vs Hyper-Threading data personally.

                     

                    I am curious what your PC's configuraton is, what CPU cooler you use, and what case you have.  You may be able to attain your goal, as long as you can keep that CPU cool.

                    • 7. Re: Turbo Boost and predicting TDP of cpu
                      Evilsizer

                      dont know why/how but i overlooked that the driver for was the graphics speed, thought it was linked to the cpu speed.

                       

                      i have seen industrial boards using mobile cpus for some time, just not many users buy them to use in desktop pc or SFF pc's.  the board im using is a Jetway NF98 using the cooler that was provided from logic supply. 

                      http://www.logicsupply.com/products/nf98_qm57

                      there are two others listed on logicsupplies web site but i dought they would be work the extra cost.  with the heatsink and fan included it will maintain a 56ish load.  the temps you see above in the screen shot are with a 120mm fan above to also get some air to the PCH heatsink, which has a really small heatsink.  as well using the 120mm even at its full speed, think 42cfm, yateloon low speed 120x25mm and the fan in the heatsink is a weee bit loud for me at full speed.  i attempted to use a Zalman chipset heatsink, blue flowering type design, however the arms and pushpins could not go out far enough.  to get the other arm/pushpin to line up it needed about another 1-2mm of side to side space.  i keep looking at different heatsinks that might work but they are all chipset heatsinks.  my only other choice for this box is to epoxy on the zalman heatsink as it has about 4 times the surface area of the current one.  it would be my passive sink setup with a 80 or 60mm fan blowing air over the motherboar, still a WIP.  the rest of the system as you can see has 2x2gig DDR3-1066, seagate LP 80gig 7200rpm drive, LG blueray/dvd/cd reader, Seasonic 300Watt TXF PSU.  right now there is no case, there is one i would like to get but does not support full size 5.25in drives.  looking at doing a custom one for this pc and thinking about another build on the same board as well. 

                       

                      while the i3-330m does not have turbo boost the other cpus im looking to get does, the i7 640m so i can get two cores at 3.46ghz.  the i7 setup will use a more mid-range gpu or possibly highest that works on the silverstone SFX-450 power supply.  it will also get a SSD one i have on hand from about a 1.5yr ago or longer, easy to lose track. 

                       

                      now before i start going off into a different direction, go back and look at that screen shot i posted, notice the 25watt TDP.  Well im wondering how they figured it out, from the sensor your referencing or some other way.  as it is the cpu is rated on the ark site as 35watts, now it coud be the other 10watts is from the graphics.  if that is the case in the second machine, i will not be using the on package graphics but a add-in board.  now the bios has no way to control the cpu's multiplier like on desktop parts, what im trying to do is basicly have a OS on/off switch for the cpu's turbo boost.  as it looks like even with 4 threads running the cpu should be under or near rated TPD for turbo boost to work.  now im not sure about this time limit thing since as i recall intel never said anything about it being time limited.  the only reference was as long as the cpu stayed under/near rated TDP, which takes me back to that section i pointed out.  where that footnote is it to me seems that you can flip it on/off by having the R0 state sent to the os and passed to the cpu.  that is the way i read it since it is not in the turbo boost section for the graphics.

                       

                      im sure rosetta is not fully using all the parts of dual core + HT, i may need to run Wprime or some other same type app.  this way i can see what effect HT has on TDP.  while no HT does show a lower cpu temp, using temp decrease is not a way to calculate cpu TDP.

                      • 8. Re: Turbo Boost and predicting TDP of cpu
                        parsec

                        Regarding the Turbo boost time limitations, as I said that is used only with (AFAIK)  Turbo 2.0, which is currently used only with Intel Second Generation  Core i5 and i7 CPUs.  I seen several threads where people were concerned  that their Turbo 2.0 "over clock's" were not sustained indefinitely  under heavy loads.  My i7 2600k and mother board does not behave  in that manner, I never lose the maximum Turbo 2.0 multiplier regardless  of how long the heavy load is sustained.  But this is not a factor with  first generation Core i7 CPUs with the original Turbo feature.

                         

                        An on/off switch for Turbo boost is something I have  never seen, and I have spent a fair amount of time visiting many of the  well known PC hardware review sites, and PC enthusiast forums.  If such a  thing were simple or even possible, I'm sure it would be well known to  the various enthusiast forums and web sites.  While many functions  within a CPU are controlled by signals and data in special registers, if  Turbo boost was available via those usual methods, an "Engage Turbo"  button would be available in everything from a mother board's BIOS, to  their utility programs, and programs created by developers that give us  CPU-Z, etc.  Alas, the most I'm aware of is enabling/disabling Turbo in  the BIOS, and Turbo boost frequency monitoring programs, one of which is  available from Intel.

                         

                        Frankly, what is the point of  manually switching to Turbo frequencies in a CPU, when it starts on it's  own when demand on a CPU is such that it is needed?  But then, I find  no reason to over clock a CPU via it's base clock frequency and/or base  multiplier so that the CPU constantly operates at a high frequency  regardless of the demands on the CPU.  IMO, SpeedStep and Turbo work  fine, adjusting the CPU frequency as appropriate.

                         

                        The  i7-640M CPU you are considering is a multi-chip package (MCP), composed  of the processor and graphics cores, and memory controller.  The memory  controller is combined with graphics core, producing one TDP spec.  The  TDP spec is the total of all three of these components.  The good news  is the processor as a whole will be thermally safe with it operating at  Tjmax and at the maximum TDP.  When not using the graphics core of the  processor, as you stated is your plan, the processor core should gain  some TDP capability, as the graphics core will be switched to it's  lowest power C-state.  This is from the technical document, and are  called extreme design points:

                         

                        • The processor core operating at maximum thermal power level (which is greater

                        than its component TDP) and the integrated graphics and memory controller

                        operating at its minimum thermal power.

                        • The integrated graphics operates at its maximum thermal power level, while the

                        processor core consumes the remaining thermal power budget.

                         

                        After  reading the technical document for this CPU, I see where you got the  idea about the Turbo driver.  The following is taken directly from that  document:

                         

                        Intel Turbo Boost Technology and Graphics Dynamic Frequency are implemented via a
                        combination of Intel silicon capabilities, graphics driver and the Intel Turbo Boost
                        Technology driver.
                        If Intel provides Intel Turbo Boost Technology support or Graphics
                        Dynamic Frequency support for the target operating system that is shipped with the
                        customer’s platform and Intel Turbo Boost technology or Graphics Dynamic Frequency
                        is enabled, the Intel Turbo Boost Technology driver and graphics driver must be
                        installed and operating to keep the product operating within specification limits.

                         

                        What the actual "Turbo Boost technology driver" is  I am uncertain of.  The terminology may not be what we are expecting,  and I would like some clarification of it.

                         

                        Tables 18 and 19 in this document list the TDP of the two components of an MCP, which will be of interest to you if you have not already seen it.

                         

                        If we had control over sending the P0 state request  to the processor, the thermal constraints of the processor and the CPU  cooler's cooling capability still determine whether or not the the  processor core Turbo frequency will be initiated and sustained.  This is  discussed in section 5.1.4 of that document.  So IMO, maintaining the  higher Turbo boost frequency of the processor core is completely  dependent upon maintaining the entire processor (MCP) at or below it's  TJmax and max TDP.  Do you believe there is a factor that is causing the  P0 state request to not be sent to the processor, thus the need to do  it manually, or on demand?

                        • 9. Re: Turbo Boost and predicting TDP of cpu
                          snapper

                          hi,

                           

                          Sorry to jump in, but did you use the BIOS default settings for voltage etc for the i3-330m in your Jetway?

                          I have just brought the same setup, but cannot load Windows installer - I just get BSOD.

                          I can't find any optimum settings anywhere, and need to work out if this is faulty hardware or user error...

                           

                          Cheers

                          • 10. Re: Turbo Boost and predicting TDP of cpu
                            Evilsizer

                            i didnt change any settings, other then making sure the cd/dvd/bd drive was primary boot device.  i did disable com ports but other then that i can think of anything else i disabled.

                            • 11. Re: Turbo Boost and predicting TDP of cpu
                              Sunshineguy

                              Hi , Can you tell me where can I got CPU parameter lists (TDP,rca) . Thanks so much.