1. Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Incorrect cpu freq

Discussion in 'RealTemp' started by abax2000, Mar 11, 2012.

  1. Bambooz

    Bambooz New Member

    Joined:
    Feb 29, 2012
    Messages:
    143 (0.13/day)
    Thanks Received:
    79
    Location:
    Germany
    Trolling much?! Wearing out capacitors!?! What bush did you come from all of a sudden?
    Capacitor wear... well.. you can start talking about that in about 10 years time, considering pretty much all halfway decent boards use solid polymer capacitors in the CPU VRM (and elsewhere).
    And "more wattage".. come on.. as long as you don't go berserk with the VCore, that ain't gonna hurt anything, unless it's a cheapo board like Asrock or ECS, in which case you'd deserve it frying the VRM on you. Any decent board will have a well overspec'd VRM area.

    edit:
    portable programs cluttering your registry?
    The BIOS Chip isn't being written to (ever) unless you flash the BIOS. All settings are stored in CMOS memory, which is volatile anyways (hence the need for a battery and 5VSB)
    That doesn't even make sense. He suggested using prime95 to load up the CPU to see if the multiplier in CPU-Z is changing, nothing else.

    Troll.
     
    Last edited: Mar 14, 2012
  2. unclewebb

    unclewebb RealTemp Author

    Joined:
    Jun 1, 2008
    Messages:
    970 (0.40/day)
    Thanks Received:
    433
    Here's an interesting screen shot while the CPU is fully loaded running two separate instances of CPU-Z.

    [​IMG]

    What should I believe?

    To be honest, they are both correct. I have set each instance of CPU-Z to read a separate die within the CPU. With the Core 2 Quad you can run each half of your Quad at a different multiplier and frequency so one half of this CPU is running with the 11 multi while the other side is running with the 6 multi even though all 4 cores are fully loaded at 100%.

    This is far from an everyday occurrence but it does demonstrate that a processor can do some funny things internally that a single instance of CPU-Z might overlook.

    Here's an example using a Core i processor.

    [​IMG]

    CPU-Z reporting the 24 multiplier for this CPU does not accurately reflect the speed that this CPU is running at. Both ThrottleStop and RealTemp use the same code and both follow the same Intel recommended monitoring method. They may not always agree with CPU-Z but I stopped worrying about that a long time ago. :)
     
    Last edited: Mar 14, 2012
  3. unclewebb

    unclewebb RealTemp Author

    Joined:
    Jun 1, 2008
    Messages:
    970 (0.40/day)
    Thanks Received:
    433
    Here's another example of what CPU-Z doesn't tell you on the newer Core i CPUs.

    [​IMG]

    This CPU is fully loaded. If you look at CPU-Z and trust that, you would be convinced that the multiplier is holding steady at 25 and everything is great. No problems.

    Now have a look at what ThrottleStop shows. All 8 threads of this CPU are reporting that they are throttling, likely due to power consumption. The core temperatures are high but they haven't reached the thermal throttling point yet. The C0% across all 8 threads is not able to maintain the full 100.0% so that is another sign that this CPU is doing some throttling.

    If the monitoring method that ThrottleStop and RealTemp uses was something I made up off the top of my head, I could understand why people would be afraid to trust it but that's not the case. The method that I use comes directly from Intel. When Intel releases a white paper for developers that goes into detail about the proper way to determine the multiplier, I tend to believe them because I think they know what they are talking about. The results speak for themselves.

    Edit: And here's an example from a Core 2 Duo mobile CPU.

    [​IMG]

    ThrottleStop is being used to lock this CPU into SLFM mode. SLFM stands for Super Low Frequency Mode and was a feature of these Core 2 CPUs. When a CPU enters SLFM mode, the bus speed gets cut in half and these CPUs set the multiplier to 8.0 so you end up with your CPU running at approximately 8.0 x 100 MHz = 800 MHz. The VID voltage also drops below 1.0 volts to its lowest value.

    Many Dell laptops used a bus speed of 199.44 MHz so half of that is 99.72 MHz. ThrottleStop gets that right, a reported multiplier of 7.99 is about as close to 8.00 as you can get and it also reports the VID correctly at 0.85V. CPU-Z doesn't get any of those values right and reports a VID voltage of 1.175 which is impossible when a Core 2 mobile CPU is locked in SLFM mode.
     
    Last edited: Mar 15, 2012
  4. unclewebb

    unclewebb RealTemp Author

    Joined:
    Jun 1, 2008
    Messages:
    970 (0.40/day)
    Thanks Received:
    433
    I decided to run a couple of wPrime benchmarks and before each benchmark I decided to also get a CPU-Z Validation so I would have a record of how fast my CPU was performing.

    Here is the official CPU-Z Validation as well as the benchmark results.
    Average time to complete wPrime 32M was 11.256 seconds.

    http://valid.canardpc.com/show_oc.php?id=2300535

    [​IMG]


    Here is another example with the same CPU.

    http://valid.canardpc.com/show_oc.php?id=2300557

    [​IMG]

    Average time to complete wPrime 32M was 72.862 seconds.


    Based on what CPU-Z is telling me, my QX9650 is running at the exact same speed in both examples. In the second example it is taking more than 6 times as long to complete the wPrime 32M benchmark. Why is my QX9650 running like such a slug? CPU-Z is telling me that there is absolutely nothing wrong with my CPU and in both examples it is running at the same speed. I even have an official CPU-Z Validation which proves my CPU is running at full speed.

    In the above example, the CPU-Z Validation is meaningless. At over 72 seconds for a wPrime 32M run, it is obvious that this CPU is being throttled to death and is running internally at a fraction of its rated speed. If you have a look at RealTemp, it correctly shows that Clock Modulation throttling is set to 12.5% and the C0% based Load number is also nowhere near 100% which confirms that this fully loaded CPU is not running at anywhere near its rated speed. RealTemp tells you when your CPU is being severely throttled. CPU-Z chooses to ignore this problem.

    The end result is that many of the major laptop manufacturers like Asus, Dell, Lenovo and HP have been able to get away with significantly throttling their Intel laptop processors for years because based on CPU-Z data, most consumers are completely unaware of CPU throttling and how that can kill CPU performance.
     

Currently Active Users Viewing This Thread: 1 (0 members and 1 guest)

Share This Page