• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

FEATURED Real Temp shows different CPU speed than CPUID?

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
I sent him a PM thanks. I guess I am not sure if I am overclocking properly. I shut off turbo and speed step and set the multiplier to 40 instead of stock 34.
 
RealTemp T|I Edition
http://www.overclock.net/t/1330144/realtemp-t-i-edition

The latest version of RealTemp will show you what C States the CPU cores or the entire CPU package are using.

CPU-Z is a tool designed for consistent MHz validation numbers. When a CPU is lightly loaded, depending on what C States your CPU is using, CPU-Z might be ignoring what your CPU is really up to internally.

RealTemp uses high performance timers within the CPU and follows the monitoring method recommended by Intel to tell you exactly what your CPU is doing. If your multiplier is hunting up and down when lightly loaded, the cores are probably rapidly entering and exiting various C States like C1E, C3 or C6.

The Control Panel - Power Options - Minimum processor state percentage will also determine what your multiplier does when it is lightly loaded.

Also keep in mind that just because you selected something in the bios, does not mean that is how your CPU was ultimately set up. The bios compares all your requests and turns on or turns off whatever needs to be turned on or off. The default multiplier for a Core i7-3770K is 35. The Intel Turbo Boost feature has to be enabled if you want the multiplier to go higher than that. You might think that you "shut off" turbo but as soon as you select a 40 multiplier and click on Save in the bios, it determines that you are going to need Turbo Boost enabled so it turns it back on without your knowledge.

Same thing with EIST - SpeedStep. On my Asus P8Z77 board there is an option in the bios to disable SpeedStep but when you boot up, SpeedStep is always enabled no matter what you selected for this bios option. Why? Because disabling SpeedStep can cause problems for the average Joe that doesn't really understand what SpeedStep is all about. You can end up with a CPU locked at the minimum multiplier. It is safer for Asus to just leave SpeedStep enabled no matter what a user requests in the bios. I am not sure about Gigabyte but I know that my Asus board is not the only board that is doing this.

On a Sandy or Ivy motherboard, there is no need to disable EIST. If you want your multiplier locked at the maximum value then disable C1E, C3 and C6 and go into Windows and make sure the Minimum processor state is set to 100%.

RealTemp is telling you that you do not have your CPU set up the way you think you have it set up. You can trust what RealTemp is telling you.

Edit: Older versions of CPU-Z used to tell me that my CPU was steady at the 16 multiplier when idle. The recent versions tell me that my multiplier is steady at 40. Neither version is correct. RealTemp correctly shows the multiplier is wandering around between these 2 values.

http://img836.imageshack.us/img836/6682/cpuzoldnew.png
 
Last edited:
Where are these settings for Minimum Processor State? I am looking under the advanced settings for my power scheme but I do not see it.

I went into the bios and disabled c1e c3 and c6 now the CPU is locked in at 4 ghz according to real temp. It no longer wanders. Is this okay? I mean do most people disable c1 c3 and c6 when they over clock? Does it have any effect on the computer going to sleep?

Thanks
 
Last edited:
Control Panel - Power Options - Processor power management - Minimum processor state

poweroptions.png


Is this okay?

That depends on what you are trying to accomplish. Your original goal was to have your CPU locked at 4000 MHz so disabling C1E, C3 and C6 and using the Windows High Performance power profile is the best way to achieve that goal.

Personally, I overclock my 3570K and I leave C3/C6 enabled. It reduces power consumption significantly with no noticeable difference in performance. An Ivy Bridge CPU can go from C6 to full speed in about 200 microseconds. No human is going to notice any lag because of C6 when surfing the internet. I like to combine C3/C6 with the Windows High Performance power profile. When a core enters C6, the voltage going to that core is virtually zero so using the Balanced power profile to save power has become pointless. C6 is kind of like a little guy in your CPU that automatically turns the lights off when no one is in the room and turns them back on instantly when needed.

For a test, run a single threaded benchmark like Super Pi mod to put a load on your CPU.

http://www.techpowerup.com/downloads/366/Super_PI_Mod_v1.5.html

If you enable C3/C6 in the bios, run RealTemp T|I, open up the C States window and you will see that even when your CPU is working on a task, the unused cores will automatically go into the C6 low power state. Whenever your CPU is running at less than 100% load, enabling C6 will allow you to reduce power consumption without any noticeable difference in performance.

Are you starting to understand that RealTemp is not erroneous? CPU-Z is fine for validations but if you need to know what your CPU is really doing, I would suggest that you use RealTemp.
 
Yes I understood that real temp was not erroneous after you replied. Strange though I do not have that setting in advanced options of my power plan. Just something for cooling which is set to active.
 
You probably do not have that setting in Windows because you disabled SpeedStep - EIST in the bios. At least this confirms that this setting in your bios is working correctly.

While we are on the topic of what monitoring software to believe, here's a good example of what various monitoring apps do not tell you.

http://img255.imageshack.us/img255/6/monitortest.png

For a test I decided to fully load my Core i5-3570K with Prime95 - Small FFTs.

In the top left corner, the Windows Task Manager shows that this CPU is indeed running with all 4 cores at full load. If you look at Core Temp, it agrees and shows all 4 cores running at a Load of 100%. CPU-Z nods its head in agreement and shows all 4 cores running at 4000.0 MHz. HWiNFO 64 also confirms that all 4 cores are running flat out at 4000 MHz.

Now have a look at the Prime95 results so far. Worker #1 has already completed 12 tests while Worker #2, #3 and #4 have only been able to complete 2 tests each. There is obviously a very serious problem with those 3 cores but the majority of trusted monitoring apps are completely oblivious to what is going on inside this CPU.

Now check out RealTemp. Right away it is obvious that there is a problem because the Load is only being reported at 35%. RealTemp does things differently. The Load meter does not use some crappy Windows API function to get its load data from. It digs deep into the CPU and accurately measures what percentage of time a CPU core is in the C0 state. This is the state that a core will be in when it is working. If you don't see 100%, the cores are not fully working. In this situation they are being throttled by what is called Clock Modulation throttling. If you look at the Thermal Status area of RealTemp, it reports this throttling.

If you look at ThrottleStop, you get a clearer picture of what is really going on. Sure the first core is running at full speed and it is spending 100% of its time in the C0 state like it should be but if you look at the other 3 cores, you can see that they are all being throttled and are only spending 13.3% of their time in the C0 state. You can also see what cores are using Clock Modulation throttling to reduce performance. After looking at this data, finally a user has an explanation why the Prime95 results are so skewed. Three of the cores are being throttled to death.

A few years ago, one of the world's largest laptop manufacturers "accidentally" released a bios that used Clock Modulation throttling on all of the threads except for the first thread. On a Core i7, 7 of the 8 threads were being throttled with various amounts of clock modulation so users were prevented from getting full performance out of their laptop. Did they do this deliberately because they knew most monitoring apps would not be able to detect a problem? Maybe. Since ThrottleStop was released, they don't do that anymore. :)
 
Unclewebb, thank you for the excellent explanation and demonstration. I did not know about this issue. I've been disabling Speedstep for a long time, but I'll be re-enabling it the next time I have to restart.

I've marked this thread as "featured" in the hopes that others read it and understand this issue, as well.
 
Back