PDA

View Full Version : Wrong CPU Speed Reported by RealBench



DoGood
02-01-2014, 01:38 PM
I've run RealBench numerous times and each time it reflects the speed that the CPU is running at the time it obtains the system information and not the max speed. I thought I would out smart it by running another stress test software when I started RealBench. When it obtained the system info, I got it to show 4598....MHz, but at the end it must have gotten the info again because it when back to idle speed of 799.8MHz.

What needs to be done to get the accurate speed?

CPU:Intel Core i7 4770K - Speed:799.8 MHz - Multi:8.0 - Bus:100.0 MHz
RAM:1199.8 MHz (1:9) - 16GB - 10-12-12-31-2T
MB:MAXIMUS VI HERO - FW:1301 - Desktop
OS:Windows 7 Professional
GPU1:NVIDIA GeForce GTX 770

http://rog.asus.com/realbench/show_comment.php?id=338
http://rog.asus.com/realbench/show_comment.php?id=364

Myk SilentShadow
02-01-2014, 01:50 PM
Are you using the balanced power plan, or high performance one?

DoGood
02-01-2014, 02:11 PM
Are you using the balanced power plan, or high performance one?

My power plan does have the minimum processor state set to 5%. I like to have it scale. I suppose I could set the min to 100% for the test.

I guess my next question would be why it would go by the current speed value instead of the target CPU speed?

DoGood
02-01-2014, 03:38 PM
Having my minimum processor state below 100% was the issue. Thank you.

IM2L844
02-03-2014, 02:05 PM
Having my minimum processor state below 100% was the issue. Thank you.

Yes, but I wonder if there is an a priori programming solution that would solve this to make it a non-issue. I mean could the instantiation of detection be delayed until a ~100% load portion of the testing is under way? I don't know. Maybe something for the RealBench v3 Suggestions Thread (http://rog.asus.com/forum/showthread.php?40497-RealBench-v3-Suggestions-Thread-Formerly-v2/page3).

Praz
02-03-2014, 02:09 PM
Having my minimum processor state below 100% was the issue. Thank you.
Not sure why changing this is seen as a solution. All that is needed is to change the power plan to High Performance when running a benchmark.

IM2L844
02-03-2014, 05:43 PM
Not sure why changing this is seen as a solution.

Well it's not an issue for me. It doesn't matter to me one way or the other. My concern was for those responsible parties who may get tired of seeing and answering the same question over and over again, unnecessarily, if it's just a simple matter of 5 minutes to move some code around to mitigate that potential waste of time and space. That's all I was saying.

ChikNoods
02-04-2014, 04:15 AM
I was going to ask this too. I'm on high performance in power settings and my report said 798.1 MHz

DoGood
02-04-2014, 02:28 PM
It is possible to have the power options set to high performance and have the minimum processor state below 100%. I did it manually, but I suppose software could have done automatically.

Go to Power Options > "Change plan settings" on the high performance row > Change advanced power settings > Processor power management > minimum processor state. Make sure that is set to 100% for plugged in.

Was it lower in your case?

aaronne
02-05-2014, 06:46 AM
Or "simply"disable all C states and Intel speedstep from bios settings. (So your cpu doesn't change its power state)

ChikNoods
02-05-2014, 06:48 AM
It is possible to have the power options set to high performance and have the minimum processor state below 100%. I did it manually, but I suppose software could have done automatically.

Go to Power Options > "Change plan settings" on the high performance row > Change advanced power settings > Processor power management > minimum processor state. Make sure that is set to 100% for plugged in.

Was it lower in your case?

all are set to 100% in processor power. I ran the test again and re-upped and its reporting 2593.8 MHz now -- not sure what happened

yes lower before. most of the G74sx report 798

Ambidexter
02-20-2014, 03:42 AM
Whether or not RB2.1 reports the CPU freq correct or not is a bit odd. But, what I wonder is does it affect the score? Do the same times and openCL score give us the same overall score?

With Intel Extreme Tuning Utility after benchmarks run, the score changes with the same test time, if the voltage is lower or the CPU is higher or using some other logic.

Actually, a post on this thread seems to get close to what IETU does. IETU's "benchmark" is just a 30 second or so test with 4 or 5 up/down cpu tests and a score that is returned which considers the max CPU, voltage, etc. So it's possible RB could do an initial "what's your CPU's speed?" test then use that in reporting results. I'd prefer this rather than having RB disable any max/min CPU I've selected in my power plan.

Thanks again for a great program!

Cheers,

-Ambi