Power Management Efficiency Sandra 2007 with X2 3800 means?

[chinna_n]

Guys,

I built a system with X2 3800 with MSI K8NGM2-FID MB and have been playing around with Undervolting as well as Overclocking. I have been using RMClock to undervolt and power management.

I saw one feature in SiSoftware Sandra 2007 lite, which is Power Management Efficiency. So I started testing this and tried with different RMClock settings.

Luckily, My CPU can run at 10x (2000Mhz) with even 1.1v, but my MB does not give stable vCore ( it varies as much as 80mv, and esp goes as low as 50mv less than set voltage, and fluctuates a lot). So I was able run two cores fully loaded at 1.125v at 10x, but I kept a margin and running at 1.15v. And also my MB has problem with dynamically changing voltage by RMClock( Have not seen anyone who has success using RMClock on this MB with X2 proc ).

So, finally I have settled for 5x to 10x all steppings at 1.15v core. So as per my calculations at this voltage max wattage for my CPU is 47watts.

With these settings, Sandra Power Management Factor shows as 1.96 for this CPU and it is more than Core Duo or even P-M.!!? I am puzzled with this result. So, with this undervolting is it more efficient than even Core Duo? or eve P-M( for that matter all rerefence CPUs in that dropdown)!!?


Note: With only 5x and 7x and 10x steppings, it is giving a factor of 1.87.

Can someone explain what it means? and how come it is more than Core-Duo which take much less power than this( even with undervolting).?

See the image here

http://www.imageigloo.com/viewer.php?id ... x23800.PNG

[merlin]

To put it simply, it means absolutely nothing.

You have one major issue here, you're not measuring power used as a comparison at all. Sisoft Sandra cannot measure power usage, therefore you're really not gaining anything looking at that data. Get a hardware measurement tool to find out your actual power.

[chinna_n]

To put it simply, it means absolutely nothing.

You have one major issue here, you're not measuring power used as a comparison at all. Sisoft Sandra cannot measure power usage, therefore you're really not gaining anything looking at that data. Get a hardware measurement tool to find out your actual power.

I think you go it wrong here. Sandra can calculate CPU power. It actually measure voltage and clock speed at each stepping and calculates.

Ex: My other MC CPU P4 2.4B which uses 60watts at default voltage, the moment I OC, then Sadra changes it estimates which are verymuch in line with other power calculators for OC as well as Undervolting.

Yes, I know total power consumption is definitely more than typical PM platforn because MB and many other components. But CPU power Efficiency factor is calculated at different load factor and check what is CPU stepping and Voltage at that load. It also may have many other factors, but this is mainly about CPU power efficiency.

[Devonavar]

Yes, it can calculate CPU power. It can't measure it. That's a very key difference. If you want an accurate idea of how much power your CPU is consuming, there is no substitute for measurement.

There are many factors beyond just clock speed and voltage that determine how much power a processor uses, most of which cannot be taken into account without some form of measurement beyond just voltage and clock speed. Computers are simply not equipped to measure current, which is what you need to know with voltage to calculate power.

It is well known that AMD CPUs vary signficantly in the amount of power they consume, clock for clock and voltage for voltage. Intel's chips probably vary at least as much. Temperature also affects how much power is lost (since it affects current), as does exact load and the total number of transistors that flip per clock cycle.

What I'm trying to say, and I think what merlin is getting at, is that calculated values for CPU power consumption are far too unreliable to trust as more than a ballpark figure. You may get values that are in the correct order relative to each other, but you won't get an a result with any absolute accuracy.

For the record, ATI Tool has a similar feature that can "calculate" the current consumed by ATI's graphics cards. During our testing of the Aerocase Condor, we tested an X1900XTX that showed as consuming 30A, which, at 12V would mean it was consuming 360W. Total system power at that time, measured at the wall, was 230W. I'll believe my watt meter, thank you very much.

[chinna_n]

Yes, it can calculate CPU power. It can't measure it. That's a very key difference. If you want an accurate idea of how much power your CPU is consuming, there is no substitute for measurement.

There are many factors beyond just clock speed and voltage that determine how much power a processor uses, most of which cannot be taken into account without some form of measurement beyond just voltage and clock speed. Computers are simply not equipped to measure current, which is what you need to know with voltage to calculate power.

It is well known that AMD CPUs vary signficantly in the amount of power they consume, clock for clock and voltage for voltage. Intel's chips probably vary at least as much. Temperature also affects how much power is lost (since it affects current), as does exact load and the total number of transistors that flip per clock cycle.

What I'm trying to say, and I think what merlin is getting at, is that calculated values for CPU power consumption are far too unreliable to trust as more than a ballpark figure. You may get values that are in the correct order relative to each other, but you won't get an a result with any absolute accuracy.

For the record, ATI Tool has a similar feature that can "calculate" the current consumed by ATI's graphics cards. During our testing of the Aerocase Condor, we tested an X1900XTX that showed as consuming 30A, which, at 12V would mean it was consuming 360W. Total system power at that time, measured at the wall, was 230W. I'll believe my watt meter, thank you very much.

Actually, I was wondering if you guys even try to completely read and understand the Main thread, and what I am trying to understand.

I did not say X2 3800 is consuming less than Core Duo. I am talking about Power Management Efficiency. This is calculated by comparing the workload and current CPU speed and voltage and someother params by Sandra 2007. I did not makeup the Core Duo figures, it is the reference values given by Sandra.

My Question howcome it is showing more powermanagement efficiecy than a CoreDuo? That is question, not a statement. If guys knows anything about it, I would like to know about it. Not someother diversion to the actual thread purpose.

Thanks
Chinna

[chinna_n]

Yes, it can calculate CPU power. It can't measure it. That's a very key difference. If you want an accurate idea of how much power your CPU is consuming, there is no substitute for measurement.

There are many factors beyond just clock speed and voltage that determine how much power a processor uses, most of which cannot be taken into account without some form of measurement beyond just voltage and clock speed. Computers are simply not equipped to measure current, which is what you need to know with voltage to calculate power.

It is well known that AMD CPUs vary signficantly in the amount of power they consume, clock for clock and voltage for voltage. Intel's chips probably vary at least as much. Temperature also affects how much power is lost (since it affects current), as does exact load and the total number of transistors that flip per clock cycle.

What I'm trying to say, and I think what merlin is getting at, is that calculated values for CPU power consumption are far too unreliable to trust as more than a ballpark figure. You may get values that are in the correct order relative to each other, but you won't get an a result with any absolute accuracy.

For the record, ATI Tool has a similar feature that can "calculate" the current consumed by ATI's graphics cards. During our testing of the Aerocase Condor, we tested an X1900XTX that showed as consuming 30A, which, at 12V would mean it was consuming 360W. Total system power at that time, measured at the wall, was 230W. I'll believe my watt meter, thank you very much.

Actually, I was wondering if you guys even try to completely read and understand the Main thread, and what I am trying to understand.

I did not say X2 3800 is consuming less than Core Duo. I am talking about Power Management Efficiency. This is calculated by comparing the workload and current CPU speed and voltage and someother params by Sandra 2007. I did not makeup the Core Duo figures, it is the reference values given by Sandra.

My Question howcome it is showing more powermanagement efficiecy than a CoreDuo? That is question, not a statement. If guys knows anything about it, I would like to know about it. Not someother diversion to the actual thread purpose.

Thanks
Chinna

[Devonavar]

As it happens, I did read the whole thread.

Could you explain to me what "Power Management Efficiency" is if not power consumption? And, if it is not, why should I care about it? What exactly is SiSoft calculating here? Unless it's power consumption, I agree with merlin, it's absolutely meaningless.

And, if it *is* power consumption, I stand by my original statement. Measure power, don't try to calculate it.

For the record, actual measurements do seem to show that AMD's chips consume less power than Core 2 Duo chips in idle, so it doesn't surprise me to learn that SiSoft considers the X2 more efficient in idle. Unfortunately, without knowing more about exactly which factors are considered in calculating "Power Management Efficiency", I can't really say more than that.