Penryn dual-core power figures preview - looking very good

All about them.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

Post Reply
smilingcrow
*Lifetime Patron*
Posts: 1809
Joined: Sat Apr 24, 2004 1:45 am
Location: At Home

Penryn dual-core power figures preview - looking very good

Post by smilingcrow » Wed Aug 08, 2007 9:58 am

HKEPC has a preview of Penryn dual-core and the power and temp figures look strange.

Idle - No EIST
Core 2 Duo E6550 - Power - 61W , Temp - 41C
Penryn 2.33GHz ES - Power - 43W , Temp - 31C

Max Power
Core 2 Duo E6550 - Power - 83W, Temp - 49C
Penryn 2.33GHz ES - Power - 59W, Temp - 37C

Gigabyte GA-P35T-DQ6 (P35 + ICH9R + DDR3)
Gigabyte GA-G33-DS3R (G33 + ICH9R)
ADATA DDR3-1066 CL 6-6-6-18 1GB x 2
MSI GeForce 8800Ultra VGA Card
Gigabyte ODIT 800W Power Supply

The figures sound too good to be true and looking at the list of components used its unclear if both systems were identical.

krille
Posts: 357
Joined: Thu Mar 23, 2006 4:56 am
Location: Sweden

Post by krille » Wed Aug 08, 2007 12:49 pm

It's not unclear. They were identical. The power consumption figures are from a Gigabyte Odin PSU. The system used was a G33 based one. That's why they're so low. Actual CPU power consumption is even lower. It's unclear however whether power consumption includes PSU inefficiencies. If so, remove some 20% from those figures.

tehfire
Posts: 530
Joined: Mon Jan 01, 2007 9:57 am
Location: US

Post by tehfire » Wed Aug 08, 2007 1:16 pm

Well with the low load wattage I'd imagine they're only talking about the wattage of the actual CPU. How they did that I have no idea. If in fact they're only talking about the CPU load than the setup of the rest of the system would have no bearing on the wattage levels. On performance, yes, but shouldn't have an effect on the CPU wattage.

krille
Posts: 357
Joined: Thu Mar 23, 2006 4:56 am
Location: Sweden

Post by krille » Wed Aug 08, 2007 2:17 pm

tehfire wrote:Well with the low load wattage I'd imagine they're only talking about the wattage of the actual CPU. How they did that I have no idea. If in fact they're only talking about the CPU load than the setup of the rest of the system would have no bearing on the wattage levels. On performance, yes, but shouldn't have an effect on the CPU wattage.
No, read my post. It's for the entire system.

tehfire
Posts: 530
Joined: Mon Jan 01, 2007 9:57 am
Location: US

Post by tehfire » Wed Aug 08, 2007 4:06 pm

An entire system including an 8800Ultra using 59W at full tilt? Are we talking about two different things? B/c I don't see that happenning...

accord1999
Posts: 84
Joined: Thu Dec 01, 2005 7:39 pm

Post by accord1999 » Wed Aug 08, 2007 4:54 pm

They switched to the G33 system with onboard video for the power consumption tests.

Matzon
Posts: 35
Joined: Sun Jul 22, 2007 12:31 am

Post by Matzon » Wed Aug 08, 2007 10:16 pm

I dont understand why the idle temperatures are so high. The 6550 is 41C, my 6750 idles at 24C right now (ambient ~22)?

smilingcrow
*Lifetime Patron*
Posts: 1809
Joined: Sat Apr 24, 2004 1:45 am
Location: At Home

Post by smilingcrow » Wed Aug 08, 2007 11:21 pm

Matzon wrote:I dont understand why the idle temperatures are so high. The 6550 is 41C, my 6750 idles at 24C right now (ambient ~22)?
They disabled EIST which is one reason. If you just focus on the difference between the two CPUs that they tested you’ll see that the difference is significant. It can be meaningless to compare temps between systems that use different coolers and temp sensors.

I still don’t know what the data relates to but it confirms what Intel has been claiming in that they have dramatically reduced static power leakage. With EIST disabled the difference in power draw is 18W which seems very significant. It helps to mitigate against the fact that these 1333MHz FSB desktop chips idle at 2GHz even with Speedstep enabled.
I wonder how low the power draw will be with the mobile Penryn that idles at 600MHz and has the new C6 sleep state.

The 59W figure for load can’t relate to system power draw, it’s just too low.

Matija
Posts: 780
Joined: Sat Mar 17, 2007 3:17 am
Location: Croatia

Post by Matija » Wed Aug 08, 2007 11:24 pm

Matzon wrote:I dont understand why the idle temperatures are so high. The 6550 is 41C, my 6750 idles at 24C right now (ambient ~22)?
Because your temperature measurement is incredibly wrong ;)

jojo4u
Posts: 806
Joined: Sat Dec 14, 2002 7:00 am
Location: Germany

Post by jojo4u » Thu Aug 09, 2007 3:35 am

smilingcrow wrote:They disabled EIST which is one reason.
Don't forget about C1E.

krille
Posts: 357
Joined: Thu Mar 23, 2006 4:56 am
Location: Sweden

Post by krille » Thu Aug 09, 2007 3:41 am

Matija wrote:
Matzon wrote:I dont understand why the idle temperatures are so high. The 6550 is 41C, my 6750 idles at 24C right now (ambient ~22)?
Because your temperature measurement is incredibly wrong ;)
Not necessarily. It's all about ambient temps and cooling. You can't compare different settings here, just different results under virtually the same settings.

smilingcrow
*Lifetime Patron*
Posts: 1809
Joined: Sat Apr 24, 2004 1:45 am
Location: At Home

Post by smilingcrow » Thu Aug 09, 2007 4:22 am

jojo4u wrote:Don't forget about C1E.
I deduce that C1E is also disabled judging by the data. For an E6550 to only consume an extra 22W at load compared to idle suggests this to me.
Matzon wrote:I dont understand why the idle temperatures are so high. The 6550 is 41C, my 6750 idles at 24C right now (ambient ~22)?
I assume your 24C relates to the motherboard CPU sensor rather than the DTS unless you have top end cooling! The DTS readings are the only ones that are directly comparable across different systems.
I don’t remember seeing a DTS reading so close to ambient. I tested an E4400 at 1.2GHz 1.136V and it was more than 10C over ambient at idle, with admittedly a low end cooler.

Matija
Posts: 780
Joined: Sat Mar 17, 2007 3:17 am
Location: Croatia

Post by Matija » Thu Aug 09, 2007 4:39 am

For a CPU temperature of 2 degrees over ambient, and using a figure of 0.2 C/W (Ninja with a fast fan), we get something simple:

2C / 0.2 C/W = 10W

Mmmh... Not really possible.

ryboto
Friend of SPCR
Posts: 1439
Joined: Tue Dec 14, 2004 4:06 pm
Location: New Hampshire, US
Contact:

Post by ryboto » Thu Aug 09, 2007 7:40 am

Well, a user in the overclocker.net forum is telling everyone Penryn can do 3+ghz on ~0.98vcore. So, if the article he sites is correct, and not a typo, these power figures fit.

Matzon
Posts: 35
Joined: Sun Jul 22, 2007 12:31 am

Post by Matzon » Thu Aug 09, 2007 1:40 pm

I am using a ninja for cooling and 2x120mm Silent Eagle (push / pull) ~1000 rpm + 2x92mm nexus intake in a solo case

Image

I would love to know why I get so low temperatures my self!!

If TAT is the wrong tool, then fine - how do I then measure the correct temperature ?

smilingcrow
*Lifetime Patron*
Posts: 1809
Joined: Sat Apr 24, 2004 1:45 am
Location: At Home

Post by smilingcrow » Thu Aug 09, 2007 2:19 pm

Matzon wrote:I am using a ninja for cooling and 2x120mm Silent Eagle (push / pull) ~1000 rpm + 2x92mm nexus intake in a solo case
I would love to know why I get so low temperatures my self!!
If TAT is the wrong tool, then fine - how do I then measure the correct temperature ?
What frequency and VCore is your system idling at? I’m also curious what temps/frequency/Vcore you get at load!

Matzon
Posts: 35
Joined: Sun Jul 22, 2007 12:31 am

Post by Matzon » Thu Aug 09, 2007 2:31 pm

it's default vcore (1.35 ?)
and using 8 multiplier @ 400 MHz

the temperature seems to top out at 53C

Image

No other components are OC'ed - I tried with my ASUS Silent 8600 GTS, but failed miserably.

cmthomson
Posts: 1266
Joined: Sun Oct 09, 2005 8:35 am
Location: Pleasanton, CA

Post by cmthomson » Thu Aug 09, 2007 3:37 pm

TAT (and all other DTS-based programs) have a built-in assumed throttling temperature for the CPU. Various TAT/CPU combinations use either 85C or 100C. It's entirely possible that your reported temperatures are 15C too low.

How is this possible and why does it matter?

Well, DTS reports the delta below throttling (ie, the headroom) so the platform can increase fan speeds as the headroom shrinks. The throttling temperature is not accessible from software, and can be different for each CPU variant, or even lot. In an attempt to provide customary reports, software such as TAT adds the DTS value (which is always negative) to the assumed throttling temperature and reports the sum as the absolute (assumed) temperature.

AMD has used a similar CPU-specific approach for a while (they call it TDP), but they make the value accessible so enthusiast software can report accurate as opposed to assumed values.

Matzon
Posts: 35
Joined: Sun Jul 22, 2007 12:31 am

Post by Matzon » Thu Aug 09, 2007 10:08 pm

So in effect, I can never know the real value? - I can add 15, but that might not be entirely true?
What about the above penryn temps, should we add 15 to those too?
Since my processors is specced to 72C max, does that mean 57C in TAT? (that would mean ~47C for the older C2D version, waay too low?)

Sorry for all of the questions, but it seems very confusing to report a value in TAT that really isn't usefull...

Using Core Temp, I consistently get a value that is 3C lower than TAT

DTS info from TAT:
=== Digital Thermal Sensor Info ===
==Processor 0 details== :
Threshold#1 : Disabled
Threshold#2 : Disabled
High Temperature Interrupt Enable : Disabled
Low Temperature Interrupt Enable : Disabled
THERMTRIP# Interrupt Enable : Disabled
FORCPR# Interrupt Enable : Disabled
Overheat Interrupt Enable : Disabled
==Processor 1 details== :
Threshold#1 : Disabled
Threshold#2 : Disabled
High Temperature Interrupt Enable : Disabled
Low Temperature Interrupt Enable : Disabled
THERMTRIP# Interrupt Enable : Disabled
FORCPR# Interrupt Enable : Disabled
Overheat Interrupt Enable : Disabled
So, what I am still confused about: Are my values correct, or should I add 15C ?

smilingcrow
*Lifetime Patron*
Posts: 1809
Joined: Sat Apr 24, 2004 1:45 am
Location: At Home

Post by smilingcrow » Fri Aug 10, 2007 12:53 am

Matzon wrote:it's default vcore (1.35 ?) and using 8 multiplier @ 400 MHz
the temperature seems to top out at 53C.
So, what I am still confused about: Are my values correct, or should I add 15C?
There’s no way you are getting an absolute temperature at idle of 2C above ambient at 3.2GHz and 1.35V so you will need to determine what the offset is and add that to the readings.
Matzon wrote:What about the above penryn temps, should we add 15 to those too?
It’s impossible to say until we know the thermal cut-off values for Penryn. This does mean that the data in the Penryn preview comparing the temperatures of Conroe v Penryn is fairly meaningless without further clarification. The one value that stands up without needing clarification is the difference in temperature between the idle and load states for each CPU. For Conroe it was 8C and Penryn was 6C.
Matzon wrote:Since my processors is specced to 72C max, does that mean 57C in TAT? (that would mean ~47C for the older C2D version, waay too low?)?
72C is the Thermal Specification for your CPU which is a separate value from that needed to calculate the actual temperature from the DTS data. The Thermal cut-off threshold for your CPU will likely be in the range 85 – 100C.

Thermal Specification: The thermal specification shown is the maximum case temperature at the maximum Thermal Design Power (TDP) value for that processor. It is measured at the geometric center on the topside of the processor integrated heat spreader.
Matzon wrote:Using Core Temp, I consistently get a value that is 3C lower than TAT?
CoreTemp should show the value that it uses to calculate the absolute temperatures. What is it showing for your CPU?
Matzon wrote:So in effect, I can never know the real value? - I can add 15, but that might not be entirely true?
It’s time that people rethink how they perceive CPU temperatures and GPU temps for that matter. The absolute value is not as important as how much headroom there is before thermal throttling occurs. This thermal threshold can vary considerably so rather than risking getting the correct value for your CPU wrong it’s easier to just work with the offset value.
I wish Speedfan had the option to display the offset value and use that as the basis for controlling fan speeds.

You can calibrate your system by allowing it to run hot to the point where it thermally throttles and note what the temperature is; RMClock can report when this occurs. If throttling occurs at 85C and your system when using its normal cooling hits 65C under load, then you know you have 20C of headroom.

Post Reply