![[H]ardOCP](/images/blank.gif){kind=small}
- Date:
- Sunday , October 28, 2007
- Author:
- Kyle Bennett
- Google +1
Intel's New 45nm Yorkfield QX9650
Intel marches forward with another groundbreaking processor. Four cores of 45nm goodness. At this rate, you have to wonder whether or not desktop software and AMD will ever catch up. How good does it overclock? All signs point to, "Wow!"
Penryn Power
Of course when moving from a 65nm to a 45nm microprocessor architecture process, one of the first thing comes to an enthusiast’s mind is the power savings….or more importantly a cooler running part that can possibly deliver better overclocking headroom.
On our ASUS motherboard, our QX9650 defaulted to 1.2v core voltage. Keep in mind that we are now powering 820 million 45nm transistors in comparison to the QX6850’s 582 million 65nm transistors.
For load variations we used a program called OCCT. This program easily loads all four cores. Here is a HardForum thread in which the software’s author is very open and answered many questions about it earlier this year. For our testing we simply toggled on and off cores through the Windows Task Manager Processor Affinity setting. For our quad core CPUs: 1 core loaded = 25%, 2 cores loaded = 50%, 3 cores loaded = 75%, and 4 cores loaded = 100%. For our X6800 dual core CPU: 1 core loaded = 50%, and 2 cores loaded = 100%. Idle is simply 0% CPU Usage with the system sitting at the desktop.
There a certainly a couple of things that stick out like a sore thumb here…
IDLE
Idle wattage usage on the QX9650 system simply blows the other two out of the water. Our QX9650 is a full 24w below the QX6850, and 30w below the X6800. Intel has learned to implement better power gating with the Penryn architecture and that will come through in our scaling discussion as well.
Power Scaling
Jumping ahead to a full 100% processor load, you would be hard pressed to identify any of our processors by simply staring at our power analyzing equipment. In fact, I would suggest that all three 100% load scores fall within the margin of error in our testing. According to Intel our TDPs are 130w for the QX9650, 130w for the QX6850, and 75w for the X6800. Going back and checking our previous X6800 power testing, we saw the exact same scaling from idle to 100% load so we are confident it is correct. Of course this might leave you wondering why the 100% load numbers for the quad-core processors is not higher. I think the fact of the matter is that the TDP numbers quoted by Intel (and AMD for that matter) have little to do with wattage loads that enthusiasts are likely to see.
One of the biggest buzz words in the industry that I have been hearing for the last couple of years is “granularity.” Unlike its older sibling, the QX9650 scales load and power in a fairly tight ratio compared to the QX6850 that looks to be either turned fully on after it loads 1 core. Certainly the QX9650 processor’s power scaling is on a finer “granular” level that what we have seen from Intel in the past.
Considering that most computers sit in a state of 0% to 20% usage the most of the time it is turned on, this lower idle power usage and finer grained power scaling is most welcome on the Penryn. Simply put, Penryn saves a good bit of power at idle compared to older high end Intel processors.
Enthusiast Power
On a bit more robustly configured system, one with an 8800 GTX, Koolance Exos Water Cooler, and 3 Western Digital Hard Drives, as well as a few high powered case fans plugged in, we are recording a wattage usage of 290w (with no 3D applications being run) being pulled from the PSU at 100% CPU load. Loading 3 of the QX9650 cores with OCCT and leaving one for the Crysis Demo’s built-in GPU test, we see the system pull of a total of 334w with the system running at 4GHz on a 1333MHz FSB and core voltage of 1.45v. (Again, this is wattage being pulled from the PSU, not wattage measured at the wall before the PSU.)