Intel moves to throttle throttle-bottle story
Wotta lotta glottals. Stop.
Intel has taken speedy steps to clarify what it believes are the issues that lie behind the phenomenon of clock throttling.
Earlier this week, we suggested that describing notebook machines as running at 1GHz was misleading, and that in reality machines using a chip running at this speed only did so for a brief period of time, even when they were plugged into the mains.
PC companies, we said, use a technique called clock throttling to prevent a notebook from performing mundane tasks like frying eggs and bacon, rather than producing complex marketing plans and roadmaps in elaborate PowerPoint presentations, which is obviously what such expensive machines are meant for.
Intel SpeedStep technology, like rival systems such as PowerNow! from AMD, and LongRun from Transmeta, use different clock speeds to maximise battery life in notebook machines.
So, for example, a notebook that uses an Intel 1GHz microprocessor in reality runs at 700MHz. (Via's got one of these steppie things too, but we forget its name, for some reason).
But, it appears, according to Intel, clock throttling is a different thing again.
An official Intel response to our story said: "Clock throttling is a system-level mechanism to decrease the clock frequency of the processor when some trigger point (eg. a temperature) is reached.
"This technique is used by PC manufacturers as a fail-safe mechanism to protect the system and data in the unlikely event that temperature rises above the normal range, such as if the system is subjected to an unusually high-temperature environment or is running a power virus or similar code designed to artificially stress the CPU beyond normal levels.
"Clock throttling should not [Intel scream, not ours - Ed] be needed under normal operation. PC manufacturers developed state-of-the-art thermal solutions to ensure that notebooks run at the peak performance level under normal conditions.
"The mobile processor has significant differences from its desktop sibling. For example, the mobile processor can operate at a junction temperature of 100°C (vs. 70°C for desktop).
"The higher junction temperature allows for a smaller-sized thermal solution while ensuring that the processor runs at its full performance level in a notebook form factor. The processor also has power management states such as deep sleep and quick start, which enable low average power to increase battery life."
Let's have a breath of air. Did you get all that? Good. Intel does take exception to some figures we quote in our original piece.
The statement continues: "The TDP stated in your article states 34 Watts, this is incorrect and is actually 24.8W. Notebooks designed to accommodate this thermal design power level do not need to clock throttle when running even the most demanding of real-life applications.
"Regarding your comments on power saving vs frequency reductions. Average power is a metric for battery life. It is measured by running an industry battery life benchmark (Ziff-Davis battery mark) in battery optimised mode The battery optimised mode is the appropriate mode for measuring battery life. In the case of the Mobile Pentium III processor at 1GHz, it consumes less than 2W average power when in battery optimised mode (which is at 700MHz)."
To make this TPD stuff somewhat clearer, we here point to a helpful Intel chart in PDF form that some reader sent us. Intel is right about the 28.4W, for which we apologise, but the power figures in this table are somewhat interesting.
We cannot but help feel that comparing a 1GHz notebook with a 1GHz desktop and disregarding video, screen and hard drives as parameters would be a very useful exercise. If any hardware site out there wants to give it a try, and has a notebook machine with a 1GHz processor inside, please let us know.
That would be, like, a real world test without all of these square waves, TPDs and the like. The validity of such a test is that the chip firms, AMD and Intel amongst them, is attempting to convince us that desktops are things of the past. ®
Sponsored: The Nuts and Bolts of Ransomware in 2016