Slashdot Mirror
Intel's Haswell Moves Voltage Regulator On-Die
MojoKid writes "For the past decade, AMD and Intel have been racing each other to incorporate more components into the CPU die. Memory controllers, integrated GPUs, northbridges, and southbridges have all moved closer to a single package, known as SoCs (system-on-a-chip). Now, with Haswell, Intel is set to integrate another important piece of circuitry. When it launches next month, Haswell will be the first x86 CPU to include an on-die voltage regulator module, or VRM. Haswell incorporates a refined VRM on-die that allows for multiple voltage rails and controls voltage for the CPU, on-die GPU, system I/O, integrated memory controller, as well as several other functions. Intel refers to this as a FIVR (Fully Integrated Voltage Regulator), and it apparently eliminates voltage ripple and is significantly more efficient than your traditional motherboard VRM. Added bonus? It's 1/50th the size."
Update: 05/14 01:22 GMT by U L : Reader AdamHaun comments: "They already have a test chip that they used to power a ~90W Xeon E7330 for four hours while it ran Linpack. ... Voltage ripple is less than 2mV. Peak efficiency per cell looks like ~76% at 8A. They claim hitting 82% would be easy..." and links to a presentation on the integrated VRM (PDF).
Considering that they've already started shipping an actual product, perhaps you should switch modes--from skeptic to sleuth. Start from the proposition that (a) it's possible or (b) they're leaving something out of the marketing jargon. There are a million ways they could do it wrong, and likely only a few ways to do it right. If you start from the proposition that Intel is shipping a working product, then it should be much easier to figure out.
Actually, no real device is ohmic at all. Even a resistor will heat up with increasing current causing an increased resistance that is non-linear. For us EEs, ohmic devices are our massless pullies and frictionless inclines.
I'm not a certified EE, but I have built electronic circuits. I know there's a lot of ways to 'cheat' on paper; switching power supplies don't get rid of ohm's law though, they're simply more efficient. Ohm's law is about the relationship between resistance, voltage, and current. Those relationships are derived from the physics about electron exchange between different materials. Now yes, capacitors and inductors both run 90 degrees out of phase between voltage and current so it can appear to be violating ohm's law, but if you apply a correction factor you'll see it's pretty close to parity. When you get down to really small discrete components, like a transitor for example, measurement inaccuracy and time domains will really start to screw with you, but ohm's law still holds even down to that scale.
Ohm's law is the reason for these changes Intel is making: An attempt to remove parasitics from the circuits, which all boil down to resistance; Whether it's phase-shifted forward because of capacitors, or backwards because of inductors, or because of components that create those effects, doesn't really matter.
Now you're right, a purely ohmic device doesn't exist. Even resistors can generate small amounts of phase shift. But that doesn't make them "massless pullies" or "frictionless inclines". Ohm's law is still useful for the same reason the OSI 7 layer model is still useful, despite no network yet having been designed that perfectly adheres to it...
#fuckbeta #iamslashdot #dicemustdie
The voltage regulation issue can easily be solved by having a feedback connection from the die to the external VRM.
There are only two benefits I can see:
1) Higher voltage in to the chip means lower current, which saves power. You I*R formula is slightly wrong, its actually I^2 * R, double the current means 4x the power loss.
2) Lower system cost. the more crap that gets stuffed on the die/in the chip, the less is required on the board. That means fewer components, smaller board area and quicker assembly.
There are of course other benefits that only benefit Intel
a) Fewer external components means they can charge more for their chip without effecting system cost.
b) smaller system = happier customer = will pay more
c) If it does actually result in lower power, then you get more performance or more battery life = customer will pay more
If you'd read at least the summary, the benefit would be less ripple. Because it takes time to get the feedback voltage to the external VRM, there would always be ripple if power demands would fluctuate fast enough. In a typical CPU on a typical load, you get a lot of power load changes, so you'd get a lot of ripple. Ripple means that ultra low power circuitry will be harder to implement and hit limits earlier, since it is more dependent on precise voltages.
Power saving wouldn't be relevant, if you are looking at the power loss in the circuit board traces to the CPU. The efficiency of the internal regulator is lower than that of external voltage regulators so it would probably consume even more power.
System cost would be higher. Other components on the main board still require regulated voltages, so no components would be saved there.
I was promised a flying car. Where is my flying car?