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).
come guys, comment, so I know how excited I should be
My suspicion(if only for die-space reasons, it isn't purely cosmetic that contemporary VRMs occupy a substantial amount of board space) is that is this a 'marketecture' summary of Intel moving some additional voltage adjustment and power gating functions on die, to support dynamic adjustment of power to the greater number of components(multiple CPU cores, possibly independently clocked, GPU, RAM controller, PCIe root, etc.); but we'll still see a bunch of chunky power silicon under serious heatsinks clustered around the CPU socket.
Given that much of the contemporary power savings are achieved by superior idling, rather than absolute gains in maximum power draw, Intel is either going to have to keep moving power regulation on die, or start dedicating even more pins to tiny voltages at nontrivial currents, with the associated resistive losses; but that won't necessarily change the fact that the circuitry that brings the 12v rail down to what the CPU wants is a pretty big chunk of board.
You can find the full slide set in PDF format here.
If I read this right, it really is a fully on-chip switching regulator, inductors and all. They already have a test chip that they used to power a ~90W Xeon E7330 for four hours while it ran Linpack. (Or a virus -- it says Linpack in the summary page.) Voltage ripple is less than 2mV. Peak efficiency per cell looks like ~76% at 8A. They claim hitting 82% would be easy, and there are "additional advancements that cannot be reported at this time" planned for the future.
The slides have bunch of other technical details about testability features, too, which is always neat to see.
Visit the
Can someone please tell me why this is a good idea
The long story is here (PDF). Motherboard will still do the heavy lifting from 12V to 2.4V, but the integrated VRM will distribute it. Advantage is extremely clean, fine-grained, low-latency and flexible power supply to deliver exactly as much power to where it's needed and probably - this is just speculation on my part - allowing the CPU to work on a wider range of voltages since there's less noise and ripple so you don't need the same tolerance limits. It sounds perfect for smart phones, tablets and laptops that are primarily battery-limited, nice to have for average machines but potentially an issue for overclockers. All you need is cooling though, it shouldn't limit overclocking if you can keep the temp down.
Live today, because you never know what tomorrow brings
you can't escape Ohm's Law.
Actually you can. It's called a switching power supply.
In other news, a Nobel Prize in Physics was awarded to Anonymous Coward of Slashdot today, after discovering that the laws of physics do not apply to switching power supplies... His next research proposal is on solving the energy crisis by designing keyboards to detect when someone is angry and then increasing the key resistance by piezoelectric effect to generate energy. While it would generate only marginal amounts of power when used by 99.975% of the population, it was recently discovered that the remainder are actually Linux and Apple fanboys who, if fed a regular diet of dismissives via their computer screen, will so furiously hit the keyboards that power for entire cities is easily achievable.
#fuckbeta #iamslashdot #dicemustdie
That's some amazing work. The current state of the art in CPU power supply designs hasn't changed in 15 years. 12V in, low voltage out, and the output voltage has been moving lower and lower for years, with designs below 1 V. If you figure you had a few percent of tolerance in the early years when everything ran off 2.5V and that few percent remains constant, then at 1 V you have almost no room for slop. So there are a lot of output capacitors there, both those electrolytics (you always hear people complaining about them but they're CHEAP) and ceramics. The ceramics cost a fortune and you need a lot of them to get your tolerance down - the first half microsecond of a load step is entirely the ceramic capacitor's response, not the controller or anything else. Moving part of the VR onboard allows them to reduce the parasitics significantly and they can probably tolerate a little higher tolerance as a result, but moreover they can get rid of some of those ceramics in the whole system - ultimately many of those on the motherboard.
So this is taking a lot of cake out of company mouths. Analog, Intersil, IRF, ON, who else - manufacturers of controllers, MOSFETs. Inductors, ceramic and 'lytic vendors are all going to lose out a bit here. Potentially Intel can reduce the platform cost vs. AMD as well, which is interesting. There is still an onboard VR but it will be 12 - 2.4 V, wherever they think the sweet spot is for efficiency and size. And the first real change in this industry for a long time. Cool work.
Just my $0.55 (US inflation, 1774-2008, for $0.02)
It's a switch-mode (Buck) regulator. You can tell from the efficiency curve and the fact that it requires an inductor. It is more efficient than a linear regulator and less efficient than a good external Buck regulator. However, being on-chip it will regulate the voltage better because there won't be significant I*R drop between the regulator output and the load. And as they mention, the cooling fan will be right on top of it, so it is more effectively cooled than an external regulator typically is.
Ohm's law is completely irrelevant to this situation *in the form you describe*. "Burning a hole through the board" would be possible and a simple function of Ohm's law only if they were using a linear regulator to generate the Vcore. But VRM's have been switching DC/DC converters since the 486 days. They achieve a voltage conversion by switching the incoming voltage on and off *very fast*, which results in an output voltage that's a function of the input voltage and the duty cycle of the on/off switching. An inductor (current-smoothing) and capacitor (voltage smoothing) give a nice clean DC voltage.
The differences between on-motherboard VRMs and this new in-package (it's technically a separate chip...) are significant. First off, physically moving it closer means that you're not sending 100+ Amps of current over the 3-4 centimeters of generally very thin copper traces on the PCB, they're sent millimeters through die-bond wires, or even through a solid substrate (no idea what Intel does at that level). There's your Ohm's law coming into play at that level, but the power losses there are relatively minimal since you're talking maybe a few tenths of an ohm. Die-bond wires are going to drop that to 10's of milli-ohms probably, so nothing major but still a positive effect.
The main reason this will generate a lot less heat is because of the *frequency* of the switching. Because this on-board VRM is so much smaller, it can switch the input faster (shorter wires, less parasitic capacitance, less ringing, etc.). This in turn means smaller value components required, e.g. the switch from the monster inductors seen on the motherboard (at maybe 1-2MHz switching) in the slide to the tiny chip-scale inductors on the FIVR (at 10's or 100's of MHz). The end result of all of this is that switching losses get significantly smaller. It's those losses that create heat local to the regulator. If they can for example go from an 80% efficient VRM to an 90% efficient FIVR for a 100W CPU load, they reduce the switching losses from 25W to 11.1W.
GStreamer - The only way to stream!
If you core requires 1V and 90 watts you need to transfer 90A through your PCB traces, up in to the chip, across the bond wires (if there are any) and on to the die.
If your die has a regulator on board and accepts 12V instead, and is 80% efficient you only need to transfer 9.4A. You've just lowered your resistive losses by about 100x. If the connection between the external VRM and die is 0.001ohms, at 90A you waste 8.1W. at 9.4A you waste 0.088W.
Intel might be the first to do it on a CPU die, but they're not the first to do on-silicon inductors by any stretch. Switching regulators with inductors on silicon have been commercially available for several years now. The R-78 and MIC33030, for example, are drop-in replacements for linear regulators, with all components on die.
The real question in my mind is why anyone still uses linear regulators for anything, but I digress.
Check out my sci-fi/humor trilogy at PatriotsBooks.