Forgot your password?
typodupeerror
Intel Hardware

Intel's Haswell Moves Voltage Regulator On-Die 237

Posted by Unknown Lamer
from the march-of-progress dept.
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).
This discussion has been archived. No new comments can be posted.

Intel's Haswell Moves Voltage Regulator On-Die

Comments Filter:
  • excited (Score:5, Funny)

    by Anonymous Coward on Monday May 13, 2013 @07:29PM (#43715815)

    come guys, comment, so I know how excited I should be

  • sinking heat? (Score:4, Interesting)

    by p51d007 (656414) on Monday May 13, 2013 @07:31PM (#43715829)
    with the on die regulator, won't that area of the chip be a tad warmer than the rest of the chip, or will the heat be a moot point?
    • by kasperd (592156)

      with the on die regulator, won't that area of the chip be a tad warmer than the rest of the chip, or will the heat be a moot point?

      The summary says 76% efficiency. That would mean 24% of the energy you put into the chip is turned into heat by the voltage regulator. Sounds like a valid concern to me.

  • Heat (Score:4, Interesting)

    by girlintraining (1395911) on Monday May 13, 2013 @07:32PM (#43715841)

    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."

    I have yet to come across a voltage regulator that doesn't run hot. Typically, it's one of the hottest components in an electrical circuit. And we're integrated this into a slab of silicon already well-known for getting so hot it can catch fire?

    Can someone please tell me why this is a good idea, because all of my experience in electrical engineering says that when things heat up, they become more unstable and prone to failure, and the one thing you do not want going critical is your voltage regulator. If that goes, the whole computer catches fire.

    • Re: (Score:3, Interesting)

      by Anonymous Coward
      Most likely, because it's integrated into the CPU itself, the voltage regulator can be made more efficiently and thus save power and heat etc. Discrete parts have their limitations, and doing it on-die might just mitigate that.
      • Re: (Score:2, Informative)

        by Anonymous Coward

        Sorry, but I design quite a lot of switch mode regulators for my own hardware design, and there are several concerns here:
        - the efficiency of 76% they claim is abysmally low, my regulators are never below 80% in their operating range, and often above 90%, with peaks in the 96-97% range.
        - switch mode regulators need an inductor, and inductors need ferrite or iron cores, which is not going to happen on a silicon die. External inductors are much better
        but low loss inductors for large currents are large, for fu

      • Most likely, because it's integrated into the CPU itself, the voltage regulator can be made more efficiently and thus save power and heat etc. Discrete parts have their limitations, and doing it on-die might just mitigate that.

        I'm not sure that follows. For transistors doing computation, the efficiency saving is by reducing the amount of current that they have to switch. In this case, all they're really doing is building a very large MOSFET onto the die itself along with a bit of other gubbins.

        Also, the eff

    • Re:Heat (Score:4, Informative)

      by Anonymous Coward on Monday May 13, 2013 @07:39PM (#43715881)

      You're going to run into that heat anyway, whether it's on the motherboard in general or on the CPU. You can't win. But at least it's better to have heat build-up near a heat-sink, so for high-power conditions it might actually be better to put it on the CPU. I'm also an electrical engineer, but thermals are really a mechanical engineer's realm, so I can't run numbers for you.

    • by rrhal (88665)

      Being 1/50th the size it will be welcome on mobile devices. Not sure that its a good thing for your gaming desktop.

      • Re:Heat (Score:5, Informative)

        by Shavano (2541114) on Monday May 13, 2013 @08:35PM (#43716183)

        Being 1/50th the size it will be welcome on mobile devices. Not sure that its a good thing for your gaming desktop.

        That 84 watts is going to rip through your mobile device's battery pretty damn fast.

        • Being 1/50th the size it will be welcome on mobile devices. Not sure that its a good thing for your gaming desktop.

          That 84 watts is going to rip through your mobile device's battery pretty damn fast.

          Don't we need to compare it to a traditional regulator implementation before we come to that conclusion? Assuming pretty damn fast means faster than current Atom based devices.

      • Being 1/50th the size it will be welcome on mobile devices.

        It's not clear how they measure "1/50th the size". I could be wrong but it sounds like marketing hype. With a switching regulator the inductors and capacitors generally take up much more real estate than the chip. If they have some magic way to reduce the inductor and capacitor sizes it isn't mentioned in the article (and that would be a much bigger deal than just putting the regulator on the die).

        • Re:Heat (Score:5, Interesting)

          by petermgreen (876956) <plugwash.p10link@net> on Monday May 13, 2013 @09:07PM (#43716331) Homepage

          You can reduce the inductor and capacitor sizes a lot by increasing the switching frequency. Of course doing so will likely increase your switching losses but it may still be worth it if it lets you put the regulator closer to the load. Especially given the ever lower voltages that modern chips are running at.

          • Re:Heat (Score:5, Interesting)

            by ebno-10db (1459097) on Monday May 13, 2013 @09:13PM (#43716373)

            You can reduce the inductor and capacitor sizes a lot by increasing the switching frequency.

            But you can do that w/ an external regulator too. Apparently the secret is on-chip inductors. Now that's impressive. I'm surprised that some of the "analog" companies making switchers didn't come up with that first. I know Intel has good fab tech, but this seems more like the sort of funky thing analog guys would come up with first.

            http://www.psma.com/sites/default/files/uploads/tech-forums-nanotechnology/resources/400a-fully-integrated-silicon-voltage-regulator.pdf

            • Re:Heat (Score:5, Informative)

              by dgatwood (11270) on Tuesday May 14, 2013 @01:09AM (#43717419) Journal

              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 [digikey.com] and MIC33030 [micrel.com], 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.

              • That R-78E3.3-0.5 looks useful... the raspberry pi uses a rather inefficient linear regulator to take the 5v input down to 3.3v. An NCP1117. Can't just replace it with an R-78, as the minimum input voltage is 6V, but if you're running it off of something higher... yes. This component may be of use to me.

            • by AmiMoJo (196126) *

              Actually the analogue guys like Texas and Linear do make switching regulators with on-board inductors that are pretty small. Small enough that they could be integrated into a typical desktop CPU form factor.

              Unfortunately TFA is light on details. 1/50th the size of what? A mains transformer? A SOT-23 IC?

            • But you can do that w/ an external regulator too. Apparently the secret is on-chip inductors. Now that's impressive. I'm surprised that some of the "analog" companies making switchers didn't come up with that first. I know Intel has good fab tech, but this seems more like the sort of funky thing analog guys would come up with first.

              They did.

              I did a semiconductor course on analog CMOS as an undergrad over 10 years ago. It's interesting because the CMOS processes are typically adapted for digital electronics,

        • Re:Heat (Score:5, Interesting)

          by viperidaenz (2515578) on Monday May 13, 2013 @10:20PM (#43716795)

          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.

    • Re:Heat (Score:5, Informative)

      by fuzzyfuzzyfungus (1223518) on Monday May 13, 2013 @07:39PM (#43715891) Journal

      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.

      • 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.

        That's the only plausible thing I could come up with as well. The control logic could go into the CPU, but I don't see how pulling 12V down to fractions of a volt is going to happen on the die itself without it burning a hole through the board; heatsink or not, you can't escape Ohm's Law.

        • Who said they're going to keep it at 12 volts? A VRM can also be the transformer, but it doesn't have to be. They could ramp down the voltage on the board just like they always did and just have a VRM on the chip that is maintaining a steady voltage. If you read the article they are barging about how little fluctuation they are getting. So it seems like what they are doing here is adding basically an extra regulator on chip so they can have extremely stable voltage. I'm guessing as small as things are getti

          • by fluffy99 (870997)

            Who said they're going to keep it at 12 volts? A VRM can also be the transformer, but it doesn't have to be. They could ramp down the voltage on the board just like they always did and just have a VRM on the chip that is maintaining a steady voltage. If you read the article they are barging about how little fluctuation they are getting. So it seems like what they are doing here is adding basically an extra regulator on chip so they can have extremely stable voltage. I'm guessing as small as things are getting now, just the trip across the motherboard can have noticeable fluctuations on supply voltages due to EM interference and temp fluctuations, so having a regulator on the chip lets them get more precise. Maybe that precision will let them do some magic on the chip to increase performance or something?

            Or they are finding those Chinese made MB motherboards have such poor regulators, that they have to do some final regulation on the die to keep things stable?

            This is probably also driven by a desire to start putting higher horsepower CPUs into smaller things like tablets.

        • Re:Heat (Score:5, Informative)

          by Omega Hacker (6676) <omegaNO@SPAMomegacs.net> on Monday May 13, 2013 @09:35PM (#43716509)

          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.

          • by fluffy99 (870997)

            . 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).

            From the linked pdf - Programmable switching frequency 30MHz to 140MHz

        • by swalve (1980968)
          Sure you can. Voltage regulators aren't just resistors any more. You divide the 12v @ 1 amp into 1v @ 1amp at 12 different spots. More or less. Think of it like TDMA, if that helps. Switching voltage regulators are super efficient. And even if there isn't an efficiency gain in the VRM, they will likely be one since the processor will be operating at tighter voltage tolerances. The VRM will be closer to the load and be able to react to load shifts quicker, meaning the processor spends less time sligh
        • The big problem with bringing 12V on-chip is not Ohm's law. It is silicon's breakdown voltage at 22nm.

          Trace-to-trace and channel/gate breakdown voltages can be addressed by patterning high-voltage areas with wider insulation gaps but that does not address insulation between metallization layers. It would be feasible but it would also make the fabrication process a fair bit more complex to accommodate the different insulation thickness in the high-voltage area vs the digital stuff.

          I'm guessing the ~2.5V happ

          • by fluffy99 (870997)

            The big problem with bringing 12V on-chip is not Ohm's law. It is silicon's breakdown voltage at 22nm.

            From the linked PDF - "90 nm technology for test devices". It looks like it's not on the same silicon as the actual processor, but rather stacked on top..

        • The control logic could go into the CPU, but I don't see how pulling 12V down to fractions of a volt is going to happen on the die itself without it burning a hole through the board; heatsink or not, you can't escape Ohm's Law.

          I think that may be the point: at some point you have to pull 100W down to fractions of a volt (i.e. hundreds of amps). By doing it on die, it is closer to where you want it to be so you don't have the problem of shipping hundreds of amps quite such long distances.

    • Re:Heat (Score:5, Interesting)

      by Virtucon (127420) on Monday May 13, 2013 @07:42PM (#43715905)

      Well even at 10W I'm wondering how they'll address the heat.
      With the density of circuits in the adjacent silicon I would wonder how they're providing enough isolation to prevent it from becoming a very small brick.

      • Re: (Score:3, Insightful)

        by Anonymous Coward

        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.

      • by swalve (1980968)
        Some of the pentium 4 chips dissipated over 100 watts. I think they know how to move heat off of silicon.
        • by Virtucon (127420)

          And I have a couple of Bulldozer AMDs that run up to 125W that still doesn't mean that you can't cook something or shorten its life. Since these will be 10W processors it will be interesting to see how well they stack up against their predicessors because if they are favorable in comparison then it will be a lot less heat and a lot less noise in desktops.

      • by fnj (64210)

        Well even at 10W I'm wondering how they'll address the heat.

        Are you serious? Essentially all of the power going into the CPU is coming out of it as heat already. That's 35-100+ watts of heat already being dissipated. And you're worried about another 10 watts?

    • I can see the logic behind shortening the length of the wire carrying 'clean' power and getting it away from all the other components (read: noise sources) on the motherboard. It also takes the thinking burden away from the chip integrator and motherboard designer (which is a non-negligible bonus for both marketing and engineering).
    • I'm not an EE so I won't pretend to fully understand this particular case but I like it when tech companies reach a bit and try something hard. This may or may not be a good idea but I'm still excited about it.

      • by citizenr (871508)

        I'm not an EE so I won't pretend to fully understand this particular case but I like it when tech companies reach a bit and try something hard. This may or may not be a good idea but I'm still excited about it.

        Thats what she said.

    • by sshir (623215)
      May be that they still use discrete FETs, it's just control circuitry is on die now. (I'm speculating)
      • by Shavano (2541114)

        May be that they still use discrete FETs, it's just control circuitry is on die now. (I'm speculating)

        That's an interesting question. They can make a FET with as big a voltage and current rating as they want by just making it many times the size of a logic FET. But they also need thicker gate oxide to prevent Vdg breakdown at multiples of the normal logic operating voltage. Not sure how they would do that. It could add a process step, which would make it expensive.

    • Re:Heat (Score:5, Informative)

      by Kjella (173770) on Monday May 13, 2013 @08:15PM (#43716091) Homepage

      Can someone please tell me why this is a good idea

      The long story is here (PDF) [psma.com]. 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.

    • by Shavano (2541114)

      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."

      I have yet to come across a voltage regulator that doesn't run hot. Typically, it's one of the hottest components in an electrical circuit. And we're integrated this into a slab of silicon already well-known for getting so hot it can catch fire?

      Can someone please tell me why this is a good idea, because all of my experience in electrical engineering says that when things heat up, they become more unstable and prone to failure, and the one thing you do not want going critical is your voltage regulator. If that goes, the whole computer catches fire.

      It's cooled by your CPU fan.

      • Thanks, that gets the overall picture.

        So, the idea is that they'll get some very nice inductors on die, capable of replacing some much more expensive external ones. Also, they can distribute the load to a lot of paralel circuits, creating the right tension for each part of the chip, and reducing the loss of each circuit.

        But really, at 90W, a embebing a 76% efficient (not really an exceptional result) conversor means that you'll need to dissipabe other 28W at peak power. Well, I can't say if this is worth it

    • by Joce640k (829181)

      I have yet to come across a voltage regulator that doesn't run hot. Typically, it's one of the hottest components in an electrical circuit. And we're integrated this into a slab of silicon already well-known for getting so hot it can catch fire?

      Gee. If only Intel had some proper engineers like you who could think of clever things like that...

  • Full presentation (Score:5, Informative)

    by AdamHaun (43173) on Monday May 13, 2013 @07:59PM (#43716005) Journal

    You can find the full slide set in PDF format here [psma.com].

    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.

    • hey 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.)

      Respectable viruses take issue with your comment that Linpack is anything like them. Viruses do useful work.

    • inductors and all

      Now that's impressive, and I suspect the real secret to this. Not to say the semi design is trivial, but without the on-chip inductors you wouldn't have much. They weren't clear about it, but perhaps it means getting rid of the big filter caps, and relying on smaller caps for each regulator. It also explains the fast response time with a bunch of smaller regulators.

    • Re:Full presentation (Score:5, Informative)

      by allanw (842185) on Tuesday May 14, 2013 @12:21AM (#43717293)
      The term "virus" in this context means a power virus -- which is an artificial workload designed to draw as much power as possible from the chip. For example, normal CPU burn stress tests might only activate 90% of the chip's power consumption, but a specially designed power virus would be able to activate all of it. In some cases designing the thermal and power integrity solution to support the chip's full power consumption under a power virus needlessly adds extra costs to a product, because it will never see that workload in real life. It's a virus because a malicious person might be able to activate this mode and melt down your CPU, so typically they _do_ have to design the system to support it.
      • re:

        The term "virus" in this context means a power virus -- which is an artificial workload designed to draw as much power as possible from the chip. ... ... It's a virus because a malicious person might be able to activate this mode and melt down your CPU, so typically they _do_ have to design the system to support it.

        Wow, thanks for the very informative post. It makes sense that being able to deal with full thermal stress would be useful. I've had my quad-core shut down on me once at 20-30 seconds into

  • by jimmyswimmy (749153) on Monday May 13, 2013 @08:30PM (#43716161)

    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.

  • ... that they can incorporate the inductor and capacitors for a 90 W switchmode regulator onto silicon. This is a breakthrough in physics, not just in semiconductor processing.
    • This is a breakthrough in physics, not just in semiconductor processing.

      What sort of a breakthrough in physics? Have they found a way around Maxwell's equations or something?

    • Not such a big breakthrough as you'd think. As you increase the switching frequency you can decrease the value of inductor and capacitors required. Last CPU supply I built - 10 years ago! - used 100 nH inductors at 300 kHz per phase. I skimmed the PSMA article but there was mention of MHz operating speeds, not at all unheard of these days, so the components ought to be much smaller. A 10 nH inductor and some hundreds of pF of capacitance seems very feasible without stretching the bounds of silicon techn

    • Nothing new about physics there. It has been known for decades that a piece of wire starts behaving like an inductor at high frequencies and parallel wires or planes behave like capacitors. Both notions have been in use for high-speed analog and RF ICs and PCBs for a long time. For capacitors, there is even a whole class called "Multi-Layer Chip Capacitor" which is basically an IC with several metallization layers connected at alternating ends.

      This is simply the somewhat unexpected but logical application o

  • Given the performance of the on-die voltage regulator, the chip could very useful for designing miniature power supplies. Unfortunately the CPU and the associated digital crap ruin what seems to be a very succesful design of an innovative power supply regulator.

Truth is free, but information costs.

Working...