Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!


Forgot your password?
DEAL: For $25 - Add A Second Phone Number To Your Smartphone for life! Use promo code SLASHDOT25. Also, Slashdot's Facebook page has a chat bot now. Message it for stories and more. Check out the new SourceForge HTML5 Internet speed test! ×

Comment Re:Heat (Score 2) 237

So what stops someone from taking the switching frequency really high, like into the hundreds of megahertz? In switching regulators, there is both conduction loss and switching loss. Conduction loss occurs from resistance in the power supply path, including switch resistance. It can be reduced by increasing the switching frequency. However, this increases the switching loss -- you have to switch the power FET gate capacitance more often. The most efficient system is achieved when conduction loss is balanced with switching loss. It is a complex engineering problem. By making a tiny package integrated solution, the inefficiencies of switching can be reduced so the frequency can go up. This cannot be easily done with a traditional discrete-based system like on current motherboards.

Comment Re:Rotten idea for performance (Score 1) 237

If they can cut down platform costs by a few tens of dollars then it is a huge win. This solution removes a lot of discrete chips sourced from a lot of outside companies. Current voltage regulators for these 100W chips are multi-phase regulators, which means something like 4 - 12 parallel voltage regulators with their own inductors, etc. Also, the required area is 50x smaller according to their presentation, which directly affects form factor and cost of these systems.

Comment Re:From a former power supply designer - Neat! (Score 1) 237

ARM chips are more like 10W, not 100W like high-performance CPU's. These power supplies are much more complicated to design. This is actually at the leading edge of R&D -- no other chip maker has made a commercial product with integrated voltage regulator in the 100W range. It's only been done in academia recently.

Comment Re:Full presentation (Score 4, Informative) 237

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.

Comment Re:Full presentation (Score 2) 237

An Intel CPU has a TDP of 90W+ running under 1V. That's 100A+ from the switching power supply. With resistive loss, and inefficiencies from multi-phasing the regulators, efficiency are worse than you say. The cost is also high -- having all of this integrated into the package saves on the platform cost.

Comment Re:Ah, to be young again... (Score 2) 92

If you keep increasing the voltage then it's likely that you can hit higher frequencies, but the power scales with voltage squared and frequency linearly, so power will go up pretty quickly. However, nowadays in advanced processes the interconnect is becoming more of a factor in the limitation on frequency scaling instead of the transistors themselves, in which case increasing the voltage will only help up to a certain point.

The trade-off that the company selling the CPU's makes is between the cost of cooling, reliability and lifetime of the device (higher voltage will wear the transistors out quicker, and high temperatures accelerates this process), and yield.

Comment Re:Every keyboard is washable (Score 3, Interesting) 205

Those ubiquitous black IC's are plastic packaging which is not moisture sealed. Not sure if it'd actually affect the silicon to soak it in water for a bit though and use normally. But if you ever order any parts, they come in moisture sealed bags with big warning labels saying that you must reflow solder the IC's within 24-72 hours of opening the package or else too much moisture from the air will seep into the packaging, causing them to act like popcorn when you bake them to 350C for soldering. So if you leave them out too long you're supposed to slowly bake them to get rid of all the moisture before reflow soldering.

Slashdot Top Deals

CCI Power 6/40: one board, a megabyte of cache, and an attitude...