Slashdot Mirror
Gigabyte N680SLI-DQ6 - A Mother Of A Motherboard
MojoKid writes "Motherboards manufacturers seem to get more exotic in their designs, with each new chipset release. HotHardware has an evaluation posted looking at the Gigabyte GA-N680SLI-DQ6; a product that seemingly out does every other current desktop motherboard in a number of key areas. The board features four Gigabit LAN controllers, 10 SATA ports, a 12-phase power array, 100% solid-state capacitors, and a unique wrap-around, passive, cooling apparatus that cools both the top and underside of the chipset and CPU socket area. And because the board is based on NVIDIA's nForce 680i SLI chipset, it also has three full-length PCI Express x16 slots for multi-GPU support. It's a good overclocker and performed well throughout the benchmarks."
From TFA:
The correct term is 'solid-core', not 'solid-state', FYI.
This refers to the power regulator onboard - i.e. internal to the motherboard itself; it's nothing to do with the 240v PSU.
The onboard power regulator is the part of the motherboard which converts the standard 3.3v to the exact voltages the CPU, RAM, etc require. The theory goes that the more phases, the cooler running, more efficient and more reliable the motherboard will be (but it's mostly about e-penis, rather than any genuine advantage).
The opposite of "solid-state" capacitors is in this case the aluminium electrolytic capacitors. Presumably they just use tantalum or some other kind instead on this board. Now, these are also polarized and technically also "electrolytic" (besides, they can still fail) but their failure mode does not include emitting nasty goop onto the board.
SIGBUS @ NO-07.308
The summary is wrong, as anyone who looked at the front page of the article could tell you.
2 PCI Express x16 slots
1 PCI Express x8 slot
1 PCI Express x1 slot
3 PCI slots
-1 Uncomfortable Truth
All the hardware faults I've had in the past few years are due to bad caps. I hope other manufacturers take the hint and ditch using electrolytics for their mobos. The only place I want to see electrolytics are in the power supply.
those damn ads kept crasing my ie:
c leid=987
http://www.hothardware.com/printarticle.aspx?arti
Sometimes, life itself is sarcasm...
Plz. ignore that nutcase below that refers to Wikipedia on 3-phase power, that's about something totally different. I suspect many /.'ers will have some understanding of electronics, but maybe less detailed than I assume. So I'll give it a go in layman's terms:
What you're looking at is a DC step-down switching regulator circuit (look that up if you want). On most mobo's, it converts 12V to around 1,5V, at many, many Amperes (fist rule: power = voltage x amps).
In it's most basic form, it consists of a coil, a (fast) switch, and a diode. The coil(s) are the thick copperwire/ceramic thingies on the board. As a switch, electronic versions known as power MOSFETs (usually black, square plastic thingies) are used. Because diodes have a small, but significant voltage drop when current passes through, this would give unacceptable losses (heat) at the high currents we have here. Therefore, another power MOSFET is used to replace the diode.
Such a pair of MOSFETs is switched on and off quickly (10s or 100s thousands of times a second), with 1 in conducting (low resistance), and 1 in non-conducting (high resistance) state at any given moment. BUT: when switching over, there is some overlap, where both are somewhat conducting, causing a momentary 'short circuit' (=losses, waste heat). Enter 3-state: switch one off, wait very short to make sure the MOSFET goes fully into non-conducting state, and only THEN switch on the other MOSFET.
My guess is this 'Quad-Triple Phase power' is a similar construction, but then 4 times, working in parallel (for more current), or alternating (to lengthen cooling periods between on-states). Basically: a high-current, energy-efficient 12V-to-CPU-voltage converter.
I have to take issue here. "n"-phase power supplies in motherboard parlance refer to different Buck-style switching regulator setups. A basic Buck regulator turns on a MOSFET (generally) to switch current into an inductor and capacitor, with a diode in parallel (you can google buck topology if you like). Thus, as the power drains out of the capacitor into the load, the switcher recharges it with little sips of current every couple of microseconds, resulting in a stable voltage from the point of view of the load. MOSFETs have a fairly hard limit on allowable pulse current and power dissipation that they can tolerate.
In order to switch more power, you can put a whole bunch of MOSFETs in parallel, or use a really big one, but then you're switching a huge amount of current all at once through your poor little inductor and capacitor, each of which also have ripple current ratings you should not exceed.
So, instead, you get a switcher IC capable of controlling multiple phases (for instance the 4-phase L6714 from ST Micro if you're interested in powering an AMD64 processor) and 4 different MOSFETs, and each time the load capacitor must be recharged (again, every 1-5 microseconds), the IC will switch on one MOSFET after the other in sequence, resulting in a more steady load voltage, and a lower ripple current on the inductors and capacitors. This has multiple advantages for voltage quality, heat dissipation, and component life.
The fact that it's subject to silly marketing does not mean they'd be stupid enough to buy 12 MOSFETs and expensive power controllers if they didn't need to for technical reasons.
Ladies and gentlemen, presenting the latest, greatest Slashdot meme!
It is funnier, at least, than imagining a you-know-what cluster of them...
Read frankschwab's link above about multiphase buck converters -- I learned some stuff from it (and I help design buck converters...) These are synchronous buck converters, so they have a high and low side mosfet, no freewheeling diode, and by putting several in parallel and then running them in round-robin style, you can reduce your power supply output ripple. It's a pretty sophisticated technique, and it's possible that they need 12 to get both the efficiency they want and the ripple they need. (If you're working with a synch buck, an efficiency limitation is the equivalent series resistance of the output capacitor, which also determines your output ripple, so by going to parallel converters you can tolerate smaller output caps without increasing output ripple.)
Nostalgia's not what it used to be.
Regular aluminum electrolytic capacitors are filled with liquid. When driven hard and hot, that liquid evaporates or boils. There are various other types of caps that have no liquid, including solid tantalum and ceramic. They generally have lower ESR (good), and last (for all intents and purposes) forever, but are definitely more expensive.
About the only advantage of an aluminum electrolytic is that it's cheap.
steve
Oh, you're not stuck, you're just unable to let go of the onion rings.