Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Data centers are transitioning from AC to DC

spectrum.ieee.org

Post reply on HN
Print view View original

71–80 of 432 posts

Re: Data centers are transitioning from AC to DC

#71

Post by shdudns »

adamking wrote: post #12
shdudns wrote: post #3

How is DC better than a three phase delta 800Vrms, at 400Hz? - Three conductors vs two, but they can be the next gauge up since the current flows on three conductors - no significant skin effect at 400Hz -> use speaker wire, lol. - large voltage/current DC brakers are.. gnarly, and expensive. DC does not like to stop flowing - The 400Hz distribution industry is massive; the entire aerospace industry runs on it. No ne…

> no significant skin effect at 400Hz -> use speaker wire, lol now run that unshielded wire 50 meters past racks of GPUs and enjoy your EMI > The 400Hz distribution industry is massive; the entire aerospace industry runs on it nothing in that catalog is rated for 100kW–1MW rack loads at 800Vrms > 3 phase @ 400Hz is x6 = 2.4kHz... Your PSU is now 14 diodes and a ceramic cap you still need an inverter-based UPS upstrea…

"now run that unshielded wire 50 meters past racks of GPUs and enjoy your EMI"

Multipole expansion scales faster than r^2.

Also, im not in the field (clearly) but GPUs cant handle 2.4 kHz? The quarter wavelength is 30km.

"nothing in that catalog is rated for 100kW–1MW rack loads at 800Vrms"

Current wise, the catalog covers this track just fine. As to the voltages, well that's the whole point of AC! The voltage you need is but a few loops of wire away.

"you still need an inverter-based UPS upstream, which is the exact conversion stage DC eliminates"

So keep it? To clarify, this is the "we're too good for plebeian power, so we'll transform it AC->DC->AC", right?

"SiC solid-state DC breakers are shipping today from every major vendor"

Of course they do. They're also pricey, have limited current capability (both capital costs and therefore irrelevant when the industry is awash with GCC money) and lower conduction, and therefore higher heat.

They're really nice though.

"wide-bandgap converters are at 95%+ with no moving parts"

transformers have no moving parts. Loaded they can do 97%+ efficiency, or 2MW of heat eliminated on a 100MW center.

Re: Data centers are transitioning from AC to DC

#72

Post by bigiain »

prezk wrote: post #32
shdudns wrote: post #3

How is DC better than a three phase delta 800Vrms, at 400Hz? - Three conductors vs two, but they can be the next gauge up since the current flows on three conductors - no significant skin effect at 400Hz -> use speaker wire, lol. - large voltage/current DC brakers are.. gnarly, and expensive. DC does not like to stop flowing - The 400Hz distribution industry is massive; the entire aerospace industry runs on it. No ne…

An advanced AI rack might use 100kW = 800V 125A, requiring gauge 2, quarter inch diameter---this isn't your lol speaker wire. Actually, I apologize, I realized I may be talking to a serious audiophile, didn't mean to disrespect your Monster cables. The skin depth by the way is sqrt(2 1.7e-8 ohm m / (2 pi 400Hz mu0))=~3mm for copper---OK for single rack, but starts to be significant for the type of bus bars that an ai…

> I realized I may be talking to a serious audiophile, didn't mean to disrespect your Monster cables.

I am a recovering audiophool.

I do own a pair of 2m long Monster Cable speaker cables (with locking gold plated banana plugs). I am fairly certain I've used welders with smaller cables.

(In my defence, I bought those as a teenager in the late 80s. I am not so easily marketed to with snake oil these days. I hope.)

(On the other hand, I really like the idea of a reliably stable plus and minus 70V or maybe 100V DC power supply to my house. That'd make audio power amplifiers much easier and lighter...)

Re: Data centers are transitioning from AC to DC

#73

Post by scheme271 »

ericd wrote: post #45
bigiain wrote: post #39

Earlier quoted context omitted.

Heh - I vaporised a fairly large soldering iron tip (probably 4mm copper cylindrical bar?), when I fucked up soldering a connector to a big 7 cell ~6000mAHr LiPo battery and shorted the terminals. Quite how I didn't end up blind or in hospital I don't know. It reinforced just how much respect you need to pay to even low-ish voltage DC when the available current was likely able to exceed 700A by a fair margin momentar…

heh man, I'm glad you got out of that easy, I definitely wore safety glasses 100% of the time after my experience. I think a lifetime of experience with dangerous wall outlets and harmless little 1.5V/9V DC cells teaches us the wrong lessons about DC safety. I've since heard stories of wrenches exploding when they fall across EV high voltage battery terminals. Wrenches aren't supposed to be explosive. The electrician…

A bunch of those big breakers require two people. One person in a flash suit and another with a 2m long pole around the first person. That way if an arc flash happens, the second person can yank the first person to safety without also getting hurt.

Re: Data centers are transitioning from AC to DC

#74

Post by ternus »

KnuthIsGod wrote: post #49

Waiting for home DC. It is silly to have AC to DC converters in all of my wall connected electronics ( LED bulbs, home controller, computer equipment etc )

The lesser-known instance of this is RV power. When you're running off small batteries and solar, you want to make the best use of the watt-hours you have, and that means avoiding the DC-to-AC-to-DC loop wherever possible. So you run 12V (or in newer models, higher voltage) versions of everything, upconverting as necessary.

Re: Data centers are transitioning from AC to DC

#75

Post by bigiain »

shdudns wrote: post #25
shiroiuma wrote: post #6

Earlier quoted context omitted.

>- no significant skin effect at 400Hz -> use speaker wire, lol. What are you talking about? There's a very significant skin effect at 400Hz. Skin effect goes up with frequency. These datacenters use copper busbars, not cable, so skin effect is an important consideration.

At 100 000 A for a 100 MW data center at 1000 V, speaker wire is a joke. You obviously need at least a dozen stands in parallel!! Clearly skin effect scales with frequency but, 400 Hz is still low, only 2.5x lines frequency (the scale is by the root); so the skin depth is 3mm. 3mm on each side makes for a pretty hefty rectangular cross-section.

If you could get that 100,000Amps flowing through your speaker wire, the vaporised copper and the plasma channel would probably keep your 100MW flowing, at least until your building caught fire.

Re: Data centers are transitioning from AC to DC

#76

Post by scarecrowbob »

KnuthIsGod wrote: post #49

Waiting for home DC. It is silly to have AC to DC converters in all of my wall connected electronics ( LED bulbs, home controller, computer equipment etc )

Well, having spent some time operating a 12VDC system last year when I moved into some shacks, I will say that I find it a lot more convenient to run 120VAC.

I end up converting stuff anyhow, because all my loads run at different voltages- even though I had my lights, vent fan, and heater fans running on 12V I still ended up having to change voltages for most of the loads I wanted to run, or generate a AC to to charge my computer and run a rice cooker.

Not to mention that running anything that draws any real power quickly needs a much thicker wire at 12V. So you're either needing to run higher voltage DC than all your loads for distribution and then lowering the voltage when it gets to the device, or you simply can't draw much power.

Not that you can't have higher voltage DC; with my newer system the line from my solar panels to my charger controller is around 350VDC and I can use 10awg for that... but none of the loads I own that draw much power (saws, instapot, rice cooker, hammond organ, tube guitar amp) take DC :D

Re: Data centers are transitioning from AC to DC

#77

Post by jacquesm »

torginus wrote: post #31
adrr wrote: post #23

Our houses should be DC. So wasteful to have all these bricks to change to AC to DC.

I've had discussed with people familiar with the matter, and they convinced me its really not worth it for many reasons, the main one being safety - DC arcs are self sustaining - AC voltage constantly goes to zero, so if an arc were to form, it gets auto extinguished when the voltage drops. With DC this never happens, meaing every switch or plug socket can create this nice long arcs and is a potential fire hazard.

The 'what is safer' question for DC and AC at the same effective current and power has a mixed set of answers depending on conditions. For instance, DC is more likely to cause your muscles to contact and not let go (bad), but AC is more likely to send your heart into ventricular fibrillation (sp?, also bad).

AC arcs are easier to extinguish than DC arcs, but DC will creep much easier than AC and so on.

From a personal point of view: I've worked enough with both up to about 1KV at appreciable power levels and much higher than that at reduced power. Up to 50V or so I'd rather work with DC than AC but they're not much different. Up to 400V or so above that I'd much rather have AC and above 400V the answer is 'neither' because you're in some kind of gray zone where creep is still low so you won't know something is amiss until it is too late. And above 1KV in normal settings (say, picture tubes in old small b&w tvs and higher up when they're color and larger) and it will throw you right across the room but you'll likely live because the currents are low.

HF HV... now that's a different matter and I'm very respectful of anything in that domain, and still have a burn from a Tronser trimmer more than 45 years after it happened. Note to self: keep eye on SWR meter/Spectrum analyzer and finger position while trimming large end stages.

Re: Data centers are transitioning from AC to DC

#78

Post by jacquesm »

bigiain wrote: post #75
shdudns wrote: post #25

Earlier quoted context omitted.

At 100 000 A for a 100 MW data center at 1000 V, speaker wire is a joke. You obviously need at least a dozen stands in parallel!! Clearly skin effect scales with frequency but, 400 Hz is still low, only 2.5x lines frequency (the scale is by the root); so the skin depth is 3mm. 3mm on each side makes for a pretty hefty rectangular cross-section.

If you could get that 100,000Amps flowing through your speaker wire, the vaporised copper and the plasma channel would probably keep your 100MW flowing, at least until your building caught fire.

Even your monster cable? ;)

Re: Data centers are transitioning from AC to DC

#79

Post by bigiain »

jacquesm wrote: post #67
bigiain wrote: post #39

Earlier quoted context omitted.

Heh - I vaporised a fairly large soldering iron tip (probably 4mm copper cylindrical bar?), when I fucked up soldering a connector to a big 7 cell ~6000mAHr LiPo battery and shorted the terminals. Quite how I didn't end up blind or in hospital I don't know. It reinforced just how much respect you need to pay to even low-ish voltage DC when the available current was likely able to exceed 700A by a fair margin momentar…

You got super lucky.

Yep. Super super lucky. I suspect my reading glasses are the only reason I can still see anything.

Re: Data centers are transitioning from AC to DC

#80

Post by adrr »

adiabatichottub wrote: post #66
jwilliams wrote: post #46

Earlier quoted context omitted.

There are niches where DC makes sense - low-voltage lighting, USB/LED ecosystems. Once you get into higher power (laptops and up), switching and distribution get harder, so the advantages fade. For bigger appliances (fridge, etc), AC is fine + practical.

Your modern fridge is probably going to have an inverter-driven motor, so you're right back to using DC.

All modern appliances and HVACs are using inverter drive motors for efficiency. Brushless DC motors are more efficient though.
Post reply on HN