[HWA] Antec interview

[Phaedrus2129]

Today I’m here to interview Christoph, one of the heads of Antec’s marketing department and former power supply reviewer for Anandtech. As we all know Antec is one of the largest names in PC hardware right now, providing cases, power supplies, and cooling products for all markets, both mainstream and enthusiast. Christoph is here to talk with us about several upcoming Antec power supplies which have been generating some excitement recently.

http://hardwareaware.com/news/antec-interview-w-christoph-katzer/



An announcement of a new budget power supply series, some details on the the High Current Pro and High Current Gamer series, and a great deal of info on how Antec works with its partners to design, test, and manufacture its power supplies. Good stuff.

[Phaedrus2129]

Bump for comments?

[ricercar]

The writing is clear and lucid. I didn't see any irrelevancies or confusing issues. There were no unanswered questions. I'm glad I didn't see anything to complain about except for one "no" that should have been a "know," making me think this was transcribed from a tape.

Unfortunately I'm not a system builder like I used to be, so I'm out of touch on the multiple 12V rail and high current PSU issues, so I can't address the completeness of the topic.

Was that the kind of ting you were looking for?

[Parak]

Needs more 80+ Platinum. I'm spoiled by all of the servers that I work with having them.

[Phaedrus2129]

The writing is clear and lucid. I didn't see any irrelevancies or confusing issues. There were no unanswered questions. I'm glad I didn't see anything to complain about except for one "no" that should have been a "know," making me think this was transcribed from a tape.

Unfortunately I'm not a system builder like I used to be, so I'm out of touch on the multiple 12V rail and high current PSU issues, so I can't address the completeness of the topic.

Was that the kind of ting you were looking for?

Sure, any comments at all are good. This wasn't transcribed from a tape, it's just that Christoph is from Germany and currently based in Taiwan, so his English isn't quite perfect and I didn't want to edit too much in case he felt I changed meaning.

Needs more 80+ Platinum. I'm spoiled by all of the servers that I work with having them.

There are two ATX PSUs that meet 80+ Platinum, one from FSP and one from Superflower, and neither of them are at retail yet, and when they do performance will probably be mediocre. Gold is good enough for now.

[Phaedrus2129]

No offense but only OCN rigs would ever need/want a 1200 watt power supply.   I have PC Power & Cooling Silent 750's and whenever I measure wattage I never even get close - more like 200 watts.

I just like quiet, reliability, single rails and long cables.

Take single rails off there, as an authority on the subject I can tell you that it usually doesn't matter, and multiple rails are sometimes better, especially with large units.

Anyway, we discussed more than just the HCP-1200. Went over the whole line-up really.

[D-EJ915]

I was going to get a Gold 750 unit but decided to get the Silver 850 just in case I get a 2nd CPU and graphics card and 4 more drives...

[ricercar]

Would multiple rails ever be better for a 750 watt unit?.

Yes, because multiple 12V rails isolate the CPU and GPUs. Multiple rails really became significant with parallel processing video cards (SLI, Crossfire). Until SLI, no one really noticed that the power dipped when launching a 3D game, because pre-SLI graphics subsystems didn't take over half the total power draw.

Back in the 2003-era, while I worked at NVIDIA, I tested a GeForce FX 5800 Ultra on a (SFF) 100W PSU. The machine would bluescreen from power demands whenever entering 3D mode over 800×600. Hooked the GPU to a separate 100W PSU (read: separate power rail), and everything was fine to the max resolution of my display.

Consider: NVIDIA GPUs back to GeForce4 were register-capable of SLI, but parallel video cards wesn't implementable to the average consumer until AGP 3.0. (AGP 3.0 supports multiple GPUs in the same bus, but PCI-Express really opened the door). NVIDIA actually made an FX 5900 Ultra board, but proved AGP SLI was not marketable. (1) SLI data saturated the AGP bus; (2) it was the size of a legal pad, and wouldn't fit into any case; and (3) there were no consumer power supplies available to power it.

[Phaedrus2129]

Yes, because multiple 12V rails isolate the CPU and GPUs. Multiple rails really became significant with parallel processing video cards (SLI, Crossfire). Until SLI, no one really noticed that the power dipped when launching a 3D game, because a single graphics system didn't take over half the total power draw, like a SLI or CrossFire subsystem did.

Back in the 2003-era, while I worked at NVIDIA, I tested a GeForce FX 5800 Ultra on a (SFF) 100W PSU. The machine would bluescreen from power demands whenever entering 3D mode over 800×600. Hooked the GPU to a separate 100W PSU (read: separate power rail), and everything was fine to the max resolution of my display. NVIDIA GPUs back to GeForce4 were capable of SLI, but the weren't implementable until AGP 3.0. (AGP 3.0 supports multiple GPUs in the same AGP bus, something only ATI ever exploited). With AGP 3.0, NVIDIA made an FX 5900 Ultra board, but proved AGP SLI was not marketable. (1) SLI data saturated the AGP bus; (2) it was the size of a legal pad, and wouldn't fit into any case; and (3) there were no consumer power supplies available to power it.

*cough*

http://hardwareaware.com/article/single-vs-multiple-rails-in-laymans-terms/





The only advantage of multiple rails on a high-+12V-amperage power supply is that in the event of a parasitic short a 30A OCP circuit will activate much sooner than a 100A OCP circuit. It's a minor concern, but valid.

Single rail PSUs have no advantage over a multi rail PSU; and some single rail PSUs lack +12V OCP entirely.

[ricercar]

>45A on the +12V (650W and higher) picking a multi-rail unit will provide you with an extra layer of protection. It isn’t essential, and it has no impact on the power supply’s performance.

material deleted, new EDIT
My "multiple rail" does not mean the article's "Multiple rail: each trace is monitored separately, so if, say, one trace goes over 25A the power supply will shut down." If that crap is what the PSU market is selling as "separate rails" since 2004, then they're selling smoke and mirrors. What I mean is that individual power sources (rails)—separate PSUs for each GPU and CPU—allow higher graphics performance than drawing all components from one power source of equivalent total amperage.

[Phaedrus2129]

Probably more likely caused by a difference in power supply performance separate from having rails or not.

Take a CWT PSH single rail and a CWT PSH quad rail and you'll find they bench exactly the same, +/-0.05%.

[Phaedrus2129]

Power supply performance is different than GPU/CPU performance. There's no data being moved around to measure. Power supply performance boils down to:

Absolute voltage regulation
Relative voltage regulation (dV/dA)
Transient response
Transient startup suppression
Ripple suppression
Hold-up time
Efficiency

As measured under a high ambient temperature (40-50*C preferable) on both 115VAC and 230VAC lines, and preferably with some low-voltage situations as well (90VAC is typical in Japan, for instance).

[ricercar]

Yah, I think we're talking apples and oranges.

I'm talking about a need for separate taps on the transformer, and the article is talking about output monitoring each tap separately or as a conglomerate output of all taps together.

[Phaedrus2129]

My "multiple rail" does not mean the article's "Multiple rail: each trace is monitored separately, so if, say, one trace goes over 25A the power supply will shut down." If that crap is what the PSU market is selling as "separate rails" since 2004, then they're selling smoke and mirrors. What I mean is that individual power sources (rails)—separate PSUs for each GPU and CPU—allow higher performance than drawing all components from one source of equivalent total amperage.

There is only a very small handful of power supplies like that; the CWT PUC platform for 1kW+ PSUs is the only common one, used in the Corsair HX1000 and Thermaltake 1500W. What you're basically asking is, "Is having two power supplies running different components better than one?" The answer being, occasionally. If the two separate power supplies happen to have better performance under a given load than a given individual one would, then sure. But it's usually cheaper and more effective to get a single quality power supply. Also you can run into crossloading issues with dual PSUs, which can result in worse voltage regulation and lowered performance.

[ricercar]

> Also you can run into crossloading issues with dual PSUs,
> which can result in worse voltage regulation and lowered performance.

Yeah, never put SLI'd GPUs on separate PSUs. You'll permanently damage them.

> But it's usually cheaper and more effective to get a single quality power supply.

Are we talking engineering reality or marketing reality? ;-)

[Phaedrus2129]

> Also you can run into crossloading issues with dual PSUs,
> which can result in worse voltage regulation and lowered performance.

Yeah, never put SLI'd GPUs on separate PSUs. You'll permanently damage them.

Sure you can. No damage at all. Guarantee it, as long as you're picking good power supplies that handle crossloading well. The GPU doesn't care where its power is coming from.



I'm thinking you've just had bad experiences where you were using crappy power supplies in ways they weren't intended to be used.

> But it's usually cheaper and more effective to get a single quality power supply.

Are we talking engineering reality or marketing reality? ;-)

Engineering reality.

[ricercar]

Sure you can. No damage at all. Guarantee it,

I have a graveyard of SLI cards to contradict that.

I'm thinking you've just had bad experiences where you were using crappy power supplies in ways they weren't intended to be used.

I'm talking 2003 consumer PSUs manufactured by companies who'd never heard of SLI outside of the 3DfX context.

[Phaedrus2129]

Back then they were mostly crap, half bridge topology was the norm, not many companies putting out good stuff; Zippy, Delta, SeaSonic, CWT. You were probably using mediocre or crappy power supplies.

[ricercar]

I'll stop editing my posts; you're too fast for me.

I don't think we disagree. You've got a modern perspective while I'm stuck in the PSU stone age.

[Phaedrus2129]

I'll stop editing my posts; you're too fast for me.

I don't think we disagree. You've got a modern perspective while I'm stuck in the PSU stone age.

Perhaps; there's been a lot of change since 2004. PSUs have changed more than CPUs since then.

[ricercar]

You were probably using mediocre or crappy power supplies.

Antec TruePower/TrueBlue, and Delta PSUs for sure; I still have them in use. Also Coolermaster, Seasonic, and whatever they put into Shuttle SFF at that time.

[Phaedrus2129]

Original Antec TruePowers weren't that great, the Fuhjyyu caps used on their secondaries tended to fail a lot.

[ricercar]

Now I'm disillusioned. I'll never buy a PSU again.

[Phaedrus2129]

lol

I'm as much into power supplies as I am in to keyboards. I know my stuff, and I share willingly. Most people know far, faaaar less about power supplies than they do about keyboards, because there's a lot more to know, and it's harder to understand.



'Course, I'll never have to worry about anything PSU related for a while because I'm using what may be one of the best consumer power supplies ever made--the Antec HCP-1200.

[kriminal]

my head just imploded

[ricercar]

Power supply design is like SCSI chain management: black magic.

I designed a power supply for a solid state audio amp I built in the early 80s, and suffered the criticism of hundreds of readers of the Speaker Builder magazine article that described it. I know a lot about being wrong about power supplies.

[Phaedrus2129]

I bet that was a linear power supply too. Then you just have a transient filter, transformer, rectifier, secondary filter. Someone posted an antique calculator a few days ago with a linear PSU, you could see it was just a transformer, schottky rectifier, capacitor, and a couple resistors.

Modern switch mode power supplies are to linear power supplies as a Core i7 is to an abacus.

You have to take the current from the lines, filter it with coils and a couple types of capacitor, plus a MOV for surges and an NTC thermistor for current inrush, rectify your current to DC, run it through a boost circuit as PFC pushing it up to 400V, then a pair of transistors switching on and off forces the polarity of the current to reverse several thousand times a second (there are about a dozen "topologies" for this stage), making an alternating current, which then passes through a transformer, and the current is then split into different paths and rectified into DC, which is then run through a magnetic amplifier to regulate the voltage (sometimes +5V and +3.3V are VRM regulated from the +12V; other times +5V and +12V share the same mag amp), then run through capacitors to filter, then into the wires.

All that disregarding powering the controller PCBs for the primary switchers, secondary, and PFC circuit, and the separate +5VSB power supply inside the PSU, plus the protections circuitry implementation, and a million niggling engineering details.


Computer PSUs are insane.

[ricercar]

my head just imploded

qft

[Phaedrus2129]

inorite?

[kip69]

Ya know, we built a power supply too, back when computers were powered by tape drives.  We added "floppy drives" to an Acorn Atom.  Talk about a speed boost!
But....  we added 2 drives and forgot to double the capacity and capacitors on the power supply.  It exploded inside our custom case.  I have never seen so much confetti suddenly appear inside a computer.  The case fan was on a separate line so it shot the confetti out into the room.  Its one of them things you look back on and laugh.  The computer still runs today.

So when your head asplodes I say: welcome to life with Phaedrus2129.