Power my System

[zr0e]

First: System.

Video: xfx 8800gtx
Processor: Q6600 (old stepping)
Thermalright Ultra 120 w/Scythe S-Flex (F)
Asus P5K-Deluxe
2x IDE HDD
2x SATA HDD
2x SATA DVDRW
2x 1GB Dominator (4-4-4-12 Corsair)
Floppy Drive
Nexus Fan Controller

I do not know how much power I actually need. I probably won't be overclocking anything, and if i do, it will not be very major overclocking. .

I've done infinite research on Power Supplies. JonnyGuru's reviews including ripple report seem to be the best but there aren't quite enough of them.

Some supplies I have been interested in but really do not know where to go, are:

Silencer Quad 750W

Enermax Infiniti 720W

Various Silverstone (Zues/Olympia) in the 650-750W range

It is possible this is *too much* power. My justification was about 300w for the GTX and 400w for the rest, i may be way off. I'll leave it to experts.

Things that I don't necessarily like are multiple 12v Rails, and I know that a couple of the aforementioned do have multiple 12v rails. I'm also not sold on the modular thing.

Anyway, I really am looking forward to your responses. This is the last thing I need to actually buy and then I can be done for a while, sort of exciting.

Thanks in advance.

p.s.

I know that there is no best anything in this [hardware] world, but I am looking for experience and opinion.

 

[John]

If you can afford to spend up to $200 on a psu then by all means do it. All of the psu's that you mention are top quality units and will give you plenty of headroom for future upgrades. Keep in mind that a quality $70 Enhance 5150GH would sufficiently power your rig. Your system will use around 300W under full burn-in & gaming load which is a 40% load on a 750W psu. Typical loads while browsing the web will be around 150-170W or a 20-23% load.

The Silencer and Infinity are 80%+ efficient at 20-100% loads. The Silverstone Zeus reaches 85% eff around a 50% load, but your rig at typical load would net a 72% eff. The Olympia is 79% (so close) eff at 40-75% loads. Ripple on any of these units is not an issue, and voltage regulation is excellent. All 3 have a 3 year warranty. Most of this data comes from looking at jonny's reviews.

 

[zr0e]

Thanks John,

Even with a GTX You think the "draw" on the power will be that low?

 

[John]

The 8800GTX (non o/c) pulls around 135W (11.25A) under load. Want to see someone using an Antec EarthWatts 380 to power a rig similar to yours?

http://www.overclockers.com/articles1452/

 

[zr0e]

So then, it is mostly all marketing and people wanting to see big numbers (i.e. 750, 850, 1000 watt)

And, also, then I could obviously be "good" to go with a smaller watt power supply, my next question is are they built as well as the higher end flagship models? The money is there to spend but I'd just as well not, if for no reason, obviously.

Silencer 610, Zues and Olympia in the 550-650 range -- are these built as well as their "bigger brothers(/sisters )"

How are the Corsairs? the 620, the 500?

I see reviews where people say "the price is high but you get what you pay for" but are those power supplies actually relatively expensive? they don't appear to be..

 

[Jhhnn]

Not that this review is terribly scientific, but they ran an overclocked processor and 2 8800 gtx cards in sli with a 400w psu- no problems at all.

http://www.madshrimps.be/?acti...tpage=2534&articID=574

I'd consider this unit to be at the upper end of sanity for your application-

http://www.newegg.com/Product/...x?Item=N82E16817104034

 

[zr0e]

That's an interesting recommendation.

Is there a problem with running low-watt power supplies at the upper-end of their rating for an extended (say, 3 years) period of time? Obviously, the reviews don't have access to that information.

 

[John]

Jhhnn, that madshrimps review is very interesting. However I do question the wall loads since they seem off (too low) by about 75W.

Originally posted by: zr0e
How are the Corsairs? the 620, the 500?

The HX series are top performers in their class. You can pick up the 520W for $89 AR from buy.com

I recently asked the Corsair PowerGuy what his thoughts were on psu utilization and here's what he had to say:

I actually recommend a 40-70% load if possible. Here's why.

http://www.corsair.com/_images...ency_graph_419x360.gif

This is the efficiency curve of the HX series, which is somewhat common for 80+ rated PSUs. The idea is that you want to be at that peak or near it as much as possible. So yes, you could be at 20% or 80%, but the problem with being very close to max load on a PSU is that it stresses components. For maximum lifespan, matching a 40-70% load is the best efficiency and will allow the PSU to last a bit longer.

** I asked the PowerGuy if he had any data to support the "a bit longer" claim since the HX series is still 81% efficient @ 80-100% loads. If those high loads do affect the life of the psu surely there would be data to support it. Then again maybe not.

 

[zr0e]

This is sort of what I was worrying about with the interesting (in the unexpected sense) aforementioned recommendations. I do appreciate your responses, though, all of you.

Thanks again

 

[Jhhnn]

Dunno why you'd question the readout from a digital commercial ammeter, John, other than to support your own notions of how much power is really required for modern rigs. Very few computers, even monster sli overclocked raid type enthusiast computers, actually consume more than 400W from the wall receptacle. At 70% efficiency for most psu's, that only 320w output...

Total overkill is currently very much in vogue with power supplies, and is also very much encouraged by the makers. Quality suppliers get to move buyers upscale, making more per sale, and unscrupulous ones get to label junque as "600WATTS!", knowing that it'll likely never be called on to put out more than 300W...

Quality power supplies are rated for full continuous output at 100K hrs mtbf, 25C ambient. Even de-rated to 50K hrs, that's still almost 6 years. If running it at 80% load significantly increases the lifespan, then it'll be a working antique when tossed into the landfill, anyway.

Either the mentioned 520w Corsair or the Fortron I recommended above will happily run an overclocked box stuffed with everything imaginable other than peltiers and very likely never hiccup, never disappoint over a normal computer lifespan, 4-5 years on the outside.

I'll grant that servers in a corporate environment are a whole different realm, but that's not the subject at hand...

 

[John]

Originally posted by: Jhhnn
Dunno why you'd question the readout from a digital commercial ammeter, John, other than to support your own notions of how much power is really required for modern rigs.

It has nothing to do with my opinion. I previously linked to mcoleg's article with an o/c'ed E6600, 8800GTX, 4x1GB, 2 HDD, etc. that pulls roughly 380-400W from the wall according to a Kill-a-watt. Since a single 8800GTX pulls around 135W, and o/c?ed 8800GTS SLI pulls at least 200W, I cannot see how the madshrimps editor is only pulling 400W from the wall. The only logical explanation is that he isn't under a true full load, or he's using defective test equipment.

Then again how does 230VAC input affect the wall loads compared to the same pc on 115V? If we take a 320W psu load @ 80% on 115V and 84% @ 230V the difference would be 20W.

Very few computers, even monster sli overclocked raid type enthusiast computers, actually consume more than 400W from the wall receptacle.

I'm sure several enthusiasts on this forum alone have rigs that pull more than 400W from the wall.

Quality power supplies are rated for full continuous output at 100K hrs mtbf, 25C ambient. Even de-rated to 50K hrs, that's still almost 6 years. If running it at 80% load significantly increases the lifespan, then it'll be a working antique when tossed into the landfill, anyway.

Don't mistake power-supply MTBF for life expectancy

Crappy psu's can also have a high MTBF. How about a nice Powmax Assassin 500W w/ MTBF @ 100K hrs 25°C ambient conditions

Either the mentioned 520w Corsair or the Fortron I recommended above will happily run an overclocked box stuffed with everything imaginable other than peltiers and very likely never hiccup, never disappoint over a normal computer lifespan, 4-5 years on the outside.

How about a modest FX-74 with Crossfire 2900XT's? In case you didn't notice, I recommended a 500W 5150GH for the OP's system. However some people like having a lot of headroom. To each his own, right?

 

[zr0e]

Once again guys, I just want to thank you for your input and suggestions. I hope you'll continue conversing so people can learn from your expertise.

I went with a Corsair from Buy.com for those interested.

 

[Jhhnn]

Heh. The fx-74 isn't a desktop processor, but rather a server unit. Apples to apples, OK?

And the mtbf ratings of cheapo psu's fall right in line with their power ratings- outright lies. The same factory in China still makes the same crap, they just slap different labels on it and rename their marketing companies- yesterday's Deer, Austin, L&C and other pseudonyms for garbage are simply replaced with different monikers... their numbers aren't from engineering, but rather from marketing...

That's not true of reputable makers, at all. Seasonic, Fortron/Sparkle, Enhance, Zippy and a few others fall into that category. Their stuff will perform as advertised. period.

Here's an interesting example of what I'm talking about-

http://techreport.com/reviews/...d-2900xt/index.x?pg=15

Their test system draws 490w at the wall, which translates into 392w output at 80% efficiency. But they had to use a 1000w psu rather than their usual 700w OCZ model... and never addressed the issue of why they had to do that, even thought their 700w unit *should* have done the job just fine... what that means, well... you tell me.

Headroom on a psu is peachy and all, but there is a point of diminished returns, obviously, and money spent on power never utilized could easily be spent on some other aspect of the system where it would make a real difference...

I just try to get people to make sense out of it, be reasonable. Yeh, sure, a .375 H&H magnum is a helluva gopher gun, but then, so is a 22LR...