Why do i need an 850w PSU to power SLI/CF

[TemjinGold]

Well,run any electronic device at 100% for extended periods and witness its downfall.

If a PSU's circuits are constantly running at capacity,it will wear down much sooner.It's inevitable.Not saying that quality PSUs will cause disasters like those I mentioned in my previous post,but still,it's a risk sometimes.

Also,I don't think many people will suggest a 1000W PSU for someone who will use a 4670(Non-k) and a 280x at stock clocks,but using a 600-650W PSU for CF/SLI is cutting it too close and asking for trouble,if it's enough at all.So I'd say exaggerate a little when buying PSUs.Just a little.

I don't dispute that. What I'm actually wondering is, do quality units get "under-rated" such that running say HX620 at 620 isn't actually "running 100%" just running "100% rated." If I were Corsair offering a 7-warranty on something like that, that would be the easiest way to keep that promise.

In other words, "running at max rating" may not be the same as "running 100%."

 

[24601]

I'm running my 1050 watt top quality PSU @ ~100watts lower than the OCP limit (I have tested it, it's 1350-1400w @ the wall).

I am drawing 1300 watts at the wall for 24/7

We'll see how long it'll last

I want an excuse to get an AX1500i 80 plus Titanium when it comes out in ~Q2 anyways

 

[KaRLiToS]

I'm running my 1050 watt top quality PSU @ ~100watts lower than the OCP limit (I have tested it, it's 1350-1400w @ the wall).

I am drawing 1300 watts at the wall for 24/7

We'll see how long it'll last

I want an excuse to get an AX1500i 80 plus Titanium when it comes out in ~Q2 anyways

Get a third 290x, then you will have an excuse. ^_^

 

[24601]

Get a third 290x, then you will have an excuse. ^_^

I have 1x290x and 2x290 right now

1x Visiontek Cryovenom 290 is coming within 2-3 weeks probably (I ordered it the day after they announced it)

So that's already the quota of 4

Right now I plan on just using my extra Antec NeoEco 620C for the Cryovenom until the AX1500i comes out in Q2

 

[KaRLiToS]

I have 1x290x and 2x290 right now

1x Visiontek Cryovenom 290 is coming within 2-3 weeks probably (I ordered it the day after they announced it)

So that's already the quota of 4

Right now I plan on just using my extra Antec NeoEco 620C for the Cryovenom until the AX1500i comes out in Q2

With the AX1500i, you would only be able to run your Quad Cards at near stock.

What CPU are you planning to use with the four cards?

 

[24601]

With the AX1500i, you would only be able to run your Quad Cards at near stock.

What CPU are you planning to use with the four cards?

There's no way I'm upgrading my CPU until Haswell-E at the earliest.

The cards are for mining

I use 1x 290 right now when I play games on a 1080p monitor.

I won't be using crossfire until I get the 120hz 1440p Asus G-Sync monitor when it comes out.

But yeah, it's true the AX1500i will not be enough for overclocks, but I'm more limited by my power circuits and room cooling capability TBH. If I overclock I can always string more PSUs onto my Franken-case

If we continue conversing we should really move the conversation to your quad-fire thread though, as we're quickly veering off-topic

 

[Torn Mind]

Sorry, there is no case study to refer you to for your piece of mind. Nevertheless, the principles of physics do apply and they have been proven time and time again via experimentation. Temperature and high voltage/current deteriorate electronic components. You can piece the rest together for yourself.

The laws of physics would tell me the specific mechanisms of failures and it would also not muddy two different factors into causing component failure(Heat energy and electrical energy coursing through the component).


Temperature is a variable dependent on the ability of the fan to remove energy from the system and the PSU's ability to "generate" heat due to its efficiency losses. The matter of temperature killing components is a separate matter from pulling max amperage of a unit and irrelevant to the discussion. That's because temperature should not even be a issue even if the PSU is running near max capacity because the fan would also be running full blast and moving the heat energy out of the system anyway.

What has been proven is that running things out of spec is what kills components. A simple resistor will likely last and last so long as it dissipates the heat it generates whether it is in a 1000W unit or a 300W. A transistor could die from reverse bias or overvoltage, but once again, applying within spec voltages will not accelerate component death.

 

[Torn Mind]

Well,run any electronic device at 100% for extended periods and witness its downfall.

If a PSU's circuits are constantly running at capacity,it will wear down much sooner.It's inevitable.Not saying that quality PSUs will cause disasters like those I mentioned in my previous post,but still,it's a risk sometimes.

Also,I don't think many people will suggest a 1000W PSU for someone who will use a 4670(Non-k) and a 280x at stock clocks,but using a 600-650W PSU for CF/SLI is cutting it too close and asking for trouble,if it's enough at all.So I'd say exaggerate a little when buying PSUs.Just a little.

The current delivered through the system is subject to variance in amperage or voltage, which can peak above acceptable specifications. It is not a pure constant with home user. It is these brief moments that the PSU goes OUT OF SPEC that is of concern when running at around 100% of the rated wattage. Going out of spec is the real concern of running something at 100%, not degradation of non-defective components, which wouldn't do anything significant since the caps will mostly likely die first.

Otherwise, I suspect that most failures that appear to be due to electrical degradation are actually due to latent defects present on the unit itself that didn't get caught during burn-in and only after the extended burn-in that is called "normal use", not necessarily at max wattage, did those components show their true colors. Could have been a cold solder joint, ESD, some chemical reaction, botched circuit board traces. For example, PSUs that had bulging caps during the cap plague on some Dell that pulled about 100-150 watts on a Dell bought by Average Joe consumer didn't die because they were worn out. They were dead because something in the manufacturing process was screwed up, be it the electrolyte formula or something else...


Whether a PSU is enough for SLI/CF does depend on the power draw of the cards themselves. 650 is enough for cards like 7870s, but not for R9 290s since the 290s could pull up to 300 watts.

 

[DigDog]

That is crazy. If i buy a PSU rated at 620w and warrantied for 7 years then id expect that PSU to run at 100% for 7 years and not die or cause me issues.

this makes me giggle. :biggrin:

620W *peak* power. meaning, anything even slightly over 620 and it will explode, taking most of your PC with it.

Also, PSU capacitors degrade over time, more so with higher temperatures.

Also Part2: Revenge of Also, PSUs vary in efficiency according to the load you place on them. As they degrade, they will draw more and more power from the wall, costing you more money, and further accelerate their degradation.
(my Tagan actually went *spark*-bang-hisss-*smoke*, back when i OCd my C2D and had no idea how much power i was drawing because of that)


Finally, Also 3: The final Also.
PSU makers lie, plain and simple.


read:


http://www.jonnyguru.com/modules.php?name=NDReviews&op=Review_Cat&recatnum=13

 

[JEDIYoda]

Ive had a PC Power & Cooling 500watt since 1996 and it is going like the Eveready bunny!!

 

[coffeejunkee]

Coffeejunkee he is correct.

And that other guy is clearly showing his inexperience in building.

You do not run a PSU at maximum wattage if you want it to last.

This is common knowledge.

Use google to find the information its out there!

Well, when I use Google I either find threads like these where people say it won't work but have no real numbers to back it up or threads where people say that much to their surpise it does work.

What I'd like to see is something like the ssd endurance testing at some sites. Just run a power supply at max load and see how long before it breaks down. Probably gonna cost a lot of electricity but it's for science.

 

[Makaveli]

Well, when I use Google I either find threads like these where people say it won't work but have no real numbers to back it up or threads where people say that much to their surpise it does work.

What I'd like to see is something like the ssd endurance testing at some sites. Just run a power supply at max load and see how long before it breaks down. Probably gonna cost a lot of electricity but it's for science.

lol when you are ready to buy the psu's and run the test i'll be happy to read your results :biggrin:

 

[Slomo4shO]

The laws of physics would tell me the specific mechanisms of failures and it would also not muddy two different factors into causing component failure(Heat energy and electrical energy coursing through the component).


Temperature is a variable dependent on the ability of the fan to remove energy from the system and the PSU's ability to "generate" heat due to its efficiency losses. The matter of temperature killing components is a separate matter from pulling max amperage of a unit and irrelevant to the discussion. That's because temperature should not even be a issue even if the PSU is running near max capacity because the fan would also be running full blast and moving the heat energy out of the system anyway.

What has been proven is that running things out of spec is what kills components. A simple resistor will likely last and last so long as it dissipates the heat it generates whether it is in a 1000W unit or a 300W. A transistor could die from reverse bias or overvoltage, but once again, applying within spec voltages will not accelerate component death.

Just because I said temperature and high voltage/current deteriorate electronic components doesn't mean that the two variables need to be conjoined... The initial statement was that capacitor or fan failures are the main factors in PSU deaths.

With that said, I don't have a PhD in electromagnetics so I will leave the discussion of the exact process of how capacitors fail to someone who is more qualified to shoulder such a discussion.

 

[Pandamonia]

You would be wrong....Groove's system and mine are very similar and you can search for the thread on mine shutting down a 1300 watt platinum rated psu during 3DMark runs. A 1600 watt Lepa solved that. Also, the power a 4.6GHz SB-e pulls is miles away from what a 5.0+GHz will be pulling.

I wouldnt be wrong because those are tested Power at the wall for that system.

I call BS your system used more than 1300w

 

[DigDog]

Well, when I use Google I either find threads like these where people say it won't work but have no real numbers to back it up or threads where people say that much to their surpise it does work.

well, the world is full of idiots. But you are lucky to have found AT. Guaranteed 99% idiot-free.

To be honest, information about PSUs is scarce on the web, as they don't make nearly as much of an interesting conversation point compared to stuff like CPUs or GPUs. I wouldn't blame people too much for not knowing what a capacitor or a resistor is, nor why some PSU are good and some are so bad they can be dangerous.

And yes, there is a lot of marketing hype. just like with any PC part, just .. when there's hype around a "popular" PC part, there will always people ready to tell it like it is, but PSUs are too "dull" to get that kind of attention.

 

[Torn Mind]

this makes me giggle. :biggrin:

620W *peak* power. meaning, anything even slightly over 620 and it will explode, taking most of your PC with it.

Also, PSU capacitors degrade over time, more so with higher temperatures.

Also Part2: Revenge of Also, PSUs vary in efficiency according to the load you place on them. As they degrade, they will draw more and more power from the wall, costing you more money, and further accelerate their degradation.
(my Tagan actually went *spark*-bang-hisss-*smoke*, back when i OCd my C2D and had no idea how much power i was drawing because of that)


Finally, Also 3: The final Also.
PSU makers lie, plain and simple.


read:


http://www.jonnyguru.com/modules.php?name=NDReviews&op=Review_Cat&recatnum=13

620W is not a common wattage rating. Unless he has the 620 W from Cooler Master(their somewhat decent Silent Pro M line), chances are that it is a Corsair, Antec, or Seasonic. Even Cooler Master's units typically can deliver their rated wattage, but with worse ripple control. These units have continuous ratings, not peak ones, or have you ignored on the JohnnGURU's reviews that praise certain units in order to propagate the falsehood that all PSU companies lie?

A lower wattage unit might have smaller heatsinks, but the fans would be going faster at higher loads. If the same Chemicon capacitor is stuffed in a 620W or a 1000W unit, chances are the rate of heat dissipation is similar.

Also, capacitors do "degrade" but the effects are not such that the maximum amperage that of the whole rail decreases. What happens is that either the waveform becomes more and more distorted to the point the unit simply cannot function properly.

The heat output of the PSU is determined simply by the formula (AC watts-DC watts). Aside from a slight increase in the caps generating heat due to dealing with worse ripple, the other components more or less remain the same thermally and hence any increase in watts due to components in the PSU would be slight. A transistor is not going suddenly become a 30 watt guzzler because it'd be dead, and so would a resistor.

A spark likely means the fuse blew and the MOSFETs might have been taken out in the process. This has nothing to do with capacitors degrading and everything to do with unintentionally running the unit out of spec, most likely by drawing too much amperage such that the MOSFETs went over their amperage spec.

 

[Torn Mind]

well, the world is full of idiots. But you are lucky to have found AT. Guaranteed 99% idiot-free.

To be honest, information about PSUs is scarce on the web, as they don't make nearly as much of an interesting conversation point compared to stuff like CPUs or GPUs. I wouldn't blame people too much for not knowing what a capacitor or a resistor is, nor why some PSU are good and some are so bad they can be dangerous.

And yes, there is a lot of marketing hype. just like with any PC part, just .. when there's hype around a "popular" PC part, there will always people ready to tell it like it is, but PSUs are too "dull" to get that kind of attention.

No, the info is not that scarce. Its just that computer PSUs are not the only PSUs out there and the search terms need to be more generic in some aspects(don't only search for computer related things) and more specific when it comes dealing researching the actual components. Searching for "PSU block diagram" is a far more productive search term than many other terms, but if you haven't heard of block diagrams, you wouldn't know to make that search. I didn't know until I tried a project to fix a freaking TV that I finally came across that term.

You won't find a schematic of an ATX unit, but you'll find some PSU schematics for TVs if you look for service manuals of HDTVs.

 

[Slomo4shO]

But you are lucky to have found AT. Guaranteed 99% idiot-free.

Rofl? Every forum is full of idiotic ideologues, this forum is no exception...

 

[SimianR]

Just wanted to chime in here and say that I got my Tri-X 290 today, and I already had an R9 290 reference. Decided to leave it in there and throw the Tri-X in as well, on my Seasonic 750W Gold power supply

With crossfire on during gaming I'm hitting about 600-620W from the wall. I'm not even going to attempt mining though, don't want to blow the thing up

 

[YBS1]

I wouldnt be wrong because those are tested Power at the wall for that system.

I call BS your system used more than 1300w

You can believe whatever you like. I suggest you check out the chart listed in post #21 showing an overclocked 2600k, then keep in mind a 5GHz 3930k will consume ~50% more power than a 2600k. A +5GHz SB-e system will be north of 500 watts before the GPUs even come into play. This is what I didn't realize before it happened to me. It's also well known a heavily overclocked/overvolted skyn3t 780 Classified can consume north of 400 watts.

 

[Torn Mind]

Just because I said temperature and high voltage/current deteriorate electronic components doesn't mean that the two variables need to be conjoined... The initial statement was that capacitor or fan failures are the main factors in PSU deaths.

With that said, I don't have a PhD in electromagnetics so I will leave the discussion of the exact process of how capacitors fail to someone who is more qualified to shoulder such a discussion.

16 V caps can be found on the secondary of PSUs, whether the PSU is 1000 W or at a lowly 400 watt. Capacitors are not limited by the current flowing through them. Capacitors are limited by voltage; too much voltage and their lifespan will shorten signifciantly. It is too much current that is the concern when running near max specs, which means components that are actually sensitive to too much current are the components to worry about. The voltage applied to secondary capacitors is approximately the amount the rail is supposed to deliver, such as 12V, since they do filtering, as the voltage has already been reduced to its necessary valued.

 

[Pandamonia]

You can believe whatever you like. I suggest you check out the chart listed in post #21 showing an overclocked 2600k, then keep in mind a 5GHz 3930k will consume ~50% more power than a 2600k. A +5GHz SB-e system will be north of 500 watts before the GPUs even come into play. This is what I didn't realize before it happened to me. It's also well known a heavily overclocked/overvolted skyn3t 780 Classified can consume north of 400 watts.

Video on youtube showed a peak of 570w in benchmarks on both GPU and CPU.

So id say your 1300w is WAY off.

 

[Pandamonia]

620W is not a common wattage rating. Unless he has the 620 W from Cooler Master(their somewhat decent Silent Pro M line), chances are that it is a Corsair, Antec, or Seasonic. Even Cooler Master's units typically can deliver their rated wattage, but with worse ripple control. These units have continuous ratings, not peak ones, or have you ignored on the JohnnGURU's reviews that praise certain units in order to propagate the falsehood that all PSU companies lie?

A lower wattage unit might have smaller heatsinks, but the fans would be going faster at higher loads. If the same Chemicon capacitor is stuffed in a 620W or a 1000W unit, chances are the rate of heat dissipation is similar.

Also, capacitors do "degrade" but the effects are not such that the maximum amperage that of the whole rail decreases. What happens is that either the waveform becomes more and more distorted to the point the unit simply cannot function properly.

The heat output of the PSU is determined simply by the formula (AC watts-DC watts). Aside from a slight increase in the caps generating heat due to dealing with worse ripple, the other components more or less remain the same thermally and hence any increase in watts due to components in the PSU would be slight. A transistor is not going suddenly become a 30 watt guzzler because it'd be dead, and so would a resistor.

A spark likely means the fuse blew and the MOSFETs might have been taken out in the process. This has nothing to do with capacitors degrading and everything to do with unintentionally running the unit out of spec, most likely by drawing too much amperage such that the MOSFETs went over their amperage spec.

I have a 620w HX from Corsair. Its been solid for my last 3! PC's

 

[TemjinGold]

Video on youtube showed a peak of 570w in benchmarks on both GPU and CPU.

So id say your 1300w is WAY off.

Err... that video is way off then. Heavily OCed SLI 780 and 5+ Ghz cpu consumes way more than 570...

 

[lakedude]

A few points:

Power supplies tend to be most efficient in the middle of their range. Ideally you want to be drawing about half of the max rating.

http://www.overclock3d.net/gfx/articles/2006/11/22163203695l.jpg

I looked up that article and near as I can tell the Unigine's Heaven Benchmark version 4.0 benchmark they ran is not a complete system stress test. It is a gaming style benchmark that simulates a gaming GPU load. It does not really stress the CPU and it does not stress the GPU as hard as something like Furmark or OCCT would. It does seem like a realistic gaming load, it is just not a max power draw stress test.

If you can afford 2x 290s or 2x 780s you can afford a new PS!