Topic: 4x 5850 Setup, Minimum PSU needed? (Read 6699 times)

Sweet thx bud

In theory, that should handle 2x5830 and 1x5770... *Maybe* even another 5830 (but won't leave much headroom), depending on which MB/CPU you're using.

But 99% sure you're fine, even if you OC the 5830's and 5770's.

Cheers,
Kermee

Its this one http://www.newegg.com/Product/Product.aspx?Item=N82E16817182191
Thx

Depends on which 700W PSU it is

Cheers,
Kermee

So is 700w good enuff for 2x5830 and a 5770

Quite a few of them have similar problems actually. However for most users, checking up the 80+ list is the first step. If it's not even there, then obviously it's not even going to cut it. After that, it's quite difficult to determine which are consistently good and which are not. Which was why in one of my previous jobs, I was in charge of testing various PSU before committing the company to buying one or the other. Wish bitcoin existed then, I could had figured out a way to convince them to use computers instead of load testers and mined quite a few BTC while doing that job

There was at least one very well known manufacturer which had PSU that did not pass muster to my surprise then. So keep in mind that just because a site says a PSU is good, doesn't exclude the possibility that it was a golden sample, or in some cases I had another reputable manufacturer switching to cheaper parts on products (not a PSU) and a not so reputable one doing the same for PSU after the first few excellent batches.

However despite all these, just because a review site (or end user testing) claims that a particular PSU does not meet 80+ standard does not mean the PSU isn't designed and manufactured up to the certified standard. IIRC 80+ tests PSU in very strict environmental conditions, which does not match most review sites conditions. E.g. environment temperature controlled at 25C. Due to thermal derating (manufacturers will provide a chart for their products if you have good reasons to ask), the way most sites test PSU by putting it in a realistically hot environment will decrease the max load and efficiency of the PSU.

Furthermore, 80+ also calculates wire losses in their certification. So modular PSU as well as longer cables means the calculated efficiency is higher than measured. Although this should be a less than 1% difference.

So the only thing the review sites can tell you is at best the relative performance of PSU to one another on their test system using their methodologies. For most of us, it would be too much time spent just to read up on every available PSU, compare between them for each site to come up with the best buy.

So easiest way is still to refer to the 80+ chart, take a quick look at what people are using here without problems and ignore whatever 1~2% difference things might make. It'd probably be made up by the BTC you could mine while doing all that research

Yes, what you are saying is right, in theory. However, if you test them in practice some will actually not meet the said ceritification. F.ex the in-win 1200W psu have a slight problem at full load not delivering more than 78-79% efficiency (Add to that it has problem even delivering 1200Watts...). Might be the certification, might be a bit faulty parts here and there, but the fact is that in real use they do not deliver advertised 80+ standards. Now that might not be a problem at all for some, but to me it's a red flag for build quality and reliability.

1000W would be safe with some headroom, better safe then sorry if its a lower quality PSU and can't generate 1000W reliably after year+ of dedicated mining.

It could just be the motherboard not being happy with two cards. Have you tried different slots and different slot order? How old is the TX650 anyway? PSU capacity will degrade over time, faster if the environment is hotter.

That's not how the 80+ certification works. In order to get a real 80+ certification, the PSU must pass efficiency tests at 20%, 50% (high efficiency is required at this load) and 100% load. If a PSU advertises gold efficiency only at one load level, it's a fake certification. There are some brands that had advertised and put the 80+ logo on PSU that were NEVER actually certified because they know it sells better.

To verify the model you are looking at is really certified, check out the certification website
http://www.plugloadsolutions.com/80PlusPowerSupplies.aspx

Honestly,I don't believe any of the hype in which multiple videocards can work in a system.I have just recently gotten a Radeon 6950 and was trying to get my old 5770 to work in the same system as a seperate videocard but it never wanted to boot whenever I tested the system.I have a TX650W Corsair PSU which didn't seem to like the fact of both videocards being in the system and had to take it out as I thought the power requirements wouldn't be that high for both cards.

I would recommend getting at least a 1000 watt PSU to be on the safe side as it may also have more PCI-E 6 pin connectors for multiple GPU combinations.

I'm running 3 x 5850s overclocked, and my entire rig is taking up about 750W, so I would really recommend a 1000W PSU minimum.

The ABSOLUTE minimum I would go for would be a good single rail 850W. That is with no GPU overclock OR no CPU/MB overclock. 5850's draw roughly no more than 150watts per card really, not clocked. 150W*4=600W, 200W for the other components (unless you have 10x0.8A fans and 10xHDD's, 200W is much for a system - GPU, even overclocked).

If you however want to be safe go for a 1050W PSU (Less likely to hit a bad brand vs 1000W).

Just to add to the PSU wave ~

Get a PSU which is Gold to Bronze certified or whatever it is called. Always read reviews about them. They are advertised with high efficiency, BUT BUT BUT you can advertise with Gold efficiency on a 1000W PSU while meeting the Gold efficiency requirements when the PSU deliver 500W and not at 1000W, and sometimes the other way around!!!

Find 2-3 PSU's you would be willing to get, then reconsider and choose better ones (if money is no hinder), then read the reviews for all 4-6 PSU's to find what fits your purpose the best.

I can also recommend having 2x PSU if your cabinet have the space for it. F.ex. 2 gold efficiency 650W should be cheaper than 1 gold efficiency 1200W PSU, delta temp should be lower (meaning lower temp per PSU, but higher average temp because of 2 units), noise could become higher as dB stack and generally you will have a better (read; more reliable) way of supplying your computer with power.

What's the breakage on gold vs bronze though? About 5% difference in efficiency?
So for a 800W load

• 85% efficiency = 941W
• 90% efficiency = 889W (delta of 52W)

At 0.15/kwh, this means $0.1872/day in savings for a total of ~$68/year
At 0.10/kwh (average where I live), that means $0.1248/day or ~$46/year

Price difference between Coolermaster 80PLUS Bronze and Gold is about 75, so you break even in over a year or two depending on your power costs. Given that a PSU will last 3-5 years or longer, this is probably a smart investment!

Bingo. Excellent point. Didn't realize it at the time of my post.

Cheers,
Kermee

You could probably get away with an 850watt, but I would buy at least 1000watt.
You don't want to be running your power supply anywhere near 100%.... it'll just act like a hair dryer and will no doubt have a short life.
 

http://www.google.com

whats a PSU

Most PSU are built with some tolerance, especially the lower capacity models in the same series based off the same PCB/design. But it's not ideal to push a PSU constantly at 90% or higher, especially if the ambient temperature is hot due to the cards.

If the PSU dies in a bad way, it can damage connected equipment.

~20W difference downclocking only memory down to 300Mhz on both my 5870 and 5850