Topic: RX480(580) cards real power draw, 1000 watts psu for a 5-6-7 RX 480 builds ? (Read 975 times)

Really great analysis but it is safe to say that Corsair Rm1000i which I am using is safely running from January with 4 RX 480 8GB Sapphire Nitro+ and 1 RX 580 8GB from Aorus. I have not got into much details as you have but I have only increased memory clock and core clock from the MSI Afterburner utility and slowing down power limit to 90%. The rig is running from January as I say without any problem and is on a Gigabyte Gaming Z97 chipset. I am using 5 risers so I am not using directly the massive PCIEX16 slot. Total consumption in MSI is about 700 watt for the cards and I guess the board and other parts another 100 watt, so in total I am using 800 watt with 5 cards. Probably this PSU can keep another card to make 950 but its lifespan would be greatly reduced then.

So yes I insist a 1000 watt PSU is more than capable of running 5 RX 4XX series (should do 5XX too but they consume a bit more) without problem running 24/7 from 10 months almost now.

Yeah, I guess you got me pegged pretty well. I do have plenty of rigs with 6 cards and every single card has a slightly different voltage set, something like: 880, 885, 902, 887, 890, 892. I strive to run each card at its most efficient while at the same time providing the max stability. I don't like rebooting rigs everyday either and will give up a percentage of hashrate or power efficiency if it means more long term stability. I also like to use high efficiency PSUs, run them in their most efficient band, and even make my own heavy gauge PSU cables. I look at RAM efficiency, don't need 8 GB sucking power needlessly, CPU efficiency, and an often overlooked area, disabling unneeded motherboard accessories such as sound cards, serial, printer ports, extra USB or sata controllers, and so on.  It may only knock off a watt or two per rig, but for me that is part of the thrill. How efficient can I make this entire rig. Gaining 10-20 watts per rig in efficiency times a large number of rigs is like running a few rigs for free.

Getting back to stability, I learned a long time ago while it may be nice seeing all your RX580's displaying 32 Mh/s, if it is only good for a few hours and often hangs or reboots you lose more during the resets than you do by dropping the clocks a bit and stabilizing the rig. I have many rigs with 1600+ hours of run-times and I probably would be higher except I did a mass upgrade to the blockchain drivers a couple of months back. I think my highest run-time before that was like 3600 hours or 150 days and I think that was only cut short due to an early morning power outage.

So while I do like efficiency, I like stability more, but that doesn't mean you just toss some high voltages at it and forget it. I think the problem more has to do with lack of patience, as you said, with most newcomers. They just want someone to give them boilerplate settings and be done with it. I kind of think it more of a challenge and similar to tinkering with a car to get an extra 5 horsepower out of the engine, I tinker with GPU settings to get the most from each and every one of them. Over the years as my rig collection has grown I cannot tune them all as much as I would like, but each new rig spends a week or two on the test bench and I like to think they are pretty well optimized before they leave for the farm racks.

You show the results of the GPU with the original BIOS. I have all the GPU is a modified BIOS. The speed of mining Ethereum 29 Mh/s. The maximum consumption I have 110 watts. This is when dual mining ETH+DCR. If only mine ETH the energy consumption drops to 88 watts. My rig is a 6 GPU and I use a 1000 watt power supply.

True but most of the time the card is not drawing anywhere near 225W anyway. Here are two separate tests where they measured the power draw from the power connector and motherboard separately:

http://www.tomshardware.com/reviews/amd-radeon-rx-480-polaris-power-fix,4668.html

https://www.pcper.com/reviews/Graphics-Cards/AMD-Radeon-RX-480-Power-Consumption-Concerns-Fixed-1671-Driver/Power-Testing-

In both tests the total power draw is around 160W - 170W. That's much closer to the AB readings xxcsu was showing earlier. If the card was properly undervolted then the power draw should be even lower than 170W. On the other hand the above tests measured the reference version of RX480. The card that xxcsu has could be a custom version and might draw more power.

Many miners or newcomers starting to mine right now are not paranoid as you or me. So most of them leave their cards with plenty of volts cause they want maximum uptime and less restarts, to have an ultra efficient mining computer you need patience to actually understand that every 0.01 mv is important to save or give to make it stable. I dont mine but my friends pay me a lot to make their computers as efficient as possible in every sense. For them i'm a freak of nature in efficiency.

Good overview of the process and I am sure this post will prove to be valuable resource for a lot of newcomers. I think the key takeaway is you can certainly run 5 or 6 GPUs with a 1000 watts PSU, but if you have no idea what you are doing start out with only 3 of 4 GPU's installed until you make the mods and voltage adjustments as shown above. Straight out of the box with no adjustments, even 4 cards will be pushing the limits a 1000 watt PSU.

One last item, I think in your testing you left off at a kind of a high GPU voltage (at least for ETH only mining). I have many cards running successfully as low as 875 mv or 900 mv while still maintaining a healthy 30 Mh/s hashrate (RX580 8 GB). This will reduce the overall rig power consumption even further as each card will be measured a bit over 100 watts at the wall and 66-70 watts in afterburner or GPU-Z. That is with ETH only, for dual mining you will probably need to keep it higher and leave it around 950 mv and you will probably be drawing closer 125 Watts per card from the wall, which is close to your second from the last example.

for your 1 point, your answer might be your point 3, 253w at wall however the card is using much less. The 225w limit is not at wall, the 225w limit is 100% efficiency, 225w is the max the card uses, so your point 3 gives a good example of psu efficiency.

Good analysis and while I don't disagree with AB power draw numbers being misleading, there are a couple of points to consider:

1) Your RX480 has a single 8-pin power connector. According to spec the max power draw via 8-pin power connector is 150W. On top of that the card can draw 75W via PCI-E slot. So in total 225W.  In your test you calculated your card is pulling 253W, which is impossible because it would go way over the spec.

2) You can't calculate the power draw of the card by subtracting 60W (idle consumption) from the total power draw because the rest of the system is not idle when mining. When you start mining, your CPU, ram, motherboard, etc. is also working harder and pulling more power vs. idle power draw.

3) You didn't take into account the efficiency of your power supply. If Kill A Watt shows 271 W and the efficiency of your PSU for that load is (for example) 85% then your system is only pulling 271W * 0.85 = 230W.

I have done some pro efficiency on my friends mining computers and 6 rx 480 or 580 using only 1020w at wall which means, all of them have same voltage settings applied, cards differ so most cards i had to underclock them to make them stable without crashes. The good thing all use same voltage and all have pretty much same power at wall, mining computers are optimized to use the bare minimum of power x efficiency, so 1020 divided by 6 = 170w each dual mining, eth only 130w each, dual mine uses 33% more electricity. That was my best without affecting too much on hashrates. So in truth 1020w at wall means 910w, I still have 90w to use if i want to on the 1000w, just to clarify a 1000w power supply at 100% usage means 1100W at wall, so I was well below that.

xxcsu example was good, usually noobs only do few modifications and forget all other important things, they think they are using 120 watts but in reality each of their gpus might be using 300 watts, so that is why hashrates drop so much cause the power supply is not enough for such load. On the rx 480 was not much however on rx 580 hashrate drops is more frequent cause at same voltage as rx 480, the rx 580 consumes a lot lot more power.

And the funny thing is as they are noobs, they do not account for the 2x power, they think 120w and the real is almost 300w. So even their roi is wrong and in the end they end up paying 2x more on the electricity bill and their roi jumps to 6 months more or even more, plus there are many other things, so if noobs see a return of investment of 18 months, that means around 24 months minimum which could prolong to 30 months and even 36 months to get your money back.

Remember that net profit is very low at moment, electricity is the key here, so if your system is not efficient then forget it. 36 months minimum and your roi might never hit it cause your gpus might die before your get the money back and most warranties are valid only for 12 months, remember that if one gpu dies then you need all other gpus to get the money back of that gpu that just died, see where this is going?