Originally posted by: SteelSix
Originally posted by: taltamir
is ATI still the underdog? they now have cards that outperform nvidia for less money, they have a whole platform, while NV is left out in the cold, they are locking into their own platform, while intel is bringing larrabelalalablee or whatever to fortify their platform.
NV is up in the creek with no escape.
If anything, nvidia is the new underdog.
I keep looking at the 280's power advantages over ATI's winners though. Hell, I was going to CF 4870's until I saw that power graph. I couldn't believe it actually! Talt, you're an expert on power, what's your take when comparing the two in that aspect?
alright.
http://techreport.com/articles.x/14990/15
4850 CF - 4850 = idle wattage of 4850 = 55watt idle
4850 total idle wattage - 4850 only idle wattage = bare system wattage = 72 watts
GTX 280 - 72 watts = 50 watts idle.
Load wattages:
4850 CF - 4850 = load wattage of 4850 = 128 watts
4850 total wattage - 4850 wattage = base system wattage in load = 108 watts
Verification: 4850 CF - 108 watts = 256 watts, exactly twice that of a single 4850. The math checks out. Meaning that they measured correctly. (some sites the math does not check out, which means that there was something wrong with their measuring methology).
GTX 280 load - bare system load = GTX only Load = 191 watts in load.
So:
4850: 55 watt idle / 128 watt load
4850 CF: 110 watt idle / 256 watt load
GTX 280: 50 watt idle / 191 watt load
Three scenarios:
1. A person who leaves the system on 24/7/365, idling all the time, plays games 2 hours a day on average (on average including ALL days)
2. A person who uses the system for 10 hours a day for 2d work, and 2 hours a day gaming (me), turns it off for the rest.
3. A person who runs folding at home on both GPU and CPU 24/7/365
4850:
2 hours per day of load: 2 hours/day * 365 days/year * 128 watts * 0.001 KW/watt = 93.44 kwh/year
10 hours per day of idle: 10 hours/day * 365 days/year * 55 watts * 0.001 KW/watt = 200.75 kwh/year
22 hours per day of idle: 22 hours/day * 365 days/year * 55 watts * 0.001 KW/watt = 441.65 kwh/year
24 hours per day of load: 24 hours/day * 365 days/year * 128 watts * 0.001 KW/watt = 1121.28 kwh/year
Scenario 1: 2 hours of play + 22 hours of idle = 93.44 kwh/year + 441.65 kwh/year = 535.09 kwh/year
Scenario 2: 2 hours of play + 10 hours of idle = 93.44 kwh/year + 200.75 kwh/year = 294.19 kwh/year
Scenario 3: 1121.28 kwh/year
4850 CF:
Scenario 1: 1070.18 kwh/year
Scenario 2: 588.38 kwh/year
Scenario 3: 2242.56 kwh/year
280GTX:
2 hours per day of load: 2 hours/day * 365 days/year * 191 watts * 0.001 KW/watt = 139.43 kwh/year
10 hours per day of idle: 10 hours/day * 365 days/year * 50 watts * 0.001 KW/watt = 182.5 kwh/year
22 hours per day of idle: 22 hours/day * 365 days/year * 50 watts * 0.001 KW/watt = 401.5 kwh/year
24 hours per day of load: 24 hours/day * 365 days/year * 191 watts * 0.001 KW/watt = 1673.16 kwh/year
Scenario 1: 2 hours of play + 22 hours of idle = 139.43 kwh/year + 401.5 kwh/year = 540.93 kwh/year
Scenario 2: 2 hours of play + 10 hours of idle = 139.43 kwh/year + 182.5 kwh/year = 321.93 kwh/year
Scenario 3: 1673.16 kwh/year
KWH hours range from 6 cents per KWH to 24 cents per KWH. The average is supposedly 9. In texas we pay 14. (after all the hidden charges and taxas that is).
I will calculate for texas since this is where I live.
I will start with scenario 1 and 3, because they are rare, a common user will be most likely scenario 2, scenario 2 is based on me in real life.
Scenario 1:
4850: 535.09 kwh/year * 0.14$/kwh = 74.9126$/year
4850 CF: 149.8252 $/year
GTX 280: 540.93 kwh/year * 0.14$/kwh = 75.7302 $/year
Scenario 3:
4850: 1121.28 kwh/year * 0.14$/kwh = 156.9792$/year
4850 CF: 313.9584 $/year
GTX 280: 1673.16 kwh/year * 0.14$/kwh = 234.2424 $/year
Scenario 2 @14 cent/KWH (this is me):
4850: 294.19 kwh/year * 0.14$/kwh = 41.1866$/year
4850 CF: 82.3732 $/year
GTX 280: 321.93 kwh/year * 0.14$/kwh = 45.0702 $/year
Scenario 2 @ 7 cent/KWH:
4850: 294.19 kwh/year * 0.07$/kwh = 20.5933$/year
4850 CF: 41.1866 $/year
GTX 280: 321.93 kwh/year * 0.07$/kwh = 22.5351 $/year
Scenario 2 @ 24 cent/KWH:
4850: 294.19 kwh/year * 0.24$/kwh = 70.6056$/year
4850 CF: 141.2112 $/year
GTX 280: 321.93 kwh/year * 0.24$/kwh = 77.2632 $/year
Delta of scenario 2 @ 14 cent/kwh: GTX 280 is 37.303$ per year cheaper than 4850CF per year in direct electricity cost for running the card itself.
Delta of scenario 2 @ 7 cent/kwh: GTX 280 is 18.6515$ per year cheaper then 4850CF per year in direct electricity cost for running the cards themselves.
Delta of scenario 2 @ 24 cent/kwh: GTX 280 is 63.948$ per year cheaper then 4850CF per year in direct electricity cost for running the card itself.
This is under normal operation, not under something crazy like 24/7 operation.
The reason I said in electricity cost per card itself, is because they also generate heat equivalent to an electric heater of a similar wattage, generally speaking, cooling with an AC is extremely inefficient compared to electric heating, it should cost 3 times the listed dollar amounts to cool the room (in the winter, it decreases your heating cost by almost that amount. Or less if you use something other then electricity to heat up your room.
Delta accounting for AC, for me. I cool for about 9 months of the year, i heat up for about 3. (texas is hot as hell, and I have Russian blood). So I cool for 9/12 of the year, aka 0.75, and I heat for 0.25 of the year. It costs 3 times as much to cool.
So it is 0.75 * 3 - 0.25 = 2
I need to add a 1 for the cost of running the card itself in additional to AC savings + cost increase.
End result. total cost for me is 3x the electricity demand of the card alone. It is probably a bit higher still,
because electricity costs more money in the summer then in the winter.
But really, no need to get that specific. Lets just take the delta and multiply it by 3.
So for me personally: the GTX would be cheaper then the 4850CF to operate by:
37.303 * 3 = 111.909$/year in electricity for the card and the AC, minus the reduced costs of heating during the winter, not accounting for the price fluctuation of energy during summer and winter (which will make it slightly higher)
Thinking again, it will probably be even higher still.. since I spend 12 hours a day next to the computer, I will feel the heat more strongly, that means that during winter i am less likely to turn on heating, and during summer i am more likely to turn the AC cooling lower, decreasing the temperature in the entire apartment, because I am getting too hot from sitting right next to the heat chucker.
I reduced my overall power consumption by 50$+ a month ever since I started using the computer in my underwear and with the doors to the computer room open wide (for air circulation with the rest of the house) and increasing the house temp by 4 degrees. (during the winter i wear a coat and heavy clothes instead).