HD 6800 Crossfire Scaling

[Arkadrel]

*cough* cost of electricity *cough*

Are you refering to the fact that 2x 6850 use less electricity than a single 470, when compaired underload?

2x 6870s will use more than a single 470 though, but again less than 1x 480.

 

[blastingcap]

Are you refering to the fact that 2x 6850 use less electricity than a singel 470, when compaired underload?

I am referring to GTX 470 eating more electricity than HD 6870 = higher cost of ownership overall. (Although it's almost wintertime and a GTX 470 could theoretically lower heating expenses, winter only comes once per year, and summer air conditioning expenses offset wintertime savings on heating expenses.)

HD 6870 = GTS 450 die size, GTX 460 power consumption, and GTX 470 performance.

The GTX 470 does have extra RAM, though, as well as CUDA/PhysX.

The HD6870 has single-GPU Eyefinity.

Edit because of your edit: I thought we were comparing CF6870 to SLI470, not CF6870 to single 470.

 

[tviceman]

*cough* cost of electricity *cough*

Just out of my own curiosity -

Anand's review shows a 100 watt difference at load between the two cards in the same setup. So 200 watts difference with SLI vs. crossfire, that is 0.2 kwh. I don't know how much energy costs in different parts of the US or the rest of the world but where I live, in the most expensive time of the year (summer), the cost of energy is 7.91 cents per kilowatt hour. So, at most, it's costing someone in my area 1.582 cents more per hour to run a gtx470 SLI setup while playing games than an hd6870.

Just food for thought. If my math is off there, please tell me how I am wrong.

 

[blastingcap]

Just out of my own curiosity -

Anand's review shows a 100 watt difference at load between the two cards in the same setup. So 200 watts difference with SLI vs. crossfire, that is 0.2 kwh. I don't know how much energy costs in different parts of the US or the rest of the world but where I live, in the most expensive time of the year (summer), the cost of energy is 7.91 cents per kilowatt hour. So, at most, it's costing someone in my area 1.582 cents more per hour to run a gtx470 SLI setup while playing games than an hd6870.

Just food for thought. If my math is off there, please tell me how I am wrong.

Luckily for me, I merely have to direct you to the explanatory post of a handsome, dashing, intelligent, and generally all-around good guy, here: http://forums.anandtech.com/showpost.php?p=30507671&postcount=15

Punch in your usage numbers to arrive at your total cost, assuming that the card will not bump you up or down a tier.

In my area, assuming Tier 1 rates, it's more like $.03 more per hour at load, and that's at rock-bottom Tier 1 rates.

This also assumes that one would not have to upgrade a PSU or get an SLI mobo; for some people that number is not zero cost.

Edit, actually it's more like 2.5 cents not 3 cents in tier 1, my bad. We're on the verge of a bump up in rates though, so it will soon be closer to 3 cents. And also, at-the-wall watts are used in calculations of actual bill impact.

 

[Skurge]

Just out of my own curiosity -

Anand's review shows a 100 watt difference at load between the two cards in the same setup. So 200 watts difference with SLI vs. crossfire, that is 0.2 kwh. I don't know how much energy costs in different parts of the US or the rest of the world but where I live, in the most expensive time of the year (summer), the cost of energy is 7.91 cents per kilowatt hour. So, at most, it's costing someone in my area 1.582 cents more per hour to run a gtx470 SLI setup while playing games than an hd6870.

Just food for thought. If my math is off there, please tell me how I am wrong.

Just for reference, electricity here is 90cents per kwh. Thats over 11x more than where you live.

You can do the math with that and see why I always find power consumption a big deal. Also why I never EVER use crossfire or SLI.

 

[tviceman]

Luckily for me, I merely have to direct you to the explanatory post of a handsome, dashing, intelligent, and generally all-around good guy, here: http://forums.anandtech.com/showpost.php?p=30507671&postcount=15

Punch in your usage numbers to arrive at your total cost, assuming that the card will not bump you up or down a tier.

In my area, assuming Tier 1 rates, it's more like $.03 more per hour at load, and that's at rock-bottom Tier 1 rates.

This also assumes that one would not have to upgrade a PSU or get an SLI mobo; for some people that number is not zero cost.

Call me dumb - but how are those formulas derived? Am I then wrong in assuming that 100 more watts at load does equal 0.1 kwh? And why does the computer idling use a different formula for finding power draw in KWH than when the computer is at load?

EDIT: 19.5 + 4.5 is hours in a day. I see. So I don't think there is anything wrong per se with my understanding, I just simplified it down to how much more does it cost when gaming.

 

[Skurge]

Call me dumb - but how are those formulas derived? Am I then wrong in assuming that 100 more watts at load does equal 0.1 kwh? And why does the computer idling use a different formula for finding power draw in KWH than when the computer is at load?

You computer is at idle more than it is at load, but I can't say how it came to those ratios of idle:load.

 

[blastingcap]

Call me dumb - but how are those formulas derived? Am I then wrong in assuming that 100 more watts at load does equal 0.1 kwh? And why does the computer idling use a different formula for finding power draw in KWH than when the computer is at load?

EDIT: 19.5 + 4.5 is hours in a day. I see. So I don't think there is anything wrong per se with my understanding, I just simplified it down to how much more does it cost when gaming.

The poster I was responding to in that thread said he gamed 4-5 hours per day and left his PC on 24/7. hence 4.5 and 19.5. I was responding to his particular numbers but you can modify it according to your own usage patterns (and cost of electricity).

My point about potentially costly PSU and mobo upgrades remains.

 

[tviceman]

The poster I was responding to in that thread said he gamed 4-5 hours per day and left his PC on 24/7. hence 4.5 and 19.5. I was responding to his particular numbers but you can modify it according to your own usage patterns (and cost of electricity).

My point about potentially costly PSU and mobo upgrades remains.

Yeah I see that you did a good job of breaking it down for him.

 

[Skurge]

Just out of my own curiosity -

Anand's review shows a 100 watt difference at load between the two cards in the same setup. So 200 watts difference with SLI vs. crossfire, that is 0.2 kwh. I don't know how much energy costs in different parts of the US or the rest of the world but where I live, in the most expensive time of the year (summer), the cost of energy is 7.91 cents per kilowatt hour. So, at most, it's costing someone in my area 1.582 cents more per hour to run a gtx470 SLI setup while playing games than an hd6870.

Just food for thought. If my math is off there, please tell me how I am wrong.

Oh and congratulations on your 1000th post!

 

[tviceman]

Just for reference, electricity here is 90cents per kwh. Thats over 11x more than where you live.

You can do the math with that and see why I always find power consumption a big deal. Also why I never EVER use crossfire or SLI.

That is incredibly high. Wow. On the same note, I've never had an SLI or crossfire setup myself.

Oh and congratulations on your 1000th post!

Look at me go!

 

[nyker96]

Comparing to 5xxx ATI has done a great job improving the Xfire scaling with the 6xxx series. It's on par with NV now. Excellent optimizations.

And a big thanks to OP to put in the hard work of collections stats. Wonderful work!

 

[MobiusPizza]

Referring to:
http://www.hexus.net/content/item.php?item=27053&page=16

It seems that with the good scaling with Crossfire with HD6850, HD6850 CF is a very attractive alternative to a single high end card. If you want performance now that surpasses HD5870, but not as much as dual CF HD5870:

Economy of scale determines that higher end products has less bang for bucks. Hexus shows that HD5870 is at 0.726. Given that HD6850 increases performance by 83% going Crossfire, the price/performance ratio of cross fire dual HD6850 is 1.006*0.83 = 0.835. This is very good value for a high performing combination that beats single GTX480.

 

[Leyawiin]

This is great news. Have a crossfire mobo and if I ever get a monitor that makes it worthwhile it'll be nice to have a couple of good, relatively inexpensive choices (particularly the HD 6850) to drop in.

 

[notty22]

Anyone find a site that might dare to be different and test the new cards in crossfire on a AMD chipset, obviously with a AMD cpu. One of the new 890fx 16x 16x boards. I didn't spot one.

 

[Grooveriding]

Anyone find a site that might dare to be different and test the new cards in crossfire on a AMD chipset, obviously with a AMD cpu. One of the new 890fx 16x 16x boards. I didn't spot one.

I don't think I ever see many sites bother using AMD cpus. They're not all that great, the difference is small between AMD and Intel once you push your settings to GPU bottleneck level, but leading up to that point AMD cpus will hold back some multi-gpu settings at lower resolutions and settings.

 

[tannat]

For more data:
Another site with crossfire/SLI results (unfortunately in swedish):

http://www.sweclockers.com/recension/12873-amd-radeon-hd-6870-och-hd-6850

Note that the EVGA GTX460 FTW for some reason show up in this review also. Although the card isn't even sold in Sweden

 

[Xzyrus]

For more data:
Another site with crossfire/SLI results (unfortunately in swedish):

http://www.sweclockers.com/recension/12873-amd-radeon-hd-6870-och-hd-6850

Note that the EVGA GTX460 FTW for some reason show up in this review also. Although the card isn't even sold in Sweden

Thanks for the link! It also has a nice collection of 6850s so we can compare their coolers being tested via the same method in the same environment!:thumbsup:

 

[cusideabelincoln]

For more data:
Another site with crossfire/SLI results (unfortunately in swedish):

http://www.sweclockers.com/recension...70-och-hd-6850

Note that the EVGA GTX460 FTW for some reason show up in this review also. Although the card isn't even sold in Sweden

Ok those have been added and their scores are reflected in the Global Averages too.

In summary they show the same trend as every other review. HD 6800 scaling > GTX 400 scaling > HD 5000 scaling.

 

[MrK6]

I've suspected for quite awhile that the 5xxx series scaling was a hardware limitation. The 4xxx series scaled much better, and it looks like the 6xxx was rebalanced so it scales better as well. I'm tempted to pick up two 6850's to see if AMD did anything to fix multi-GPU microstuttering (probably not), but with Cayman so close I'll just be getting the 6970. Thanks OP for doing all the homework and posting it :thumbsup:.

 

[Grooveriding]

I've suspected for quite awhile that the 5xxx series scaling was a hardware limitation. The 4xxx series scaled much better, and it looks like the 6xxx was rebalanced so it scales better as well. I'm tempted to pick up two 6850's to see if AMD did anything to fix multi-GPU microstuttering (probably not), but with Cayman so close I'll just be getting the 6970. Thanks OP for doing all the homework and posting it :thumbsup:.

Yes very nice info. Looks like the 6990 is going to be even faster now with this boost. Likely 6970 CF will still best it due to power restraints on the 6990, but this improved scaling is going to give the high end a nice boost like it has the mid-range with the 68XX series.

 

[Arkadrel]

Mrk6 actually I think the mircostuttering was fixed in one of the last few catalyst drivers.


6870x2 draws less power than 1x480 = lol
6850x2 draws less power than 1x470 = lol

fermi has some serious power issues.... hopefully those will be gone when nvidia move on to 28 nm.

 

[darckhart]

thx for compiling. loving how 6 series is turning out. cant wait until next month!

 

[Sylvanas]

Thanks for the excellent work in compiling all of this cusideabelincoln :thumbsup: Bookmarked for later reference.

 

[Grooveriding]

http://www.kitguru.net/components/graphic-cards/zardon/his-hd6870-crossfirex-review/7/

Some more scaling data in this review. This is the first review I've seen showing the 850core 460 FTW models against stock 6870s in Crossfire. The 6870s at stock are still faster in 6 out of the 7 games benched.