670 Reviews are up

[Gloomy]

Really? I don't remember seeing two of those three titles in Tom's review, which is who I'm referring to if you look back at the beginning of the discussion.

Now, TPU includes a lot of games, which is why I praised them for it.

Okay, that makes a lot more sense.

 

[LOL_Wut_Axel]

I think it'd be more due to memory but in any case, the whole 7970 energy guzzler reptuation thing is overblown. If you don't overvolt much or at all, the 7970's power draw doesn't grow nearly as much. (Stock vs stock a 7970 only draws 19 more watts at load than a gtx670 despite the HPC silicon and extra 1GB of GDDR5 VRAM.) The 7970 also has lower idle wattage due to ZeroCore so if you leave your PC idling a lot, it will offset your load power somewhat. http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_670/26.html The 670 is still more energy efficient but it's not like a night-and-day comparison stock voltage vs stock voltage.

It's a pretty big difference.

24%, stock. Overclocked, the difference will be much bigger, especially if overvolting is included into the equation.

But again, the HD 7950 redeems itself in this metric. You'd need a 975MHz HD 7950 to match a stock (925MHz) HD 7970, and power consumption should only increase some 15-20W from that because you do not need to over-volt for that. It'd still be some 35-40W lower than a stock HD 7970, so yes, it's mostly a case of the 7970 in particular using a good bit amount of power. The HD 7950 is a lot more efficient, even overclocked.

 

[IlllI]

anyone have an opinion about which 'short' 670 is the best? None really seemed to stick out from the rest from what I saw, but I might be thinking about removing the stock heatsink and shroud and trying to fit into a sff. The extra tiny PCB might come in handy.

 

[blastingcap]

It's a pretty big difference.

24%, stock. Overclocked, the difference will be much bigger, especially if overvolting is included into the equation.

But again, the HD 7950 redeems itself in this metric. You'd need a 975MHz HD 7950 to match a stock (925MHz) HD 7970, and power consumption should only increase some 15-20W from that because you do not need to over-volt for that. It'd still be some 35-40W lower than a stock HD 7970, so yes, it's mostly a case of the 7970 in particular using a good bit amount of power. The HD 7950 is a lot more efficient, even overclocked.

Use average not peak. See their definitions for why peak only matters when picking p s u size

 

[lavaheadache]

It's a pretty big difference.

24%, stock. Overclocked, the difference will be much bigger, especially if overvolting is included into the equation.

But again, the HD 7950 redeems itself in this metric. You'd need a 975MHz HD 7950 to match a stock (925MHz) HD 7970, and power consumption should only increase some 15-20W from that because you do not need to over-volt for that. It'd still be some 35-40W lower than a stock HD 7970, so yes, it's mostly a case of the 7970 in particular using a good bit amount of power. The HD 7950 is a lot more efficient, even overclocked.

I don't understand why you guys are even making a fuss about this.

We are taking about low double digit wattage, as if it makes any difference in anything but a nit pick fight. Who honestly cares if one card consumes 35 more watts while gaming? What is that going to equate to during regular usage?

Do you guys penny pinch like this when it comes to gas mileage and cars since that can actually make a noticeable impact on your wallet?

 

[Stuka87]

anyone have an opinion about which 'short' 670 is the best? None really seemed to stick out from the rest from what I saw, but I might be thinking about removing the stock heatsink and shroud and trying to fit into a sff. The extra tiny PCB might come in handy.

The short cards are reference, so they should all be identical. So it comes down to warranty and such.

What would you replace the stock cooler with? I cannot think of any after market coolers that would fit, with the exception of a water block.

 

[3DVagabond]

You are 100% correct here. The issue is if you decide to OC your cards. Kepler will not ramp-up power usage much, while you gain 100w on the 79xx cards. Not a big deal to some people, and definitely not a issue at all for those keeping their cards at stock.

AMD also has better long-idle power consumption too, IIRC.

The 7970 uses that kind of power when overvolted. I'm sure that if Kepler's voltage was unlocked we could dump the kind of voltage into it to cause it's power numbers to go through the roof. That's if it would hold together with that kind of power running through it. Considering nVidia doesn't allow overvolting on them, I have my doubts they could take it.

 

[Piano Man]

Is it a problem that my pants are moist after reading TPU's review of the ASUS's 670 dcu ii?

http://www.techpowerup.com/reviews/ASUS/GeForce_GTX_670_Direct_Cu_II/1.html

GTX 680 performance out of the box @ 25 dba and an additional 10% OC performance headroom.

COME TO DADDY!

 

[N4n45h1]

Is it a problem that my pants are moist after reading TPU's review of the ASUS's 670 dcu ii?

http://www.techpowerup.com/reviews/ASUS/GeForce_GTX_670_Direct_Cu_II/1.html

GTX 680 performance out of the box @ 25 dba and an additional 10% OC performance headroom.

COME TO DADDY!

I have the same reaction. I can't wait for this to appear on Newegg or Amazon

Anyone have any idea when Asus is coming to the party? I will admit that Diablo 3 has "justified" this buy to me though... :biggrin:

 

[jzmagic]

I'm not sure if I should return my Asus 7850 which just arrived today for the 670. Asus made another monster.

 

[LOL_Wut_Axel]

Use average not peak. See their definitions for why peak only matters when picking p s u size

Peak: Crysis 2 at 1920x1200, Extreme profile, representing a typical gaming power draw. Highest single reading during the test.

How is that unrealistic???

Also, using average the difference is still about the same in percentage.

 

[Don Karnage]

I'm not sure if I should return my Asus 7850 which just arrived today for the 670. Asus made another monster.

Asus isn't available anywhere so hold on for a few days

 

[vhx]

Waiting for the 4GB model then I'm going to pounce.

 

[blastingcap]

How is that unrealistic???

Also, using average the difference is still about the same in percentage.

Do you seriously need me to spell this out for you? Average is what matters as that is over time.

Taking your logic to the extreme, if a card averages 1 watt for an hour but has one spike up to 100 watts for a fraction of a second before settling back down to 1 watt for an hour again, then that card consumes more power than another card that averages 90 watts for an hour with no variation.

No. It doesn't work that way. The first card consumes ~89 watts less than the second card, regardless of any random spikes here or there. I don't care if you don't believe the math. Your electric company surely agrees with my math, and they are who matters.

It's 19w difference average for the cards they measured for the game they measured. Peak only matters to determine what size PSU you need. When comparing power draws, you need to compare averages.

By the way: power draw on average across many many games will be different than what a one-game comparison says. Power draw could be more or less depending on the particular card and binning of the chip, as there isn't even a single "stock" voltage anymore with these newfangled GPUs). See, e.g., HardOCP for how energy usage varies depending on the game: http://www.hardocp.com/article/2012/03/22/nvidia_kepler_gpu_geforce_gtx_680_video_card_review/13

 

[exar333]

Do you seriously need me to spell this out for you? Average is what matters as that is over time.

Taking your logic to the extreme, if a card averages 1 watt for an hour but has one spike up to 100 watts for a fraction of a second before settling back down to 1 watt for an hour again, then that card consumes more power than another card that averages 90 watts for an hour with no variation.

No. It doesn't work that way. The first card consumes ~89 watts less than the second card, regardless of any random spikes here or there. I don't care if you don't believe the math. Your electric company surely agrees with my math, and they are who matters.

It's 19w difference average for the cards they measured for the game they measured (could be more or less depending on the particular card and binning of the chip, as there isn't even a single "stock" voltage anymore with these newfangled GPUs). See, e.g., HardOCP for how energy usage varies depending on the game, too: http://www.hardocp.com/article/2012/03/22/nvidia_kepler_gpu_geforce_gtx_680_video_card_review/13

Peak only matters to determine what size PSU you need. When comparing power draws, you need to compare averages.

Little bit or a reach there. Next thing you are going to say is that AMD users tend to shut off their computers more at night and they actually use LESS energy than NV.

 

[OILFIELDTRASH]

It's a pretty big difference.

24%, stock. Overclocked, the difference will be much bigger, especially if overvolting is included into the equation.

But again, the HD 7950 redeems itself in this metric. You'd need a 975MHz HD 7950 to match a stock (925MHz) HD 7970, and power consumption should only increase some 15-20W from that because you do not need to over-volt for that. It'd still be some 35-40W lower than a stock HD 7970, so yes, it's mostly a case of the 7970 in particular using a good bit amount of power. The HD 7950 is a lot more efficient, even overclocked.

Holy smoke. My 670 is gonna use 120w less than my 480 & kick the 480's ass all over the place. Yes!

 

[blastingcap]

Little bit or a reach there. Next thing you are going to say is that AMD users tend to shut off their computers more at night and they actually use LESS energy than NV.

If you don't understand my explanation of wattage and why it makes more sense to look at average and not peak, I don't even know what to say. Because this has nothing to do with AMD vs NV. I'm sorry if you view the world in a AMD vs NV way. Not all of us do.

You never responded to my counterpoint to your wrong statement that the gtx 670 beats the 7970 at 5760x1200; you also did not respond to my other comment. http://forums.anandtech.com/showpost.php?p=33417201&postcount=180 Apparently you think adding more VRAM, all else equal, DECREASES power consumption. Lmao. Do you mind if I quote you in my signature line?

 

[exar333]

Holy smoke. My 670 is gonna use 120w less than my 480 & kick the 480's ass all over the place. Yes!

Yeah, thinking along those same lines. Power should be very close to my 5870, but probably will use a little less because my 5870 is overclocked and uses more than stock anyway. Worst-case, a wash for power and a lot more FPS.

 

[ocre]

I don't understand why you guys are even making a fuss about this.

We are taking about low double digit wattage, as if it makes any difference in anything but a nit pick fight. Who honestly cares if one card consumes 35 more watts while gaming? What is that going to equate to during regular usage?

Do you guys penny pinch like this when it comes to gas mileage and cars since that can actually make a noticeable impact on your wallet?

Why was it ever? It started a few yrs back. performance per watt became a selling point. Why are you surprised by this???

to put it in contrast, using the same averages the gtx 570 uses 8 more watts (4%) than the the 6970. Efficiency was repeatedly used as a selling point. It was a big deal not so long ago. Now the shoe is on the other foot, and the difference is at least 20% and your asking why?

Why was it ever then?

Personally i could care less about it. I never bought into that stuff. It was an everyday topic for the past couple yrs. How are you surprised it is today?

 

[LOL_Wut_Axel]

Do you seriously need me to spell this out for you? Average is what matters as that is over time.

Taking your logic to the extreme, if a card averages 1 watt for an hour but has one spike up to 100 watts for a fraction of a second before settling back down to 1 watt for an hour again, then that card consumes more power than another card that averages 90 watts for an hour with no variation.

No. It doesn't work that way. The first card consumes ~89 watts less than the second card, regardless of any random spikes here or there. I don't care if you don't believe the math. Your electric company surely agrees with my math, and they are who matters.

It's 19w difference average for the cards they measured for the game they measured. Peak only matters to determine what size PSU you need. When comparing power draws, you need to compare averages.

By the way: power draw on average across many many games will be different than what a one-game comparison says. Power draw could be more or less depending on the particular card and binning of the chip, as there isn't even a single "stock" voltage anymore with these newfangled GPUs). See, e.g., HardOCP for how energy usage varies depending on the game: http://www.hardocp.com/article/2012/03/22/nvidia_kepler_gpu_geforce_gtx_680_video_card_review/13

Um, what?

So what you're saying is that it doesn't matter because people aren't gonna use the cards for gaming but for letting the desktop sit at idle? Most people just put their computers on sleep or turn it off if they're not gonna use it, and load power consumption matters because not only does it raise the power bill depending on how much you game, but it also raises ambient temperature meaning increased room cooling is needed. That, and the heat on the other components.

Unrelated: don't listen to anything HardOCP says because they're full of crap, especially now-a-days.

 

[RussianSensation]

You guys are forgetting something, GTX670 is only cheaper than 7970 in THE USA. In other parts of the world they are the same price and 7950 is significantly cheaper.

"At £330 for basic models, the GTX 670 2GB also asks some hard questions of AMD when it comes to value; the HD 7970 3GB still sells for around £400, while the HD 7950 3GB will set you back around £330 at time of writing; we’d regard the GTX 670 2GB as superior to both at single-screen resolutions, and even equal to the HD 7970 3GB when it comes to three-screen performance." Bit-Tech Review in UK

It's also about $100 cheaper in Canada than the 7970. After market 7970s are going for > $500 for the most part here.

If you can find HD7950 for significantly less than 670 where you live, that changes things, obviously.

BTW, almost no one plays the PC versions of DiRT 3 and Batman:AC, yet you mention them in your comment knowing that.

You keep repeating this from time to time and I disagree. Dirt 2 and 3 are 2 of the better racing games on the PC at the moment. A lot of PC racing fans play them. It is because of the success of Dirt games that Codemasters became strictly focused on racing games. You stating no one plays them on the PC is ridiculous. If you don't like racing games or don't play PC racing games, then you might not care for them. They are way better games than NFS series for example. Sure there are other more hardcore PC racing games such as Project CARS, but the EGO engine is used in F1 2010/2011 and will be used in F1 2012, as well as in Dirt 2, 3. That's at least 5 racing games! Covering Dirt 3 in benchmarks covers at least 5 other racing games on the PC which is why it's important.

I am not arguing that Dirt games are better than GranTurismo 5 or Forza or Project Gotham Racing but you make it sound like no one cares about them on the PC, which is not true.

Secondly, Batman AC is one of the best third person action games on the PC, as was Batman AA. You keep dismissing it as if no one plays it. Are we supposed to be playing Batman AA/AC on Xbox 360 or PS3 then? Or you think Batman games are awful?

Actually if you want to talk about a game hardly anyone plays, it's AvP. Dirt 3 and Batman AA/AC are actually good games and AvP is meh.

While I think Metro 2033 is a cool game, it actually ended up on the Top 10 Games You've Never Played List. To suggest that more people care about Metro 2033 than Batman games on the PC is ludicrous.

Batman AC is a far more accomplished game than Metro 2033 or AvP are. Additionally, game-rental service GameFly announced that it was the most requested game of 2011, beating out Call of Duty: Modern Warfare 3. The Official PlayStation Magazine named Batman:Arkham City their Game Of the Year 2011. If you don't like 3rd person action games on the PC, that's a different story. But to say PC gamers don't care to play one of the best 3rd person action games of 2011 is odd to say the least.

Also, when GTX670 wins against the 7950, it smashes it by 20-40%+. Metro 2033 requires Quad-SLI 690s to approach 60 fps at 2560x1600. That's why for most people GTX670 performing poorly in Metro is a non-issue since it'll take GTX900 series to max that game.....

Thanks to discounts that are not available to most people, I got a new Sapphire OC 7970 for ~$410 after rebate, and actually more like ~$390 if you count the value of the miniDP->DVI adapter as $20, which is what I did when deciding between the 7950 and 7970, since I actually needed that adapter (unlike most people).I got the 7970 for more than just gaming, but even if I had got it for just gaming, ~$390 for a 7970 ain't bad.

That's a great deal. $410 for Sapphire Dual-X is a different story. Right now it's $480 though.

 

[blastingcap]

Um, what?

So what you're saying is that it doesn't matter because people aren't gonna use the cards for gaming but for letting the desktop sit at idle? Most people just put their computers on sleep or turn it off if they're not gonna use it, and load power consumption matters because not only does it raise the power bill depending on how much you game, but it also raises ambient temperature meaning increased room cooling is needed. That, and the heat on the other components.

What are you going on about? Did you read TPU's definition of average? It's average gaming load while playing Crysis 2. It's not average among all usage.

 

[IlllI]

The short cards are reference, so they should all be identical. So it comes down to warranty and such.

What would you replace the stock cooler with? I cannot think of any after market coolers that would fit, with the exception of a water block.

you are on the right mark

 

[cmdrdredd]

I have the same reaction. I can't wait for this to appear on Newegg or Amazon

Anyone have any idea when Asus is coming to the party? I will admit that Diablo 3 has "justified" this buy to me though... :biggrin:

I wouldn't be surprised if Diablo 3 doesn't even come close to needing the 670. Still...any reason to upgrade is a good reason I guess.

 

[MustangSVT]

I am amazed how small the actual card is, could be the perfect card for future matx builds with custom cooling.