GTX260 55nm Preview/Review

[Nasgul]

Why would you think it would revolutionize everything. RV670 and RV770 are at 55nm already and there's nothing magical about them. A 4870 draws about as much power as a GTX 260 65nm.

Well, if you read the previous post before mine. You'd see that EVERYONE "thought" that nVidia would revolutionized the GPU industry by moving from 65nm to 55nm.

It's rather amusing that everyone believes that with a reduction in processing, such major improvement "should" occur.

I for once don't see anything magical about the reduction of the GPU. But everyone else "expects" this magical accomplishment with a mere change.

___________________________________

And something tells me you did not look at the 2nd picture, where a 260 draws 265W vs the 4870's 312W and even at idling the 4870 draws more too. So, no such thing as "as much" from what you can see.

[trinibwoy]

Ah, gotcha. Well "everyone" needs to temper their expectations so as to avoid being disappointed over and over. G80 and 4870 are two recent examples of us being pleasantly surprised though. But that can't happen every time.

[Blkout]

Kinda like "G200-103-B2" is faster & more efficient than "G200-103-A2"?
Oh, wait?


G200-103-B2 = 55nm
G200-103-A2 = 65nm

The 55nm will consume less power and produce less heat than the 65nm if all settings are equal. This is not even debatable, simple physics.

[SNiiPE_DoGG]

Ah, gotcha. Well "everyone" needs to temper their expectations so as to avoid being disappointed over and over. G80 and 4870 are two recent examples of us being pleasantly surprised though. But that can't happen every time.

very true, a great example of 65nm to 55nm that wasn't a g80 or a 4870 is clearly the 3870. It was a low power card for sure but it set the 4870 up for greatness.

I think nvidia would have been better off going for a risk on the 280/260 being 55nm in the first place rather than leaving the midrange with nothing to run off of.

[Blkout]

Looks like the 55nm for Nvidia has leakage problems. It makes scene to dump this stepping at GTX260 and not the 270/290. My guess is there hoping a new stepping will have less leakage but you never know till the new stepping is tested.

You sure are drawing a lot of conclusions from ONE half-assed preview.

[LexDiamonds]

Something that everyone should keep in mind when judging measured heat on these 55nm GPUS is thermal density. I dont know if nVidia lowered vGpu on these new cards or not, but assuming they did not there is going to be a similiar amount of heat being dissapated thru a smaller contact area with the HSF- therefore equal or higher measured temperature. (Thermal density could even be higher with a decrese in vgpu...)

This being said, I think what everyone is still hoping to see is a decrease in net power consumption.

I also subscribe to the theory that nVidia is closely binning these first 55nm chips for use in the GX2s. Releasing a "new" 260 216 is a great way to use up chips that leak more current than they should on the new process. They are also more than likely binning current chips for the upcoming 285 as well. If I had to guess, these 285 bins probably fall somewhere between the GX2 chips and the "recycled" 55nm 216 parts for leakage.

[spursindonesia]

Fact that a 470 mm^2 chip endowed card with 896 MB 448 bit specs must be sold at around 250 US$, i think nVidia would make sure every darn chips count. But i do also think, when people get lucky with this card, they can get a golden chip that clocks amazingly, because i don't think the sales of GTX 295 and 285 would be that high in this kinda economic environment, while the demand of nicely priced GTX 260 core 216 would be quite high.

[Glow9]

looks like this is gonna be a slooooow year for Nvidia, hopefully ATI will make a killing in the midrange with the 740 and arrive at its upward trend equal with the GT300 in October.

whens this going to happen? the 740

[ChinStrap]

the only one i could find online was the eVGA and the dang things sold out almost everywhere it's offered.

[Kingcarcas]

Wow, i was expecting another 9800GTX+, slightly lower power consumption but OCs higher. Once you OC it the power consumption is actually the same or more than the old one.

[SNiiPE_DoGG]

whens this going to happen? the 740

well since its already taped out it should be out before the end of Q1 '09

nvidia's new midrange doesn't come till at least q3

[Blacklash]

I want to see max core and shader clock for these.

[Rattle]

I got a stock clocked evga 55nm coming midweek...

[largon]

And something tells me you did not look at the 2nd picture, where a 260 draws 265W vs the 4870's 312W and even at idling the 4870 draws more too. So, no such thing as "as much" from what you can see.

expreview's power consumption measurements were done during Furmark which is known to cause RV770s to draw power more than in any other benchmark. Thus using Furmark for powr consumption testing makes no sense as it doesn't represent any real world scenario.

Xbitlabs measures the exact power draw of the card itself (during loop of 3DMark06 SM3.0) and their figures show HD4870 consumes 6 watts more than 65nm GTX260. And since 55nm GTX260 draws 10W more than 65nm GTX260...


Idle figures?
Reference HD4870 bios has quite conservative PowerPlay settings for the hungry GDDR5 - users can downclock their memories and cut ~40W... That's a "no-can-do" on GTX260.

[dan7777]

I just wonna know will it clock better than 65nm gtx260 216sp...

got mine 260-216 clocked to 750-1540-1250 im happy with that.

[ColonelCain]

Something that everyone should keep in mind when judging measured heat on these 55nm GPUS is thermal density. I dont know if nVidia lowered vGpu on these new cards or not, but assuming they did not there is going to be a similiar amount of heat being dissapated thru a smaller contact area with the HSF- therefore equal or higher measured temperature. (Thermal density could even be higher with a decrese in vgpu...)

This being said, I think what everyone is still hoping to see is a decrease in net power consumption.

I also subscribe to the theory that nVidia is closely binning these first 55nm chips for use in the GX2s. Releasing a "new" 260 216 is a great way to use up chips that leak more current than they should on the new process. They are also more than likely binning current chips for the upcoming 285 as well. If I had to guess, these 285 bins probably fall somewhere between the GX2 chips and the "recycled" 55nm 216 parts for leakage.

Good point on the thermal disapation!

I to "subscribe" to the theory that nvidia as binning the hell out of these new chips. At this point, nvidia cares a heck of a lot more about the performance crown than mainstream parts such as this new model.

expreview's power consumption measurements were done during Furmark which is known to cause RV770s to draw power more than in any other benchmark. Thus using Furmark for powr consumption testing makes no sense as it doesn't represent any real world scenario.

Xbitlabs measures the exact power draw of the card itself (during loop of 3DMark06 SM3.0) and their figures show HD4870 consumes 6 watts more than 65nm GTX260. And since 55nm GTX260 draws 10W more than 65nm GTX260...


Idle figures?
Reference HD4870 bios has quite conservative PowerPlay settings for the hungry GDDR5 - users can downclock their memories and cut ~40W... That's a "no-can-do" on GTX260.

Thank you for pointing this out largon.

[LaMpiR]

got mine 260-216 clocked to 750-1540-1250 im happy with that.

What cooling method are you using?

[Blkout]

expreview's power consumption measurements were done during Furmark which is known to cause RV770s to draw power more than in any other benchmark. Thus using Furmark for powr consumption testing makes no sense as it doesn't represent any real world scenario.

Xbitlabs measures the exact power draw of the card itself (during loop of 3DMark06 SM3.0) and their figures show HD4870 consumes 6 watts more than 65nm GTX260. And since 55nm GTX260 draws 10W more than 65nm GTX260...


Idle figures?
Reference HD4870 bios has quite conservative PowerPlay settings for the hungry GDDR5 - users can downclock their memories and cut ~40W... That's a "no-can-do" on GTX260.

Why would it make no sense to use Furmark just because it stresses a video card more than using 3DMark06? If you want to see the maximum current draw, you would choose a benchmark that stresses a card to it's maximum. Disregarding Furmark just for the sake of showing ATI in a more favorable light is just as bad as disregarding any other benchmark to show favorable results for one piece of hardware over another.

While Furmark may not be typical real world scenario today, who's to say that it won't be with tomorrow's games, therefore Furmark is just as valid as any other benchmark assuming one manufacturer is not somehow optimizing their power circuitry to show favorable results, which in this case isn't happening.

[jaredpace]

got mine 260-216 clocked to 750-1540-1250 im happy with that.

55 or 65nm?

[Ket]

expreview's power consumption measurements were done during Furmark which is known to cause RV770s to draw power more than in any other benchmark. Thus using Furmark for powr consumption testing makes no sense as it doesn't represent any real world scenario.

Err.. actually Furmark is kinda MORE relevent as it represents a future real world scenario. Just look at Crysis, if game software keeps evolving at that rate Furmark will be outdated in a year.

[largon]

Why would it make no sense to use Furmark just because it stresses a video card more than using 3DMark06? If you want to see the maximum current draw, you would choose a benchmark that stresses a card to it's maximum.

Why would make sense to benchmark power consumption and actual performance with different standards?
RV770 totally and absolutely dominates G200 in Furmark. So does RV770 dominate in real world performance?

Furmark is kinda MORE relevent as it represents a future real world scenario.

But not present. And I like to deal with present. But that's just me.


Btw, what's it like in the future?

[Blkout]

Why would make sense to benchmark power consumption and actual performance with different standards?
RV770 totally and absolutely dominates G200 in Furmark. So does RV770 dominate in real world performance?
But not present. And I like to deal with present. But that's just me.


Btw, what's it like in the future?

We're not judging performance, we're judging power consumption. Did you get confused?

[largon]

2×
What's the relevance of power consumption in some non-real world situation? One cannot draw any perf/watt concusions based on such figures.

[onethreehill]

the only one i could find online was the eVGA and the dang things sold out almost everywhere it's offered.

In stock at Newegg

EVGA 896-P3-1255-AR GeForce GTX 260 Core 216 @$254.99
http://www.newegg.com/Product/Produc...82E16814130434

EVGA 896-P3-1257-AR GeForce GTX 260 Core 216 Superclocked Edition @$264.99
http://www.newegg.com/Product/Produc...82E16814130433

[moiraesfate]

the only one i could find online was the eVGA and the dang things sold out almost everywhere it's offered.

Newegg wasn't supposed to have any until Jan 5, but they got some in yesterday.