GF100 Previews from Digital Experience

[Janooo]

a single gpu Fermi will NOT consume 300 watts. thats just what a lot of ignorant people are saying because it has a 6 and and 8 pin connector. most of those same ignorant people dont realize that other cards have had 6 and and 8 pin connectors too. it just means that it could have a tdp thats close to 225 watts so those connectors are there just for reassurance.

Consider the following. NV knows what performance they want to get out of Fermi. They would like to top the 5870. If close to 300W is required to do that what would you do? Wouldn't you go for it?
The 448 SP Tesla is 225W. More SP's and higher clock could be close to 300W.

 

[Janooo]

The damn thing crashed at the end :thumbsdown:

Drivers, or heat, or power, or ... ? Take your pick.

 

[Lonyo]

How serious and reliable is PCWorld usually?

If we are to believe the article, there's a rumor that GF100 has the same performance as a 5870. However, from other posts it seems that other rumors say it's 35% faster and yet other saying more like 15% faster only.

I find the "Fermi certified" case cooling for SLI quite scary as well.

It could be all three.
36% best case, 15% average, equal in too many cases.
Of course it could be all three for other reasons, like at certain resolutions it's only equal, or with certain settings (AA/AF etc).

There's no point in taking any % faster numbers to heart because there are too many variables. They could all be right, and yet they can also all be wrong.

 

[Wreckage]

It's not about support for three displays, it's stretching games on 3 - 6 high resolution (2560x1600) monitors on an affordable part made for the masses.

Because 6 monitors (2560x1600) 1 or 2 of which require display port are "affordable for the masses.

You are grasping at straws here.

 

[toyota]

Consider the following. NV knows what performance they want to get out of Fermi. They would like to top the 5870. If close to 300W is required to do that what would you do? Wouldn't you go for it?
The 448 SP Tesla is 225W. More SP's and higher clock could be close to 300W.

I said no SINGLE gpu Fermi will consume that much and theres no way it will. mark my words that their fastest single gpu card will have a tdp of less than 225 watts. they will likely have a dual gpu card to compete with the 5970 and that will approach 300watts. their dual gpu card will of course have to be a cut down version and not just two gtx380s.

 

[akugami]

@Zstream

The crash is likely to be driver related, although we can't rule out early engineering samples not being fully stable. In fact, that's a likely reason why they're not showing performance numbers of any sort at this time, probably still a lot of work to do on the driver side of things even if the hardware side has been finalized and in production. Now, if it is unstable hardware due to something like heat on a water cooled case then, "Houston, we have a problem." But I'd say it's more likely a driver issue.

Dunno if this has been mentioned but it seems that production of the GF100 has been confirmed. Assuming it started recently, right about the time or just before CES 2010, we should be seeing a release of late Feb to early Mar. This would point to Charlie's claim of the A3 respin being problematic as false.

Overclockers.net Here's an interesting link with info that I haven't seen posted here. If it has already been posted here and discussed, I apologize. Goes without saying, these are unverified.

*Cross out the frame rate numbers because we have no metric to properly measure them, hardware config, resolution, etc.

*The 46% average faster than 5870 is a relevant rumor. As well as only 8% faster than a 5970. The 5870 would likely need a price adjustment to compete at those numbers.

*GF100 load temp being 55C is very nice if performance holds true. Also a GF104 with unmentioned temps. GF100 is likely not clocked as high as Radeon 5 series to reach these temps or one killer cooling solution but either way it would point to overclocking headroom.

*GF100 has a 32x anti-aliasing mode. Dunno how relevant this is. At a certain point, with these increasingly high resolutions, aliasing makes a decreasing impact on image quality.

*GF100 has 100% 1080p Blu-Ray or H246 playback. Not that relevant IMHO and the reason is if you are thinking of picking up a GF100 or 5870, you got a pretty high end system that can handle 1080p playback already.

*Relevant to the pricing is "GF100 promises to deliver at least 40% more performance than the GTX295 for less money."

*Not Fermi related but I have expressed being interested in Tegra2 and RIM seems to have ordered a million Tegra2 chips for Blackberries. Apple ordered a few million for a future iPhone! I use an iPhone so this looks like it might be in my phone upgrade plans. And once again, the Nintendo DS upgrade using Tegra2 rumor comes back.

 

[cganesh75]

how is production of GF100 is confirmed? because Jen Hsun said so? its the same guy who showed a wood screw card and said "this puppy is Fermi"..

 

[Schmide]

I didn't see a crash. WTF

 

[toyota]

I didn't see a crash. WTF

me either

 

[akugami]

how is production of GF100 is confirmed? because Jen Hsun said so? its the same guy who showed a wood screw card and said "this puppy is Fermi"..

That was a presentation and while it doesn't put nVidia in a good light to put a mockup and try to pass it off as an actual card, when the CEO of a publicly traded company says their latest major product is in production, it usually is. Putting out false info like that could be grounds for a suit by shareholders.

 

[Zstream]

I didn't see a crash. WTF

You are blind as a bat then...

Doh I am retarded, it was the video with the rocket sled

 

[SirPauly]

http://www.tvtech.it/video/546/nvidia-fermi-al-ces/

This video has more of the Supersonic Sled demo and shows part of the destructible bridge and in the distance one can see Batman being run on 3d vision Surround.

 

[Zstream]

http://www.tvtech.it/video/546/nvidia-fermi-al-ces/

This video has more of the Supersonic Sled demo and shows part of the destructible bridge and in the distance one can see Batman being run on 3d vision Surround.

Why not show the crash?

http://www.youtube.com/watch?v=3-1GMbqzGX0

 

[T2k]

Overclockers.net Here's an interesting link with info that I haven't seen posted here. If it has already been posted here and discussed, I apologize. Goes without saying, these are unverified.

*Cross out the frame rate numbers because we have no metric to properly measure them, hardware config, resolution, etc.

*The 46% average faster than 5870 is a relevant rumor. As well as only 8% faster than a 5970. The 5870 would likely need a price adjustment to compete at those numbers.

*GF100 load temp being 55C is very nice if performance holds true. Also a GF104 with unmentioned temps. GF100 is likely not clocked as high as Radeon 5 series to reach these temps or one killer cooling solution but either way it would point to overclocking headroom.

*GF100 has a 32x anti-aliasing mode. Dunno how relevant this is. At a certain point, with these increasingly high resolutions, aliasing makes a decreasing impact on image quality.

*GF100 has 100% 1080p Blu-Ray or H246 playback. Not that relevant IMHO and the reason is if you are thinking of picking up a GF100 or 5870, you got a pretty high end system that can handle 1080p playback already.

*Relevant to the pricing is "GF100 promises to deliver at least 40% more performance than the GTX295 for less money."

*Not Fermi related but I have expressed being interested in Tegra2 and RIM seems to have ordered a million Tegra2 chips for Blackberries. Apple ordered a few million for a future iPhone! I use an iPhone so this looks like it might be in my phone upgrade plans. And once again, the Nintendo DS upgrade using Tegra2 rumor comes back.

If you look at the entire original discussion and the pathetic patronizing tone he's trying to use I think he was told to spill the beans - and kept in dark many ways, to make sure he won't spill anything he shouldn't and that he will be good at this "gyezinformacija" operation... this kind of misinformation is quite normal, AMD used it on more than one occasion as well.

Also noticed:

I revealed this info because I am no longer employed by nVidia and several of the technical NDAs expired at midnight, seeing as how Jan 4th was my last day.

Ummm correct me but - let's assume for a sec that it's true - obviously if he were high enough to know key details his NDA would probably stand for years to come...

 

[SirPauly]

Why not show the crash?

http://www.youtube.com/watch?v=3-1GMbqzGX0

That video has been offered numerous times and very welcomed, too. Was curious if there was anything after what looks like a crash, and found another video that had a destructible bridge -- which I thought was neat. Then noticed 3d vision surround in the back-ground -- considering how few videos they are -- all welcomed to me and try to share.

 

[Zstream]

@Zstream

*GF100 load temp being 55C is very nice if performance holds true. Also a GF104 with unmentioned temps. GF100 is likely not clocked as high as Radeon 5 series to reach these temps or one killer cooling solution but either way it would point to overclocking headroom.

Anyone who believes that a card capable of 300W TDP is at 55C is not only silly but incapable of understanding the workings of a GPU. FUD was at the show and stated the FAN was on really high, he stated it was quite noticeable.

 

[cganesh75]

That was a presentation and while it doesn't put nVidia in a good light to put a mockup and try to pass it off as an actual card, when the CEO of a publicly traded company says their latest major product is in production, it usually is. Putting out false info like that could be grounds for a suit by shareholders.

no it doesnt.. he can always comeback and say.. we encountered technical diffiulties.. or whatever to cover up the lie.. nobody would know..

 

[toyota]

Anyone who believes that a card capable of 300W TDP is at 55C is not only silly but incapable of understanding the workings of a GPU. FUD was at the show and stated the FAN was on really high, he stated it was quite noticeable.

again just because it has a 6 and 8 pin connector doesnt mean its going to be anywhere near 300watts. it could could easily just have a TDP of around 225 and still have those connectors.

55C under some load would not be all that unrealistic if the fan was spinning pretty fast and it had a good cooler. my 65nm gt260 has a TDP of 182 watts and doesnt go over mid to upper 50s in some games even while overclocked and fan at stock settings.

remember too that Nvidia TDP ratings are no where near the actual power usage which is much lower.

 

[Zstream]

again just because it has a 6 and 8 pin connector doesnt mean its going to be anywhere near 300watts. it could could easily just have a TDP of around 225 and still have those connectors.

55C under some load would not be all that unrealistic if the fan was spinning pretty fast and it had a good cooler. my 65nm gt260 has a TDP of 182 watts and doesnt go over mid to upper 50s in some games even while overclocked and fan at stock settings.

remember too that Nvidia TDP ratings are no where near the actual power usage which is much lower.

Huh? Who said I was claiming my opinion was based on just a 6+8 pin adaptor?

Oh and apparently you are not well informed on the PCIE 2.0 specs.

75+150+75 = ????????????????????? Yeah you guessed it.... 300! 75+150 comes from the 6 & 8 pin alone, add 75 from the slot and well once again it equals 300.

Ok, let me be a bit more clear. The current ATI offerings is hovering around 180-200 Watts, 190W for a surface area of 334mm2. Now IF the 50% more transistor count is accurate then it will certainly creep up to the 250+ range no question.

190W divided by 334mm2 is around .56. If we took the exact same transistor count from ATI and increase the surface area to 530mm2 is around .35.

.35 + .17 (increase of 50%) then we reach around 275watts at full load.

 

[toyota]

Huh? Who said I was claiming my opinion was based on just a 6+8 pin adaptor?

then what is it being based of off? sorry but there is NO WAY their single gpu card will have a TDP of anywhere near 300 watts. again mark my words on that.

 

[lopri]

I said no SINGLE gpu Fermi will consume that much and theres no way it will. mark my words that their fastest single gpu card will have a tdp of less than 225 watts.

How will you live up to your words? Just out of curiosity.

GTX 280's TDP is 236W, BTW.

 

[lopri]

Ha, I saw you sneaked in "around 225". A good one.

 

[Zstream]

then what is it being based of off? sorry but there is NO WAY their single gpu card will have a TDP of anywhere near 300 watts. again mark my words on that.

Read my post again.

 

[toyota]

How will you live up to your words? Just out of curiosity.

GTX 280's TDP is 236W, BTW.

you can ban me. lol.

yeah and the GTX285 had faster clocks and dropped the TDP from 236 to 183 just from going to 55nm from 65nm. hell the ATI 5870 doubled the SP/TMU/ROP aspect of their 4890 card and had faster memory and still got a 2watt lower TDP going frm 55nm to 40nm. Nvidia will be lowering their bus width and going with a 4:1 shader/tmu so really it will not even be doubling all the aspects that would consume power like ATI did. again it will not be remotely close to 300 watts even with 512sp.

 

[toyota]

Ha, I saw you sneaked in "around 225". A good one.

I wasnt sneaking that in. as soon as I typed it I realized that it could certainly be over 225 so i changed it. you can see that even me going back and adding other stuff to my comment was done in less than 2 minutes.