NVIDIA Kepler speculation thread

[Alexko]

Interesting. Wonder why it's so much better than 680-SLI. Maybe I'll get one of these to kick around until the real Kepler shows up.

Lower-latency inter-GPU communication?

 

[boxleitnerb]

Interesting. Wonder why it's so much better than 680-SLI. Maybe I'll get one of these to kick around until the real Kepler shows up.

Driver perhaps?

 

[CarstenS]

Or just a very regular forward movement which is being optimized to hell and back?
http://www.pcgameshardware.de/aid,8...Grafikkarte-der-Welt/Grafikkarte/Test/?page=3
For Crysis Warhead, we're observing the same as Computerbase when moving forward in basically a straight line. When dodging and strafing, as is usual in first person shooters, things start to get nasty: GTX 690 seems to be able to smooth five to seven frames before one drastic outlier occurs.

 

[Kaotik]

Hmh, 300W limit? It has 2x8pin, it's not following the limitations, the real-life consumption (according to TPU) is about same as 6990. (sure, it's under 300W, but TDP != consumption)

 

[SirPauly]

Another micro-stutter investigation from HT4U.net:

http://ht4u.net/reviews/2012/nvidia_geforce_gtx_690_im_test/index3.php

Good to see sites investigate this.

 

[ECH]

Another micro-stutter investigation from HT4U.net:

http://ht4u.net/reviews/2012/nvidia_geforce_gtx_690_im_test/index3.php

Good to see sites investigate this.

It seems that the lower the frame rate the more pronounced the MS gets (although higher frame rate results are missing). It would be nice to see what the results are at 90 FPS to get an idea of whats going on (even if you have to reduce IQ and/or resolution to get it). I also have to bring into question how frames rendered ahead is being used as well. What would be the difference between 0 and 1 at 90 FPS for example.

 

[trinibwoy]

Another micro-stutter investigation from HT4U.net:

http://ht4u.net/reviews/2012/nvidia_geforce_gtx_690_im_test/index3.php

Good to see sites investigate this.

Whoa, on second thought.....

 

[silent_guy]

Hmh, 300W limit? It has 2x8pin, it's not following the limitations, the real-life consumption (according to TPU) is about same as 6990. (sure, it's under 300W, but TDP != consumption)

So here's a card that, according to Anandtech, consumes at load 100 to 120W less than the GTX590, yet retains the same power supply capacity. IOW: a large safety margin over what's strictly required.

And somehow you consider that bad? What do you suggest they do instead? Remove those 2 GND pins to get 8+6 pin connectors and reduce the operating margin? Just to satisfy an footnote in the PCIe specification that's not even required to pass certification?

 

[AlphaWolf]

http://www.newegg.ca/Product/Product.aspx?Item=N82E16814130781

Wonder if we'll ever see one for anything near $999. On the bright side, you can buy one.

 

[Alexko]

So here's a card that, according to Anandtech, consumes at load 100 to 120W less than the GTX590, yet retains the same power supply capacity. IOW: a large safety margin over what's strictly required.

And somehow you consider that bad? What do you suggest they do instead? Remove those 2 GND pins to get 8+6 pin connectors and reduce the operating margin? Just to satisfy an footnote in the PCIe specification that's not even required to pass certification?

I still have no idea why this footnote is even there.

 

[trinibwoy]

http://www.newegg.ca/Product/Product.aspx?Item=N82E16814130781

Wonder if we'll ever see one for anything near $999. On the bright side, you can buy one.

Instantly sold out. Lots available for $2000 on eBay though

 

[Dooby]

http://www.newegg.ca/Product/Product.aspx?Item=N82E16814130781

Wonder if we'll ever see one for anything near $999. On the bright side, you can buy one.

They are the equiv of $1400+ in the UK. Insanity Wolf.

 

[DuckThor Evil]

Maybe I'll get one of these to kick around until the real Kepler shows up.

After having seen my GTX 570 struggle a bit in Crysis 2, I bit the bullet and purchased an EVGA GTX 680 SC+ w/Backplate. Flesh is so weak

 

[no-X]

Tridam's just published an article about GPU Boost, revisited: http://www.hardware.fr/focus/65/gpu-boost-gtx-680-double-variabilite.html

Basically, his press sample was qualified up to 1110MHz while retail cards may be limited to 1097, 1084, 1071, or perhaps as low as 1058MHz. So he benched a random retail card against his press sample, measured a 1.5% difference on average, up to 5% in Anno 2070.

reminds me when TomsHardware got a X800XT-PE sample with memory clocked at 575 instead of 560MHz...

 

[CarstenS]

Jeez, you have a looooong cat's memory.

 

[DSC]

http://vr-zone.com/articles/k-ngp-n-achieves-1442mhz-core-clock-on-air-cooled-gtx-680/15776.html

Not bad at all for air cooling..

 

[Man from Atlantis]

GTX 670

Spoiler

http://we.pcinlife.com/thread-1896133-1-1.html

 

[DSC]

That is really small, even smaller than the GTX 460 PCB.

http://images.anandtech.com/doci/3809/460Clean.jpg for comparison.

 

[Man from Atlantis]

very small indeed, it's smaller than my 65W 9800GT*..

 

[psolord]

That Gigabyte OC card is very close to the stock GTX 680. Nice.

If these are graphics scores and not overall system that is.