NVIDIA Kepler speculation thread

[AnarchX]

More @ GT 640M:

http://static.acer.com/up/Resource/Acer/Docs/TW/20120224/201203Pricebook.pdf
found at: http://forum.notebookreview.com/acer/649057-detail-spec-acer-timeline-ultra-aspire-m3-5800-tg.html

GT 640 1GB DDR3

進 化 效 能
搭載NVIDIA ® GeForce ® GT640M玩家級獨立顯卡，支援DirectX 11，完美呈現如影片般
栩栩如生的影像細節和更完美光滑的曲線。3DMARK實測達P6412，PCMARK
VANTAGE實測分數達7317 4 ，效能超進化。

translated:
Feed of the effectiveness of
NVIDIA ® GeForce ® GT640M enthusiast discrete graphics, support for DirectX 11, the perfect presentation such as video
Lifelike image detail and a more perfect and smooth curves. 3DMARK measured up to P6412, PCMARK,
The VANTAGE measured fraction of the 73174, the performance of super-evolution.

Seems to be P6412 @ 3DMark Vantage.
GT 540M was ~P3500.

 

[trinibwoy]

The 640m is being benchmarked and still no word on specs? 85% more performance than GF108 on the same DDR3 is impressive but is it Kepler or Fermi based?

 

[chavvdarrr]

GK107 GT640M (unknown specs) tested: http://translate.google.com/transla...er+timeline+ultra&cd=7&hl=zh-TW&ct=clnk&gl=hk

Comparable game benchmarks with GT 630M (GF108 96SPs/128-Bit @ 675MHz and 1.6Gbps DDR3): http://translate.googleusercontent....-630m/&usg=ALkJrhhmL_M2Rb5zhatsr4E4PUM50QeHhA

GT 640M seems to be arround ~50% faster.

Damn , it took me 5min to understand that I'm looking at review on my own language which was translated to chinese and then to english
Looks like these guys really have 640M and put the benches online by error... its removed from the site atm... - I hope they won't be punished for this :/

 

[ninelven]

In short: doing away with separate texturing, using shaders to do that.

I did read the link. Without knowing the implementation details (unit count etc..), I fail to see how you can conclude that would lead to less "gaming prowess." Worse performance / mm^2 I might buy, but the GTX 560 TI 448 vs GTX 560 TI results don't even support that.

EDIT: (Even in the patent, the changes are described as advantageous with regard to performance/area.)

 

[Kaotik]

Is it confirmed that the 640M is infact Kepler?
I mean, it's just strange that there would be completely mixed Kepler / Fermi line-up on the mobile 6xx-line (630M Fermi, 670M & 675M are apparently confirmed GF114 Fermis)

 

[Man from Atlantis]

Is it confirmed that the 640M is infact Kepler?
I mean, it's just strange that there would be completely mixed Kepler / Fermi line-up on the mobile 6xx-line (630M Fermi, 670M & 675M are apparently confirmed GF114 Fermis)

Since there is only 2 Kepler GPUs(GK104 and GK107) is alive atm, it seems normal to me.. GK106 was scheduled in May or early Q3 but maybe it's changed idk..

 

[CarstenS]

I did read the link. Without knowing the implementation details (unit count etc..), I fail to see how you can conclude that would lead to less "gaming prowess." Worse performance / mm^2 I might buy, but the GTX 560 TI 448 vs GTX 560 TI results don't even support that.

EDIT: (Even in the patent, the changes are described as advantageous with regard to performance/area.)

I was talking about the proposed topic about no changes in functional units. If there's the same number of shaders in two products,...

 

[ninelven]

You may as well be speculating about the fast food industry on Mars.

 

[mczak]

Since there is only 2 Kepler GPUs(GK104 and GK107) is alive atm, it seems normal to me.. GK106 was scheduled in May or early Q3 but maybe it's changed idk..

I wonder if gt640m and gt650m will turn out to be almost exactly the same just like gt550m and gt540m? Those were only separated by 10% core clock - certainly nothing which would warrant a different model number. Or possibly the gt640m might have one SM disabled out of four (the option of disabling SMs on the gf108 were a bit limited...)

 

[rpg.314]

Wrong guy, Ninelven was making the claim that a HPC only part should just remove the graphics pipeline. I'm arguing that including the graphics pipeline is much more cost effective in a part that is tailor made for the HPC market yet still viable for the professional market with some appeal to enthusiasts in the consumer market.

The same hypothetical idea that Carsten proposed.

Regards,
SB

It wasn't a critique of what you said. It was an addition.

 

[AnarchX]

More @ GT 640M:

http://static.acer.com/up/Resource/Acer/Docs/TW/20120224/201203Pricebook.pdf
found at: http://forum.notebookreview.com/acer/649057-detail-spec-acer-timeline-ultra-aspire-m3-5800-tg.html

GT 640 1GB DDR3

Seems to be P6412 @ 3DMark Vantage.
GT 540M was ~P3500.

P3500 is GPU score, full score is a bit over P4000 for GT 540M. But GT 640M was coupled with a 1,6/2,3T GHz ULV SNB.

 

[Picao84]

CharLIE and OBR have started their engines

CharLIE is saying Kepler is launching next week:
http://semiaccurate.com/2012/03/05/nvid ... in-a-week/

OBR is saying CharLIE is wrong:
http://www.obr-hardware.com/2012/03/sem ... about.html

Who will win the race?

 

[Miksu]

CharLIE and OBR have started their engines

CharLIE is saying Kepler is launching next week:
http://semiaccurate.com/2012/03/05/nvid ... in-a-week/

OBR is saying CharLIE is wrong:
http://www.obr-hardware.com/2012/03/sem ... about.html

Who will win the race?

I'm puzzled why people care about what OBR says? Hasn't it been proved over and over that the guy is just some random poster, without any connections?

 

[whitetiger]

CharLIE and OBR have started their engines

CharLIE is saying Kepler is launching next week:
http://semiaccurate.com/2012/03/05/nvid ... in-a-week/

OBR is saying CharLIE is wrong:
http://www.obr-hardware.com/2012/03/sem ... about.html

Who will win the race?

For once I hope that Charlie is right, and knows what he's talking about!


Either way, it's going to be soon....

 

[trinibwoy]

I'm giving OBR the odds on this one. As of a few days ago the guys at [H] still weren't under Kepler NDA. The event on March 12th could be just a press briefing and NDA start.

 

[Picao84]

I'm puzzled why people care about what OBR says? Hasn't it been proved over and over that the guy is just some random poster, without any connections?

Nevertheless you have to admit it is fun having those two challenging eachother (or OBR challenging Charlie.. whatever ).

I'm giving OBR the odds on this one. As of a few days ago the guys at [H] still weren't under Kepler NDA. The event on March 12th could be just a press briefing and NDA start.

Makes more sense, yes.
Appopin from Alienbabeltech hinted at some "travelling" soon some days ago.

 

[trinibwoy]

Prices on GTX 580/570/560 are also just beginning to move south and not dramatically so. A public launch in a week is really unlikely.

 

[Alexko]

I'm giving OBR the odds on this one. As of a few days ago the guys at [H] still weren't under Kepler NDA. The event on March 12th could be just a press briefing and NDA start.

This is pretty much what Charlie is saying:

Since they mentioned the GTX680 by name in the invites, and sites don’t have cards yet, it looks to be about as ‘real’ as recent AMD launches.

 

[trinibwoy]

AMD's recent paper launches at least had reviews.

 

[DarthShader]

wait, so OBR is saying two things:
1) Review sites DO have the cards already
2) But launch is NOT soon.

Huh?

Anyways, another screenshot to my collection, when the blog post gets pulled. Remember what the guy said about Cape Verde?