NVIDIA Kepler speculation thread

[mczak]

Slightly off-topic maybe: "Nvidia shipping Kepler to notebook manufacturers", according to http://www.fudzilla.com/home/item/25747-nvidia-shipping-kepler-to-notebook-manufacturers

Interesting. Does that mean nvidia would have a whole family of chips ready at approximately the same time, but release the slower ones for mobile first?
Though not sure what I should think about Kepler in ultrabooks. Ivy Bridge should have decent IGP (probably around the performance of a GT610M), so I think a potential Kepler chip for ultrabooks would need to be faster while having a significantly smaller TDP (10W or so?) at the same time. I'm just not sure nvidia really holds a significant advantage in perf/power to make it feasible (considering the inherent power disadvantage of such an approach, two powered memory buses, powered pcie interface).

 

[rpg.314]

When were the patents for it filed?

You tell me.

 

[trinibwoy]

I could be wrong but I think the RF cache work is too recent to have been included in Kepler.

Possibly, I wouldn't be surprised at that either. One interesting thing is that the RF cache paper made references to much shallower pipeline depths than Fermi, instead comparable to AMD's stuff. That could be a hint of the direction they're heading. A shorter pipeline would dovetail nicely with the rumours of no more hot-clock. Pure speculation though.

Well, considering the source…

Not that hard to believe. That seems to be a reasonable lead time if Ivy Bridge notebooks will start appearing first week of April.

 

[AnarchX]

http://forums.laptopvideo2go.com/topic/29042-v29518-windows-7vista-64bit-asus-mobile/

[DEV_0FDB&SUBSYS_10AC1043] NVIDIA GK107-ESP-A1

Also other SKU strings included for DEV_0FD... devices, which are rumored to be GT 640M, GT 650M and GTX 660M.

BSN saw GK107 - NVIDIA N13P-GT - GT 650M?

Really strange specification output by the driver. This would put it below a 40nm GT 630M.

Could Nvidia include some "stealth" code for Kepler GPUs in the driver panel? The smaller Fermis were all leaked through such screenshots, so they are forced to did this?

My guess:
The driver could detect Kepler GPUs as Fermi GPUs:
- base-clock = shader-clock / 2 (but Kepler is speculated to have no shader domain)
- SMs number is calculated by x 32 or x 48 (but Kepler could have 96SPs per SM)

So GT 650M could look like this:
- 192SPs@810MHz
- 288SPs@810MHz

The second one seem more likely, because first one would have only 10% more raw FLOPs than GT 550M.

 

[xDxD]

Demerjian continues with cryptic, but imho interesting, phrases: http://semiaccurate.com/forums/showpost.php?p=150266&postcount=392

 

[AnarchX]

Cost down PCBs and coolers => reduced prices

If you are looking at GPUs like HD 6800 series and after, price reductions were only based on cheaper self designed PCBs and coolers and a very small bit of vendor and retailer competition.

Reduction of GPU prices by AMD and Nvidia seem to be not happen anymore (remember GT200 and RV770 with card prices ~$130).

 

[xDxD]

http://errorlevelanalysis.com/permalink/d2d6e56/

edit

 

[Ailuros]

You tell me.

It was an honest question since I obviously don't know.

 

[SimBy]

http://news.techeye.net/chips/nvidias-kepler-suffers-wobbly-perturbations

Ageia hardware on Kepler?!

CD did mention Kepler has a secret...

 

[trinibwoy]

Right cause dedicated hardware makes a lot of sense when dedicated software can't even get traction.

I completely believe that nVidia has been canvassing reviewers hating on the 7950 though.

 

[Kaotik]

Right cause dedicated hardware makes a lot of sense when dedicated software can't even get traction.

While true, and while I don't believe there's any "dedicated physx hardware", there's no denying that there clearly need for something, since apparently GPU Rigid bodies, promised ages ago via APEX 1.2, still aren't available (well, weren't in November last year anyway)

 

[Rangers]

Right cause dedicated hardware makes a lot of sense when dedicated software can't even get traction.

I completely believe that nVidia has been canvassing reviewers hating on the 7950 though.

Hating?

TechEye has heard Nvidia has been asking hacks why the reviews have been so good.

Nvidia absolutely denies the rumour which suggests representatives have been asking journalists to file a short report on what is exactly so great about the 7950.

Read more: http://news.techeye.net/chips/nvidias-kepler-suffers-wobbly-perturbations#ixzz1l68AbsFQ


Sounds more like "what about the 7950 are you guys so positive on so we can learn from it?" or something of that nature, to me.

 

[trinibwoy]

Lol no, the Techeye article is obviously saying that nVidia is incredulous at the positive reviews of the 7950 and is trying to downplay it. Why would they need reviewers to teach them anything about a video card?

what exactly is so great about the 7950

Does that sound like sincere curiosity to you?

 

[AlphaWolf]

I read it the same way as trinibwoy. They were asking reviewers why they thought it was good as a way of suggesting they didn't think much of it. I'm sure Nvidia has done their own investigations into Tahiti, I don't know what they would really learn from reviewers that would be useful to them outside of casting aspersions.

 

[Arty]

Lol no, the Techeye article is obviously saying that nVidia is incredulous at the positive reviews of the 7950 and is trying to downplay it. Why would they need reviewers to teach them anything about a video card?


Does that sound like sincere curiosity to you?

Maybe they are curious as to why last year's performance for $50 less is amazing.

 

[madyasiwi]

Maybe they are curious as to why last year's performance for 10% less is amazing.

Fixed.

 

[CarstenS]

Maybe the amazing things are it's enormous OC potential and, if you aren't into that, the ridiculously low power consumption compared to other cards with „last year's performance“? Power is down to 57% in our test case for example

 

[no-X]

Maybe they are curious as to why last year's performance for $50 less is amazing.

Why they didn't ask themselves the same question during the launch of GTX 460?

 

[trinibwoy]

Which card did the 460 match in performance while being only $50 cheaper? The GTX 285 launched 18 months before the 460 at $400. The 460 was $230 and was a few percent faster than the 285.

 

[Richthofen]

Maybe the amazing things are it's enormous OC potential and, if you aren't into that, the ridiculously low power consumption compared to other cards with „last year's performance“? Power is down to 57% in our test case for example

While this is a nice thing. However in the past this has never been really an amazing argument in the past for 400$ or higher products. Especially with a manufacturing node shrink. Amazing things with new generations have always been performance improvements and nice new features. I think it is nice that the power draw has been reduced significantly. However this is something i would take as a given on a node shrink without huge performance or feature improvements. So it is nice yes - but amazing is something different for me.

Why they didn't ask themselves the same question during the launch of GTX 460?

Well the GTX460 was a performance product with very attractive pricing. This is not comparable to the HD7950.