NVIDIA Kepler speculation thread

There's just not enough to that story. No amount of understanding on display or any meat to speculate over.
It's almost pure rarified uncertainty, and there's now more text saying "I'm not shorting the stock, honest" than there is story.

The Intel chipset comparison was a poor one and possibly an attempt to conjure up legitimacy by mentioning a problem we all know was true. There are areas I can think of that could have a similarity in an abstract sense, such as a fault that trips the retry logic of the GDDR5 channels or something breaking down with the power monitoring circuitry.

Most of the GPU would not have a similar failure mode with electromigration as some are speculating (because of that Intel comparison). The lack of ECC or RAS for GK104 means errors aren't found and don't cause a retry that can sap performance. Cumulative electromigration wouldn't slow it down, rather the chip would run improperly or die.
 
I'd keep an old 8400GS over a faster IGP with worse drivers.
want to run some CAD or 3D program? then the gt620 surely will work better than a HD 4000.

want to build a socket 2011 system for music production, 2D graphics, some video editing etc.? a gt 610 is ideal, as long as it's fanless or has a fan controller.

Would you please explain in what exactly one may observe a difference between if he runs AutoCAD on Ivy Bridge or GT 610?

Rebranding is a great for product line clarity, the contrary of 'digging your own grave'. If it were the latter, Nvidia would have stopped doing it long time ago. After all, they've been doing it for, what, more than 4 years now?

I don't think rebranding in this particular moment is that great an idea. They simply stay at one and the same level too long, with next-generation Haswell, NV should prepare something really special in order to convince all those OEMs that it's worth it to build-in one of those additional cards. Customers will buy what is modern, fashionable and what's available then.
I'm not saying that at present the situation is not similar, just that it should be more pronounced by then.
 
Of course not, but people looking to upgrade from something older might pick the GT610 over, say, some HD 6000, on account of the former being perceived as "newer"; which is probably why AMD has renamed some low-end HD 6000s to HD 7000s.


Peoples who buy OEM PC with this type of GPU inside, anyway dont care about the gpu who is in the box ... its sad, but this is like that ... thoses are dedicated to cheapo OEM pc...

Ethic wise, this is miserable, but this have allways been there sadly..
 
I have become rather ambivalent on the renaming issue. There isn't even much of a featureset variation. The only time there really was an instance of that was back with Radeon 9xxx and GeForce 4.

Actually I kind of wonder what could be done with a G92 on 28-40nm. ;)
 
Alexko said:
Easy, maybe, but it's also misleading.
IMHO, the best system would simply have the following "GeForce Fast", "GeForce Faster", "GeForce Fastest". With fancier names, of course "Tall",
"Grande", "Venti"), but you get the idea. Just like nobody in Starbucks cares about the actual oz. of each cup, nobody cares about the silicon or the number of cores inside.

Not really practical because there are more than 3 performance segments, but the current system is very close to this. (The only real confusing part is the plethora of GTX560 SKUs.)

I was just as WTF when the renaming of G92 started, but I've obviously made a 180 degree turn on this. It's extremely consumer friendly. Leaves zero confusion for people who are in the market who want to decide between different performance levels. It also leaves zero confusion for those who move from one generation to another, as long as they know that the first digit is the series number: a GT640 is always going to be faster than a GT530. Is a GT630 faster than a GT540? Who knows, but that's not something you would solve by not rebranding anyway...

The remainder, the ever so small set of consumers who don't fall in the groups listed above, maybe confused. Well, tough: you can't satisfy everybody. I'm sure a handful of blokes were screwed by GPU perf when Apple upgraded its MacBook Airs from an Nvidia GPU to an Intel.

But misleading? On the contrary. Unethical? BS. Digging their own grave? <redacted>
 
never mind ... long explication i will do for nothing so finally i retain..

But i got real mixed feeling about renaming.. in many case it do sense, technically and cost wise ( why developp a 7670 in 28nm, for end with the same performance of what allready exist ( maybe why not reuse a 5770 )... In some other case, it is really for satisfied the OEM who need new cards on new laptop/pc series for make believe they sold new hardware, new pc... ( the 630 is a good example of it )
 
Last edited by a moderator:
IMHO, the best system would simply have the following "GeForce Fast", "GeForce Faster", "GeForce Fastest". With fancier names, of course "Tall",
"Grande", "Venti"), but you get the idea. Just like nobody in Starbucks cares about the actual oz. of each cup, nobody cares about the silicon or the number of cores inside.

Not really practical because there are more than 3 performance segments, but the current system is very close to this. (The only real confusing part is the plethora of GTX560 SKUs.)

I was just as WTF when the renaming of G92 started, but I've obviously made a 180 degree turn on this. It's extremely consumer friendly. Leaves zero confusion for people who are in the market who want to decide between different performance levels. It also leaves zero confusion for those who move from one generation to another, as long as they know that the first digit is the series number: a GT640 is always going to be faster than a GT530. Is a GT630 faster than a GT540? Who knows, but that's not something you would solve by not rebranding anyway...

The remainder, the ever so small set of consumers who don't fall in the groups listed above, maybe confused. Well, tough: you can't satisfy everybody. I'm sure a handful of blokes were screwed by GPU perf when Apple upgraded its MacBook Airs from an Nvidia GPU to an Intel.

But misleading? On the contrary. Unethical? BS. Digging their own grave? <redacted>

I think it's misleading because having a GXX 6XX line-up makes it seem as though they had the same features, the same level of power-efficiency, etc.

People may not care about the number of so-called CUDA cores, but perf/W, number of supported displays, video decoding/encoding engines do matter.
 
I think it's misleading because having a GXX 6XX line-up makes it seem as though they had the same features, the same level of power-efficiency, etc..

Sigh, why do we rehash this same thing every few months? There is no rule in consumer products that says all members of a generation share the same features. It's quite the opposite.

Take a look at Panasonic's 2012 plasma lineup. Do the XT50, UT50, ST50 and GT50 check all the same boxes as the VT50? No, no, no and no :) Don't know where these higher standards for graphics cards are coming from.
 
Sigh, why do we rehash this same thing every few months? There is no rule in consumer products that says all members of a generation share the same features. It's quite the opposite.

Take a look at Panasonic's 2012 plasma lineup. Do the XT50, UT50, ST50 and GT50 check all the same boxes as the VT50? No, no, no and no :) Don't know where these higher standards for graphics cards are coming from.

Simply because when they're truly based on the same architecture, then they do share most of the same features.
 
Simply because when they're truly based on the same architecture, then they do share most of the same features.

Right but I'm not a tv enthusiast so I don't know or care about the panel manufacturing tech, microprocessors or any of that crap in Panasonic's tv's.

I shop based on features, price and reviews. Same goes for people shopping for a graphics card. They don't know what architecture or chip or TSMC process is under the hood nor do they give a damn.
 
Right but I'm not a tv enthusiast so I don't know or care about the panel manufacturing tech, microprocessors or any of that crap in Panasonic's tv's.

I shop based on features, price and reviews. Same goes for people shopping for a graphics card. They don't know what architecture or chip or TSMC process is under the hood nor do they give a damn.

Never said they did, but they might care about power-efficiency and features, regardless of how it's achieved.
 
We need a thread for this topic. One that can be resurrected whenever somebody feels the need. Could use some threads like that for a few other topics too, actually. ;)
 
Back
Top