The owner of the copyrighted material (TomsHardware Guide) that was embedded and linked in this thread has requested that the AnandTech Forums remove all embedded instances and off-site links to it.
In accordance with the AnandTech Forum Guidelines, members are allowed to talk about misappropriated information (as in "leaks") but when we are notified that the material in question is resident on our forums (be it links to pirated software or pirated benchmarks) we have little choice but to do the ethical/conscionable thing and remove those links and embedded pics.
Just as you are allowed to talk about pirated movies or pirated software, you are allowed to discuss the leaked benches which are copyrighted Tom's Hardware property, just don't link to those leaked results (even if hosted on a 3rd party site) and do not embed them in your posts.
But you can speak to the numbers in them, you can discuss the bench results, etc. This is an unusual position for us to be in, but only because we so rarely get contacted with a take-down request.
So long as the other leaked content remains without a take-down request (the VR zone report for example) then you are free to continue linking and embedding those materials until such time that the content owners contact us with similar requests.
If you have questions or concerns on this moderator action we ask that you submit your inquiries through the proper channels, i.e. use the Moderator Discussions subforum.
An for the record, the OP of this thread did nothing wrong, this is not a sanction of their actions, its just the reality of melding the enthusiasm of our valued members with the business world.
Admininistrator Idontcare
For those with a 30" monitor we don't care about 1080p performance.
I suppose that removes any doubts on the veracity of the pictures we saw.
I suppose that removes any doubts on the veracity of the pictures we saw.
>>>
....
....
...Until i see a sign that the gk100 exist, i wont adopt such an imagined scenario as truth. I am only presenting multiple other reasons. Its far to early to jump the gun
I would hope that the people buying these high-end cards are self-respecting enough anyway to not be running a piddly little 1080p monitor!
So many people running such powerful hardware and outputting it to a cheap TN panel trying for 60 fps with everything maxed on a cruddy monitor.
***shudders***
Anyone see this? I know the source isn't all that great but interesting. http://www.theinquirer.net/inquirer...hed-amds-mid-range-radeon-hd-7870-gpu-compute
Anyone see this? I know the source isn't all that great but interesting. http://www.theinquirer.net/inquirer...hed-amds-mid-range-radeon-hd-7870-gpu-compute
After reading this, whether it is true or not, I wonder how fast the 7970 would be if there were driver optimizations and game engine tweaks to run faster on the new architecture.
As a side note, what did the performance patch do for bf3 and 7970?
I saw that already and have sort of though something similar since the THG leaks: Nvidia have a more efficient die with GTX680 for the first time in ages and it seems largely because they've sacrificed gCompute for gaming performance (which is fine for 95%+ of the market).
Now AMD are stuck with having used a larger die with good DP gCompute and having to compete with a faster, simpler and smaller design, esp with Tahiti. The logical thing for them would be to release a higher clocked 7870 (say 1300MHz+ and with faster RAM (whatever it takes to match GTX680)), and price it against GTX680.
But where would that leave Tahiti? Ok, 384-bit bus, 3GB, good gCompute performance but a die larger than needed for features gamers don’t want. I had wondered what the gCompute die size tax for the 7000 series was and I’d guess even for 77x0 it’s around 25%+.
The 7870OC strategy might make more sense but I suspect corporate bride would prevent it. Just look at Bulldozer – Stars is a better design by almost every metric but AMD will not make a proper 28nm-redesinged Stars CPU. Fusion is seen to be so important at AMD that they are willing to put a lot of effort and die space into that. I’m unsure if they are currently willing to design an APU with very poor DP gCompute performance (although Trinity is still VLIW4).
The poor gCompute performance of GTX680 leaves me with little doubt that a larger Kepler is in the works.
Anyway, while these discussions are fun to follow, what I wanted from 28nm looks like being elusive: a sub £100 GPU which idles and sips power like an APU (I now realise it would probably have to use an onchip frame buffer and maybe a big.LITTLE design with a separate mobile 2D chip).
I suppose that removes any doubts on the veracity of the pictures we saw.
They were hosted on their webpage. While their testing methods are highly questionable, the leak itself seems legitimate.
Yeah, and the takedown notice just assured us that they were legit.
Hopefully they realized their error and corrected them, although I imagine they still might face a firestorm.No, no it didn't. It was already assured that they were when people from TH confirmed the leak. It still doesn't mean they are final graphs.
RussianSensations theory is good, but what if GK104 is an intentional misnomer to counteract the unchecked sales of the 7000 series? Touting the '680' as a midrange card means the rest of the line must be far more powerful.
Nvidia could also be preparing '685'/'690'/'695' exploiting AMDs current price gouging and pushing prices higher.
But if your theory is true, Nv would purposefully gimp the lower cards (660/670 esp. If they are very ocable) to maintain the price difference b/w the cards. Then kepler 2 vs 8000 would be the actual battle.
Confusing.