780?
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
** The Official Nvidia GeForce 'Pascal' Thread - for general gossip and discussions **
- Thread starter Boomstick777
- Start date
780?
probably means 780Ti but the 780Ti wasn't a slightly cut down Titan unlike the 980Ti lol.
Soldato
- Joined
- 30 Nov 2011
- Posts
- 11,358
Yes, 780, came out just after the original titan
I made a post detailing how NV might rush to market and then some hours later that Italian bloke steals it as "news"!
Nice delusions of grandeur.
He's been saying for over a year that Pascal is basically die shrunk Maxwell with DP hardware bolted back on (and therefore less efficient for gaming), with still only the ability to emulate asynchrous compute and shaders.
The fact that he's published this article now suggests that he's received reliable confirmation of such.
Pascal absolutely was a rush job, with less than 2 years of work between commission and tape out, due to the massive delays to Volta and abandonment of HMC memory. They had to knock something together quickly to hold them over. The longer it takes them to get it out of the door, the more obvious its weaknesses will be.
Either way, they look dead in the water for forthcoming AAA console titles .... especially as the rumoured PS and XB 4K machines will use GCN 4.0 and leverage asynch and other AMD specific architectural advantages much more heavily.
15 percent quicker at best for the new cardsWell, it may well be...
Caporegime
Oops, original already posted
Last edited:
Hmm nVidia had the design kit for 20nm over 2 years ago and likewise building the hardware for architecture emulation well over 2 years ago - some aspects may be a little more rushed but the overall design certainly isn't.
Last edited:
Caporegime
What are you talking about? This isn't 20nm and TSMC's 20nm UHP process was designed specifically for NVIDIA, but it was going to be so expensive and likely have such poor yields that the two parties mutually agreed to cancel it. NVIDIA never designed anything else at 20nm, and Pascal was put on the drawing boards after 20nm was already dead.
Also, like I say, it was well under 2 years from commission to tape out for Pascal. It was a huge rush for them. They cancelled a GPU architecture I forget the name of, extended the life of Maxwell and hoped to shift Volta forwards. Instead they then ended up delaying Volta even more and plugging the gap with Pascal.
Last year, I thought that Pascal would be a significant upgrade and that Fottemberg (the guy who writes bits & chips) was wrong. I mean it had to be more than die shrunk Maxwell with DP hardware added back? Right? Wrong, seemingly. He's not the only one who's now adamant that this is the case.
Last edited:
You are missing the implications of what I'm saying - there are big changes to things like the lithography going from 28nm to 20nm planar that also apply to finfets (same design rules just some differences to transistor structure), the emulator lab for designing and testing architectures and design elements at this level have been up and running even longer than that. Sure there has been some switching around of designs and a few elements are going to be a little more rushed than ideal but in the scheme of things 2 years is quite a lot of time given the base that already exists to work off. Pascal is more than just a shrunk Maxwell with DP added back in.
Last edited:
Soldato
gpu architectures are normally in development for 4+ years end to end
Soldato
Associate
- Joined
- 14 Aug 2014
- Posts
- 1,070
Haven't seen this posted anywhere:
http://www.guru3d.com/news-story/micron-starts-sampling-gddr5x-memory-to-customers.html
Lends credence to the suggestions of 8GB cards.
http://www.guru3d.com/news-story/micron-starts-sampling-gddr5x-memory-to-customers.html
Lends credence to the suggestions of 8GB cards.
Associate
Don't you know how much trouble Nvidia and developers would get in if they removed async from console ports to degrade AMD'S performance?
I will tell you, huge amounts. It's highly illegal. It's anti competitive behaviour at its worst.
Nvidia know the score already. They have lost this round. There is not much they can do about it. Even offering parts of game works for free means nothing now.
Associate
Nvidia are going to sell nowhere near as much this time round. AMD have the superior hardware this time round. End of story. Also, for those waiting to see what's best for VR _ again AMD.
