LinkBack URL
About LinkBacksnow the 16nm is nailed I pray they relaunch an older but effective model on the newer fab and make a super low power, slim line quiet card (or two).
I've got a 9800GT Green Edition which is a superb card.... using that rough theory.
now the 16nm is nailed I pray they relaunch an older but effective model on the newer fab and make a super low power, slim line quiet card (or two).
I've got a 9800GT Green Edition which is a superb card.... using that rough theory.
Originally Posted by Advice Trinity by Knoxville
I'm not sure if the economics of that make sense on current nodes. They would often use die shrinks of existing designs as pipecleaners for a new node, but with the cost of designing for the likes of 16nm being so high it only really makes sense if there's a big market for the product - a slightly more efficient existing GPU probably wouldn't cut it, especially given that the cost/transistor isn't decreasing rapidly like it used to. And once you have a new architecture which in itself might be more power/space efficient than the existing one, you're probably better off just using a smaller version of that.
For instance, we'll likely see smaller dies on 16nm using the new architectures - Polaris11 for example, and I'm sure Nvidia have something similar lined up for the mobile market.
PCGamer just put up this article. http://www.pcgamer.com/what-exactly-...nders-edition/
Summary is that the founders cards are just reference cards renamed, with a new cooler and no overclock. Difference being they won't be limited edition apparently. The higher price is supposedly to give the AiB partners some wiggle room for price.
Now that even Intel's integrated graphics is getting reasonably powerful, I wonder if AMD and Nvidia will bother to make anything lower powered than the Polaris 11.
But yeah, usually the lowest end cards don't get updated for many years, I guess there isn't a real need for them to be on the latest process.
As watercooled mentioned, it's not worth them doing exactly that.. however in general that's sort of what they'll be doing with some releases since they can take advantage of the minimum redesign necessary and still get a nice new card in the new process. The Polaris 11 range from AMD look especially well suited to this low power, slim line card - their arch seems especially well suited to scaling up or down. nVidia won't ignore that (very lucrative) market for long though - even if they currently seem to be focusing on low-volume parts.
GP107 and GP106 appear to be Q3 2016 parts. We already have seen the GP106 in a Drive PX2 unit and it appears to be around 200MM2 in size.
Forget the 1080. It is shockingly bad at DX12, as it is only 6fps faster than the Fury-X.
http://www.dsogaming.com/news/nvidia-gtx-1080-first-ashes-singularity-dx12-benchmarks-leaked/
The real facts about 1080 -
https://www.youtube.com/watch?v=tQaKdN_Wlck
The real facts about the 1080 -
https://www.youtube.com/watch?v=tQaKdN_Wlck
The launch event and coverage were very compelling - great job by Nvidia. Their budget for this card was massive too.
I might bite, not until I know all the competition, though! Especially with the SMP improvement for VR rendering. Let's see how this year goes.
hexus trust : n(baby):n(lover):n(sky)|>P(Name)>>nopes
Be Careful on the Internet! I ran and tackled a drive by mining attack today. It's not designed to do anything than provide fake texts (say!)
Much nicer breakdown of it here, IMO
http://wccftech.com/nvidia-geforce-g...12-benchmarks/
Videocardz also has a pretty nice breakdown, but comparing directly on those ones is somewhat more difficult as you can't know the specifics of the system behind it.
http://videocardz.com/59725/nvidia-g...x12-benchmarks
Regardless, Fury X seems to pull better performance on Crazy 4k for example, which suggests GDDR5X is still not necessarily ideal for 4k, but the new cards seem to be killing it on lower resolution. It also seems like on Extreme 4k the new Nvidias far exceed the power of the fury. I suspect that means that they are actually significantly faster than the Fury, but that Crazy 4k simply uses up too much memory bandwidth to where GDDR5/X suffers heavily, while fury can soldier on at it's usual performance due to HBM. There are a lot of oddities in the results though so it's really pretty hard to say.
Still, gives us a bit of an idea what to expect when official reviews and benchmarks come out.
Last edited by jag272; 11-05-2016 at 01:51 PM.
kalniel (11-05-2016)
Looks to be more or less exactly where they said it would be - up to 25% faster than a 980ti. Whether that's through brute force or improvements in DX12 we'll have to wait for more benchmarks to compare.
NDA lifts today so we'll here more soon. Slide deck for reviews (from editors day?) is available at
http://videocardz.com/59962/nvidia-g...h-presentation
Specs in line with estimates. Still looks like they don't have async compute but they've improved the premption. GPU boost also improved and there is slightly better mem compression tech (20% better). A new display sync mode called fast sync is mentioned, specifically for VR, though will be interesting to see if there are any benefits for the desktop - no tearing but also no flow control kind of looks like it's low/zero? frame buffer mode.
Nothing about the 1070 though, which is a shame since that's the one I've got questions about and whether there's a chance polaris can get close.
Interesting read, looking forward to benchmarks and also 1070 info. I'm somewhat deciding between them atm, as I could afford a 1080 but I don't think I could justify it over a 1070 so we'll see.
The memory difference seems to offer very little benefit, it serves 4k a little better but the ashes benchmarks seem to suggest that 4k still caps out the 1080 on memory, so it can probably be treated the same as GDDR5 and then we wait for HBM2.
Think NDA lift is at 6pm, though Hexus usually sneak in a little early, we'll see how much they play nVidia's game this time
There are currently 1 users browsing this thread. (0 members and 1 guests)
Posting Permissions
Reply With Quote