Reputable Contributor AbnRanger Posted September 20, 2014 Reputable Contributor Report Share Posted September 20, 2014 Basically, the GTX 980, at roughly $570USD, performs slightly better than the $750 GTX 780Ti...drawing considerably less power http://www.anandtech.com/show/8526/nvidia-geforce-gtx-980-review Might have to get one, when the manufacturers start coming out with non-reference card designs (usually include a decent factory overclock + MUCH better cooling solutions). 1 Quote Link to comment Share on other sites More sharing options...
Reputable Contributor AbnRanger Posted September 20, 2014 Author Reputable Contributor Report Share Posted September 20, 2014 Toms Hardware review (always enjoy these): http://www.tomshardware.com/reviews/nvidia-geforce-gtx-980-970-maxwell,3941.html Quote Link to comment Share on other sites More sharing options...
Advanced Member Aleksey Posted September 20, 2014 Advanced Member Report Share Posted September 20, 2014 still only 256bit memory bus so they just refuse to make something that is better for CUDA than a 580? Quote Link to comment Share on other sites More sharing options...
Advanced Member Grimm Posted September 20, 2014 Advanced Member Report Share Posted September 20, 2014 I'm looking at a 970 very closely. I use Octane so the CUDA performance doesn't look good with the first round of testing that I have seen. But even with the lackluster performance, that might get better with new drivers, etc., I just might get it. Even now it's twice as fast as my old 460, uses less power, has double the memory and at $330 I don't know if I will be able to resist. Both the Octane and the Blender (Cycles) guys have noticed that with the last round of drivers from Nvidia the Kepler series has gotten a lot faster than Fermi. Maybe the same will happen with Maxwell? Quote Link to comment Share on other sites More sharing options...
PolyHertz Posted September 20, 2014 Report Share Posted September 20, 2014 The mobile improvements are much more impressive then the desktop improvements this time around: Quote Link to comment Share on other sites More sharing options...
Advanced Member Aleksey Posted September 21, 2014 Advanced Member Report Share Posted September 21, 2014 I'm looking at a 970 very closely. I use Octane so the CUDA performance doesn't look good with the first round of testing that I have seen. But even with the lackluster performance, that might get better with new drivers, etc., I just might get it. Even now it's twice as fast as my old 460, uses less power, has double the memory and at $330 I don't know if I will be able to resist. Both the Octane and the Blender (Cycles) guys have noticed that with the last round of drivers from Nvidia the Kepler series has gotten a lot faster than Fermi. Maybe the same will happen with Maxwell? define: "a lot" =) I think i'll just wait for a 6gb 9 series now... But im curious if this makes the gtx770 for example beat the 580... Quote Link to comment Share on other sites More sharing options...
Advanced Member Grimm Posted September 21, 2014 Advanced Member Report Share Posted September 21, 2014 Here is a Cycles test for Blender 2.71. You can see that the 580 gets beat by the 680 and trounced by the 770. http://blenderartists.org/forum/showthread.php?342000-The-new-Cyles-GPU-2-71-Benchmark/page3 My two little 460's can't compete anymore. Quote Link to comment Share on other sites More sharing options...
Advanced Member Aleksey Posted September 21, 2014 Advanced Member Report Share Posted September 21, 2014 Here is a Cycles test for Blender 2.71. You can see that the 580 gets beat by the 680 and trounced by the 770. http://blenderartists.org/forum/showthread.php?342000-The-new-Cyles-GPU-2-71-Benchmark/page3 My two little 460's can't compete anymore. well that's some interesting info. ah.. decisions decisions... thnks for that. will wait and see how the 9's go. but this also means for the 7's and 9's to out perform the 5's, the developers have to update to the new cuda toolkit... Quote Link to comment Share on other sites More sharing options...
Advanced Member L'Ancien Regime Posted September 22, 2014 Advanced Member Report Share Posted September 22, 2014 (edited) Might have to get one, when the manufacturers start coming out with non-reference card designs (usually include a decent factory overclock + MUCH better cooling solutions). http://www.techpowerup.com/ And word is that the 990 will be here in early 2015; that'll be 16nm with 6 or even 8 gigs of ddr5 RAM.. Now DDR5 RAM is expensive but it'll be worth the price. I'm guessing something in the vicinity of $1000. http://www.tweaktown.com/news/38953/nvidia-will-reportedly-launch-its-geforce-gtx-990-in-early-2015/index.html It won't stop there, as we should see NVIDIA unveil the GeForce 1000 series sometime in the next 18-24 months, too. The Volta??? Edited September 22, 2014 by L'Ancien Regime Quote Link to comment Share on other sites More sharing options...
Reputable Contributor AbnRanger Posted September 22, 2014 Author Reputable Contributor Report Share Posted September 22, 2014 They did mention something about the 900's having 40% better CUDA performance than the Kepler cards (including the 780). I may have to go ahead and grab a 970, at that price. I can still use the 580's on the render node, when rendering with Thea or Moskito, may have to get acquainted with Cycles and the Blender's smoke/fire simulation system. The latest NVidia drivers have seemed to bump up the overall tumbling and wireframe performance in the Sculpt room, even on the 580. But some on here have said it improved on the Kepler cards, especially. I don't like them going back to a smaller memory bus. Still cannot figure out why they are taking this approach. They went to 256bit with the 600-770 series (little Kepler). Then went back up the Titan and 780. Now, they are going backwards again. I understand they are using some other means to offset that, but I still don't get why you engineer to go backwards on the memory bus. They must REALLY want to skimp on the gold those bus lanes are made with. Quote Link to comment Share on other sites More sharing options...
Advanced Member L'Ancien Regime Posted September 22, 2014 Advanced Member Report Share Posted September 22, 2014 http://www.geeks3d.com/20140919/game24-the-journey-of-maxwell-gtx-980/?utm_source=feedburner&utm_medium=feed&utm_campaign=Feed%3A+TheGeeksOf3d+%28The+Geeks+Of+3D%29 It's not the gold..it's your electricity consumption they're lowering. They're using the new system to dramatically reduce electricity consumption, a not inconsiderable thing particularly for academics and industry building CUDA supercomputers. Quote Link to comment Share on other sites More sharing options...
Reputable Contributor AbnRanger Posted September 23, 2014 Author Reputable Contributor Report Share Posted September 23, 2014 I guess it wouldn't hurt to go ahead an buy one, if one is so inclined to do so, and if you notice that performance drops from what one has (with a 580 or so), in a given area....most retailers will let you return the item. My problem with the GTX 670 4GB I bought over a year ago, was the NVidia rep I spoke with, said they were working on a driver. I waited past the 30day point, and nothing changed. So, I ended up having to sell it on eBay for about a $75 loss. This time around, I'd make sure to get it returned before that 30 day deadline. I'm actually thinking of snagging a 780 6GB off of eBay, because they have dropped about $100 or more now that the 900 series is out. But the 970 4GB sure looks tempting at $330. Unlike the 770 vs the 780, the 970 is the same card as it's sibling, just fewer CUDA cores. Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.