Message boards :
Number crunching :
GPU Wars 2016: GTX 1050 Ti & GTX 1050: October 25th
Message board moderation
Previous · 1 · 2 · 3 · 4 · 5 · 6 · 7 · 8 . . . 19 · Next
Author | Message |
---|---|
Al Send message Joined: 3 Apr 99 Posts: 1682 Credit: 477,343,364 RAC: 482 |
Jason, Thank You for that most excellent response. |
shizaru Send message Joined: 14 Jun 04 Posts: 1130 Credit: 1,967,904 RAC: 0 |
Not me, after watching them release the 980 then waiting a few months to release the 980Ti as short time later with specs that equalled or passed the Titans Xs Not arguing with ya or anything, but the Ti was something like 9 months after the vanilla 980. As for the Titan, we now know that it's a niche GPU and likely "useless" for SETI. In the sense that most of the extra money you pay is for DP performance (which for SETI-only crunchers is useless)... Which brings me to my next post... |
Zombu2 Send message Joined: 24 Feb 01 Posts: 1615 Credit: 49,315,423 RAC: 0 |
The real question is Who's getting one on the 27th and does it work with seti ...i'd hate to buy 4 and then no workie I came down with a bad case of i don't give a crap |
woohoo Send message Joined: 30 Oct 13 Posts: 972 Credit: 165,671,404 RAC: 5 |
Only works in eight way sli |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
The real question is Who's getting one on the 27th and does it work with seti ...i'd hate to buy 4 and then no workie Some testing would be needed to make sure. I have to wonder about the speed. In that we've seen invalids with over locked gpus. How would gpus running at twice the current levels fair? Would the current apps have to be changed to take advantage of them? |
shizaru Send message Joined: 14 Jun 04 Posts: 1130 Credit: 1,967,904 RAC: 0 |
(continued from previous post) ....So comparing the Titan X to the 1070 was a bad idea. The price difference clouded my judgment. Sorry. However the 1070 should be noticeably better than the vanilla 980 @Seti crunching. Maybe 20% better realistically? Too early to tell of course... But that brings us to Juan's question: - Will be the 1070 a real winner or is better to go to the 1080? Nah... maybe for gaming there'll be an edge but not for Seti. Evidence so far suggests that dollar-for-dollar they're likely almost completely equal. Actually it's the 1080 that may be marginally better in terms of dollar-to-throughput. But again, likely not enough of a dif to be important. So I guess the real answer is "whichever suits you best". By that I obviously mean budget, existing power supply, case, number of wanted cards etc. However the above does NOT take power consumption into consideration. But difference there likely won't be significant either. Because these are same-generation cards. But difs may likely be considerable to previous gen cards. Probably worth the change-over for 24/7 crunchers of multiple high end cards as power savings in dollars could run into the hundreds annually. But we still have time to figure all that out:) Wiki specs are up now BTW: https://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units#GeForce_10_Series |
Zombu2 Send message Joined: 24 Feb 01 Posts: 1615 Credit: 49,315,423 RAC: 0 |
well since these cards are not overclocked i'd think there should be no invalids but i'm a lil concerned about the seti app not playing well EDIT: The specs are wrong according to nvidias release presentation the cards run over 2 ghz stock plus boost I came down with a bad case of i don't give a crap |
Al Send message Joined: 3 Apr 99 Posts: 1682 Credit: 477,343,364 RAC: 482 |
Just a guess about speeds and over clocked vs. invalids. My guess is that they are caused by the card running right up to the bleeding edge of it's capabilities, and any little thing that pushes it over that edge, even momentarily could toss those errors (heat, bit errors, who knows). Since at least at stock speeds they are running within their design specs, I would _hope_ that they would do anything untoward. But then, I've been wrong before, and certainly will be again. *edit* Yeah, what everyone else said. lol (slow typer sometimes...) |
Zombu2 Send message Joined: 24 Feb 01 Posts: 1615 Credit: 49,315,423 RAC: 0 |
Nvidia said yesterday if you get the 699$ card that you get enormous OC potential over 2ghz clock and did you guys see the 10ghz they mentioned on the ram compared to the titan it will be more then twice as fast as the titan I came down with a bad case of i don't give a crap |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
I always keep in mind these are designed for gaming first, crunching is a secondary (if ever) concern. |
Zombu2 Send message Joined: 24 Feb 01 Posts: 1615 Credit: 49,315,423 RAC: 0 |
actually if anyone watched the presentation yesterday they talked about sience crunching a lot and the speed of these cards for it I think/hope that this card is gonna be the holy grail of crunching I came down with a bad case of i don't give a crap |
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
I'll have to watch later tonight. Stuck at work with only a cell phone and no charger lol. |
Zombu2 Send message Joined: 24 Feb 01 Posts: 1615 Credit: 49,315,423 RAC: 0 |
HAH never leave the house with some form of charger I came down with a bad case of i don't give a crap |
shizaru Send message Joined: 14 Jun 04 Posts: 1130 Credit: 1,967,904 RAC: 0 |
Going to wait until they decide to show us all of their products, not just a first few that come out. Actually... I just remembered there's a GP102 rumor flying around. IF true then that'll likely be a Titan but maybe even a 1080Ti too :) So if maximum throughput per card is the goal then yes, it may actually show up sooner rather than later. But again in dollars-to-throughput I doubt the scales will tip dramatically... However next to no-one thinks there's a chance the GP100 ("Big" Pascal in the Tesla card) is going to show up in a GeForce this gen. So maybe around this time next year... Again all rumors. |
shizaru Send message Joined: 14 Jun 04 Posts: 1130 Credit: 1,967,904 RAC: 0 |
Quoting myself: Actually it's the 1080 that may be marginally better in terms of dollar-to-throughput. Damn. My bad. Too tired to think straight these days and was relying on math I did on the back of a napkin last night right before going to bed... and got the fractions the wrong way 'round. Oops :) Anyway: GTX 1070 6500 GFLOPS @$379 GTX 1080 9000 GFLOPS @$599 So 6500/379 equals 17.15 GFLOPS per dollar for the 1070 And 9000/599 equals 15.025 GFLOPS per dollar for the 1080 So you'll probably get at least 10% more real-world Seti performance per dollar with the 1070. Theoretically maybe even 15%... |
shizaru Send message Joined: 14 Jun 04 Posts: 1130 Credit: 1,967,904 RAC: 0 |
OK so let's do some number crunching in the Number Crunching forum: Wiki doesn't have the full 1070 specs up yet so I'll use the 1080 specs and Wiki's more conservative 8228 GFLOPS. But let's stay on the optimistic side and pretend for now that this generation is compute-oriented (because of VR requirements and those insane core frequencies) until/unless we have evidence that indicates otherwise. So here goes nothing: 8228 GFLOPS divided by GTX 460 907.2 GFLOPS equals 9.069 times faster. GTX 480 1344.96 GFLOPS - GTX 1080 6.117 times faster - GTX 560 Ti 1263.4 GFLOPS - 6.512 times faster - GTX 580 1581.1 GFLOPS - 5.204 times faster - GTX 660 1881.6 GFLOPS - 4.373 times faster - GTX 680 3090.43 GFLOPS - 2.662 times faster - GTX 770 3213.3 GFLOPS - 2.560 times faster - GTX 780 Ti 5045.7 GFLOPS - 1.630 times faster - GTX 970 3494 GFLOPS - 2.354 times faster - GTX 980 Ti 5632 GFLOPS - 1.461 times faster So if Pascal really is compute oriented and there's a loose real-world GFLOPS correlation with Fermi gen cards then you get: Roughly 6x the compute power of a 480 or 560 Ti for about 180 Watts. Pas mal ;) |
shizaru Send message Joined: 14 Jun 04 Posts: 1130 Credit: 1,967,904 RAC: 0 |
"We know from NVIDIA’s previous Tesla P100 announcement that the company has opted to pay special attention to CUDA Core efficiency with Pascal, improving the throughput of the architecture as opposed to adding a significant number of additional CUDA cores. As a result, like previous architectural shifts, core-for-core comparisons are going to be tricky thanks to GP104 hitting a higher percentage of its theoretical throughput in practice." http://anandtech.com/show/10304/nvidia-announces-the-geforce-gtx-1080-1070 Awesome :) |
arkayn Send message Joined: 14 May 99 Posts: 4438 Credit: 55,006,323 RAC: 0 |
|
Zalster Send message Joined: 27 May 99 Posts: 5517 Credit: 528,817,460 RAC: 242 |
Ok Zombu2, I watched the entire presentation. He gave a passing mention to scientists. I had hoped for something a little more in depth as to how it would help. Looks like we will have to wait until someone gets one and puts it thru it's paces. Volunteers? lol.. Edit... I'm holding out for hybrid models. Previous experience tells me that I need those due to the noise level with multiple GPUs in 1 case, and I'm never going for ACX ever again, lol.. |
Grant (SSSF) Send message Joined: 19 Aug 99 Posts: 13727 Credit: 208,696,464 RAC: 304 |
I will be waiting for the GTX 1060. GTX 1060Ti or 1050Ti for me. EDIT- similar power consumption to my GTX 750Tis, but double the performance (with a bit of luck). That'd make it worthwhile for me. Grant Darwin NT |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.