Loading...
> >

Video card choices: GTX 1080?

2»

Comments

  • marblemarble Posts: 7,500
    TheKD said:

    If they both fit, I would keep both. My mobo only has only two PCIE and they are way too close together to fit two cards :(

    Good point - I can't check until my PC arrives after being shipped to my new home here in New Zealand. May be only days away now so looking forward to getting back into my hobby. I wish I'd made a note of the model numbers of all the components - I bought the PC as a custom build 8 months ago.

  • TheKDTheKD Posts: 2,711

    Kicking myself for not forking out another $90 and getting a fullsize board lol. More RAM slots, and more room for PCIE stuff. My first PC I ever built, so I have to go easy on me haha.

  • luisvrayluisvray Posts: 3

    I installed yesterday the GTX1080 to improve my rendertimes, but it does'nt render anything. The image is black. I don't know what have i done wrong. Can someone help me. I had firs a Nvidia Geforce 680 installed and it worked well but to slow. and now a put it away and installed de GTX1080 but it renders the image black.

  • hphoenixhphoenix Posts: 1,335
    luisvray said:

    I installed yesterday the GTX1080 to improve my rendertimes, but it does'nt render anything. The image is black. I don't know what have i done wrong. Can someone help me. I had firs a Nvidia Geforce 680 installed and it worked well but to slow. and now a put it away and installed de GTX1080 but it renders the image black.

    Update your nvidia driver.  The new Iray requires 372.xx or higher.

     

  • nDelphinDelphi Posts: 1,922
    edited January 2017
    luisvray said:

    I installed yesterday the GTX1080 to improve my rendertimes, but it does'nt render anything. The image is black. I don't know what have i done wrong. Can someone help me. I had firs a Nvidia Geforce 680 installed and it worked well but to slow. and now a put it away and installed de GTX1080 but it renders the image black.

    Are you rendering with DAz Studio? If so, make sure you have the latest version. If you are using Poser 11, there is still an issue with Poser 11 and newer NVIDIA drivers. No word yet from SM on the issue that I can see. I am just waiting for an update to Poser 11 that might have that fixed.

    Post edited by nDelphi on
  • mcorrmcorr Posts: 1,104
    edited January 2017

    I was told that if I have a 4GB and a 2GB nvidia card, DS will render/use the card with the highest GB (not the combined GB of both cards, nor the lower GB card) and the combined CUDAs of both cards ... as long as ... the scene fits into the highest GB card. Mentioned above is that the scene must fit into the LOWER GB card (or it reverts back too CPU rendering). That doesn't square with what I was told. As I said, I was told it won't revert to CPU rendering as long as the larger GB card can hold the scene. Which is right?

    Post edited by mcorr on
  • mcorrmcorr Posts: 1,104
    edited January 2017

    .

    Post edited by mcorr on
  • Richard HaseltineRichard Haseltine Posts: 108,351
    mcorr said:

    I was told that if I have a 4GB and a 2GB nvidia card, DS will render/use the card with the highest GB (not the combined GB of both cards, nor the lower GB card) and the combined CUDAs of both cards ... as long as ... the scene fits into the highest GB card. Mentioned above is that the scene must fit into the LOWER GB card (or it reverts back too CPU rendering). That doesn't square with what I was told. As I said, I was told it won't revert to CPU rendering as long as the larger GB card can hold the scene. Which is right?

    Every card that the scene fits into will be used, any card that can't hold the scene will not be used at all.

  • mcorrmcorr Posts: 1,104
    Richard Haseltine said:

    Every card that the scene fits into will be used, any card that can't hold the scene will not be used at all.

    Thanks Richard!

     

  • mcorrmcorr Posts: 1,104

    Thanks Richard!

  • kyoto kidkyoto kid Posts: 41,928
    Artini said:

    I was surprised how little performance wise in iray GTX 1070 differs from GTX 1080 - look at the thread

    http://www.daz3d.com/forums/discussion/53771/iray-starter-scene-post-your-benchmarks#latest

    One can buy almost two GTX 1070 for the price of GTX 1080 and GTX 1070 has also 8GB of VRAM, only slightly slower.

    Just wonder, if that is not a better solution for iray rendering if you wish to spend similar amount of money,

    but would like to see some benchmark first.

     

    ...the 1070 at peak draws 10W less than my old Fermi 1GB GTX 460 as well as runs 10C cooler.  Yeah, far more efficient.

  • kyoto kidkyoto kid Posts: 41,928
    Muon Quark said:

    My brother mentioned that using two 1070 cards would give you more cuda cores than one 1080.  Don't know if that makes a difference or not.

    ..it would even give you more cores than a single Titan-X (P)

  • ArtisanSArtisanS Posts: 209

    Personally I always wait for the last one in the series, in my case the 980TI in yours the 1080TI.......and maybe just maybe I will add a 1080TI to my 980TI......

    Greets, ArtisanS

  • kyoto kidkyoto kid Posts: 41,928
    edited January 2017
    ColinFrench said:

    Note that nVidia will be releasing a GTX 1080 Ti and from the leaked specs it seems to be very close to a Titan X (12 GB memory). Hopefully it won't be as expensive as a Titan though, and it's release may drive down prices of regular 1080's somewhat.

    Yeah I know, there's always something better on the horizon and all that, but if you're just in the "looking around" stage, this might be something to keep an eye out for.

    ...hard to say at this time what the memory spec will actually be. I remember when the 980 Ti was at this stage, most sources were saying it would have a full 8 GB (and there was even supposed to be a 970 Ti as well) instead of the 6 GB that it actually had. 

    Currently I am hearing two numbers, depending on the source, of 10 GB and 12 GB.  As with the 980 Ti, we may very well not know how much memory it will have until it actually hits the streets.

    Post edited by kyoto kid on
  • boisselazonboisselazon Posts: 458

    price seem to drop a bit on 1070/1080 here in france: 350 for a 1070 and 499 for a 1080. It starts to be interesting (or more reasonable)

  • Takeo.KenseiTakeo.Kensei Posts: 1,303
    kyoto kid said:
    Muon Quark said:

    My brother mentioned that using two 1070 cards would give you more cuda cores than one 1080.  Don't know if that makes a difference or not.

    ..it would even give you more cores than a single Titan-X (P)

     

    One way to get a comparison would be to look at benchmarks but that usually doesnt say a lot if you are not trying to compare two cards

    The good way to have an Idea is to get numbers ie single precision processing power for rendering

    The GTX 1070 is around 5,7 Tflops processing power (and 6,4 at boost clock)

    The GTX 1080 is around 8,2 Tflops processing power (and 8,9 at boost clock)

    The Pascal Titan X is around 10,1 Tflops processing power (and 11 at boost clock)

    The GTX 1070 is 75% the speed of the GTX 1080 for around 75% it's price. You get what you pay

    If you get 2 x GTX 1070 you beat a Pascal Titan X for less money but 50 W more consumption,4GB less memory and one PCIe Slot occupied

    If we go on with comparison : a VCA is 8 x Quadro M6000. Making it have a theoritical processing power of 48 Tflops. That thing costs around 50K $

    With four Titan X you're not far in term of power and its price seems almost cheap in comparison (4 x 1500 $)

    An other interresting number is that the previous Titan X only had 6 Tflops processing power. That's what you get now with the GTX 1070 for a lot less $

     

    For people having slots that are too close, there was an interresting news in December about a single slot GTX 1070 http://www.anandtech.com/show/10926/galax-shows-off-single-slot-geforce-gtx-1070-graphics-card

  • mtl1mtl1 Posts: 1,508

    Just a heads up that Zotac came out with a 1080 Mini, so this will be perfect for those of us with smaller rigs.

  • boisselazonboisselazon Posts: 458
    Takeo.Kensei said:
    kyoto kid said:
    Muon Quark said:

    With four Titan X you're not far in term of power and its price seems almost cheap in comparison (4 x 1500 $)

    to be totally honest, the VCA is a really to plug machine, with MB+dual xeon. so, the comparisaon is closer to a 10k or 12K machine with 4 titan (6k)+MB/xeons/PSUetc..

    More of this, new VCA is around the P6000 quadro with 24GB VRAM. (7K each!) Some computer can have 8 of them !!! WOW

     

  • ArtiniArtini Posts: 10,342
    edited January 2017

    And this new laptop prototype at CES 2017 with three 4K screens and double GTX 1080 cards - wow.

    Post edited by Artini on
  • kyoto kidkyoto kid Posts: 41,928
    Takeo.Kensei said:
    kyoto kid said:
    Muon Quark said:

    My brother mentioned that using two 1070 cards would give you more cuda cores than one 1080.  Don't know if that makes a difference or not.

    ..it would even give you more cores than a single Titan-X (P)

     

    One way to get a comparison would be to look at benchmarks but that usually doesnt say a lot if you are not trying to compare two cards

    The good way to have an Idea is to get numbers ie single precision processing power for rendering

    The GTX 1070 is around 5,7 Tflops processing power (and 6,4 at boost clock)

    The GTX 1080 is around 8,2 Tflops processing power (and 8,9 at boost clock)

    The Pascal Titan X is around 10,1 Tflops processing power (and 11 at boost clock)

    The GTX 1070 is 75% the speed of the GTX 1080 for around 75% it's price. You get what you pay

    If you get 2 x GTX 1070 you beat a Pascal Titan X for less money but 50 W more consumption,4GB less memory and one PCIe Slot occupied

    If we go on with comparison : a VCA is 8 x Quadro M6000. Making it have a theoritical processing power of 48 Tflops. That thing costs around 50K $

    With four Titan X you're not far in term of power and its price seems almost cheap in comparison (4 x 1500 $)

    An other interresting number is that the previous Titan X only had 6 Tflops processing power. That's what you get now with the GTX 1070 for a lot less $

     

    For people having slots that are too close, there was an interresting news in December about a single slot GTX 1070 http://www.anandtech.com/show/10926/galax-shows-off-single-slot-geforce-gtx-1070-graphics-card

    ...yes but I am not doing deep learning, climatological or geological modelling, just rendering pretty pictures, so the Tflop rate is kind of meaningless.

  • kyoto kidkyoto kid Posts: 41,928
    boisselazon said:
    Takeo.Kensei said:
    kyoto kid said:
    Muon Quark said:

    With four Titan X you're not far in term of power and its price seems almost cheap in comparison (4 x 1500 $)

    to be totally honest, the VCA is a really to plug machine, with MB+dual xeon. so, the comparisaon is closer to a 10k or 12K machine with 4 titan (6k)+MB/xeons/PSUetc..

    More of this, new VCA is around the P6000 quadro with 24GB VRAM. (7K each!) Some computer can have 8 of them !!! WOW

     

    ..the VCA is intended for use with multi system networks not an individual workstation. Unless Daz decides to support Linux, it is kind of useless.

  • TaozTaoz Posts: 10,269

    Here's something about some GTX 1070 cards having Micron VRAM of poor quality, causing issues in some cases. Have anyone here had problems with this?

    http://www.pcgamer.com/msi-releases-geforce-gtx-1070-bios-update-to-fix-micron-memory-issue/

  • nicsttnicstt Posts: 11,715
    edited January 2017
    marble said:
    nDelphi said:
    Muon Quark said:

    Would having two nvidia cards be better than one for rendering?  Advantages?  Disadvantages?

    There are pluses and minuses. The pluses are faster renders. The minuses are minimal, like one card having less memory than the other. If a scene doesn't fit in the one with less memory it will only render in the one that has enough memory to hold the scene.

    I have two cards, a GTX 960 (4 GB) and a GTX 1080 (8 GB). I only use the GTX 1080 most of the time and keep my GTX 960 free to watch videos, etc, while I wait for the render to finish.

    I have a GTX 970 4GB but am constantly hitting the VRAM limit so am thinking of getting a 1070. The question I have is should I try to sell the 970 or keep both? I am currently using the 970 for Iray and the built-in GPU on the Skylake CPU for my monitor, but that is a hassle because it is a BIOS setting. So I'm thinking of keeping the 970 for the display and adding the 1070 for Iray. I'm not even sure that I have all the connectors coming from my power supply for another GPU although I'm pretty sure I have a strong enough PSU.

    One advantage might be that for small scenes - such as an animation with just one figure and very little other content - I could check both cards for quicker rendering of an image series for animation.

    The 970 would be great for driving your monitors; i use one to drive three x 2560x1440

    Post edited by nicstt on
  • nicsttnicstt Posts: 11,715
    kyoto kid said:
    ColinFrench said:

    Note that nVidia will be releasing a GTX 1080 Ti and from the leaked specs it seems to be very close to a Titan X (12 GB memory). Hopefully it won't be as expensive as a Titan though, and it's release may drive down prices of regular 1080's somewhat.

    Yeah I know, there's always something better on the horizon and all that, but if you're just in the "looking around" stage, this might be something to keep an eye out for.

    ...hard to say at this time what the memory spec will actually be. I remember when the 980 Ti was at this stage, most sources were saying it would have a full 8 GB (and there was even supposed to be a 970 Ti as well) instead of the 6 GB that it actually had. 

    Currently I am hearing two numbers, depending on the source, of 10 GB and 12 GB.  As with the 980 Ti, we may very well not know how much memory it will have until it actually hits the streets.

    10GB seems more likely, imo. It's going to be rediculously expensive though, just like the 1070 and 1080.

  • kyoto kidkyoto kid Posts: 41,928
    edited January 2017

    ...well considering I paid 295$ for my 1 GB 336 Core 460 when I built my system several years ago, another 100$ or so (GTX 1070) for eight times the VRAM and more than five and a half times the CUDA cores is not bad particularly since it runs cooler and draws a bit less power than my current card (I wouldn't need a new PSU).

    Back then Nvidia's flagship, the Quadro 6000, only had 6 GB, 448 CUDA cores and cost a whopping 4,999$  Meanwhile today, it's successor, the P 6000, retails for about 400$ more, has four times the VRAM (24 GB) and about eight and a half times the number of CUDA cores (3,840).

    In many ways the performance of what we are getting for the money has outweighed the increase in price.

     

    Post edited by kyoto kid on
Sign In or Register to comment.