Loading...
> >

Semi-OT - nVidia Pascal cards debut finished, 1080GTX and 1070GTX announced!

1…89111314…16

Comments

  • MEC4DMEC4D Posts: 5,249

    Seeing is believing .. I will give it fat 6 months first and then we see but my intuition tells me something else .

    StratDragon said:

    all this also assumes Nvidia gets it right by August, delivers a working set of tools and they actually do what they are supposed to without issues. Until that happens all performance gains are a matter of speculation and conjecture.

     

  • MEC4DMEC4D Posts: 5,249

    Exactly  +1 .. that are not speculations that are pure gossip full of hope just for $699 .. and there are the others  that spent $3600 for a card and still need 40 hours to finish the render 

    hphoenix said:
    lgrim72201 said:

    The thing to keep in mind is the 1080 and 1070s will be compatible with the new version of CUDA coming out in August. I believe if you have 2 or more cards it will use the combined vram of muliple cards. I read somewhere that CUDA 8 will not work with older Nvidia cards. I will wait for the dust to settle and upgrade my 980tis to 1080s. 3 1080s would give me 24 gb of vram and at that point I would never have to worry about the size of my scenes again.

    As others have noted, the Unified Memory Model in CUDA 6 does NOT mean you add up the memory.  The full scene still has to fit in the memory of any single card.

    The Unified Memory Model is simply an abstraction layer between the host program and the CUDA layer code.  Prior to CUDA 6, developers had to manually create the memory block allocations in both main memory, and on the CUDA devices, then copy it across.  "Unified Memory" is a simple (and very marketing friendly) way of saying the library code now does this for you.

     

     

  • nicsttnicstt Posts: 11,715
    MEC4D said:

    Exactly  +1 .. that are not speculations that are pure gossip full of hope just for $699 .. and there are the others  that spent $3600 for a card and still need 40 hours to finish the render 

    hphoenix said:
    lgrim72201 said:

    The thing to keep in mind is the 1080 and 1070s will be compatible with the new version of CUDA coming out in August. I believe if you have 2 or more cards it will use the combined vram of muliple cards. I read somewhere that CUDA 8 will not work with older Nvidia cards. I will wait for the dust to settle and upgrade my 980tis to 1080s. 3 1080s would give me 24 gb of vram and at that point I would never have to worry about the size of my scenes again.

    As others have noted, the Unified Memory Model in CUDA 6 does NOT mean you add up the memory.  The full scene still has to fit in the memory of any single card.

    The Unified Memory Model is simply an abstraction layer between the host program and the CUDA layer code.  Prior to CUDA 6, developers had to manually create the memory block allocations in both main memory, and on the CUDA devices, then copy it across.  "Unified Memory" is a simple (and very marketing friendly) way of saying the library code now does this for you.

     

     

    If anyone thinks NVidia is going to give a consumer card, the same or even really close to the same performance and functionality as a card costs five to ten times as much, well I've got some valuable crap stuff in my attic I'd be glad to sell em.

  • PlacidFalconPlacidFalcon Posts: 16
    edited July 2016

    I'll easily skip the 1070 and 1080.  The 1080 Ti is a no-brainer for me, but we're talking August 2017 at the earliest.

    Right now, I'm just locking down the CPU spec.  I've had the 4790K for a year and I'm struggling to find any weaknesses.  I can watch two streams of American Dad and Pawnstars, while rendering 5GB scenes in iray in a few minutes.  I'm curious if I even need X99.  It'll just be cores I may or may not even use.  I mean, when I use CPU as a photoreal device, it renders slower than GPU-alone...

    Post edited by PlacidFalcon on
  • kyoto kidkyoto kid Posts: 42,395
    nicstt said:
    MEC4D said:

    Exactly  +1 .. that are not speculations that are pure gossip full of hope just for $699 .. and there are the others  that spent $3600 for a card and still need 40 hours to finish the render 

    hphoenix said:
    lgrim72201 said:

    The thing to keep in mind is the 1080 and 1070s will be compatible with the new version of CUDA coming out in August. I believe if you have 2 or more cards it will use the combined vram of muliple cards. I read somewhere that CUDA 8 will not work with older Nvidia cards. I will wait for the dust to settle and upgrade my 980tis to 1080s. 3 1080s would give me 24 gb of vram and at that point I would never have to worry about the size of my scenes again.

    As others have noted, the Unified Memory Model in CUDA 6 does NOT mean you add up the memory.  The full scene still has to fit in the memory of any single card.

    The Unified Memory Model is simply an abstraction layer between the host program and the CUDA layer code.  Prior to CUDA 6, developers had to manually create the memory block allocations in both main memory, and on the CUDA devices, then copy it across.  "Unified Memory" is a simple (and very marketing friendly) way of saying the library code now does this for you.

     

     

    If anyone thinks NVidia is going to give a consumer card, the same or even really close to the same performance and functionality as a card costs five to ten times as much, well I've got some valuable crap stuff in my attic I'd be glad to sell em.

    ...we already have 12GB with the Maxwell Titan-X, I don't see where a 16 GB Pascal Titan-X woud be out of the question particularly since they just doubled the memory of their flagship Maxwell Quadro GPU to 24GB. 

    When I built my system several years ago people were saying 2 GB in a GTX card was the most that would ever be needed, today we have 6 times that.  I'm patient to wait and see what develops.

  • MEC4DMEC4D Posts: 5,249

    if you use 2 cards you don't need X99 , I had to since I am running 4 cards and i7490K was too slow for that and there was huge delay compared to only 2 cards. 

    hf1986_b8762a3ee6 said:

    I'll easily skip the 1070 and 1080.  The 1080 Ti is a no-brainer for me, but we're talking August 2017 at the earliest.

    Right now, I'm just locking down the CPU spec.  I've had the 4790K for a year and I'm struggling to find any weaknesses.  I can watch two streams of American Dad and Pawnstars, while rendering 5GB scenes in iray in a few minutes.  I'm curious if I even need X99.  It'll just be cores I may or may not even use.  I mean, when I use CPU as a photoreal device, it renders slower than GPU-alone...

     

  • MEC4DMEC4D Posts: 5,249

    we talking about this topic already, when the games are ready so will be  the GTX card but not soon as for now gamers not even need 12GB yet so this getting slowly . for  the rest rendering  enthusiast customers Quadro is the proper choice  according to Nvidia not GTX .. that is enough reason

    Don;t get me wrong I would love too, but not seeing it happening soon

     

    kyoto kid said:
    nicstt said:
    MEC4D said:

    Exactly  +1 .. that are not speculations that are pure gossip full of hope just for $699 .. and there are the others  that spent $3600 for a card and still need 40 hours to finish the render 

    hphoenix said:
    lgrim72201 said:

    The thing to keep in mind is the 1080 and 1070s will be compatible with the new version of CUDA coming out in August. I believe if you have 2 or more cards it will use the combined vram of muliple cards. I read somewhere that CUDA 8 will not work with older Nvidia cards. I will wait for the dust to settle and upgrade my 980tis to 1080s. 3 1080s would give me 24 gb of vram and at that point I would never have to worry about the size of my scenes again.

    As others have noted, the Unified Memory Model in CUDA 6 does NOT mean you add up the memory.  The full scene still has to fit in the memory of any single card.

    The Unified Memory Model is simply an abstraction layer between the host program and the CUDA layer code.  Prior to CUDA 6, developers had to manually create the memory block allocations in both main memory, and on the CUDA devices, then copy it across.  "Unified Memory" is a simple (and very marketing friendly) way of saying the library code now does this for you.

     

     

    If anyone thinks NVidia is going to give a consumer card, the same or even really close to the same performance and functionality as a card costs five to ten times as much, well I've got some valuable crap stuff in my attic I'd be glad to sell em.

    ...we already have 12GB with the Maxwell Titan-X, I don't see where a 16 GB Pascal Titan-X woud be out of the question particularly since they just doubled the memory of their flagship Maxwell Quadro GPU to 24GB. 

    When I built my system several years ago people were saying 2 GB in a GTX card was the most that would ever be needed, today we have 6 times that.  I'm patient to wait and see what develops.

     

  • kyoto kidkyoto kid Posts: 42,395

    ...not being either a game person or a pro graphics designer, I kind of feel like the redheaded stepchild of the CG world.  Yeah the toys we want will finally come to us, but only after the others have played with them.

    Maybe it's time to start looking more seriously at Octane as it doesn't require a boatload of graphics memory to render a scene in a reasonable amount of time.

  • PlacidFalconPlacidFalcon Posts: 16
    edited July 2016
    MEC4D said:

    if you use 2 cards you don't need X99 , I had to since I am running 4 cards and i7490K was too slow for that and there was huge delay compared to only 2 cards. 

     

    Thanks for the tip.

    In that case, I will wait for Kaby Lake then (the 6770k?). I still want to upgrade for quad-channel, but the X99 boards are too much for me (i will never have 4x1080 Tis). On a side note, have you noticed Daz Studio benefitting from Quad-channel?

    Post edited by PlacidFalcon on
  • kyoto kidkyoto kid Posts: 42,395

    ...when I built my system several years ago I went with a DDR3 Tri Channel setup and it seemed to render pretty quick.  When I would mention the render times people who had similar specs (but dual channel memory) were surprised at the results I was getting.

  • nicsttnicstt Posts: 11,715
    edited July 2016
    kyoto kid said:
    nicstt said:
    MEC4D said:

    Exactly  +1 .. that are not speculations that are pure gossip full of hope just for $699 .. and there are the others  that spent $3600 for a card and still need 40 hours to finish the render 

    hphoenix said:
    lgrim72201 said:

    The thing to keep in mind is the 1080 and 1070s will be compatible with the new version of CUDA coming out in August. I believe if you have 2 or more cards it will use the combined vram of muliple cards. I read somewhere that CUDA 8 will not work with older Nvidia cards. I will wait for the dust to settle and upgrade my 980tis to 1080s. 3 1080s would give me 24 gb of vram and at that point I would never have to worry about the size of my scenes again.

    As others have noted, the Unified Memory Model in CUDA 6 does NOT mean you add up the memory.  The full scene still has to fit in the memory of any single card.

    The Unified Memory Model is simply an abstraction layer between the host program and the CUDA layer code.  Prior to CUDA 6, developers had to manually create the memory block allocations in both main memory, and on the CUDA devices, then copy it across.  "Unified Memory" is a simple (and very marketing friendly) way of saying the library code now does this for you.

     

     

    If anyone thinks NVidia is going to give a consumer card, the same or even really close to the same performance and functionality as a card costs five to ten times as much, well I've got some valuable crap stuff in my attic I'd be glad to sell em.

    ...we already have 12GB with the Maxwell Titan-X, I don't see where a 16 GB Pascal Titan-X woud be out of the question particularly since they just doubled the memory of their flagship Maxwell Quadro GPU to 24GB. 

    When I built my system several years ago people were saying 2 GB in a GTX card was the most that would ever be needed, today we have 6 times that.  I'm patient to wait and see what develops.

    Likewise, I can see the memory getting added; the Titan series operate on a slightly different market, but even with the extra memory, are not offering even similar performance to the Quadros. Will a ten series card offer something closer? Sure the performance gains that seems to be offered will make it closer, but drivers won't be optimised, like the commercial variants are, for different tasks.

    kyoto kid said:

    ...when I built my system several years ago I went with a DDR3 Tri Channel setup and it seemed to render pretty quick.  When I would mention the render times people who had similar specs (but dual channel memory) were surprised at the results I was getting.


    I still have my 920, but I have upgraded since; I went for a dual channel i7; however, now I am considering an 8 core.

    ... Sadly two 22 core Xeons are a little out of my budget. :)

    Post edited by nicstt on
  • kyoto kidkyoto kid Posts: 42,395

    ...yeah my build as changed a bit. Going with a fast Hex Core i7 (3.6 GHz) as it calls for just two Titan Xs and 128 GB DDR4 2400 memory in Quad Channel mode for Ray Trace rendering (3DL and Carrara).

    Most likely by the time I get the funds up to build it, the new HBM 2 Pascal Titan series will probably be out. They could still be just 12 GB (four 3 stacks) or possibly 16 just to stay that step ahead of AMD. Quadros will most likely get the boost first.  I could see the 5200 equivalent having 16 GB and 6000 with 32.

    ..yeah all speculation at the moment, but from the different reports I have been reading also a decent possibility.

  • MEC4DMEC4D Posts: 5,249

    I will say keep your i74790K rig as long you can , there is not benefit from any upgrade unless you build all new system what will be faster with the system but for iray not really as with 2 cards it don't need more than 4 cores CPU and it is not going to use the full system power unless you render with 3DL , with the new DDR4 and X99 mobo and CPU you getting higher temperatures , I had to add double radiator for CPU that never again run @ 25C idle since it is hotter by default , I had to add DDR4 cooling or it will be very hot too so  don't do it unless you really need it .

    hf1986_b8762a3ee6 said:
    MEC4D said:

     

    Thanks for the tip.

    In that case, I will wait for Kaby Lake then (the 6770k?). I still want to upgrade for quad-channel, but the X99 boards are too much for me (i will never have 4x1080 Tis). On a side note, have you noticed Daz Studio benefitting from Quad-channel?

     

  • MEC4DMEC4D Posts: 5,249

    But sadly it is not with iray rendering, and you need to full upgrade your system and not just memory

    my DDR3 &i74790K with 2 GPU rendered faster in iray than X99 i75960x DDR4 with 2 GPU , since my CPU cores run @5.0Ghz and i75960x at 3.5Ghz so with 2 cards my older rig was faster and $2000 cheaper 

    but that is 2 cards scenario only

    but of course for other tasks it is 4 times faster on default than my old rig but not because of  the DDR4 only that run faster and use less power but produce more heat , then you have new CPU that use more power and run hotter by default you will need better case more airflow or your rig will not last long or you have to switch for water cooling . So you have some benefits and negative side effects 

    If my old MoBo's PCI-e never broke I would not upgraded my rig yet and for sure not until next year as that was my perfect 2 GPUs iray rig with DDR3 on board and that is a fact .

     

    kyoto kid said:

    ...when I built my system several years ago I went with a DDR3 Tri Channel setup and it seemed to render pretty quick.  When I would mention the render times people who had similar specs (but dual channel memory) were surprised at the results I was getting.

     

  • kyoto kidkyoto kid Posts: 42,395

    ...actually my case has excellent airflow as it is large (lot's of "breathing room" inside) with 6 fans including a large one right where the GPUs are positioned. The aftermarket CPU cooler I have is huge as well, with a set of heat pipes that I like to say resemble a V8 automobile engine manifold. 

    Agreed, to support Maxwell GPUs (980TI and Titan-X) would require a new system as the MB is an old X58 with an LGA 1366 socket and PCI 2.0 x 16 slots. As I also work in Carrara 64 bit and 3DL, upgrading the memory to DDR4 quad channel would have a benefit.

  • MEC4DMEC4D Posts: 5,249

    That''s for sure, you will benefit from upgrade more than iray users since it is the best for biased rendering 

    I just got  myself 4th new Titan X 12GB SC  Hydro for just $660 .. yeah!!!  goodbye Pascal !!!  my rig is now FINALIZED  now I am good for a very long time ..

    I just noticed that EVGA discontinued Titan X in general , I think it is the end for Titan X so I grabbed this offer as I don't want to screw my rig with Pascal card and not expect I  will ever get it at this price again especially that they do not offer it anymore , just $100 more than 980ti and faster than 1080,  I could not resist 

    kyoto kid said:

    ...actually my case has excellent airflow as it is large (lot's of "breathing room" inside) with 6 fans including a large one right where the GPUs are positioned. The aftermarket CPU cooler I have is huge as well, with a set of heat pipes that I like to say resemble a V8 automobile engine manifold. 

    Agreed, to support Maxwell GPUs (980TI and Titan-X) would require a new system as the MB is an old X58 with an LGA 1366 socket and PCI 2.0 x 16 slots. As I also work in Carrara 64 bit and 3DL, upgrading the memory to DDR4 quad channel would have a benefit.

     

  • mtl1mtl1 Posts: 1,508
    MEC4D said:

    But sadly it is not with iray rendering, and you need to full upgrade your system and not just memory

    my DDR3 &i74790K with 2 GPU rendered faster in iray than X99 i75960x DDR4 with 2 GPU , since my CPU cores run @5.0Ghz and i75960x at 3.5Ghz so with 2 cards my older rig was faster and $2000 cheaper 

    but that is 2 cards scenario only

    but of course for other tasks it is 4 times faster on default than my old rig but not because of  the DDR4 only that run faster and use less power but produce more heat , then you have new CPU that use more power and run hotter by default you will need better case more airflow or your rig will not last long or you have to switch for water cooling . So you have some benefits and negative side effects 

    If my old MoBo's PCI-e never broke I would not upgraded my rig yet and for sure not until next year as that was my perfect 2 GPUs iray rig with DDR3 on board and that is a fact .

     

    kyoto kid said:

    ...when I built my system several years ago I went with a DDR3 Tri Channel setup and it seemed to render pretty quick.  When I would mention the render times people who had similar specs (but dual channel memory) were surprised at the results I was getting.

     

    Cath, what's the power consumption for your 4790k at 5GHz? I have a 4790k undervolted, and it runs stock clocks pretty well... keeps the heat down in the room when I'm rendering ;)

    Also, you mentioned 4 cores: do you mean 4 physical cores or are 4 threads good enough? I usually turn off two threads in task manager so I can multitask/browse the web when I'm rendering...

  • MEC4DMEC4D Posts: 5,249

    I switched to i75960X one month ago when I upgraded all my system so right now with the new CPU the max I am getting is 140w at 3.5 Ghz for 8 cores 16 threads and 175W for 8 cores with 16 threads at 3.7 , @4.9Ghz it consumed too much power for my likening  and they can suck to 500W when OC but the sweet spot I found was 8 cores at 3.7Ghz for good performance overall and when I made CPU rendering tests only using 16 threads it never heat up more than 52C what is fantastic , but while idle 35C what is normal for the size of CPU and water will not cool it down bellow 35C where GPU are 25C on idle water cooled but are much smaller.

    I74790K was something between 75W to 114W at 5.0Ghz and produced nice heat from the radiator but was 25C on idle , however while rendering 65C so hooter than my 8 core CPU , however @ stock just perfect , it is my favorite CPU all times , I have i74790K  i74820K , i75820K and i75690X , and so far i74790K and i75960X are my favorites CPUs for 2 and 4 way GPU SLI or not

    and with cores I mean physical cores not threads so for 2 GPUs you can disable 2-4 threads if you run 4 core CPU as it will never use more than 45% anyway while GPU rendering , but for CPU rendering it will take all it have  of course .

     

    mtl1 said:
    MEC4D said:

    But sadly it is not with iray rendering, and you need to full upgrade your system and not just memory

    my DDR3 &i74790K with 2 GPU rendered faster in iray than X99 i75960x DDR4 with 2 GPU , since my CPU cores run @5.0Ghz and i75960x at 3.5Ghz so with 2 cards my older rig was faster and $2000 cheaper 

    but that is 2 cards scenario only

    but of course for other tasks it is 4 times faster on default than my old rig but not because of  the DDR4 only that run faster and use less power but produce more heat , then you have new CPU that use more power and run hotter by default you will need better case more airflow or your rig will not last long or you have to switch for water cooling . So you have some benefits and negative side effects 

    If my old MoBo's PCI-e never broke I would not upgraded my rig yet and for sure not until next year as that was my perfect 2 GPUs iray rig with DDR3 on board and that is a fact .

     

    kyoto kid said:

    ...when I built my system several years ago I went with a DDR3 Tri Channel setup and it seemed to render pretty quick.  When I would mention the render times people who had similar specs (but dual channel memory) were surprised at the results I was getting.

     

    Cath, what's the power consumption for your 4790k at 5GHz? I have a 4790k undervolted, and it runs stock clocks pretty well... keeps the heat down in the room when I'm rendering ;)

    Also, you mentioned 4 cores: do you mean 4 physical cores or are 4 threads good enough? I usually turn off two threads in task manager so I can multitask/browse the web when I'm rendering...

     

  • kyoto kidkyoto kid Posts: 42,395
    edited July 2016
    MEC4D said:

    That''s for sure, you will benefit from upgrade more than iray users since it is the best for biased rendering 

    I just got  myself 4th new Titan X 12GB SC  Hydro for just $660 .. yeah!!!  goodbye Pascal !!!  my rig is now FINALIZED  now I am good for a very long time ..

    I just noticed that EVGA discontinued Titan X in general , I think it is the end for Titan X so I grabbed this offer as I don't want to screw my rig with Pascal card and not expect I  will ever get it at this price again especially that they do not offer it anymore , just $100 more than 980ti and faster than 1080,  I could not resist 

    kyoto kid said:

    ...actually my case has excellent airflow as it is large (lot's of "breathing room" inside) with 6 fans including a large one right where the GPUs are positioned. The aftermarket CPU cooler I have is huge as well, with a set of heat pipes that I like to say resemble a V8 automobile engine manifold. 

    Agreed, to support Maxwell GPUs (980TI and Titan-X) would require a new system as the MB is an old X58 with an LGA 1366 socket and PCI 2.0 x 16 slots. As I also work in Carrara 64 bit and 3DL, upgrading the memory to DDR4 quad channel would have a benefit.

     

    ...oh bugger.  I like the EVGA ones. Don't want to have to settle for only 8 GB considering the types of scenes I do, and don't care for used gear especially if the previous owner overclocked the daylights out of it without liquid cooling..

    Well I guess that is that, so much for the new workstation design.

    There are days I really really hate the tech curve.

    Post edited by kyoto kid on
  • MEC4DMEC4D Posts: 5,249

    well who knows what the 2017 bring us again , last year in japan Nvidia stated that Pascal will be up maximum  to 16GB but they did not do it so maybe they preparing another Titan P for 2017 lol

    Also Titan X servers are no more offered they replaced Titan X with 1080 so it may be the dusk of Titan X, so quick and the Titan X SC are higher in price as they realized the card is still better and best choice for workstation for a ok price compared to Quadro that everyone hate as that is just another Titan X at stock speed without magic iray driver or 30% better performance , it is just for offices than can scrap it off from the taxes anyway and  pretending PRO users, but all smaller studios I know run Titan X(s) as first choice especially when working with iray and studio max .

    So don't give up , there may be something on the corner we all not even expecting , and  the prices will be down for other components you need soon like good CPU and Memory since now it is crazy overprice 

    kyoto kid said:
    MEC4D said:

    That''s for sure, you will benefit from upgrade more than iray users since it is the best for biased rendering 

    I just got  myself 4th new Titan X 12GB SC  Hydro for just $660 .. yeah!!!  goodbye Pascal !!!  my rig is now FINALIZED  now I am good for a very long time ..

    I just noticed that EVGA discontinued Titan X in general , I think it is the end for Titan X so I grabbed this offer as I don't want to screw my rig with Pascal card and not expect I  will ever get it at this price again especially that they do not offer it anymore , just $100 more than 980ti and faster than 1080,  I could not resist 

    kyoto kid said:

    ...actually my case has excellent airflow as it is large (lot's of "breathing room" inside) with 6 fans including a large one right where the GPUs are positioned. The aftermarket CPU cooler I have is huge as well, with a set of heat pipes that I like to say resemble a V8 automobile engine manifold. 

    Agreed, to support Maxwell GPUs (980TI and Titan-X) would require a new system as the MB is an old X58 with an LGA 1366 socket and PCI 2.0 x 16 slots. As I also work in Carrara 64 bit and 3DL, upgrading the memory to DDR4 quad channel would have a benefit.

     

    ...oh bugger.  I like the EVGA ones. Don't want to have to settle for only 8 GB considering the types of scenes I do, and don't care for used gear especially if the previous owner overclocked the daylights out of it without liquid cooling..

    Well I guess that is that, so much for the new workstation design.

    There are days I really really hate the tech curve.

     

  • kyoto kidkyoto kid Posts: 42,395
    edited July 2016

    ...the issue I have is the system specs and compatibility is all set around the Titan-X and DDR4 Quad Channel memory. The Hex Core GPU I selected has the fastest "out of the box" clock speed at 3.6GHz and like I mentioned, is 500$ less than the 8 core 3.2 GHz one. Memory prices are bottoming right now, I can get 128GB for just a little more than what 24 GB would have cost me back when I built my current system.  If I could get a couple Titan-X Hydros for 600$ each, that would be a real deal.

    Sadly I don't have the resources at this time to even just purchase the GPUs before they completely disappear from the market.  Yeah, Pascal may very well turn out to be the next best thing since sliced bread, but if they hold back to satisfy only the gaming and Pro markets and don't offer an 3D enthusiasts' card with more than the memory resources of the 1070/1080 (other than the Quadro line), it's a "no sale" for me.  If am to continue I rendering in Iray, I want it to occur on the GPU.

    Post edited by kyoto kid on
  • mtl1mtl1 Posts: 1,508

    Oh no, EVGA discontinued the Titan X? I suppose I could always pick up a few used ones next year when the next Titan comes out...

  • kyoto kidkyoto kid Posts: 42,395

    ...just make sure they are the water cooled ones.  Like I mentioned, the previous owner might have overclocked them so if they were the air cooled version there is a good chance they won't hold up as long.

  • ANGELREAPER1972ANGELREAPER1972 Posts: 4,826

    that explains why they are so hard to find in Australia and many that did offer them no longer seem to in their builds 980, 980Ti, 1070, and 1080s are the best on offer now though many here don't even offer those

  • Kevin SandersonKevin Sanderson Posts: 1,643

    Plenty of water cooled Titans on ebay.

     

  • MEC4DMEC4D Posts: 5,249
    edited July 2016

    The Titan X Hybrid run almost 1400Mhz out of the box ( better than SC ) so there is less chance they are going to be overclocked often and the temp never reach 67C  ( my usual 37-42 max ), so second hand like Titan X stock or SC are the worse as they are running at 80C with games or rendering ( Nvidia stock cooling)

    Titan X consume only 9watt while idle  , never reach 250W even with games or 170W while rendering so power efficient compared to other cards from preview generation 

    I could not go with Pascal as it would alter my rig speed , but if you don't have high end cards yet I would wait and see the Pascal performance in iray and decide on that.

    mtl1 said:

    Oh no, EVGA discontinued the Titan X? I suppose I could always pick up a few used ones next year when the next Titan comes out...

     

    Post edited by MEC4D on
  • MEC4DMEC4D Posts: 5,249
    edited July 2016

    Small tip , Just pay attention it is original Hybrid not SC or stock with Hybrid kit as they look exactly the same and I just don't trust purchases on ebay too much, original hydro run much faster as they have upgraded firmware , SC reach max 1277Mhz with boost ,  Hydro 1377Mhz to 1400Mhz  , I have now 2 new SC that I upgraded to Hydro and 2 original new Hydro all with 3 years warranty , the latest Hybrids have LED text on the side (EVGA GTX Hybrid ) the old one not  only small text Geforce GTX Titan X , so if you get the one with LED it was not used more than 6 months as they made them later . Anyway after you register your card with EVGA you wil see the truth anyway based on the serial number .

    Kevin Sanderson said:

    Plenty of water cooled Titans on ebay.

     

     

    Post edited by MEC4D on
  • MEC4DMEC4D Posts: 5,249

    I found only 10 Titan X Hybrid on ebay , from $400 to $2.500

    that is not much and the 2 cheap one for $400 using the Hybrid kit not original ( I would skip it ), original start at $700 

    not much folk selling it as the card run faster and cooler than Pascal 1080 so no reason to get rid off it yet and a lot of people use it in workstations in first place  

    I see a lot of air cooled 980Ti 6 GB for $255 so 50% less than new one , not bad deals here even for an open box with 3 years warranty it is still a stiller 

    whatever you decide make sure it is not the stock Nvidia one as they run too hot you will lose on performance and the heat will drive you crazy

  • Oso3DOso3D Posts: 15,115
    edited July 2016

    It's getting annoyingly hard to pick out new machines. :/

    I keep kicking my machine decisions down the street hoping things will be more clear in a month, two months, a year...

    Post edited by Oso3D on
  • ANGELREAPER1972ANGELREAPER1972 Posts: 4,826

    that's the thing with tech there's always going to be something newer better and sometimes cheaper being released usually not long after you invest in something you think is the next best thing

Sign In or Register to comment.