Loading...
> >

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

1…1011121416

Comments

  • kyoto kidkyoto kid Posts: 41,925

    ...but why not just put the fan over the where the GPU and memory chips are? That would make more sense. 

    Or do like AMD does with the Fury X and go to a fluid based cooling system.

  • fredmusicfredmusic Posts: 29
    kyoto kid said:

    ...but why not just put the fan over the where the GPU and memory chips are? That would make more sense. 

    Or do like AMD does with the Fury X and go to a fluid based cooling system.

     

    You can but it may reduce the surface area of the cooling fins, so the air cooling is less effective.  That means either higher fan speeds or less cooling ability.  Below Titan X heatsink, which is pretty large with one blower fan to the side.

    http://www.puregaming.net/wp-content/uploads/2015/12/titanx_heatsink_w_600-730x430.jpg

    Below is a twin fan cooler placed over the top of the components.  Here the cooler fins are not as deep due to the thickness limitation of the PCIe slot, so you need more than one fan.  Water cooling needs much less surface area, so it doesn't have to have the space for the cooling fins.

    https://dlcdnimgs.asus.com/websites/global/products/sjiKJMMbjgY73Ell/img/burger.jpg

  • kyoto kidkyoto kid Posts: 41,925

    ...so by that token, if a Hydro version of an HBM "Titan P" were released we might then see a smaller form factor similar to AMDs Fury-X.

  • ZooriZoori Posts: 6
    edited July 2016

    Some interesting results for the older cards:

    https://www.pugetsystems.com/labs/articles/NVIDIA-Iray-GPU-Performance-Comparison-785/

    Post edited by Zoori on
  • hphoenixhphoenix Posts: 1,335

    Well, the new Titan-X (what we have called the Titan-P) has been announced.  Details in this thread:  http://www.daz3d.com/forums/discussion/101816/new-pascal-titan-announced-1200

    $1200 US.  3,584 CUDA cores at 1.53GHz.  12 GB of GDDR5X memory.    Available from nVidia directly on August 2nd.

    No further need for speculation.....

     

  • MEC4DMEC4D Posts: 5,249

    I saw it this morning , http://arstechnica.com/gadgets/2016/07/gtx-titan-x-pascal-specs-price-release-date/

    well I expected it to be better with more VRAM.. 24% faster than 1080 .. meeeeh ! for the new stock Titan X 11 ... however manufacturers are going to make something better from that stock card soon but of course at higher power consumption than 250W what is out of my interest.

    I am glad I have what I have for now and waiting for something better in the next 2 years to arrive .. 

  • alexhcowleyalexhcowley Posts: 2,403
    edited July 2016
    MEC4D said:

    I saw it this morning , http://arstechnica.com/gadgets/2016/07/gtx-titan-x-pascal-specs-price-release-date/

    well I expected it to be better with more VRAM.. 24% faster than 1080 .. meeeeh ! for the new stock Titan X 11 ... however manufacturers are going to make something better from that stock card soon but of course at higher power consumption than 250W what is out of my interest.

    I am glad I have what I have for now and waiting for something better in the next 2 years to arrive .. 

    Cath,

    A while back, I worked out the cost per cuda core of the Maxwell cards. Best value for money were the 970 and the 980ti.  Poorest value were the 980 and the Titan X, although the Titan X did have more memory, of course.  The Pascal Titan seems to follow this, roughly twice the price of a 1080 but significantly less than twice the cuda cores.  It looks like my plan to buy a 1080 in a few months time and another one next year could be the way to go.

    Cheers,

    Alex.

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

    Hi Alex, totally agree , 

    alexhcowley said:
    MEC4D said:

    I saw it this morning , http://arstechnica.com/gadgets/2016/07/gtx-titan-x-pascal-specs-price-release-date/

    well I expected it to be better with more VRAM.. 24% faster than 1080 .. meeeeh ! for the new stock Titan X 11 ... however manufacturers are going to make something better from that stock card soon but of course at higher power consumption than 250W what is out of my interest.

    I am glad I have what I have for now and waiting for something better in the next 2 years to arrive .. 

    Cath,

    A while back, I worked out the cost per cuda core of the Maxwell cards. Best value for money were the 970 and the 980ti.  Poorest value were the 980 and the Titan X, although the Titan X did have more memory, of course.  The Pasal Titan seems to follow this, roughly twice the price of a 1080 but significantly less than twice the cuda cores.  It looks like my plan to buy a 1080 in a few months time and another one next year could be the way to go.

    Cheers,

    Alex.

     

  • nicsttnicstt Posts: 11,715
    edited July 2016
    alexhcowley said:
    MEC4D said:

    I saw it this morning , http://arstechnica.com/gadgets/2016/07/gtx-titan-x-pascal-specs-price-release-date/

    well I expected it to be better with more VRAM.. 24% faster than 1080 .. meeeeh ! for the new stock Titan X 11 ... however manufacturers are going to make something better from that stock card soon but of course at higher power consumption than 250W what is out of my interest.

    I am glad I have what I have for now and waiting for something better in the next 2 years to arrive .. 

    Cath,

    A while back, I worked out the cost per cuda core of the Maxwell cards. Best value for money were the 970 and the 980ti.  Poorest value were the 980 and the Titan X, although the Titan X did have more memory, of course.  The Pasal Titan seems to follow this, roughly twice the price of a 1080 but significantly less than twice the cuda cores.  It looks like my plan to buy a 1080 in a few months time and another one next year could be the way to go.

    Cheers,

    Alex.

    It is of course true, and valid.

    That extra memory though is different and has a different value or set of values applied to it.

    There are work arounds to using a card with half the memory.

    I could buy 3 980ti for not much more than one titan; but if it won't fit in memory, I waste time splitting up parts of the scene, or other work arounds, or I put up with massively slower renders.

    ... So it is: what price is that extra memory is worth?

    After the hastle with the current scene, I'm reevaluating. :)

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

    To load 17 minions poly into DS scene in open GL it will use around 1GB vram and DS will use 4.4GB of system RAM and will last very long to load.
    17 milion poly = 692 Genesis on sub 1 level without textures or 17 Stonemason Fern Lakes

    you will have trouble even smooth operate Studio after you load it as DS is not build for that
    now you switch to iray , it will use 4.3GB of VRAM from your card and over 5GB of Vram total to render it

    so that is where 6 GB card come handy at max

    or

    You can load uncompressed  95 x 4000x4000 color textures with 6 GB card in iray or 190 at half size .

    so hiding your poly and removing object parts really no matter , what matter is the amount of textures used and its resolution and as you already know compressing jpg in your photo editor will not help you as Iray read data per pixel so the resolution counts

    So having a very low poly model with a lot detailed textures for replace some missing parts will take more space in iray than having high detailed model with less textures and correct shader   

    A lot of poly in your scene will not take much vram but will slow the viewport performance, textures will not slow down the performance of the viewport but will take all vram 

    nicstt said:
    alexhcowley said:
    MEC4D said:

     

    It is of course true, and valid.

    That extra memory though is different and has a different value or set of values applied to it.

    There are work arounds to using a card with half the memory.

    I could buy 3 980ti for not much more than one titan; but if it won't fit in memory, I waste time splitting up parts of the scene, or other work arounds, or I put up with massively slower renders.

    ... So it is: what price is that extra memory is worth?

    After the hastle with the current scene, I'm reevaluating. :)

     

    Post edited by MEC4D on
  • kyoto kidkyoto kid Posts: 41,925
    nicstt said:
    alexhcowley said:
    MEC4D said:

    I saw it this morning , http://arstechnica.com/gadgets/2016/07/gtx-titan-x-pascal-specs-price-release-date/

    well I expected it to be better with more VRAM.. 24% faster than 1080 .. meeeeh ! for the new stock Titan X 11 ... however manufacturers are going to make something better from that stock card soon but of course at higher power consumption than 250W what is out of my interest.

    I am glad I have what I have for now and waiting for something better in the next 2 years to arrive .. 

    Cath,

    A while back, I worked out the cost per cuda core of the Maxwell cards. Best value for money were the 970 and the 980ti.  Poorest value were the 980 and the Titan X, although the Titan X did have more memory, of course.  The Pasal Titan seems to follow this, roughly twice the price of a 1080 but significantly less than twice the cuda cores.  It looks like my plan to buy a 1080 in a few months time and another one next year could be the way to go.

    Cheers,

    Alex.

    It is of course true, and valid.

    That extra memory though is different and has a different value or set of values applied to it.

    There are work arounds to using a card with half the memory.

    I could buy 3 980ti for not much more than one titan; but if it won't fit in memory, I waste time splitting up parts of the scene, or other work arounds, or I put up with massively slower renders.

    ... So it is: what price is that extra memory is worth?

    After the hastle with the current scene, I'm reevaluating. :)

    ...this is exactly my point.  Once the scene file exceeds VRAM and drops to the CPU/Physical Memory, those 8,448 CUDA cores are absolutely useless.

    Considering my GPU has only 1 GB, I'm already on Physical Memory (total 11 GB after Windows) having Daz with that railway station scene open. So now as the Daz application and scenefile is taking up over a third of my avavilable Physical Memory it leaves even less for rendering. Once Physical Memory is exceeded, the process drops into Virtual Memory which is even slower.   This is what I am having to deal with rendering in Iray.

  • MEC4DMEC4D Posts: 5,249
    edited July 2016

    That is all correct but your 11 GB System RAM is not what will be in your VRAM ,  the process is little different , Example : My system use 4.8 GB when DS is Open  but vram only 999MB 

    and with 1 GB of Vram you can't do anything in iray , rendering an image will take more so not even space for anything else , and openGL ? monitor etc..  no possible 

    and I adore people that render in iray with CPU for their patience , I could never do it my CPU rendering is over since release of Octane and I never want to come back to it even if I had to eat Chinese noodle in a cup for a month . 

    kyoto kid said:
    nicstt said:
    alexhcowley said:
    MEC4D said:

     

    ...this is exactly my point.  Once the scene file exceeds VRAM and drops to the CPU/Physical Memory, those 8,448 CUDA cores are absolutely useless.

    Considering my GPU has only 1 GB, I'm already on Physical Memory (total 11 GB after Windows) having Daz with that railway station scene open. So now as the Daz application and scenefile is taking up over a third of my avavilable Physical Memory it leaves even less for rendering. Once Physical Memory is exceeded, the process drops into Virtual Memory which is even slower.   This is what I am having to deal with rendering in Iray.

     

    Post edited by MEC4D on
  • mtl1mtl1 Posts: 1,508
    MEC4D said:

    To load 17 minions poly into DS scene in open GL it will use around 1GB vram and DS will use 4.4GB of system RAM and will last very long to load.
    17 milion poly = 692 Genesis on sub 1 level without textures or 17 Stonemason Fern Lakes

    you will have trouble even smooth operate Studio after you load it as DS is not build for that
    now you switch to iray , it will use 4.3GB of VRAM from your card and over 5GB of Vram total to render it

    so that is where 6 GB card come handy at max

    or

    You can load uncompressed  95 x 4000x4000 color textures with 6 GB card in iray or 190 at half size .

    so hiding your poly and removing object parts really no matter , what matter is the amount of textures used and its resolution and as you already know compressing jpg in your photo editor will not help you as Iray read data per pixel so the resolution counts

    So having a very low poly model with a lot detailed textures for replace some missing parts will take more space in iray than having high detailed model with less textures and correct shader   

    A lot of poly in your scene will not take much vram but will slow the viewport performance, textures will not slow down the performance of the viewport but will take all vram 

    nicstt said:
    alexhcowley said:
    MEC4D said:

     

    It is of course true, and valid.

    That extra memory though is different and has a different value or set of values applied to it.

    There are work arounds to using a card with half the memory.

    I could buy 3 980ti for not much more than one titan; but if it won't fit in memory, I waste time splitting up parts of the scene, or other work arounds, or I put up with massively slower renders.

    ... So it is: what price is that extra memory is worth?

    After the hastle with the current scene, I'm reevaluating. :)

     

    Yes, a lot of it has to do with optimization and well-built models. Large and small textures should be used when appropriate, as well as high and low poly items...

  • kyoto kidkyoto kid Posts: 41,925
    MEC4D said:

    That is all correct but your 11 GB System RAM is not what will be in your VRAM ,  the process is little different , Example : My system use 4.8 GB when DS is Open  but vram only 999MB 

    and with 1 GB of Vram you can't do anything in iray , rendering an image will take more so not even space for anything else , and openGL ? monitor etc..  no possible 

    and I adore people that render in iray with CPU for their patience , I could never do it my CPU rendering is over since release of Octane and I never want to come back to it even if I had to eat Chinese noodle in a cup for a month . 

    kyoto kid said:
    nicstt said:
    alexhcowley said:
    MEC4D said:

     

    ...this is exactly my point.  Once the scene file exceeds VRAM and drops to the CPU/Physical Memory, those 8,448 CUDA cores are absolutely useless.

    Considering my GPU has only 1 GB, I'm already on Physical Memory (total 11 GB after Windows) having Daz with that railway station scene open. So now as the Daz application and scenefile is taking up over a third of my avavilable Physical Memory it leaves even less for rendering. Once Physical Memory is exceeded, the process drops into Virtual Memory which is even slower.   This is what I am having to deal with rendering in Iray.

     

    ...I've been losing my patience with Iray CPU rendering.  All I can really do are exterior scenes as interior ones take almost as long as Reality/Lux to render "clean". I have yet to even bother with increasing the render quality. Octane would probably be the best solution for my current system but the 600$ price tag + the cost of even a 6 GB 780 TI GPU is way too much for my budget.

    Sat down and estimated to cost for a system that would be the most optimal for Iray, and I'm looking at over 6,000$ (based on dual Titan-Ps with hydro cooling which I estimate will probably run about 1,400$ each  1,400$ is about what it cost me  to build my current system sans display).

    I don't need a render to finish in 30 - 90 seconds, I just don't want it tying up my system for a day or more.

    I don't have a 50,000$ a year job that I can afford the exorbitant rents and living costs where I live while keeping pace with the tech curve at the same time. This was supposed to be fun, a means of relaxation, but I find myself getting more stressed about it every time something changes and I cannot afford to adapt.

  • MEC4DMEC4D Posts: 5,249

    I think you can get less than 5K  for super system if you choice right .. build it slow as I did in 12 months .. started with $2000 rig and build up slowly 

    Octane use by default the same mode as Interactive in iray  , it is Direct Light with Ambient Occlusion that why it render faster ?..no it does not , it render faster with multiple GPUs thanks for great GPU scaling ..

    If you use Indirect light in Octane it is not faster than Photoreal in iray .. and that is common mistake that make people thinking it is faster ..they use the wrong rendering mode

    kyoto kid said:
    MEC4D said:

    That is all correct but your 11 GB System RAM is not what will be in your VRAM ,  the process is little different , Example : My system use 4.8 GB when DS is Open  but vram only 999MB 

    and with 1 GB of Vram you can't do anything in iray , rendering an image will take more so not even space for anything else , and openGL ? monitor etc..  no possible 

    and I adore people that render in iray with CPU for their patience , I could never do it my CPU rendering is over since release of Octane and I never want to come back to it even if I had to eat Chinese noodle in a cup for a month . 

    kyoto kid said:
    nicstt said:
    alexhcowley said:
    MEC4D said:

     

    ...this is exactly my point.  Once the scene file exceeds VRAM and drops to the CPU/Physical Memory, those 8,448 CUDA cores are absolutely useless.

    Considering my GPU has only 1 GB, I'm already on Physical Memory (total 11 GB after Windows) having Daz with that railway station scene open. So now as the Daz application and scenefile is taking up over a third of my avavilable Physical Memory it leaves even less for rendering. Once Physical Memory is exceeded, the process drops into Virtual Memory which is even slower.   This is what I am having to deal with rendering in Iray.

     

    ...I've been losing my patience with Iray CPU rendering.  All I can really do are exterior scenes as interior ones take almost as long as Reality/Lux to render "clean". I have yet to even bother with increasing the render quality. Octane would probably be the best solution for my current system but the 600$ price tag + the cost of even a 6 GB 780 TI GPU is way too much for my budget.

    Sat down and estimated to cost for a system that would be the most optimal for Iray, and I'm looking at over 6,000$ (based on dual Titan-Ps with hydro cooling which I estimate will probably run about 1,400$ each  1,400$ is about what it cost me  to build my current system sans display).

    I don't need a render to finish in 30 - 90 seconds, I just don't want it tying up my system for a day or more.

    I don't have a 50,000$ a year job that I can afford the exorbitant rents and living costs where I live while keeping pace with the tech curve at the same time. This was supposed to be fun, a means of relaxation, but I find myself getting more stressed about it every time something changes and I cannot afford to adapt.

     

  • kyoto kidkyoto kid Posts: 41,925
    edited July 2016

    ....2,000$ would barely get me the MB I need, 6 core CPU, dual SSDs, and memory (you want to get your memory in a full kit from the same die batch for stability and reliability).  Tack on another 1,200$ for a single Titan P (air cooled)  as the Titan-X is now obsolete and can't be found "new" (not going to put used components in the system).  Replacing components to "expand" or update the system is not cost efficient particularly when on a tight budget so I need the full base core of the system as designed.

    Post edited by kyoto kid on
  • ArtiniArtini Posts: 10,340
    edited July 2016

    Since now the prices of new Titan X cards is dropping down, is it better to buy a Titan X card (only 30 % more expensive than 1080gtx)

    or wait for the iray drivers for 1080gtx and buy it instead. Titan X has 12 GB GDDR5 vram. What do you think?

    Probably I will wait to the release of new Titan P, to see how it will affect the prices of Titan X.

    Even better deal is on 980ti - on can buy a new one for the 50% of the price of Titan X card, but it only has 6 GB GDDR5 vram.

    Since now I have only 2 GB of GDDR5 vram on my Nvidia graphic card, any of the above cards will make a big difference.

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

    The old Titan X cards are already out of production and not sold anymore by the manufacturers for that reason the higher prices as it is hard to get and will be even harder soon .

    I think for you will be the best choice to wait and see how the new cards perform in iray and make the right decision .. and forget about the 9 series already 

    I paid for my 2 Titan X last year $1229 for each card when they was released as I need faster rig to do my job .. if I had nothing now I would wait for the new cards and see the performance and based on that decide what I want and what is the best choice for the money .. 

    Artini said:

    Since now the prices of new Titan X cards is dropping down, is it better to buy a Titan X card (only 30 % more expensive than 1080gtx)

    or wait for the iray drivers for 1080gtx and buy it instead. Titan X has 12 GB GDDR5 vram. What do you think?

    Probably I will wait to the release of new Titan P, to see how it will affect the prices of Titan X.

    Even better deal is on 980ti - on can buy a new one for the 50% of the price of Titan X card, but it only has 6 GB GDDR5 vram.

    Since now I have only 2 GB of GDDR5 vram on my Nvidia graphic card, any of the above cards will make a big difference.

     

  • nicsttnicstt Posts: 11,715

    I wouldn't buy the old titan at the price, it's way too much.

    still about £1000 over here, and a little more. Presuming (I hate doing that) that there are significant gains for the new titan, as the - admittedly - meagre evidence suggests, then there is going to be a bigger performance boost than the price increase suggests. Even then, waiting for the titan with HBM2 memory might be better still. The price though, oh the price. :(

  • kyoto kidkyoto kid Posts: 41,925
    edited July 2016

    ..at 1,200$ US, a GDDR5 Titan P air cooled is too expensive for what it offers. I would imagine another 200$ for the EVGA Hydro when that comes out would not be out of the question.

    Used Titan X's might be cheaper, but then...it's used.

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

    old Titan X Hybrid was around $1500 , it is not just the $99 cooling but also different bios to make it much faster at the base clock . Anyway still not worth the price itiswaste of money

    as you r

    kyoto kid said:

    ..at 1,200$ US, a GDDR5 Titan P air cooled is too expensive for what it offers. I would imagine another 200$ for the EVGA Hydro when that comes out would not be out of the question.

    Used Titan X's might be cheaper, but then...it's used.

     

  • kyoto kidkyoto kid Posts: 41,925

    ....true, but as I mentioned, any of the new Pascal series GPUs wil require I buld an entrirely new system as my current hardware is out of date and therefore incompatible.

    So what did the EVGA Titan-X Hydros in your system cost?

  • ArtiniArtini Posts: 10,340

    Well, I will wait then. In the meantime I have tested gtx1080 on my son's computer in Blender.

    When trying to render in Cycles on GPU, it gives only the "CUDA kernel" error.

    Right now this graphic card works only in Unity and then it gives more than twice as much of performance boost

    over his previous Nvidia card 970.

  • MEC4DMEC4D Posts: 5,249

    You mean Titan X Hybrid  ( Hydro is with only water block for central water cooling )  last year when I purchased my first 2 cards, Titan X was around $1000 , Titan X SC $1229 and Titan X Hybrid $1500

    My first was Titan X SC $1229 each , then I upgraded it with Hybrid kit $99  so each card cost me $1328 each still cheaper as original Hybrid and run at the same speed . This year I got additional 2 original Titan X Hybrids 1 for $1099 and the other for $1000  so I spent $4755 for my 4 cards 

    The new Titan X P base price is $1200 so it is not going to cost less for a long while , now you add Hydro kit for $99 ending in the same price range as me last year with Titan X SC + Hydro kit but whwn EVGA change the bios , set it at 1800Mhz or more and add more power it will cost no less than $1600-$1700 so in this case everyone will be happy to render with 2 x 1080 at the price of $1200 , 4 core CPU at max with 32GB of RAM .. best perfomance for the money 

    you need upgrade your PC 

    $1200 for 2 x 1080

    $400 for CPU

    $250 for Motherboard 

    $250 for RAM

    1000W PSU or little less if not gaming 

    so you have super fast rendering rig for less than $2500 if build yourself and render in couple of  minutes and not hours

    I am not even using my 4 cards on a daily basic but max 2 as the viewport works wonderful and it render pretty quick + I can do all other stuff at the same time without even notice something  is rendering 

    However you want the power of CPU rendering as well so that is different scenario and different planing egarding you CPU and RAM so the costs will be much higher

     

    kyoto kid said:

    ....true, but as I mentioned, any of the new Pascal series GPUs wil require I buld an entrirely new system as my current hardware is out of date and therefore incompatible.

    So what did the EVGA Titan-X Hydros in your system cost?

    o

     

  • MEC4DMEC4D Posts: 5,249

    The application need to support the new architecture as well for full potential  , DAZ could make Pascal run using currently codas in DS but the performance would be very drag so make no sense to torturing your new cards and waste time , Nvidia said they need to make everything else run perfect before they go into iray what is on the end of this 3 years long process and most important parts are still in Beta so you need to wait a little longer  I guess .

    Artini said:

    Well, I will wait then. In the meantime I have tested gtx1080 on my son's computer in Blender.

    When trying to render in Cycles on GPU, it gives only the "CUDA kernel" error.

    Right now this graphic card works only in Unity and then it gives more than twice as much of performance boost

    over his previous Nvidia card 970.

     

  • kyoto kidkyoto kid Posts: 41,925
    edited July 2016
    MEC4D said:

    You mean Titan X Hybrid  ( Hydro is with only water block for central water cooling )  last year when I purchased my first 2 cards, Titan X was around $1000 , Titan X SC $1229 and Titan X Hybrid $1500

    My first was Titan X SC $1229 each , then I upgraded it with Hybrid kit $99  so each card cost me $1328 each still cheaper as original Hybrid and run at the same speed . This year I got additional 2 original Titan X Hybrids 1 for $1099 and the other for $1000  so I spent $4755 for my 4 cards 

    The new Titan X P base price is $1200 so it is not going to cost less for a long while , now you add Hydro kit for $99 ending in the same price range as me last year with Titan X SC + Hydro kit but whwn EVGA change the bios , set it at 1800Mhz or more and add more power it will cost no less than $1600-$1700 so in this case everyone will be happy to render with 2 x 1080 at the price of $1200 , 4 core CPU at max with 32GB of RAM .. best perfomance for the money 

    you need upgrade your PC 

    $1200 for 2 x 1080

    $400 for CPU

    $250 for Motherboard 

    $250 for RAM

    1000W PSU or little less if not gaming 

    so you have super fast rendering rig for less than $2500 if build yourself and render in couple of  minutes and not hours

    I am not even using my 4 cards on a daily basic but max 2 as the viewport works wonderful and it render pretty quick + I can do all other stuff at the same time without even notice something  is rendering 

    However you want the power of CPU rendering as well so that is different scenario and different planing egarding you CPU and RAM so the costs will be much higher

     

    kyoto kid said:

    ....true, but as I mentioned, any of the new Pascal series GPUs wil require I buld an entrirely new system as my current hardware is out of date and therefore incompatible.

    So what did the EVGA Titan-X Hydros in your system cost?

    o

     

    ....however, only about a 70% chance that a scene would stay in VRAM.  Oh, and there is still the Pascal/IRay driver issue.

    Not about to go back to duo channel memory either.   One of the reasons I had better render times with 3DL was because of the way tri channel memory works. If I am going to shell out to build a new system, I'd like it to be a reasonable step up from what I now have, just like my current one was over the last.  Again, I am looking to render in higher resolution for gallery quality printing.  For ray trace (3DL & Carrara), that means a more memory and CPU cores. 

    There are still things that 3DL is capable of (and which I already have utilities/plugins for) which Iray does not support.

    Post edited by kyoto kid on
  • hphoenixhphoenix Posts: 1,335
    edited July 2016

    Well, nVidia just announced the new Pascal-based Quadro line at SIGGRAPH.  http://www.anandtech.com/show/10516/nvidia-announces-quadro-pascal-family-quadro-p6000-p5000

    P6000 has 3840 cores and 24GB of GDDR5X VRAM.  P5000 has 2560 cores (same as 1080) and 16GB GDDR5X VRAM.  No clock speeds or pricing yet.....

    Set to be available in October.....

     

     

    (edit:  nVidia did announce the pricing would be similar to last gen, so expect around $5k for a P6000, and around $2k for a P5000.)

     

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

    So focus on the base first what you need CPU, memory etc.. and build it up slowly as your budged allow you or you will never have it.. unless you hit the lotto of course so you can buy everything at once 

    next year new stuff come out as usual and you will calculate again .. I did this some time ago and wasted 3 years lol  waiting for cheaper components not working as before ..as you see they stopped the production and everything vanished at lower price , and everyone that was hoping for lower prices of old Titan X wasted 12 months , it is all moving too fast and on some point you need to make the decision and go for it and no matter what you do you going to lose your money after .. that is how it works , it is bad investment anyway , the moment you open the box the value sliding down already .

    remove the snake from your wallet and start build from the basic ..cheeky

    kyoto kid said:
    MEC4D said:

     

    ....however, only about a 70% chance that a scene would stay in VRAM.  Oh, and there is still the Pascal/IRay driver issue.

    Not about to go back to duo channel memory either.   One of the reasons I had better render times with 3DL was because of the way tri channel memory works. If I am going to shell out to build a new system, I'd like it to be a reasonable step up from what I now have, just like my current one was over the last.  Again, I am looking to render in higher resolution for gallery quality printing.  For ray trace (3DL & Carrara), that means a more memory and CPU cores. 

    There are still things that 3DL is capable of (and which I already have utilities/plugins for) which Iray does not support.

     

  • MEC4DMEC4D Posts: 5,249

    I just said it yesterday, they keept the 3840 cores for Pascal Quadro ...  and I suspect the core clock will be  lower as usual ...  nice one anyway but the price indecision

    hphoenix said:

    Well, nVidia just announced the new Pascal-based Quadro line at SIGGRAPH.  http://www.anandtech.com/show/10516/nvidia-announces-quadro-pascal-family-quadro-p6000-p5000

    P6000 has 3840 cores and 24GB of GDDR5X VRAM.  P5000 has 2560 cores (same as 1080) and 16GB GDDR5X VRAM.  No clock speeds or pricing yet.....

    Set to be available in October.....

     

     

    (edit:  nVidia did announce the pricing would be similar to last gen, so expect around $5k for a P6000, and around $2k for a P5000.)

     

     

  • kyoto kidkyoto kid Posts: 41,925
    edited July 2016

    ....just looking to plan for expendability without needing to fully replace components.

    Yeah I guess I could use the 1060 or 1080 to drive the displays when I was able to afford a Titan-P WC (or Titan-HBM when that comes out).  Still need the x 99 MB as part of the core to get the memory and PCI slots needed for expansion.  Again it is recommended to purchase a full memory kit from the same silicon batch than to mix & match so for expanding memory, it would mean effectively "tossing" say the 32 GB for a 64 or 128 gb kit when I can afford it.  The CPU choice also needs to be part of the core. as that is an expensive replacement.  PSUs on the other hand are not that terribly expensive so no reason to scrimp there. PCI SSDs are much faster than SATA III and tend to be more solid as for reliability and service life.

    Basically I would like to minmise the possibility of having to effectively build yet another new machine just to upgrade when the money comes available, as like I mentioned, depreciation on computer parts is hell.

     

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