Loading...
> >

General GPU/testing discussion from benchmark thread

1…5681011…18

Comments

  • Daz Jack TomalinDaz Jack Tomalin Posts: 13,127
    bluejaunte said:
    ebergerly said:

    So let's see...

    Despite all the hype, the 2080ti gives, effectively, for the last 9 months, zero net gain over the 1080ti (twice as fast, twice the price). And now, they're suddenly introducting yet another line of "Super" cards which might actually be a better deal. And at this point all we have is a promise of some still unknown degree of future improvement in the RTX line sometime this year. 

    Yeah, I think I'll wait another 6 months to see what shakes out. laugh

    Zero net gain? That sounds a little strange, seeing as it's twice as fast (at least for us Irayers). You can now have one card instead of two for the same performance. You could also have 2x 2080 TI instead of 4x 1080 TI. the latter usually would be impossible on most mainboards and tough to cool.

    With a 2080 TI I also got better performance in games and viewports of various software, as these usually only ever use the main graphics card. It's plenty net gain for me. Did I mention I all got it for about 0 bucks because I sold 2x 1080 TI to cover the cost? And all of this is before full RTX support for Iray?

    There was no 2080 TI Super announced, by the way.

    Yea it's kinda where I'm at.. while it's cheaper to get another 4 1080ti's (in addition to the 4 already),density becomes an issue.. so while more expensive, having 4 2080ti's at least means you can have it in a fairly standard case... and not some GPU mining case where it's all open etc.

    And yes, I know you can get 8 GPU 4U servers but they have their own considerations.

  • ebergerlyebergerly Posts: 3,255
    edited July 2019

    I totally agree. You can go to great lengths to prove a point is wrong by citing "fringe" examples that few users here or anywhere else would care about, but yes, the statement that it's zero gain is not 100% correct for everyone on the planet. Which is the case for most statements about technology. 

    I was referring to stuff that most users here care about, which is Iray rendering on a single or maybe dual cards. Though not two cards that cost over $3,000. 

    Post edited by ebergerly on
  • bluejauntebluejaunte Posts: 1,861
    ebergerly said:

    I totally agree. You can go to great lengths to prove a point is wrong by citing examples that few users here or anywhere else would care about, but yes, the statement that it's zero gain is not 100% correct for everyone on the planet. Which is the case for most statements about technology. 

    I was referring to stuff that most users here care about, which is Iray rendering on a single or maybe dual cards. Though not two cards that cost over $3,000. 

    It's all a matter of perspecitve. I'd understand your argument better if we were not talking about Iray. The internet at large is certainly not too excited about RTX, simply because it costs too much. And I agree it does cost too much. But we here are in the lucky Iray circle where we get twice the performance before RTX is implemented, and if you look at the specs this is pretty unexpected. When do we ever get 2x more performance for anything? Talking about a fiasco here in our little niche is just perplexing Not to mention, everytime you do use such language, you are implying that everyone who did buy one of these cards is a complete idiot.

  • outrider42outrider42 Posts: 3,679
    edited July 2019
    ebergerly said:

    So let's see...

    Despite all the hype, the 2080ti gives, effectively, for the last 9 months, zero net gain over the 1080ti (twice as fast, twice the price). And now, they're suddenly introducting yet another line of "Super" cards which might actually be a better deal. And at this point all we have is a promise of some still unknown degree of future improvement in the RTX line sometime this year. 

    Yeah, I think I'll wait another 6 months to see what shakes out. laugh

    For someone who constantly asks people to back up their claim with facts, I find your constant mockery of the 2080ti to be lacking. The 2080ti is indeed expensive and probably overpriced, but your constant "twice the price" tagline is not accurate. The 1080ti MSRP is $700 and always has been. The 2080ti MSRP is $1000. These two figures do not line up with your statement. Even the more expensive Founder's Edition price of $1200 is still not twice the price...it is $200 short. The 1080ti almost never sold for less than its MSRP, as the crypto boom shot prices sky high. So much of the 1080ti's lifespan saw it well above MSRP, often in the $1000 range.

    Going beyond MSRP for real world pricing, if you attempt to find a 1080ti right now, you will find them listed for over $1000 almost everywhere. Even refurbs!

    https://www.newegg.com/p/pl?d=gtx+1080+ti

    You will find 2080tis cheaper than these.

    https://www.newegg.com/p/pl?d=rtx+2080+ti

    Obviously there are other retailers, but the 1080ti is no longer produced, so will have a hard time finding any.

    You could attempt to argue the secondary market, where used 1080tis may be bought for roughly $500 or so, but arguing used prices is never fair, and a lot of people refuse to buy used (I know this, I've tried to talk people into buying used GPUs). I purchased a used 970 for $150 during a time when the 1070 was $380 new, the previous generation card was quite a bit less than half the price of the new one. Did that make the 1070 a terrible deal? No, it did not.

    So yes, the 2080ti is priced high, and I would love to see that change. I would love to see AMD compete here and force the issue, I have long stated that I believe the 1080ti released as it was because of the threat of AMD Vega. Keep in mind that AMD had just shocked the world with its 1st gen Ryzen months before the 1080ti launched, so I believe that got Nvidia's attention. Nvidia was concerned about what Vega might be when they finalized pricing for the 1080ti. But Vega sort of fell flat, and by the time RTX launched, AMD had nothing to compete anywhere on the horizon as Navi was still way off. So without any competition Nvidia priced their RTX line accordingly.

    But twice the price as the 1080ti? No. It never has been.

    Post edited by outrider42 on
  • bluejauntebluejaunte Posts: 1,861

    Speaking of used, I never followed up what happened with my defective 2080 TI. Zotac eventually found is was indeed kaputt and offered to replace it. The store was kind enough to refund instead as I had already bought a replacement from them, which works fine so far.

  • outrider42outrider42 Posts: 3,679

    Zotac has to diagnose it before offering to replace it?

    One of my 1080tis actually died a while back, it was not a defect of the card, but rather a lightning surge. My lights flashed and everything at that moment, so it is quite clear it was a surge. My main 1080ti completely died.

    As you may know, I bought my 1080ti used on eBay from a regular person. It so happened that my dead card was EVGA. One of the reasons I bought it used was that I knew EVGA still honored warranties on 2nd hand purchases. The way it works is very simple, the go by the original shipping date and the warranty is good for 3 years, no matter who owns the card at that time. I could give to somebody else and it would still be under warranty until 2020.

    Anyway, I registered my 1080ti and contacted EVGA. I spoke to a fellow who did not hesitate to start the RMA process. The dude was very clear and helpful. All I had to do was pay for shipping to California where EVGA is located. I shipped it off, and the very same day that EVGA received my dead 1080ti they shipped out a replacement. They only check to see that the card is legit and not physically damaged. They emailed me the status along the way, so I knew when the card was received and the replacement shipped.

    I got my replacement a few days later. It looks brand new, and actually performs slightly cooler than the one I bought. The original EVGA would hit 74 during an Iray render. My replacement only hits 66!

    All in all, it was about 2 weeks for this process. And all of that time was from shipping.

    Needless to say I am pleased with EVGA, and I am not even a paying customer! They never got a dime from me.

    I have always heard that EVGA had great customer service, but now I have first hand experience in just how good their service really is.

    Next time I plan on buying EVGA. A new one, LOL. Plus EVGA offers 5 and 10 year extended warranties on new purchases, which can be from any retailer. These extended warranties are crazy cheap. Like for a 1080ti or 2080ti, the 10 year warranty is only $60. And EVGA will honor that warranty, replacing the card as many times as needed for the entire 10 years.

    For $60!!!

    I seriously doubt I would even use the same GPU for 10 years. Which on that note, the extended warranty also can be transferred, which I bet could greatly increase the resell value of GPU if you sell it.

  • RayDAntRayDAnt Posts: 1,120

    Btw Iray RTX has officially made its way into the Daz Studio private build channel:

    • Source maintenance

    • Extended DzCreateNewItemDlg SDK API; added setNewItemName(), getNewItemName(); added parameter to addOption()

    • Extended DzObject public API; added isBuildingGeom(), isBuildingGeomValid()

    • Updated SDK version to 4.12.0.10; SDK min is 4.5.0.100

    • Updated SDK API documentation; DzCreateNewItemDlg

    • Update to NVIDIA Iray RTX 2019.1.1 (317500.2554)

    • Fixed a timing/update issue in DzMeshSmoother

    DAZ Studio : Incremented build number to 4.12.0.10

  • ArtRouladeArtRoulade Posts: 20
    edited July 2019

    Here a little guide if anybody wants to test Iray RTX 2019.1.1 in Iray Server 2.53 with a trial license.

    - first you need to download Iray Server 2.53 from irayplugins.com. - free 30 days trial and install everything

    - start Iray Server, open your webbrowser and enter the ip adress/port of the Iray Server web gui (eg. default - 127.0.0.1:9090). The adress may vary. Look at the command-line of Iray Server to get the correct one.

    - next step: login to Iray Server web gui: default name: admin default pw: admin - change your password

    - under tab Resources you find a benchmark scene which you can add to your render queue and also select a different Iray Version for benchmarking

    - the current bridge protocol in Daz Studio doesn't support Iray 2019.1.1 and thus any exported scene from Daz Studio to Iray Server is limited by the supported Iray version of the current bridge protocol.

    Now the solution to export and import a custom scene to test Iray RTX.

    - load a scene in Daz Studio which you want to test, set your desired render options  and export it as .mi (mental images scene format) and save it somewhere with a short name like eg. test1.

    - now go the folder: eg. C:\Program Files\NVIDIA Corpotration\Iray Server\benchmarks and make a copy of the folder 'cornell_box' within the benchmark folder. You can rename the new folder. In my example i don't change it - 'cornell_box - copy'

    - go into the copied folder and delete everything except main.mi and the folder called 'shaders'. in folder shaders you can delete everything.

    - now go to the following path and copy three folders called daz_3d, nvidia and resources into the shader folder of 'cornell_box - copy'. Path: eg. C:\Program Files\DAZ 3D\DAZStudio4\shaders\iray - here you find the three mentioned folders.

    - next step: open the main.mi file with your text editor and delete everything and insert the following line of code '$include "test1.mi"' and save/overwrite it. It references to your saved scene .mi files : 

    - next step: copy your saved .mi files into the folder 'cornell_box - copy'

    - the folder 'cornell_box - copy' should now consists of: folder 'shaders' with three subfolders 'daz_3d', 'nvidia' and 'resources' and the three files: 'main.mi', 'test1.mi' and 'test1.shader_decl.mi'

    - now go back to your webbroswer and go to the mentioned 'Resources' tab. Under benchmarks you should find your custom scene for benchmarking. add it to the queue and choose an Iray Version for testing. eg. 2018.1.3 (is one of the version which is used by Daz Studio - non RTX) and 2019.1.1 (RTX).

    - after adding the render jobs to the queue you have to start the queue in the identically named tab. Have fun with your benchmarks :)

    If you want to test vram pooling you need two Geforce RTX cards with nvlink ports, a nvlink brdige and you have to activate SLI via NVIDIA system options. Check your command line/logs of Iray Server if your two cards are correctly recognized and pooled as one render group.

    My guide is based on https://www.daz3d.com/forums/discussion/132951/tutorial-iray-server-render-farm-batch-rendering-for-daz-studio/p1

    Post edited by ArtRoulade on
  • bluejauntebluejaunte Posts: 1,861

    Thanks, though I'd rather know how much faster stuff is. Have you done no testing yet? laugh

  • nicsttnicstt Posts: 11,714
    outrider42 said:

    Even Daz themselves state outright that the Daz Studio beta SHOULD NOT be used in mission critical machines. You linked it yourself. End of debate. You can guys can argue all you want, but there are people who are effected by this. And once again, you guys assume that regular users are tech oriented enough to understand all of this. That regular users even KNOW there is a beta, as there is no notification of a beta outside the walls of this forum, and even then you will only find talk of the beta in specific places. There are a lot of people who barely know much more than how to power on their computer here. This is a forum of artists and people use Daz because it is generally more user friendly than other similar software, and you can buy a preset for just about anything.  We also have a lot of Apple users here! (This is a joke...maybe.)

    The irony about this is that I have seen far more people complain about the 4.11 full release than anybody ever complained about the betas...what is up with that? I would make a suggestion of my thoughts as to why that is, but I would get into trouble. But I can say this, that so many people are having issues points to how many people still used 4.10 over the beta, since the full release has many of the features that the beta has had for some time. So you can keep arguing about the quality of the beta, but we have solid proof that a fair number of people never used it.

    You forgot how many complained about 4.9; I don't have any real issues myself and love the dforce and/or strand based features.

  • ArtRouladeArtRoulade Posts: 20
    edited July 2019
    bluejaunte said:

    Thanks, though I'd rather know how much faster stuff is. Have you done no testing yet? laugh

    My trial license on my pc expired a long time ago before the release of Iray Server 2.53 sad. I could only test it on my old notebook with an old GeForce NVDIA GTX 980M and the speed difference for such an old card is insignificant laugh I would like to test it as i have two 2080 Ti's and an active nvlink bridge frown

    Post edited by ArtRoulade on
  • Daz Jack TomalinDaz Jack Tomalin Posts: 13,127
    edited July 2019
    neumi1337 said:
    bluejaunte said:

    Thanks, though I'd rather know how much faster stuff is. Have you done no testing yet? laugh

    My trial license on my pc expired a long time ago before the release of Iray Server 2.53 sad. I could only test it on my old notebook with an old GeForce NVDIA GTX 980M and the speed difference for such an old card is insignificant laugh I would like to test it as i have two 2080 Ti's and an active nvlink bridge frown

     

    Comparing the standard cornell_box benchmark scene, 2019.1.1 rendered it in 26 seconds and 2018.1.3 in 28 seconds (on 3 2080ti's). 

    Upping the resolution to 4k, comes in at 1m 41 and 1m 50 respectively.  Not sure to what extent (if any) RTX is being leveraged in the new Iray 2019.1.1.. but times seem a tad quicker.  I guess you'd need something longer to really see if it stretches its legs.

     

    Post edited by Daz Jack Tomalin on
  • RayDAntRayDAnt Posts: 1,120
    neumi1337 said:

    Here a little guide if anybody wants to test Iray RTX 2019.1.1 in Iray Server 2.53 with a trial license.

    - first you need to download Iray Server 2.53 from irayplugins.com. - free 30 days trial and install everything

    - start Iray Server, open your webbrowser and enter the ip adress/port of the Iray Server web gui (eg. default - 127.0.0.1:9090). The adress may vary. Look at the command-line of Iray Server to get the correct one.

    - next step: login to Iray Server web gui: default name: admin default pw: admin - change your password

    - under tab Resources you find a benchmark scene which you can add to your render queue and also select a different Iray Version for benchmarking

    - the current bridge protocol in Daz Studio doesn't support Iray 2019.1.1 and thus any exported scene from Daz Studio to Iray Server is limited by the supported Iray version of the current bridge protocol.

    Now the solution to export and import a custom scene to test Iray RTX.

    - load a scene in Daz Studio which you want to test, set your desired render options  and export it as .mi (mental images scene format) and save it somewhere with a short name like eg. test1.

    - now go the folder: eg. C:\Program Files\NVIDIA Corpotration\Iray Server\benchmarks and make a copy of the folder 'cornell_box' within the benchmark folder. You can rename the new folder. In my example i don't change it - 'cornell_box - copy'

    - go into the copied folder and delete everything except main.mi and the folder called 'shaders'. in folder shaders you can delete everything.

    - now go to the following path and copy three folders called daz_3d, nvidia and resources into the shader folder of 'cornell_box - copy'. Path: eg. C:\Program Files\DAZ 3D\DAZStudio4\shaders\iray - here you find the three mentioned folders.

    - next step: open the main.mi file with your text editor and delete everything and insert the following line of code '$include "test1.mi"' and save/overwrite it. It references to your saved scene .mi files : 

    - next step: copy your saved .mi files into the folder 'cornell_box - copy'

    - the folder 'cornell_box - copy' should now consists of: folder 'shaders' with three subfolders 'daz_3d', 'nvidia' and 'resources' and the three files: 'main.mi', 'test1.mi' and 'test1.shader_decl.mi'

    - now go back to your webbroswer and go to the mentioned 'Resources' tab. Under benchmarks you should find your custom scene for benchmarking. add it to the queue and choose an Iray Version for testing. eg. 2018.1.3 (is one of the version which is used by Daz Studio - non RTX) and 2019.1.1 (RTX).

    - after adding the render jobs to the queue you have to start the queue in the identically named tab. Have fun with your benchmarks :)

    If you want to test vram pooling you need two Geforce RTX cards with nvlink ports, a nvlink brdige and you have to activate SLI via NVIDIA system options. Check your command line/logs of Iray Server if your two cards are correctly recognized and pooled as one render group.

    My guide is based on https://www.daz3d.com/forums/discussion/132951/tutorial-iray-server-render-farm-batch-rendering-for-daz-studio/p1

    Will try this with the publicly available benchmarking scenes on my single Titan RTX rig soonest chance I get (tomorrow probably.) Still have a couple weeks left in my Iray Server trial...

  • ArtRouladeArtRoulade Posts: 20
    edited July 2019
    Jack Tomalin said:
    neumi1337 said:
    bluejaunte said:

    Thanks, though I'd rather know how much faster stuff is. Have you done no testing yet? laugh

    My trial license on my pc expired a long time ago before the release of Iray Server 2.53 sad. I could only test it on my old notebook with an old GeForce NVDIA GTX 980M and the speed difference for such an old card is insignificant laugh I would like to test it as i have two 2080 Ti's and an active nvlink bridge frown

     

    Comparing the standard cornell_box benchmark scene, 2019.1.1 rendered it in 26 seconds and 2018.1.3 in 28 seconds (on 3 2080ti's). 

    Upping the resolution to 4k, comes in at 1m 41 and 1m 50 respectively.  Not sure to what extent (if any) RTX is being leveraged in the new Iray 2019.1.1.. but times seem a tad quicker.  I guess you'd need something longer to really see if it stretches its legs.

     

    To exclude some things. Which gpu driver version are you using and pls check in your logs which optix version is used. Also deactivate the usage of your cpu in the resources tab for the rendering process. you can open the logs as a text file and search for it. 

    Post edited by ArtRoulade on
  • Saxa -- SDSaxa -- SD Posts: 871
    RayDAnt said:

    Btw Iray RTX has officially made its way into the Daz Studio private build channel:

    • Update to NVIDIA Iray RTX 2019.1.1 (317500.2554)

     

    Jack Tomalin said:
     Not sure to what extent (if any) RTX is being leveraged in the new Iray 2019.1.1..

    Oh that is too tragically funny.  Update is labelled RTX 2019 and maybe the RTX isn't even being utilized to any extent in an Iray update in 2019 for a product released by Nvidia in 2018 with so much exclamation?   Goes to show never Ass/u/me just cos changelog shows NVIDIA Iray RTX 2019.1.1 .  Guess it's best to saran-wrap own hopes and put them in the freezer for now.  Dreams of 5-10-20 minute renders taking  2-7 minutes are on ice for next beta, until we see a Beta version that has it, or Daz gives us another rumor to enjoy.  Thanks to @JackTomalin for helping to keeping it real.  (Will just have to take longer to render your fab-environment sets for now.)

  • Daz Jack TomalinDaz Jack Tomalin Posts: 13,127
    edited July 2019
    neumi1337 said:
    Jack Tomalin said:
    neumi1337 said:
    bluejaunte said:

    Thanks, though I'd rather know how much faster stuff is. Have you done no testing yet? laugh

    My trial license on my pc expired a long time ago before the release of Iray Server 2.53 sad. I could only test it on my old notebook with an old GeForce NVDIA GTX 980M and the speed difference for such an old card is insignificant laugh I would like to test it as i have two 2080 Ti's and an active nvlink bridge frown

     

    Comparing the standard cornell_box benchmark scene, 2019.1.1 rendered it in 26 seconds and 2018.1.3 in 28 seconds (on 3 2080ti's). 

    Upping the resolution to 4k, comes in at 1m 41 and 1m 50 respectively.  Not sure to what extent (if any) RTX is being leveraged in the new Iray 2019.1.1.. but times seem a tad quicker.  I guess you'd need something longer to really see if it stretches its legs.

     

    To exclude some things. Which gpu driver version are you using and pls check in your logs which optix version is used. Also deactivate the usage of your cpu in the resources tab for the rendering process. you can open the logs as a text file and search for it. 

    CPU is disabled, yes.

    Driver version is 430.86

    From the logs, it doesnt say which version is being used, except for "Initializing OptiX for CUDA device 0".  On the 2018 benchmark, it says "Using OptiX Prime ray tracing (5.0.1)."

    Post edited by Daz Jack Tomalin on
  • Daz Jack TomalinDaz Jack Tomalin Posts: 13,127
    Saxa -- SD said:
    RayDAnt said:

    Btw Iray RTX has officially made its way into the Daz Studio private build channel:

    • Update to NVIDIA Iray RTX 2019.1.1 (317500.2554)

     

    Jack Tomalin said:
     Not sure to what extent (if any) RTX is being leveraged in the new Iray 2019.1.1..

    Oh that is too tragically funny.  Update is labelled RTX 2019 and maybe the RTX isn't even being utilized to any extent in an Iray update in 2019 for a product released by Nvidia in 2018 with so much exclamation?   Goes to show never Ass/u/me just cos changelog shows NVIDIA Iray RTX 2019.1.1 .  Guess it's best to saran-wrap own hopes and put them in the freezer for now.  Dreams of 5-10-20 minute renders taking  2-7 minutes are on ice for next beta, until we see a Beta version that has it, or Daz gives us another rumor to enjoy.  Thanks to @JackTomalin for helping to keeping it real.  (Will just have to take longer to render your fab-environment sets for now.)

    I'm not really speculating either way - I'm just reporting what I see on these two benchmarks.. they might not even be representative of anything.  I'm no privvy to any development info, either for what its worth.

    Personally, the addition of RTX wasn't the whole thing for me.. I just wanted faster cards.. and the 2080TI's are just that.   The speed increase between Iray 2018 and 2019 is welcome though, so will be curious to see how that plays out with future revisions and updates.

  • ArtRouladeArtRoulade Posts: 20
    edited July 2019

    Very odd. Could you check your current driver version of your GPUs? Quote: OptiX 6.0 requires a display driver version 418 or higher.  Correct: To get access to the RTX features you need to be running driver version 430.86 or later on Windows

    Post edited by ArtRoulade on
  • Daz Jack TomalinDaz Jack Tomalin Posts: 13,127
    neumi1337 said:

    Very odd. Could you check your current driver version of your GPUs? Quote: OptiX 6.0 requires a display driver version 418 or higher.

    Edited my post above, I'm on 430.86

  • Daz Jack TomalinDaz Jack Tomalin Posts: 13,127

    Ah i missed it.. yea, "Using OptiX version 6.1.2"

  • outrider42outrider42 Posts: 3,679

    OptiX

    Saxa -- SD said:
    RayDAnt said:

    Btw Iray RTX has officially made its way into the Daz Studio private build channel:

    • Update to NVIDIA Iray RTX 2019.1.1 (317500.2554)

     

    Jack Tomalin said:
     Not sure to what extent (if any) RTX is being leveraged in the new Iray 2019.1.1..

    Oh that is too tragically funny.  Update is labelled RTX 2019 and maybe the RTX isn't even being utilized to any extent in an Iray update in 2019 for a product released by Nvidia in 2018 with so much exclamation?   Goes to show never Ass/u/me just cos changelog shows NVIDIA Iray RTX 2019.1.1 .  Guess it's best to saran-wrap own hopes and put them in the freezer for now.  Dreams of 5-10-20 minute renders taking  2-7 minutes are on ice for next beta, until we see a Beta version that has it, or Daz gives us another rumor to enjoy.  Thanks to @JackTomalin for helping to keeping it real.  (Will just have to take longer to render your fab-environment sets for now.)

    You seem to be celebrating for some reason. The changelog is not related to this test, because the changelog is for a Studio that has not been released in any form, not general nor beta. This test was a convoluted process to get working, and as it stands there seems to be an error as demonstrated below.

     

    Jack Tomalin said:
    neumi1337 said:
    Jack Tomalin said:
    neumi1337 said:
    bluejaunte said:

    Thanks, though I'd rather know how much faster stuff is. Have you done no testing yet? laugh

    My trial license on my pc expired a long time ago before the release of Iray Server 2.53 sad. I could only test it on my old notebook with an old GeForce NVDIA GTX 980M and the speed difference for such an old card is insignificant laugh I would like to test it as i have two 2080 Ti's and an active nvlink bridge frown

     

    Comparing the standard cornell_box benchmark scene, 2019.1.1 rendered it in 26 seconds and 2018.1.3 in 28 seconds (on 3 2080ti's). 

    Upping the resolution to 4k, comes in at 1m 41 and 1m 50 respectively.  Not sure to what extent (if any) RTX is being leveraged in the new Iray 2019.1.1.. but times seem a tad quicker.  I guess you'd need something longer to really see if it stretches its legs.

     

    To exclude some things. Which gpu driver version are you using and pls check in your logs which optix version is used. Also deactivate the usage of your cpu in the resources tab for the rendering process. you can open the logs as a text file and search for it. 

    CPU is disabled, yes.

    Driver version is 430.86

    From the logs, it doesnt say which version is being used, except for "Initializing OptiX for CUDA device 0".  On the 2018 benchmark, it says "Using OptiX Prime ray tracing (5.0.1)."

    I believe you need OptiX Prme 6.0 for this to work properly. So you are NOT using RT cores right now. The speed difference you see is kind of encouraging, as this at least shows Iray 2019 is still a bit faster even without RT cores. And for those who have been following, migenius stated that Iray 2019 is faster even without RT cores. So this falls perfectly in line with that statement.

  • Daz Jack TomalinDaz Jack Tomalin Posts: 13,127
    outrider42 said:

    OptiX

    Saxa -- SD said:
    RayDAnt said:

    Btw Iray RTX has officially made its way into the Daz Studio private build channel:

     

    Jack Tomalin said:
     Not sure to what extent (if any) RTX is being leveraged in the new Iray 2019.1.1..

    Oh that is too tragically funny.  Update is labelled RTX 2019 and maybe the RTX isn't even being utilized to any extent in an Iray update in 2019 for a product released by Nvidia in 2018 with so much exclamation?   Goes to show never Ass/u/me just cos changelog shows NVIDIA Iray RTX 2019.1.1 .  Guess it's best to saran-wrap own hopes and put them in the freezer for now.  Dreams of 5-10-20 minute renders taking  2-7 minutes are on ice for next beta, until we see a Beta version that has it, or Daz gives us another rumor to enjoy.  Thanks to @JackTomalin for helping to keeping it real.  (Will just have to take longer to render your fab-environment sets for now.)

    You seem to be celebrating for some reason. The changelog is not related to this test, because the changelog is for a Studio that has not been released in any form, not general nor beta. This test was a convoluted process to get working, and as it stands there seems to be an error as demonstrated below.

     

    Jack Tomalin said:
    neumi1337 said:
    Jack Tomalin said:
    neumi1337 said:
    bluejaunte said:

    Thanks, though I'd rather know how much faster stuff is. Have you done no testing yet? laugh

    My trial license on my pc expired a long time ago before the release of Iray Server 2.53 sad. I could only test it on my old notebook with an old GeForce NVDIA GTX 980M and the speed difference for such an old card is insignificant laugh I would like to test it as i have two 2080 Ti's and an active nvlink bridge frown

     

    Comparing the standard cornell_box benchmark scene, 2019.1.1 rendered it in 26 seconds and 2018.1.3 in 28 seconds (on 3 2080ti's). 

    Upping the resolution to 4k, comes in at 1m 41 and 1m 50 respectively.  Not sure to what extent (if any) RTX is being leveraged in the new Iray 2019.1.1.. but times seem a tad quicker.  I guess you'd need something longer to really see if it stretches its legs.

     

    To exclude some things. Which gpu driver version are you using and pls check in your logs which optix version is used. Also deactivate the usage of your cpu in the resources tab for the rendering process. you can open the logs as a text file and search for it. 

    CPU is disabled, yes.

    Driver version is 430.86

    From the logs, it doesnt say which version is being used, except for "Initializing OptiX for CUDA device 0".  On the 2018 benchmark, it says "Using OptiX Prime ray tracing (5.0.1)."

    I believe you need OptiX Prme 6.0 for this to work properly. So you are NOT using RT cores right now. The speed difference you see is kind of encouraging, as this at least shows Iray 2019 is still a bit faster even without RT cores. And for those who have been following, migenius stated that Iray 2019 is faster even without RT cores. So this falls perfectly in line with that statement.

    See above, it was using it.. 

  • ArtRouladeArtRoulade Posts: 20

    Thx lol. Then i would think that the attached benchmark scene 'cornell_box' isn't profiting from the speedup of the RT cores. We need other benchmark scenes or have to wait until RayDAnt gets the scenes he meant.The prove Iray is using Optix 6.1.2 is a big step up.

  • Saxa -- SDSaxa -- SD Posts: 871
    Jack Tomalin said:
    Saxa -- SD said:
    RayDAnt said:

    Btw Iray RTX has officially made its way into the Daz Studio private build channel:

    • Update to NVIDIA Iray RTX 2019.1.1 (317500.2554)

     

    Jack Tomalin said:
     Not sure to what extent (if any) RTX is being leveraged in the new Iray 2019.1.1..

    Oh that is too tragically funny.  Update is labelled RTX 2019 and maybe the RTX isn't even being utilized to any extent in an Iray update in 2019 for a product released by Nvidia in 2018 with so much exclamation?   Goes to show never Ass/u/me just cos changelog shows NVIDIA Iray RTX 2019.1.1 .  Guess it's best to saran-wrap own hopes and put them in the freezer for now.  Dreams of 5-10-20 minute renders taking  2-7 minutes are on ice for next beta, until we see a Beta version that has it, or Daz gives us another rumor to enjoy.  Thanks to @JackTomalin for helping to keeping it real.  (Will just have to take longer to render your fab-environment sets for now.)

    I'm not really speculating either way - I'm just reporting what I see on these two benchmarks.. they might not even be representative of anything.  I'm no privvy to any development info, either for what its worth.

    Personally, the addition of RTX wasn't the whole thing for me.. I just wanted faster cards.. and the 2080TI's are just that.   The speed increase between Iray 2018 and 2019 is welcome though, so will be curious to see how that plays out with future revisions and updates.

    Thanks alot for clarifying.  Thought maybe you were trying to temper our expectations a bit?  Will still keep my hopes on ice, because you make a very good and valid point that just cos update is labelled RTX 2019, doesn't mean it will bring large degrees of RTX support to IRAY in next few beta versions.

    Am very pleased already with how much my 2080Ti speeds up DAZ work (from a GTX1070).  But won't lie that I didn't hope that we would see in the next 12 months larger scale speed increases for complex environments like your DAZ products where bouncing light is a heavy thing.

  • ArtRouladeArtRoulade Posts: 20

    Migenius had updated their benchmark overview for Iray 2019.1.1 ... https://www.migenius.com/products/nvidia-iray/iray-rtx-2019-1-1-benchmarks

  • Saxa -- SDSaxa -- SD Posts: 871
    outrider42 said:

    You seem to be celebrating for some reason. The changelog is not related to this test, because the changelog is for a Studio that has not been released in any form, not general nor beta. This test was a convoluted process to get working, and as it stands there seems to be an error as demonstrated below.

    LOL, "celebrating".  Not sure how my message though not written in dry factual has anything to do with celebrating.

    Short factual version: Jack made a really good point that hit a nerve that I need to manage my own expectations, even though shows Nvidia Iray RTX in Daz Changelog.  Made me realize am starting to assume.

  • Daz Jack TomalinDaz Jack Tomalin Posts: 13,127
    Saxa -- SD said:
    Jack Tomalin said:
    Saxa -- SD said:
    RayDAnt said:

    Btw Iray RTX has officially made its way into the Daz Studio private build channel:

     

    Jack Tomalin said:
     Not sure to what extent (if any) RTX is being leveraged in the new Iray 2019.1.1..

    Oh that is too tragically funny.  Update is labelled RTX 2019 and maybe the RTX isn't even being utilized to any extent in an Iray update in 2019 for a product released by Nvidia in 2018 with so much exclamation?   Goes to show never Ass/u/me just cos changelog shows NVIDIA Iray RTX 2019.1.1 .  Guess it's best to saran-wrap own hopes and put them in the freezer for now.  Dreams of 5-10-20 minute renders taking  2-7 minutes are on ice for next beta, until we see a Beta version that has it, or Daz gives us another rumor to enjoy.  Thanks to @JackTomalin for helping to keeping it real.  (Will just have to take longer to render your fab-environment sets for now.)

    I'm not really speculating either way - I'm just reporting what I see on these two benchmarks.. they might not even be representative of anything.  I'm no privvy to any development info, either for what its worth.

    Personally, the addition of RTX wasn't the whole thing for me.. I just wanted faster cards.. and the 2080TI's are just that.   The speed increase between Iray 2018 and 2019 is welcome though, so will be curious to see how that plays out with future revisions and updates.

    Thanks alot for clarifying.  Thought maybe you were trying to temper our expectations a bit?  Will still keep my hopes on ice, because you make a very good and valid point that just cos update is labelled RTX 2019, doesn't mean it will bring large degrees of RTX support to IRAY in next few beta versions.

    Am very pleased already with how much my 2080Ti speeds up DAZ work (from a GTX1070).  But won't lie that I didn't hope that we would see in the next 12 months larger scale speed increases for complex environments like your DAZ products where bouncing light is a heavy thing.

    I think it’s sensible to take everything with a pinch of salt until you see it with your own eyes.

     

  • Saxa -- SDSaxa -- SD Posts: 871
    Jack Tomalin said:
    Saxa -- SD said:
    Jack Tomalin said:
    Saxa -- SD said:
    RayDAnt said:

    Btw Iray RTX has officially made its way into the Daz Studio private build channel:

     

    Jack Tomalin said:
     Not sure to what extent (if any) RTX is being leveraged in the new Iray 2019.1.1..

    Oh that is too tragically funny.  Update is labelled RTX 2019 and maybe the RTX isn't even being utilized to any extent in an Iray update in 2019 for a product released by Nvidia in 2018 with so much exclamation?   Goes to show never Ass/u/me just cos changelog shows NVIDIA Iray RTX 2019.1.1 .  Guess it's best to saran-wrap own hopes and put them in the freezer for now.  Dreams of 5-10-20 minute renders taking  2-7 minutes are on ice for next beta, until we see a Beta version that has it, or Daz gives us another rumor to enjoy.  Thanks to @JackTomalin for helping to keeping it real.  (Will just have to take longer to render your fab-environment sets for now.)

    I'm not really speculating either way - I'm just reporting what I see on these two benchmarks.. they might not even be representative of anything.  I'm no privvy to any development info, either for what its worth.

    Personally, the addition of RTX wasn't the whole thing for me.. I just wanted faster cards.. and the 2080TI's are just that.   The speed increase between Iray 2018 and 2019 is welcome though, so will be curious to see how that plays out with future revisions and updates.

    Thanks alot for clarifying.  Thought maybe you were trying to temper our expectations a bit?  Will still keep my hopes on ice, because you make a very good and valid point that just cos update is labelled RTX 2019, doesn't mean it will bring large degrees of RTX support to IRAY in next few beta versions.

    Am very pleased already with how much my 2080Ti speeds up DAZ work (from a GTX1070).  But won't lie that I didn't hope that we would see in the next 12 months larger scale speed increases for complex environments like your DAZ products where bouncing light is a heavy thing.

    I think it’s sensible to take everything with a pinch of salt until you see it with your own eyes.

     

    Am at the point with life-experiences where I have pails of salt nearby just in case laugh

    But fooled myself thinking hey RTX for Iray has got to get going now anyday.  And maybe it will?  Or not.  Power of wishful thinking.  Need to go replenish salt now.

  • RayDAntRayDAnt Posts: 1,120
    neumi1337 said:

    Thx lol. Then i would think that the attached benchmark scene 'cornell_box' isn't profiting from the speedup of the RT cores. We need other benchmark scenes or have to wait until RayDAnt gets the scenes he meant.The prove Iray is using Optix 6.1.2 is a big step up.

    Yeah, keep in mind that gains from more powerful raytracing acceleration are going to be HIGHLY scene content dependent. A simplistic cornell box is likely one of the LEAST opportunistic cases for this.

  • drzapdrzap Posts: 795
    edited July 2019
    neumi1337 said:

    Migenius had updated their benchmark overview for Iray 2019.1.1 ... https://www.migenius.com/products/nvidia-iray/iray-rtx-2019-1-1-benchmarks

    So about a 17% increase for a 2080ti, which is a little smaller increase from what other renderer developers have shown with their benchmarks in this first go-around with RTX.  Of course its highly scene dependent but I am assuming everyone will be using scenes that will help embelish their products.  In every sense, it is a far cry from what Nvidia was leading us to expect.  To be fair, gaming stands to benefit the most from this technology at this point and games are receiving a much bigger boost in performance.  Glad I kept my money in my pocket this round.  I may make the jump next generation but I'm all set now with my 4 x 1080ti setup.  I don't use a gpu renderer as my main engine anyways.

    Post edited by drzap on
This discussion has been closed.