Loading...
> >

Daz Studio Iray - Rendering Hardware Benchmarking

1…3940424445

Comments

  • outrider42outrider42 Posts: 3,679
    edited June 2023

    Iray dev team just posted this, and it could be big:

    Iray 2023.0.0 final

    ..was just released.

    Major improvements over the beta:

    Performance, performance, performance. :)

    We mostly got back all the perf loss we introduced in the beta when running the new caustic sampler 2.0 (compared to cs 1.0 in 2022.1.X), plus we optimized general performance on the Ada Lovelace GPU architecture, and Iray is now faster when using rather complicated MDL materials (i.e. MDL JIT path is triggered).
    Then a lot of performance optimizations all over the place, e.g. when using fibers/hair, rounded corners or cutouts.

    In addition there is also one additional new major feature over the beta: Support for the new Mie phase function for volumes, introduced along with MDL 1.8.X.

    On top we have dozens of important bug fixes, as usual.

    The new release will also be available soon within Omniverse, of course, via the RTX Accurate rendering mode.

    That's a lot of hype for this new version of Iray. I hope it delivers. 4090 owners rejoice. BTW, when Daz Studio gets this, if you do run the benchmark, I would be very interested in knowing your power draw before and after the update. I am interested in seeing if the new performance for Lovelace changes power draw any. I'll be double checking my own power draw as well. Lovelace is not the only improvement promised here.

    It will also be interesting to see what these performance improvements mean. Do we finally get our DS 4.15 speed back? Maybe even more? Better performance in more complicated scenes is huge as well, but our bench is not that complicated.

    Anybody using Omniverse? BTW, we may be getting USD format support soon, which would be a quick pathway to Omniverse and all sorts of other apps.

    Post edited by outrider42 on
  • mwasielewski1990mwasielewski1990 Posts: 343
    edited June 2023

    Sup guys, I'm about to swap my 3090 for a 4090, so running the bench before I swap them out :)

    System Configuration
    System/Motherboard: MSI MAG B760 Tomahawk WIFI DDR4
    CPU: i5 12600K @STOCK
    GPU: Zotac Gaming Trinity RTX 3090 (undervolted, @900mv, clock speed ~1845mHz)
    System Memory: 4x 16GB = 64GB Lexar Thor DDR4 @3200mHz (XMP profile)
    OS Drive: Samsung 970 Evo PLUS 2TB M.2
    Asset Drive: same
    Power Supply: Corsair RM1000e (1000W)
    Operating System: Windows 10 Pro 10.0.19045 N/A Build 19045
    Nvidia Drivers Version: 535.98 STUDIO
    Daz Studio Version: 4.21.0.5 Official release, x64
    Optix Prime Acceleration: n/a

    Benchmark Results
    2023-06-29 08:13:21.199 [INFO] :: Total Rendering Time: 2 minutes 7.21 seconds
    2023-06-29 08:14:30.722 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (NVIDIA GeForce RTX 3090): 1800 iterations, 1.368s init, 124.187s render
    Iteration Rate: 14.49 iterations per second
    Loading Time: 3.023 seconds

    Post edited by mwasielewski1990 on
  • mwasielewski1990mwasielewski1990 Posts: 343
    edited July 2023

    Okay, here go again, but with the 4090!

    I just noticed a new studio driver came out, so I'm gonna benchmark both versions.

    POST-RENDER EDIT: Interestingly, the card only draws a maximum of 311 Watts. Any idea why? Why is the full 450W not utilized? My 3090 without undervolting topped at it's specced 350W.

    System Configuration
    System/Motherboard: MSI MAG B760 Tomahawk WIFI DDR4
    CPU: i5 12600K @STOCK
    GPU: Inno3D RTX4090 X3 OC (@STOCK = 1050mv)
    System Memory: 4x 16GB = 64GB Lexar Thor DDR4 @3200mHz (XMP profile)
    OS Drive: Samsung 970 Evo PLUS 2TB M.2
    Asset Drive: same
    Power Supply: Corsair RM1000e (1000W)
    Operating System: Windows 10 Pro 10.0.19045 N/A Build 19045
    Nvidia Drivers Version: 535.98 STUDIO
    Daz Studio Version: 4.21.0.5 Official release, x64
    Optix Prime Acceleration: n/a

    Benchmark Results
    2023-06-29 14:40:16.162 [INFO] :: Total Rendering Time: 1 minutes 27.87 seconds
    2023-06-29 14:40:43.335 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (NVIDIA GeForce RTX 4090): 1800 iterations, 6.101s init, 80.160s render
    Iteration Rate: 22.45 iterations per second
    Loading Time: 7.71 seconds (not sure why slower than 3090)

     

    Post edited by mwasielewski1990 on
  • mwasielewski1990mwasielewski1990 Posts: 343
    edited July 2023

    Okay, so now with the updated driver. No real gains or losses really.

    System Configuration
    System/Motherboard: MSI MAG B760 Tomahawk WIFI DDR4
    CPU: i5 12600K @STOCK
    GPU: Inno3D RTX4090 X3 OC (@STOCK = 1050mv)
    System Memory: 4x 16GB = 64GB Lexar Thor DDR4 @3200mHz (XMP profile)
    OS Drive: Samsung 970 Evo PLUS 2TB M.2
    Asset Drive: same
    Power Supply: Corsair RM1000e (1000W)
    Operating System: Windows 10 Pro 10.0.19045 N/A Build 19045
    Nvidia Drivers Version: 536.40 STUDIO
    Daz Studio Version: 4.21.0.5 Official release, x64
    Optix Prime Acceleration: n/a

    Benchmark Results
    2023-06-29 15:47:43.695 [INFO] :: Total Rendering Time: 1 minutes 26.59 seconds
    2023-06-29 15:47:45.288 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (NVIDIA GeForce RTX 4090): 1800 iterations, 5.806s init, 79.236s render
    Iteration Rate: 22.71 iterations per second
    Loading Time: 7.35 seconds

    Post edited by mwasielewski1990 on
  • mwasielewski1990mwasielewski1990 Posts: 343

    outrider42 said:

    Iray dev team just posted this, and it could be big:

    Iray 2023.0.0 final

    ..was just released.

    Major improvements over the beta:

    Performance, performance, performance. :)

    We mostly got back all the perf loss we introduced in the beta when running the new caustic sampler 2.0 (compared to cs 1.0 in 2022.1.X), plus we optimized general performance on the Ada Lovelace GPU architecture, and Iray is now faster when using rather complicated MDL materials (i.e. MDL JIT path is triggered).
    Then a lot of performance optimizations all over the place, e.g. when using fibers/hair, rounded corners or cutouts.

    In addition there is also one additional new major feature over the beta: Support for the new Mie phase function for volumes, introduced along with MDL 1.8.X.

    On top we have dozens of important bug fixes, as usual.

    The new release will also be available soon within Omniverse, of course, via the RTX Accurate rendering mode

    Is there any info on when this new Iray version is going to be implemented in DAZ?

  • oddboboddbob Posts: 397

    mwasielewski1990 said:

    POST-RENDER EDIT: Interestingly, the card only draws a maximum of 311 Watts. Any idea why? Why is the full 450W not utilized? My 3090 without undervolting topped at it's specced 350W.

    That's normal behaviour for a 4090 with the versions of Iray we currently have access to. It may change when better ADA support arrives but you'll still be limited by voltage. Raise the voltage to 1.1 and you'll see power draw increase but under heavy workloads HWinfo will flag vrel as the limiting factor. People who have done hardware voltage mods on 4090s have seen power draw rise at silly rates. Also look at your vram speed when rendering, for some reason it claims to be capped at 20500 when it's rated at 21000.

    I've got the 450w locked Inno3D card with the alphacool waterblock and stock performance is about what you're seeing, it settles at 2820 boost at about 290w. The Daz Studio beta is about 17% faster for me, so 26 iterations and change. Plus 100 on the core and plus 1000 on the vram will take you to about 28 iterations in the beta and should be easy enough to run. Voltage to 1.1, about 3100 on the core and plus 1500 on the vram hit 29.5 iterations in the beta, but only on one run. That said I tend to leave mine stock, it's quick enough and I prefer it cool and quiet. I'm hoping the new Iray version will get 30 iterations at sensible clocks when it arrives.

  • outrider42outrider42 Posts: 3,679

    The power draw has been discussed a lot. It is normal for now. Actually 311W is higher than what others have posted for a 4090. Part of this is that Iray is genuinely NOT as demanding as some software, because it is not passing so much data over the bus (like say a video game might.) That means that naturally there will be less power used with Iray. But at the same time it is way less than expected, and the Iray dev team themselves have acknowledged this in their blog. The post that Iray 2023 optimizes Lovelace is in reference to that.

    Daz will never tell us when something like this is coming ahead of time, that is just not something they do. It would be nice if they at least said they were testing the new Iray 2023 SDK, I do not see the harm in letting customers know that much, but Daz always plays total radio silence. You can only keep an eye on the Daz beta release threads, because the beta will get the updates first. It looks like the last beta came out in April, the new Iray just released in June. It is very possible that the next beta release still doesn't even have the newest Iray. It typically takes at least a month or two for a new Iray to get into the Daz beta. It takes even longer to find its way into the main build. The worst case was back in the Pascal days, it took many months for the main Daz Studio to get Pascal support, and at that time, no 1000 series card even functioned with Daz Daz Studio. It was a mess to put it mildly. At least today we don't have this issue and new GPUs will work at launch, even if not at their full capability.

    Still no 4060ti benchmarks I see. Also waiting for new 4060 marks. But I can understand people not wanting an 8gb card for Iray. Hopefully the 4060ti with 16gb will get more attention when it releases.

  • mwasielewski1990mwasielewski1990 Posts: 343
    edited July 2023

    Thanks people, I thought something is wrong with my card. Nevertheless, it's a 56% improvement over 3090 + the card is absolutely dead silent... My 3090 even after undervolting started to sound like a jet plane after a couple hours of batch renders.

    Also, I decided to try the BETA. Here are the results:

    System Configuration
    System/Motherboard: MSI MAG B760 Tomahawk WIFI DDR4
    CPU: i5 12600K @STOCK
    GPU: Inno3D RTX4090 X3 OC (@STOCK = 1050mv)
    System Memory: 4x 16GB = 64GB Lexar Thor DDR4 @3200mHz (XMP profile)
    OS Drive: Samsung 970 Evo PLUS 2TB M.2
    Asset Drive: same
    Power Supply: Corsair RM1000e (1000W)
    Operating System: Windows 10 Pro 10.0.19045 N/A Build 19045
    Nvidia Drivers Version: 536.40 STUDIO
    Daz Studio Version: 4.21.1.48 Public BETA, x64
    Optix Prime Acceleration: n/a

    Benchmark Results
    2023-06-30 08:42:50.546 [INFO] :: Total Rendering Time: 1 minutes 18.95 seconds
    2023-06-30 08:42:53.369 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (NVIDIA GeForce RTX 4090): 1800 iterations, 7.495s init, 69.864s render
    Iteration Rate: 25.76 iterations per second
    Loading Time: 9.08 seconds

    So we're looking at a 77% improvement over my undervolted 3090. Nice!

     

    Post edited by mwasielewski1990 on
  • mwasielewski1990mwasielewski1990 Posts: 343

    Darn, I mis-typed the specs and typed 3090 instead of 4090. I corrected all three posts.

    Another thing - does anyone keep mainteance of this thread and update the result table on the first page?

  • RayDAntRayDAnt Posts: 1,140
    mwasielewski1990 said:

    Darn, I mis-typed the specs and typed 3090 instead of 4090. I corrected all three posts.

    Another thing - does anyone keep mainteance of this thread and update the result table on the first page?

    In theory I do. However things like real life and technical limitations of the DAZ Forum itself (no actual html5 support, per-post character limits) have severely curtailed my update efforts. Plus long-term testing results have shown - time and again - that Nvidia GPU rendering performance scales EXTREMELY linearly per each chip generation. Meaning that all you really need to know is the performance of a few high end cards from eg. the RTX 4000 series and Cuda specs for the rest of the product stack to be able to extrapolate to surprisingly accurate degree the performance of the rest of the lineup.
  • RayDAntRayDAnt Posts: 1,140

    RayDAnt said:

    mwasielewski1990 said:

    Darn, I mis-typed the specs and typed 3090 instead of 4090. I corrected all three posts.

    Another thing - does anyone keep mainteance of this thread and update the result table on the first page?

    In theory I do. However things like real life and technical limitations of the DAZ Forum itself (no actual html5 support, per-post character limits) have severely curtailed my update efforts. Plus long-term testing results have shown - time and again - that Nvidia GPU rendering performance scales EXTREMELY linearly per each chip generation. Meaning that all you really need to know is the performance of a few high end cards from eg. the RTX 4000 series and Cuda specs for the rest of the product stack to be able to extrapolate to surprisingly accurate degree the performance of the rest of the lineup.

    ETA: Also, this thread is already only 8 pages away from its posting limit, so it makes more sense imo to wait for a new thread before attemtping to heavily revise/rework anything.

  • Richard HaseltineRichard Haseltine Posts: 101,516

    RayDAnt said:

    RayDAnt said:

    mwasielewski1990 said:

    Darn, I mis-typed the specs and typed 3090 instead of 4090. I corrected all three posts.

    Another thing - does anyone keep mainteance of this thread and update the result table on the first page?

    In theory I do. However things like real life and technical limitations of the DAZ Forum itself (no actual html5 support, per-post character limits) have severely curtailed my update efforts. Plus long-term testing results have shown - time and again - that Nvidia GPU rendering performance scales EXTREMELY linearly per each chip generation. Meaning that all you really need to know is the performance of a few high end cards from eg. the RTX 4000 series and Cuda specs for the rest of the product stack to be able to extrapolate to surprisingly accurate degree the performance of the rest of the lineup.

    ETA: Also, this thread is already only 8 pages away from its posting limit, so it makes more sense imo to wait for a new thread before attemtping to heavily revise/rework anything.

    This thread? It has another 48 pages to fill.

  • RayDAntRayDAnt Posts: 1,140
    Richard Haseltine said:

    RayDAnt said:

    RayDAnt said:

    mwasielewski1990 said:

    Darn, I mis-typed the specs and typed 3090 instead of 4090. I corrected all three posts.

    Another thing - does anyone keep mainteance of this thread and update the result table on the first page?

    In theory I do. However things like real life and technical limitations of the DAZ Forum itself (no actual html5 support, per-post character limits) have severely curtailed my update efforts. Plus long-term testing results have shown - time and again - that Nvidia GPU rendering performance scales EXTREMELY linearly per each chip generation. Meaning that all you really need to know is the performance of a few high end cards from eg. the RTX 4000 series and Cuda specs for the rest of the product stack to be able to extrapolate to surprisingly accurate degree the performance of the rest of the lineup.

    ETA: Also, this thread is already only 8 pages away from its posting limit, so it makes more sense imo to wait for a new thread before attemtping to heavily revise/rework anything.

    This thread? It has another 48 pages to fill.

    Lol right you are - for some reason I was thinking the limit is 50 pages. I'm gonna aim for an update/rework of the beginning of the thread in the near future. I'm mostly just reluctant to do it because it will mean a significant trimming of the collected data there.
  • outrider42outrider42 Posts: 3,679

    Perhaps a link to a separate page would be easier to maintain than trying to work within the forum limitations. Plus there would be no risk of the forum breaking the formatting.

  • mwasielewski1990mwasielewski1990 Posts: 343
    edited July 2023

    outrider42 said:

    Perhaps a link to a separate page would be easier to maintain than trying to work within the forum limitations. Plus there would be no risk of the forum breaking the formatting.

    A Google Sheets spreadsheet perhaps?

    Post edited by mwasielewski1990 on
  • RincewindRincewind Posts: 6

    Replaced my 2070S with a 4080, my 2070S was benching at slightly over 4 iterations per second with this version of Daz.

    System Configuration
    System/Motherboard: X570 Taichi
    CPU: Ryzen 3900X
    GPU: Maxsun RTX 4080 iCraft OC, EVGA RTX 3060 XC (both on PCI-E x8 4.0)
    System Memory: 64GB G-Skill DDR4 @ 3200
    OS Drive: Samsung 970 Pro 1TB NVMe
    Asset Drive: Seagate 4TB Barracuda
    Operating System: Windows 10 Home x64 10.0.19042
    Nvidia Drivers Version: 536.40 DCH
    Daz Studio Version: 4.21.0.5

    Benchmark Results
    2023-07-11 08:37:34.389 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Device statistics:
    2023-07-11 08:37:34.389 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 1 (NVIDIA GeForce RTX 3060): 486 iterations, 1.277s init, 80.024s render
    2023-07-11 08:37:34.389 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (NVIDIA GeForce RTX 4080): 1314 iterations, 1.207s init, 79.848s render

    2023-07-11 08:37:28.797 [INFO] :: Total Rendering Time: 1 minutes 23.93 seconds


    Rendering Performance:

    • 22.5 iterations per second (total)
    • 6 iterations per second (3060)
    • 16.4 iterations per second (4080)
  • RincewindRincewind Posts: 6

    And again with the beta

    System Configuration
    System/Motherboard: X570 Taichi
    CPU: Ryzen 3900X
    GPU: Maxsun RTX 4080 iCraft OC, EVGA RTX 3060 XC (both on PCI-E x8 4.0)
    System Memory: 64GB G-Skill DDR4 @ 3200
    OS Drive: Samsung 970 Pro 1TB NVMe
    Asset Drive: Seagate 4TB Barracuda
    Operating System: Windows 10 Home x64 10.0.19042
    Nvidia Drivers Version: 536.40 DCH
    Daz Studio Version: 4.21.1.48

    Benchmark Results

    2023-07-11 08:46:39.331 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Device statistics:
    2023-07-11 08:46:39.331 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 1 (NVIDIA GeForce RTX 3060): 454 iterations, 10.537s init, 70.903s render
    2023-07-11 08:46:39.332 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (NVIDIA GeForce RTX 4080): 1346 iterations, 10.488s init, 70.653s render

    2023-07-11 08:46:29.059 [INFO] :: Total Rendering Time: 1 minutes 33.14 seconds


    Rendering Performance:

    • 25.4 iterations per second (total)
    • 6.4 iterations per second (3060)
    • 19 iterations per second (4080)
  • LenioTGLenioTG Posts: 2,118

    Retlaw said:

    LenioTG said:

    Anyone with the RTX 4070 yet?

    I know it has been criticized because of the low price difference with the 4070 Ti, but in Italy the price cut is huge (600€ vs 900€). It still has plenty of VRAM and I guess it would be a big jump from my RTX 2070

    System Configuration
    System/Motherboard: MSI PRO Z790-P
    CPU: Intel® Core™ i7-13700K
    GPU: MSI GeForce® RTX 4070 12GB
    System Memory: 64GB (4x16GB) DDR5/5200mhz
    OS Drive: 2TB Solidigm P41 Plus M.2 NVMe PCIe SSD
    Asset Drive: 2TB Solidigm P41 Plus M.2 NVMe PCIe SSD
    Power Supply: Corsair RM850X 850W 80+ Gold
    Operating System: Windows 11 Home 64bit
    Nvidia Drivers Version: 531.61
    Daz Studio Version: 4.21.1.48 Pro Public Build
    Optix Prime Acceleration: STATE (Daz Studio 4.12.1.086 or earlier only)

    Benchmark Results
    2023-05-14 18:00:41.944 [INFO] :: Finished Rendering
    2023-05-14 18:00:41.977 [INFO] :: Total Rendering Time: 2 minutes 24.33 seconds
    2023-05-14 18:00:50.513 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (NVIDIA GeForce RTX 4070): 1800 iterations, 0.812s init, 141.528s render
    Iteration Rate: (12.71) iterations per second
     

    Thank you so much for answering!
    No one with a 4070 Ti yet? Or a 4060/Ti?

    I'm in need of an upgrade, but can't choose what to buy. Currently I'm still using a 2070 Super. Sadly, prices in Italy are crazy.

  • LuckyTLuckyT Posts: 4

    System Configuration
    System/Motherboard: Gigabyte X570 AORUS MASTER
    CPU: AMD RYZEN 5 3600 @ stock 
    GPU: Inno3D RTX 4070 Ti X3 @ stock
    System Memory: Corsair Vengeance 64GB/4x16GB @ 3200
    OS Drive: GIGABYTE GP-ASM2NE6500GTTD
    Asset Drive: Seagate IronWolf 4TB
    Operating System: Windows 10 Home 22H2

    Nvidia Drivers Version: 536.40
    Daz Studio Version: 4.21.1.48

    Benchmark Results
    2023-07-15 08:57:17.364 [INFO] :: Finished Rendering
    2023-07-15 08:57:17.432 [INFO] :: Total Rendering Time: 2 minutes 9.91 seconds

    2023-07-15 08:57:25.280 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : Device statistics:
    2023-07-15 08:57:25.280 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (NVIDIA GeForce RTX 4070 Ti): 1800 iterations, 10.678s init, 117.584s render

    Iteration Rate: 15.308 iterations per second
    Loading Time: 12.326 seconds

  • outrider42outrider42 Posts: 3,679

    There are some 4070ti marks buried in here, but we have another thanks to LuckyT.

    No 4060ti marks yet. I am sure people are waiting on the 16gb version to release. It releases on July 18, so only 2 more days. I expect at least a couple of people will post as soon as they get them.

    But the decision is going to come down to which is more important, the extra VRAM or the extra speed of a faster GPU. I like my 3060, but I can go over its 12gb sometimes, so I am glad I have my 3090. So I am tempted to get the 4060ti 16gb for that reason. I rarely go over 16. But I will probably wait unless things go real well for me this quarter.

    We may not see the next generation 5000 series for a long time. It may not come out until 2025, which is a longer than normal period. Things could change, of course.

  • billyben_0077a25354billyben_0077a25354 Posts: 771
    edited July 2023

    Right now, I have  3060 12GB.  Thinking about a 4090 (debating if I want to do that kind of damage to my credit card balance right now or not).  The below system was built to run a RTX 3090 but I was never able to get one.


    Current system specs are as follows:
    System Specs: AMD Ryzen 7 3700X
    Gigabyte X570 Aorus PRO WIFI
    EVGA 3060 12GB
    32 GB G.Skill Trident Z Neo DDR-4 3200 memory
     Samsung 860 EVO 1TB SSD
    2 Western Digital Blue WD20EZRZ 2TB HDDs in Raid 1
    EVGA Supernova 1200 P2 PSU
    CoolerMaster H500P Mesh Wihte case
    AOC 27G2G4 G-Sync Monitor @ 1080p
    Windows 10 Professional 64


    Other than uping my system memory to 64GB, anything else yoiu can think of that I need to upgrade?


    Also planing on getting a bigger hard drive (4TB EVO SSD) as mine is 90% full & over 2/3rds of it is DAZ content & software and make the switch to Win 11 Pro and do a general hard drive cleaning.

    Will be doing before and after comparisons after the Win11 upgrade with both cards.

     

     

    Post edited by billyben_0077a25354 on
  • outrider42outrider42 Posts: 3,679

    Only more RAM is needed. Daz Studio is not picky, except for GPUs. Any other upgrades would be potentially for other software more than DS.

    If you plan on using both GPUs at the same time, you might want to double check the sizes on them. 4090s can be so big that you might not have enough space for a 2nd card.

  • billyben_0077a25354billyben_0077a25354 Posts: 771

    outrider42 said:

    Only more RAM is needed. Daz Studio is not picky, except for GPUs. Any other upgrades would be potentially for other software more than DS.

    If you plan on using both GPUs at the same time, you might want to double check the sizes on them. 4090s can be so big that you might not have enough space for a 2nd card.

    It depends, I am looking at the Gigabyte cards and the only way I can use 2 cards is if I go watercooled,

  • Robert FreiseRobert Freise Posts: 4,464
    edited July 2023

    Not looking forward to when I have to upgrade my graphics cards. I'm currently running four water cooled GTX 1070 AERO 8G OC

    Post edited by Robert Freise on
  • outrider42outrider42 Posts: 3,679

    billyben_0077a25354 said:

    outrider42 said:

    Only more RAM is needed. Daz Studio is not picky, except for GPUs. Any other upgrades would be potentially for other software more than DS.

    If you plan on using both GPUs at the same time, you might want to double check the sizes on them. 4090s can be so big that you might not have enough space for a 2nd card.

    It depends, I am looking at the Gigabyte cards and the only way I can use 2 cards is if I go watercooled,

    It also depends on your motherboard, as spacing between pcie slots might vary a little. I could actually fit two 3090s in my rig if I wanted to, and doing a 3090+4090 would work as well, assuming these are all FE models. The 3090 FE is 57mm thick, while the 4090 FE is 61mm, yes, it is only 4mm thicker than the 3090. The 4090 FE is also slightly shorter in length than the 3090 FE. The 3rd party 4090s are completely insane, though, so that can mess up everything.

    They would only be a few millimeters apart, but the unique blower on these FE models would help. Half of the card blows air out the back just like an old school blower, and old school blowers are actually better for multiple GPU setups. The other half blows air through a heatsink, as there is no PCB on this side so the air goes right through. The 4090 would be passing some air up into the 3090, but it wouldn't be too bad. My 3090 runs in the very low 60s when rendering, and this is with a 3060 blowing all of its hot air directly onto it. A 4090 might cause things to run warmer, logically, but I think it would still remain reasonable. The 4090 would only be passing some of its heat, since a lot goes out the back.

    As it stands, I have a 3060 in the first slot on top, and my 3090 is in the 2nd slot. My 3060 is small (EVGA Black,) but I had a larger 1080ti in this same slot above my 3090 before that. The 3090 has plenty of space below it and the power supply. Meaning if I swapped my 3090 for a 4090, it would fit no problem and I can keep my 3060.

  • billyben_0077a25354billyben_0077a25354 Posts: 771

    Getting ready to upgrade m,y system including a new hard drive and a jump to WIN 11 Professional.  I am fresh reinstalling everything to clear out the clutter.  My question is does the curent 4.21 DAZ install have all of the asets to run the current benchmark file.  I am going to upgrade to a RTX 4090 and 64 GB of memory as part of this upgrade process.  Also debating of I will go from a Ryzen 7 3700X to a 5950X or not.  I am going to run a final set of benchmarks with WIN 11 with my 3060 thenrun a new set with the 4090 & 64 GB of memory.  Will see how much of a jump there is.  It should be substantial.  Been several years since I was able to buy the fastest video card on the planet (at leasdt until the next big ting comes out that is faster).

  • outrider42outrider42 Posts: 3,679

    Yes, you need all of the content the scene has in order to perform a proper benchmark. If anything is missing then it will not be valid, as it could be easier or harder to compute. Most of the content comes from G8F Starter Essentials, and radiant used an outfit from the Daz freebies. I can't think of it at the moment. It was mentioned somewhere in this thread, but we are at 42 pages now. That is a good number, but a lot of pages to sift just for that. Loading up the scene will report any missing items. When I get a chance I'll look.

  • DustRiderDustRider Posts: 2,754

    I thought someone might be interested in the results for a RTX 4090 Mobile (laptop) GPU. Keep in mind that the laptop version of the 4090 isn't a throttled down desktop RTX 4090, From the specs it looks to be a RTX 4080. So it's really more like an undervolted (lower core speed) RTX 4080. With that in mind, the performance isn't too bad. nowhere near a desktop RTX 4090, but really good for a laptop. I should also probably note that I ran the test in "Balanced" mode, not "Performance" mode. Balanced mode reduces max fan speeds, so there is no doubt some GPU throttling happening to keep temps down (62 degrees C was the max temp for the benchmark). This is the mode I am using for working on the laptop, so it's more of a real world benchmark IMHO. The fans are simply too obnoxiously loud in performance mode surprise. The system tested is a MSI Titan GT77HX 13VI with a 16Gb RTX "4090".

    System Configuration
    System/Motherboard: Laptop - Titan GT77HX 13VI
    CPU: 13th Gen Intel(R) Core(TM) i9-13980HX, 2200 Mhz
    GPU: Nvidia Mobile RTX 3090 16G
    System Memory: 128G
    OS Drive: 4TB NVME SSD Gen4
    Asset Drive: 4TB NVME SSD Gen4
    Power Supply: N/A
    Operating System: Windows 11 Pro 10.0.22621 Build 22621
    Nvidia Drivers Version: Studio Driver 536.67
    Daz Studio Version: 4.21.1 Public Beta


    Benchmark Results
    DAZ_STATS
    2023-07-31 15:18:58.470 [INFO] :: Total Rendering Time: 1 minutes 48.21 seconds

    IRAY_STATS
    2023-07-31 15:19:09.008 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (NVIDIA GeForce RTX 4090 Laptop GPU): 1800 iterations, 0.881s init, 105.716s render

    2023-01-21 22:21:42.864 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (NVIDIA GeForce RTX 3060): 1800 iterations, 2.520s init, 283.612s render
    Iteration Rate: (1800 / 105.716) iterations per second = 17.026
    Loading Time: ((0 * 3600 + 1 * 60 + 48.21) - 105.716) seconds = 2.494

  • LenioTGLenioTG Posts: 2,118

    Thank you so much for the 4070 Ti benchmark @LuckyT !

    outrider42 said:

    There are some 4070ti marks buried in here, but we have another thanks to LuckyT.

    No 4060ti marks yet. I am sure people are waiting on the 16gb version to release. It releases on July 18, so only 2 more days. I expect at least a couple of people will post as soon as they get them.

    But the decision is going to come down to which is more important, the extra VRAM or the extra speed of a faster GPU. I like my 3060, but I can go over its 12gb sometimes, so I am glad I have my 3090. So I am tempted to get the 4060ti 16gb for that reason. I rarely go over 16. But I will probably wait unless things go real well for me this quarter.

    We may not see the next generation 5000 series for a long time. It may not come out until 2025, which is a longer than normal period. Things could change, of course.

    Found those now, thank you. It appears the last update has indeed given the 4070 Ti a bit extra performance.

    The 4060 Ti is probably gonna go down in price, but right now it's €540 in Italy, definitely not worth it, in comparison to the 4070 €660.

    I render comics with 1 to 4 characters per scene, I almost never go out of my current 8GB of VRAM, so 12GB would be plenty for me. What I care the most is performance right now, I've skipped the RTX 3000 so I'm a bit behind.

    In the end, I've decided for a RTX 4070: it's already a huge jump in performance (+136%) and +50% in VRAM, and I can more easily afford it (yet, it's the most expensive GPU I've ever bought, by far). I can't justify the extra €240 for the RTX 4070 Ti.
    Everybody says in 2 years the RTX 5000 will wreck the market, and I'm confident I could get a better job by then, to afford a higher end GPU.

  • jparks123jparks123 Posts: 37
    edited August 2023

    4060 TI 16 GB

    Not as fast as expected, however not a disaster and OK if you reallly need 16 GB.

    System Configuration
    System/Motherboard: Gigabyte motherboard, details unknown
    CPU: i7-12700K
    GPU: Nvidia 4060 TI 16GB (MSI GAMING X)
    System Memory: 32GB
    OS Drive: M2 SSD
    Asset Drive: Same drive
    Power Supply: ?
    Operating System: Windows 11
    Nvidia Drivers Version: 536.67
    Daz Studio Version: 4.21.0.5
    Optix Prime Acceleration: ?

    Benchmark Results
    2023-08-01 23:21:40.442 Iray [INFO] - IRAY:RENDER ::   1.0   IRAY   rend info : CUDA device 0 (NVIDIA GeForce RTX 4060 Ti): 1800 iterations, 28.430s init, 249.087s render
    Iteration Rate: 7.23 iterations per second
    Loading Time: 28.430s seconds

    Post edited by jparks123 on
Sign In or Register to comment.