Daz 3D is part of 
Connect
DAZ Productions, Inc.
7533 S Center View Ct #4664
West Jordan, UT 84084
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
So to clafify, even though the progress bar indicates a percantage, that's not actually showing the convergence percentage, it just shows how close the render is to finishing.
So when I render the benchmark scene, it is set up to stop at 95% convergence when the progress bar gets to 100%. I see now that this is the default convergence ratio for DS.
I think I've got it now.
Correct.
Some fresh (sub-optimally cooled) Titan RTX Sickleyield benchmark numbers for the DS 4.11 Pro non-beta just released:
System Configuration
System/Motherboard: Gigabyte Z370 Aorus Gaming 7
CPU: Intel i7-8700K @ 4.7GHz (all cores)
GPU: Nvidia Titan RTX
System Memory: Corsair Vengeance LPX 32GB DDR4 @ 3,000MHz
OS Drive: Samsung 970 Pro 512GB
Asset Drive: Sandisk Extreme Portable SSD 1TB
Operating System: Windows 10 Pro 1809 17763.557
Nvidia Drivers Version: 380.86 SRD
Daz Studio Version: 4.11.0.383 64-bit
Benchmark Results
OptiX Prime Acceleration: Off
Total Rendering Time: 1 minutes 10.1 seconds
CUDA device 0 (TITAN RTX): 5000 iterations, 2.050s init, 66.969s render
Iteration Rate: 74.661 iterations per second
System Overhead: 3.131 seconds
OptiX Prime Acceleration: Off
Total Rendering Time: 1 minutes 24.61 seconds
CUDA device 0 (TITAN RTX): 5000 iterations, 2.084s init, 81.627s render
Iteration Rate: 61.254 iterations per second
System Overhead: 2.983 seconds
TL;DR Analysis: Virtually identical rendering performance as the previous 3-4 Beta releases (4.11 official is currently still using a pre-RTX version of Iray.)
DS 4.11 General Release
3 x GTX1080ti Founders Edition
Sickleyeild's test: 1 minutes 3.17 seconds
Outrider's test: 3 minutes 43.97 seconds
[Numerous edits due to bad spelling]
Just upgraded my new laptop and decided to run the good old Sickleyield benchmark.
Dell Precision 7710
Intel Xeon E3-1535 v5
32GB DDR4
Nvidia Quadro P4000 8GB
Daz Studio 4.11 General Release
OptiX Prime on
GPU only render
2019-06-14 21:19:39.805 Total Rendering Time: 4 minutes 19.37 seconds
As many users I run to get a new RTX card, one 2080 TI. I was using 2x 1070 at the time.
The results, however, were disappointing, 2080 TI achieve the same time as dual 1070, but dual 1070 are a lot cheaper than one 2080 TI.
I put the blame on the beta version and didn't published until yesterday when I saw the new 4.11 official build.
CPU: Ryzen 7 1700
RAM: 64gb DD4 2400mhz
MOB: Asrock x370 Killer SLI
Using file iray_bench_2018_b223da33.duf
GPU Only
Daz 4.10.0.123 Pro Geforce 416.34
2x 1070(Monitor On 1) + 1070 - Optix On = 4 minutes 39.4 seconds
2x 1070(Monitor On 1) + 1070 - Optix Off = 5 minutes 15.1 seconds
Daz 4.10.0.123 Pro Geforce 417.01
2x 1070(Monitor On 1) + 1070 - Optix On = 4 minutes 39.1 seconds
2x 1070(Monitor On 1) + 1070 - Optix Off = 5 minutes 21.69 seconds
Daz 4.11.0.123 Pro (Beta) Geforce 417.01
2x 1070(Monitor On 1) + 1070 - Optix On = 8 minutes 0.85 seconds
2x 1070(Monitor On 1) + 1070 - Optix Off = 8 minutes 51.79 seconds
Daz 4.11.0.236 Pro (Beta) Geforce 417.71
2080 TI (Monitor on) - Optix On = 5 minutes 26.36 seconds
2080 TI (Monitor on) - Optix Off = 5 minutes 14.74 seconds
Daz 4.11.0.383 Pro Geforce 430.86
2080 TI (Monitor in other gpu) - Optix On = 4 minutes 57.46 seconds
2080 TI (Monitor in other gpu) - Optix Off = 5 minutes 5.68 seconds
So I tried to apply the Denoiser resource, that were the variations
Denoiser Available (On) Denoiser Enable (On) Denoiser Start Iteration (8) Denoiser Denoise Alpha (On) Optix (Off)
5 minutes 5.48 seconds
Denoiser Available (On) Denoiser Enable (On) Denoiser Start Iteration (8) Denoiser Denoise Alpha (On) Optix (On)
4 minutes 46.85 seconds
Denoiser Available (On) Denoiser Enable (On) Denoiser Start Iteration (8) Denoiser Denoise Alpha (Off) Optix (On)
4 minutes 46.70 seconds
Denoiser Available (On) Denoiser Enable (Off) Denoiser Start Iteration (8) Denoiser Denoise Alpha (Off) Optix (On)
4 minutes 47.64 seconds
Edit: I was curious about how IrayTestSceneStarterEssentialsOnly.duf would be: 1 minutes 9.97 seconds Optix(on) and 1 minutes 25.19 seconds Optix(off).
Adding more details, in my previous set I had:
1x Asus 1070 Dual Oc on 1st PCIE x16 (Monitor attached)
1x Gigabyte 1070 Mini ITX on 2nd PCIE x16
Windows on dual 240GB SSD Intel 730 on RAID 0
Daz Content on 2TB HD SATA3 Seagate Barracuda
Current set:
1x Galax RTX 2080 TI OC DUAL on 1st PCIE x16
1x Gigabyte 970 Mini ITX on 2nd PCIE x16 (Monitor attached)
Windows on 512GB SSD M.2 XPG Gammix S11 Pro
Daz Content on 1TB SSD SATA3 WD Blue
Keep in mind that the Iray renderer version shipping with 4.11 General does not yet have full RTX hardware ray-tracing acceleration support. Meaning that there should be a major performance uplift for your 2080ti some time in the near future.
Your RTX 2080ti has three advantages over the dual GTX 1070 setup.
1. It only uses 1x double height PCI-e slot instead of 2x Double height slots
2. 11GB of VRAM (3 More than the GTX 1070)
3. RTX cores (when support gets added into Iray
I ended up using the second PCI-e slot to put a weaker card for Windows, as my cpu Ryzen 7 1700 doesn't have onboard graphics. My goal was let the 11gb just for render.
I notice that you are using 2x Xeon E5-2650V2, I recently bought these two from Aliexpress to replace one i5-4690k, it's hard do find benchmarks on Google. Just for curiosity, how good they perform with Daz?
Just upgraded my video cards.
2xRTX 2080ti Founders Edition
Sickleyeild's test: 45.28 seconds
Outrider's test: 2 minutes 37.72 seconds
Not too shabby. Faster renders, less electrical consumtion, and room for the upper card to breathe. And I've already got buyers for my old cards.
1070 + 1060 + ryzen 2600x(no overclocking yet) no optix
Total Rendering Time: 2 minutes 53.23 seconds
1070 + 1060 + ryzen 2600x(no overclocking yet) with optix
Total Rendering Time: 2 minutes 7.8 seconds
1070 + 1060 no CPU no optix
Total Rendering Time: 3 minutes 1.91 seconds
1070 + 1060 no CPU with optix
2 minutes 10.49 seconds
Heya folks! ili here for post some useless benchs
IrayTestSceneStarterEssentialsOnly
1070TI X1 with PCIe 2.0 x16 2:33sec
1070TI X1 with PCIe 2.0 x1 2:32sec WTF ???!!!? my brain is gone x1 get same results as x16
1070TI X2 1:18sec
1070TI X2 + 980TI X1 56sec
iray_bench_2018_b223da33
1070Ti X2 6:27sec
1070TI X2 + 980TI X1 4:33sec
RIG:
2X Zotac GTX 1070 Ti AMP!
Nvidia GTX 980 TI Reference
№1 1070TI runs with PCIe 2.0 x16 (Desktop)
№2 1070TI runs with PCIe 2.0 x1 (usb riser from mining)
№2 980TI runs with PCIe 2.0 x1 (usb riser from mining)
Windows 10 1903
DAZ 4.11 Beta
I keep telling you guys that PCIe is not much of a factor in Iray GPU rendering. There is not much data traveling across the bus during a render. The entire scene is fit into the GPU, where it pretty much stays in VRAM while the GPU processes it.
Now things might change if multiple GPUs are involved, but probably not a lot. I am not sure what point we can saturate PCIe.
Even in the case of GPU pooling VRAM with Nvlink, the Nvlink bypasses the PCIe by establishing a direct link between the GPUs.
One possiblity is if we reach a point of real time rendering in the future and the scene is animating, drawing many frames every second. Like a video game does. Then the bus speed will be a factor as the data has to move in and out more rapidly.
I think we can see the biggest impact of PCIe bandwith/lanes on the Iray viewport!
I've noticed a world of difference there upgrading my GPU
2 x 2080ti w/ NVLink and OptiX ON - Sickleyield: 28.73 seconds
I can't download Outrider's from the gallery link for some reason, link keeps redirecting to daz gallery homepage
Just upgraded the secondary card in my workstation from a Quadro K4000 to a Quadro P4000 8GB, so I decided to run the trusty old Sickleyield benchmark to see how it fares.
Now, keep in mind that this is a single slot card that only uses 105W
Daz Studio 4.11 Public release
Nvidia Display Driver ver. 431.36
Quadro P4000 8GB (Optix Prime On) (No Scene Pre-load)
2019-07-14 00:01:01.370 Total Rendering Time: 3 minutes 34.59 seconds
Quadro P4000 8GB (Optix Prime Off) (No Scene Pre-load)
2019-07-14 00:16:13.359 Total Rendering Time: 4 minutes 56.98 seconds
And just for fun:
Quadro P4000 + GTX 1080 ti (Optix Prime On) (No Scene Pre-load)
2019-07-14 00:19:43.877 Total Rendering Time: 1 minutes 23.61 seconds
vs.
GTX 1080 ti (Optix Prime On) (No Scene Pre-load)
2019-07-14 00:22:29.215 Total Rendering Time: 2 minutes 6.1 seconds
@brunocesarrrr - I will run the benchmark on my 2x Xeon E5-2650 V2 CPUs next
Dual Intel Xeon E5-2650 v2 (total 16 cores - 32 threads)
64GB quad channel REG ECC DDR3 1600MHz (8 out of 16 slots filled
2X E5-2650 V2 (Optix Prime On) (No Scene Pre-load)
2019-07-14 00:42:00.361 Total Rendering Time: 14 minutes 56.78 seconds
and just for kicks:
2X E5-2650 V2 + Quadro P4000 + GTX 1080 ti (Optix Prime On) (No Scene Pre-load)
2019-07-14 00:50:32.214 Total Rendering Time: 1 minutes 19.47 seconds
adding in the CPUs knocked 4 seconds off the render time...
For those interested in comprehensive performance statistics relating to the latest Das Studio beta release (now featuring full RTX acceleration support on capable cards) I highly recommend checking out this new thread Daz Studio Iray - Rendering Hardware Benchmarking. Through careful design and thorough testing, it aims to resolve many (if not all) of the issues surrounding accurate benchmarking which have plagued threads like this one over the years.
[double post]
Daz 4.12 with Iray RTX is out ! Here my benchmark !
Iray Test scene starter Essential.
I7 6700K 4.2Ghz + 32Go RAM 2200Mhz - GTX 1070 x 2 at 1980Mhz
GPU ONLY - No Optix - 100%
Daz Studio 4.11 No Optix: 2 minutes 5.44 seconds
Daz Studio 4.12 No Optix: 1 minutes 28.29 seconds
More than half minute GAIN !!
Iray 2019 has some pretty large performance increases even if you aren't on the latest RTX cards. Just to be warned, the first render on the latest version of Iray will take a little extra time, but all renders after that (with the same shaders) should be faster.
So that means we will never know how much performance gain came from RTX. Great
That is a really nice way of saying we'll never know how much performance came from RT cores
Right, good point that should work somewhat.
Maybe I am doing something wrong, because I am not seeing any speed increase at all with either the SY scene or my 2018 scene with Daz 4.12. I am using the newest Creative Studio Drivers, 430.86. Going into the help file I can confirm that Iray is using its 5.0 June release build. It also says it is beta, BTW.
IRAY rend info : Using iray plugin version 5.0-beta, build 317500.2554 n, 08 Jun 2019, nt-x86-64-vc14.
Which means the bugs that have been discussed are indeed all preset with this build.
My machine: 2x 1080ti. They are not the same model, so they have slightly different clocks. They tend to run between 1949 and 1964 during the course of the render, but that can depend on ambient room temps. CPU i5-4960 @3.5 Ghz 16Gb RAM.
SY scene OptiX ON
1 minute 1.87 seconds
SY OptiX OFF
2019-07-22 22:27:04.751 Total Rendering Time: 1 minutes 2.18 seconds
My 2018 scene
OptiX ON
2019-07-22 22:12:44.565 Total Rendering Time: 4 minutes 25.70 seconds
OptiX OFF
2019-07-22 22:24:17.864 Total Rendering Time: 4 minutes 25.4 seconds
So just looking at this, OptiX Prime OFF has taken a big leap compared to what it used to be, however, at least in my tests, it can only match the existing Prime ON times.
And something has gone haywire. I went back to Daz 4.11 and the render time was slower than I've ever seen with it. I also tried 4.10 Pro, and its time was not on par with what I have done in the past. I tried reinstalling my drivers, even switching to the latest game drivers instead of creator studio, and no dice. Somehow installing 4.12 beta has screwed up my other Daz installations!
OK, I think I see what's going on here, and its not what people think.
The new 4.12 loads into VRAM faster than ever, its actually a fantastic improvement, but I think some people are confusing this as a rendering speed improvement. It looks like 4.12 renders faster than 4.11, but only slightly. However, compared to 4.10, it is still behind in pure rendering speed.