Daz 3D is part of 
Connect
DAZ Productions, Inc.
224 S 200 W, Suite #250
Salt Lake City, UT 84101
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.Daz 3D is part of
Connect
DAZ Productions, Inc.
224 S 200 W, Suite #250
Salt Lake City, UT 84101
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
In my opinion I think waiting on the 4070 at this point is a very safe bet. New rumors suggest that the 4000 series could launch as soon as July, just a couple months from now.
Pretty much every rumor and leak says the 4000 series will be a huge leap, and that is just in gaming. Every generation of RTX has improved Iray performance far more than gaming.
There is plenty of history behind this for Iray. The 3060 is in the ballpark of the flagship 2080ti. The 2060 Super was faster than flagship 1080ti. So the past two generations have already shown low end cards can perform on par with the previous generation flagship. Just going by recent history, the 4070 should not only beat the 3080, but it should render faster than the 3090. That it could come with 12GB is even better.
And that doesn't account for pricing. The 4070 should be much cheaper than the 3080. There is no good reason to buy a 2000 series card for Iray, unless it is dirt cheap. And at this point I would avoid any GTX card (1600 and older) all together. The current versions of Iray seem to be buggy with GTX cards, too. The ray tracing cores in RTX are truly game changing, and the big leaps in Iray performance are directly due to the RT cores getting faster.
The only downside is that new GPUs will require new drivers and a new version of Daz Studio to run. Iray has been historically slow to update, so there is a very real chance that if you buy a 4070 when it launches that it will not work in Daz Studio for a while. Also, I sincerely hope that Iray's problems get sorted out with its 2022 release. The speed loss in Daz 4.20 is very real. The more testing I do, the worse it gets. This version of Iray in Daz 4.20 is the single worst performing version of Iray ever released. I am not saying that to be dramatic, the current version of Iray is a massive step backwards in speed. If you are on 4.16 right now, you might just need a 4090 to recover the speed you lose in making the mistake of downgrading to Daz 4.20.
So there is that possibility, that the performance loss in Iray in Daz 4.20 counters all the performance gains of new hardware. Nothing could be worse than spending all that money but the software kills performance.
4.20.1.38
EVGA 3060 RTX
RYZEN 5800X
[INFO] :: Total Rendering Time: 4 minutes 54.94 seconds
[INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 3060): 1800 iterations, 0.810s init, 292.719s render
Default/GPU Only - 6.15 iterations/s
2022-05-21 08:42:52.109 [INFO] :: Total Rendering Time: 4 minutes 14.69 seconds
[INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 3060): 1637 iterations, 0.840s init, 251.919s render
[INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CPU: 163 iterations, 0.693s init, 252.519s render
Maximum Overclock GPU +CPU - 7.13 iterations/s
Reached default convergence in 7m41s with 3280 iterations.
That's an even bigger slow-down than the previous release, hopefully it's actually fixing things.
Did you notice the mention in the notes that we should be using "guided sampling" for all but the simplest scenes? That would bring my iteration rate down to 5.77.
Turning on all the bells and whistles with 'guided' and 'caustics' for maximum realism, I'm down to 3.52 iterations/s (and that's with an overclocked GPU with CPU assist).
I stumbled together a machine and temporarily have my 1060 in there, so I wanted to run your test scene jobbie. Am I correct that it requires a prop named Shadow Thief Outfit for Genesis 8 Female(s)?
I don't own that because I don't own it, but I do like the grey blocks that appear instead. I'm keeping them!
It also seems to require the pants from Persian Beauty, but it looks like this is an old version that came with Genesis 2 Female Starter Essentials... correct? If so, I'll go ahead and install those parts of it.
Anyway, just checking. This has been a long process. Forgot how to build machines, made several errors, and now discovering I've forgotten how to set up DS itself as I want it! I'm feeling very old:-)
Everything came from the starter essentials from 2, 3, and 8. However the Shadow Thief is not in the Essentials download. I remember when I first used this scene after raydiant made it I didn't have the outfit, either. I had no idea it was included with Daz, when I made my bench scene I strictly used Genesis Starter Essentials. But I was able to download it from my product library. You just need to find it. You may need to 'purchase' either it or Daz to make it downloadable.
https://www.daz3d.com/shadow-thief-outfit-for-genesis-8-female-s
If you look at the "This product is included with..."
It shows it is included with Daz 4.x. Make sure you have "bought" Daz.x https://www.daz3d.com/daz-studio-4-pro
When I search for Daz Studio 4.x Pro in my account library, at the bottom it shows this bundle contains the following products, of which the Shadow Thief should be listed.
If the Shadow Thief is not listed there I would file a support ticket asking about it. If it is no longer a free download that causes a big problem with our benchmark. But you are the first person who says they don't have it.
Thanks, outrider-the-42nd! You saved me some time. I was searching for something similar in my props to replace it. So I DL'd the outfit and will now resume getting my machine set up.
I'm trying to limit what I install in these new content libraries. In the past I used barely 20% of what was in that cluttered mess. From now on I'll only install a non-daily-used item if & when I need it.
Now, back to work. Still much to do before I begin rendering. Have a great weekend!
No problem. The other 41 Outriders were a bunch of jokers.
But it is very important that people use exactly what is in the bench scene, anything different alters the results, even if it is seemingly similar. The stuff it uses only takes up a small amount of data in a runtime, though the Shadow Thief is the largest one at 186MB. The templates are totally optional, so they can be ignored. You could also delete the pieces of the outfit and their textures that are not used. I believe only the top is used in the scene.
I think the 4xxx are gonna be tought to get hands on. There are a lot of people that wanted a upgrade, but said screw that when they finally started being in stock at nonmoronic prices.
(There are no other NotAnArtists. In fact, I don't even exist).
Quick question to anyone out there listening. Is Step 7, "Close DS completely before continuing," the reason I don't see DEVICE_ITERATION_COUNT, nor DEVICE_RENDER_TIME, nor anything with TRT showing up in the log file?
I didn't take Step 7 seriously because it didn't mention WHY it would keep me from making a complete report. I don't follow orders unless I am forced to. I learned that long ago in the military.
I also had GPU-Z running illegally, because this was a test run of my new system with a 1060 6GB installed. Is it actually that significant?
I'll do it again if necessary. It was beautiful, btw:
CPU ave temps were 50C with very little variance
GPU ave temps were 61C with very little variance
But the hot spot hung at 69C... which worries me because I'm wondering what it will reach when I replace it with a larger, hotter GPU. I don't trust my build yet - it was like Homer Simpson, "Daw!" over and over again.
Well, it is a good idea to do so to ensure consistency. But in my testing it doesn't seem to be absolute.
This should not effect the stats you mention. I think it is a little unclear perhaps the way it is written. You need to close the render you just made. Either save it or close the render window, either one. This will finalize the render info you get in the help log file. Don't close DS itself just yet, you need to get that help log info. The help log contains a mountain of stuff, and you can quickly lose track of it in the piles of lines that the log spits out. If you close or save the render window, and then open the log file you should see all your stats at the bottom of the log. If you shut down Daz instead, those numbers might get lost.
Your temps just depend on what you have. Your case airflow is probably the biggest factor. Some cases have optional covers that actually hurt airflow. The Coolmaster case I have has a plastic cover that goes over the front of the case...where the fan intakes are! That is stupid beyond belief. The front is all mesh, so without that dumb cover it works great and looks fine. These days I have my side panel off the case, too. The 3090+3060 combo is pretty warm and even with all the vents for air in the top, bottom and back, it does run cooler with the side off. This may not be true of every case, some cases may depend on the side panel to help direct air. But often times this isn't an issue, and I think my PC is pretty sexy like this.
Props for surviving.
Completely closing Daz Studio is indeed in the instructions there to control for those stats not showing up in the log file when you look for them. Without going into too much detail, the way apps like Daz Studio access Iray for rendering is through SDK calls called "Instances" that need to be both programmatically created and (more importantly for our purposes here) destroyed by the host app in order for things like those lines of log file data to to be generated. And the only guaranteed way to make them appear going from one version of Daz Studio to another is by completely closing it.
I've never observed it making an actual difference. But there is always the possibility of something unexpected happening (GPU-Z can itself hang.) So the gneral rule is the less you have running the better.
Those all sound very well within spec to me (including that hotspot value - see the "Temperature Limit" section of GPU-Z's "Advanced tab for your GPU to see what temp limits it really cares about.) I'd expect a larger, more recent gen GPU in that same mix to perform perfectly fine. Going multi-GPU is where you may start running into trouble.
Barely! Not feeling "all here" lately. Anyway, I re-did the scene with my 1060. No GPU-Z running this time, nor anything else of course. The render began, I assume, at:
13:45:06.690 Iray [INFO] - IRAY:RENDER :: 1.8 IRAY rend info : Allocating 1-layer frame buffer.
It ended at 13:59:05.900 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend progr: Maximum number of samples reached. That's 14 minutes.
Beyond that, I don't understand your math. (Got A's in Calculus but was terrible in math). So I'm not even sure if I should submit this yet. Shall I just leave the Iteration and Loading lines blank?
Afterwards, I swapped my 1060 with a new 3080 Ti and rendered the scene. Very awkward getting that in there, but it works so far. I will submit it after I see how badly I misunderstood what you need in this first one. Total Rendering Time: 1 minutes 47.74 seconds!
Bare with me. I'm a literal reader, so when I saw "DEVICE_ITERATION_COUNT" etc yesterday, I tried to find those specific phrases in the log.
Also, I'm worried about how well I built my machine... And to finalise this whine, I don't read text well with reading-glasses which I got from the VA with a bad prescription due to cataracts. Boom, two days of crunching headache. So pity this old fart.
Oh! I just realized something - "TRT" meant total render time! I thought it was some deep, complex computer-speak. OMG, I'm too f'ing literal... I'll wait to hear from you if I should submit the data without doing the math;-)
I should have a benchmark result for an A5000 in a few days.
The log entry should look just like this:
2022-04-30 20:47:30.235 [INFO] :: Total Rendering Time: 4 minutes 31.60 seconds
2022-04-30 20:47:57.894 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 3060): 1800 iterations, 9.479s init, 259.889s render
Except of course with your GPU model listed. If you are rendering with 1 GPU, you should see the 1800 iterations as shown here. If you rendered with 2 GPUs, then there will be two lines which will total 1800 iterations. That is because the benchmark is designed to stop at 1800 iterations. These are all we need, and of course your PC build info as requested on the first page.
So if you are testing the 1060, you can just search your log file for "1800 iterations", and it should take you right to the line. But if you just open the log file after closing the render, this info is going to be at the bottom of the file and easy to spot.
If your computer has not blown up yet, it is probably OK. Each part has a thermal limit, which can be a little different. But as long as the computer has airflow it should be fine. You didn't build the GPU, so it will be fine as long as it has enough air for the fans to do their job. Most modern GPUs are built ok for their given power use. The CPU is something you want to pay attention to, if you built the PC that means you must have applied the thermal paste and cooler on the CPU. That can be a little scary at times, but as long as the hardware monitors show decent temps you did fine. These parts have pretty high tolerances, though it is good to keep them cool as possible.
The good news is that Daz Iray is not as demanding as most people think it is. It does run a GPU to 100% for the duration of a render, but the temps are actually lower than playing the latest video games, which can push GPUs as much as 10C hotter than Iray can. I saw this with my own 1080ti, which peaked at around 70C sometimes a little more with Iray, but would 84C playing demanding video games. While these are on the high side, they are in fact acceptable for the 1080ti. But you woudn't want to go above 84C. If you are in the 70s you are doing fine. If you are in the 60s you are doing super.
[I'm ready to post the 1060 results, except for the Loading Time, the formula of which makes NO SENSE to me. Iteration Rate is OK now that I realize I had to change my minutes to seconds. Please, I need to know what the Loading Time formula is actually asking for!]
I was fine with the thermal paste, etc, measuring the blob to exactly 4mm, because a tutorial said 3mm & I didn't agree. The scary part didn't come until I noticed something wrong with the cooling tower. I had meticulously put notes on it in pencil to show how it came assembled in its box. Where was the front, how did the fan latch and which way did it face, etc. I installed it and proudly walked away. Then it hit me what I had just seen - No!! But sure enough, yes. It completely loomed over the first PCI slot! I took it apart, cleaned up the paste (which was PERFECT, darn it! Edge to edge, corner to corner, not even drooling over; perfect), and started over. From then on the trust in myself was gone.
But CPU temps have been OK. In a quick check with the 3080 Ti in a partial Benchmark render, the cpu read between 31c to 49c, although it was bouncing up and down weirdly. With the 1060, it was mostly around 50-55c to occasionally just under 60c. Kind of strange difference.
The 3080 Ti's temps during a quick partial Benchmark render was 72c with a few higher peaks, and the hotspot was at 80c, but that was a very quick check. Longer rendering might have built up more heat, maybe. If so, I'll underclock it a bit. I'm just here for the CUDAs;-)
Installing the GPU was another scary part. THREE GPU cables crammed into the PSU/HD space below! I have two 3.5" drives down there stacked near the PSU's module connections. And two loose cables for eventually using a 3rd drive to transfer my gigs of props to the system. It's a cable jungle down there! But it works, so far, it's a miracle...
The total rendering time stat is your render time PLUS the loading time. That is on the first line, then a line below it will have the device info which lists the render time in seconds.
So back to the example given,
2022-04-30 20:47:30.235 [INFO] :: Total Rendering Time: 4 minutes 31.60 seconds
2022-04-30 20:47:57.894 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 3060): 1800 iterations, 9.479s init, 259.889s render
I highlighted the render time in orange, this shows it took 259.889 seconds to render. So to get the loading time you need to subtract the 259.889 from the yellow time 4 minutes 31.60 seconds. That means you need to convert the yellow time to seconds. So 4 times 60 plus 31.60. When I do that, I get 271.60 seconds. Now we can subtract the orange render time of 259.889.
That gives us a loading time of 11.711 seconds for this example.
The loading time doesn't really serve a huge purpose here. Loading times can be effected by all sorts of things, from the drive speed to RAM and CPU. But in the overall scope these loading times are low enough to NOT be an issue. Like here, the time was under 12 seconds, for a render that took over 4 minutes that is very small amount of time. What I am saying is you don't have to chase the best computer to get fast load times. You really just want to focus on getting the best GPUs you can when it comes to Daz Studio. These loading times are NOT to be confused with the time it takes to load a scene file, or to load a character preset from your library. Other factors effect them, including how large your Daz library becomes. The load times here only reflect the time it takes for Daz Studio to send the scene you have built to the GPU for rendering.
I have two Benchmarks here, one for a GTX 1060 & one for an RTX 3080 Ti. Thanks outrider42 and RayDAnt for your help!
Let me know if something here doesn't look right. I'll fix what's necessary ASAP. It's good to see my math talents at least haven't worsened in 73 years! When I run out of fingers, the numbers turn into the same old degree of dyslexic confusion.
I'll install DS 4.20 soon and do these tests again if you want, but give me time. I have lots of other issues to deal with here in these next couple of weeks. Been putting them off for too long. Thanks!
PS-If it means anything, in both cases I'm using the GPU for the display, as well. The CPU has no integrated video.
System Configuration
System/Motherboard: ASRock X570 Taichi
CPU: Ryzen 7 3700X (stock)
GPU: EVGA GeForce GTX 1060 6GB (stock)
System Memory: G.SKILL TridentZ RGB Series 64GB (2 x 32GB) DDR4 3200
OS Drive: Seagate Barracuda 1TB
Asset Drive: Same
Power Supply: EVGA SuperNOVA 1000 G+, 80 Plus Gold 1000W
Operating System: Windows 10 Pro Version 21H2
Nvidia Drivers Version: 512.59-desktop-win10-win11-64bit Studio Driver
Daz Studio Version: 4.14.0.10_Win64
Optix Prime Acceleration: N/A
Benchmark Results
2022-05-30 13:59:06.446 Finished Rendering
2022-05-30 13:59:06.478 Total Rendering Time: 14 minutes 5.51 seconds
2022-05-30 13:59:43.931 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2022-05-30 13:59:43.931 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce GTX 1060 6GB): 1800 iterations, 2.485s init, 839.758s render
Iteration Rate: 2.143475 iterations/second
Loading Time: 5.752 seconds
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
System Configuration
System/Motherboard: ASRock X570 Taichi
CPU: Ryzen 7 3700X (stock)
GPU: EVGA GeForce RTX 3080 Ti FTW3 Ultra Gaming (stock)
System Memory: G.SKILL TridentZ RGB Series 64GB (2 x 32GB) DDR4 3200
OS Drive: Seagate Barracuda 1TB
Asset Drive: Same
Power Supply: EVGA SuperNOVA 1000 G+, 80 Plus Gold 1000W
Operating System: Windows 10 Pro Version 21H2
Nvidia Drivers Version: 512.59-desktop-win10-win11-64bit Studio Driver
Daz Studio Version: 4.14.0.10_Win64
Optix Prime Acceleration: N/A
Benchmark Results
2022-05-30 15:25:08.190 Finished Rendering
2022-05-30 15:25:08.221 Total Rendering Time: 1 minutes 47.74 seconds
2022-05-30 15:25:32.221 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2022-05-30 15:25:32.221 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 3080 Ti): 1800 iterations, 6.667s init, 98.215s render
Iteration Rate: 18.327139 iterations/second
Loading Time: 9.525 seconds loading time
Gawd I hope I did these correctly;-)
I screwed up copying the log information and didn't grab the total render times for my first two renders, so I hope this is still usable.
Gigabyte TRX40 AORUS Master
AMD Ryzen Threadripper 3960X 24-core (stock)
EVGA GeForce RTX 2060 SUPER SC ULTRA GAMING
PNY RTX A5000
G.SKILL Ripjaws V Series 4x32GB DDR4 2666 RAM (stock)
Patriot SCORCH M.2 2280 512GB
Synology DS1815+ NAS with 8x Seagate Ironwolf 6TB 7200RPM hard drive
EVGA SuperNOVA 1300 G+
Windows 10 Home build 19044.1706
Nvidia Studio 512.96
DS 4.20.0.17
2022-06-01 16:45:33.283 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA RTX A5000): 1800 iterations, 6.521s init, 118.911s render
2022-06-01 17:11:06.563 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA RTX A5000): 1368 iterations, 7.270s init, 90.688s render
2022-06-01 17:11:06.564 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA GeForce RTX 2060 SUPER): 432 iterations, 7.994s init, 89.933s render
2022-06-01 17:32:35.791 [INFO] :: Total Rendering Time: 1 minutes 37.73 seconds
2022-06-01 17:32:44.793 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA RTX A5000): 1288 iterations, 7.044s init, 86.271s render
2022-06-01 17:32:44.793 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA GeForce RTX 2060 SUPER): 408 iterations, 7.084s init, 85.871s render
2022-06-01 17:32:44.793 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CPU: 104 iterations, 5.884s init, 86.987s render
Hi Gordig
That makes your iteration rates:
A5000: 15.14
A5000: 15.08 + RTX2060: 4.80 = Combined: 19.85
A5000: 14.93 + RTX2060: 4.75 + 3960X: 1.20 = Combined: 20.69
I'm not sure if its been mentioned in this thread, but I've tried two different web browsers on two different computers and both show the results this way now:
Same here I am seeing the same
I suppose the forums have made some kind of change, and that has altered the HTML that RayDant wrote to create those charts. The results are still in there, they are just hard to make out. It is a daunting process to build and maintain those things, especially now that there are so many entries over 31 pages.
It appears that the Daz forums have suddenly stopped honoring newline characters in the raw text of Code Snippets.Code Snippets are pretty much the only way the forums currently give you to fit that much columnized text in a sinlge post without humongous hypertext tag markup overhead. Unless/until Daz decides to completely update the forums to modern web standards there isn't really much that can be done about it, I'm afraid.
My best advice for right now is to go to the Quote button at the bottom of the post whose text you want to view, press it, and scroll through its contents in the WYSIWYG (without hitting Post Comment when you're done) since posts still appear properly formatted there (the WYSIWYG editor not agreeing with what's seen in submitted posts has been a growing problem on the forums here for at least 5 years now.)
I have a new test with my 3080 Ti, this time with DS4.20.
All I've done with the machine since I built it has been doing this total of three tests. There's barely been time to turn the machine on. Really would like to get back to having a hobby:-/ But these tests offer insite into what more I should be able to plan now with the new machine.
I got the 3080 Ti for the CUDAs, for render speed, nothing else. Not vast cityscapes with hi-rez creatures running about. Think animations. From what I can see so far, if the scenes are simple, with backdrops or maybe composited 3DL background anims under hi-quality Iray foreground anims, I'll have a good formula for short animations. This cost me so much of my savings (which took me the last 20 years to compile), I'm doubting my sanity now, but at least I'm going down happy.
Meanwhile, I have a request to provide simple, "cute", images for a children's book. No pay, it's my sister's fascinating new book idea, so I'm all into this anyway. So much potential! Now I just need to find the time to ACT!
*Hope my math was done correctly. My cataract eyes can't focus on it any more and my brain shuts down at the mere thought of trying.
NEW TEST WITH 3080 Ti and DS4.20.0.17:
System Configuration
System/Motherboard: ASRock X570 Taichi
CPU: Ryzen 7 3700X (stock)
GPU: EVGA GeForce RTX 3080 Ti FTW3 Ultra Gaming (stock)
System Memory: G.SKILL TridentZ RGB Series 64GB (2 x 32GB) DDR4 3200
OS Drive: Seagate Barracuda 1TB
Asset Drive: Same
Power Supply: EVGA SuperNOVA 1000 G+, 80 Plus Gold 1000W
Operating System: Windows 10 Pro Version 21H2
Nvidia Drivers Version: 512.59-desktop-win10-win11-64bit Studio Driver
Daz Studio Version: 4.20.0.17_Win64
Optix Prime Acceleration: N/A
Benchmark Results
Daz Stats:
2022-06-12 15:44:50.876 [INFO] :: Finished Rendering
2022-06-12 15:44:50.907 [INFO] :: Total Rendering Time: 1 minutes 52.38 seconds
IRAY_STATS:
2022-06-12 15:45:04.925 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2022-06-12 15:45:04.925 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 3080 Ti): 1800 iterations, 2.449s init, 107.292s render
Iteration Rate: 1800/107.292 = 16.776647
Loading Time: 1m 52.38s (112.38s) - 107.292s = 5.088s
I saw NewEgg was having a sale. And that they had 30x0s in stock. So I went for it. Unless something goes horribly wrong, I have a 3080 on its way to me.
That led me to do a set of unofficial benchmarks so that I can reassure myself that I didn't just waste a bunch of cash.
A few notes about my system and the tests below:
- Internal 3080 Super.
- External 3070 Super, on its own PS.
- Essential content packages are installed to one directory. All other content is installed to another directory.
With all of that in mind, what I wanted to test was:
1. Does having the extra video card powered on or off, but not used, have an impact on render time. (No.)
2. Does having a bunch of content installed impact render time. (No.)
3. Am I seeing a significant difference with both video cards utilized. (Yes, about 2x speed.)
4. Is there a benefit to having the external card just render solo. (Surprisingly, yes, but no where near having both utilized.)
5. Do the resulting images have any noticable differences. (No. They were identical to my eye.)
When the new video card comes in, I plan to run more tests:
1. 3080 external, with 2080 internal.
2. 3080 internal, with 2080 external.
3. 3080 internal solo.
======================
001 -
Extra Content Directories Hidden
Extra Video Card Powered Off
Just NVIDIA GeForce RTX 2080 SUPER
======================
2022-06-23 14:34:07.849 [INFO] :: Finished Rendering
2022-06-23 14:34:07.882 [INFO] :: Total Rendering Time: 4 minutes 54.16 seconds
2022-06-23 14:35:51.440 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2022-06-23 14:35:51.440 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 2080 SUPER): 1800 iterations, 8.316s init, 283.538s render
======================
002 -
Extra Content Directories Hidden
Extra Video Card Powered ON
Just NVIDIA GeForce RTX 2080 SUPER
======================
2022-06-23 14:53:12.904 [INFO] :: Finished Rendering
2022-06-23 14:53:12.934 [INFO] :: Total Rendering Time: 4 minutes 47.94 seconds
2022-06-23 14:53:21.383 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2022-06-23 14:53:21.383 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 2080 SUPER): 1800 iterations, 2.065s init, 283.609s render
======================
003 -
Extra Content Directories Hidden
Extra Video Card Powered ON
Just NVIDIA GeForce RTX 2070 SUPER
======================
2022-06-23 15:02:32.493 [INFO] :: Finished Rendering
2022-06-23 15:02:32.534 [INFO] :: Total Rendering Time: 4 minutes 25.55 seconds
2022-06-23 15:02:57.916 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2022-06-23 15:02:57.916 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA GeForce RTX 2070 SUPER): 1800 iterations, 2.980s init, 259.971s render
======================
004 -
Extra Content Directories Hidden
NVIDIA GeForce RTX 2080 SUPER and
NVIDIA GeForce RTX 2070 SUPER
======================
2022-06-23 15:09:04.712 [INFO] :: Finished Rendering
2022-06-23 15:09:04.742 [INFO] :: Total Rendering Time: 2 minutes 27.47 seconds
2022-06-23 15:09:22.823 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2022-06-23 15:09:22.823 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 2080 SUPER): 889 iterations, 2.139s init, 142.529s render
2022-06-23 15:09:22.823 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA GeForce RTX 2070 SUPER): 911 iterations, 3.363s init, 141.095s render
======================
005 -
Extra Content Directories Visible
NVIDIA GeForce RTX 2080 SUPER and
NVIDIA GeForce RTX 2070 SUPER
======================
2022-06-23 15:15:56.717 [INFO] :: Finished Rendering
2022-06-23 15:15:56.746 [INFO] :: Total Rendering Time: 2 minutes 26.96 seconds
2022-06-23 15:16:11.113 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2022-06-23 15:16:11.113 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 2080 SUPER): 888 iterations, 2.165s init, 142.287s render
2022-06-23 15:16:11.113 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA GeForce RTX 2070 SUPER): 912 iterations, 2.915s init, 141.202s render
The 3080 arrived and good news / bad news.
Good news, 3080 worked into system easily. No technical issues.
Bad news, it's longer than pictures and measurements would suggest. So putting it internal would take a bunch of angling that I am not comfortable doing at the moment.
2022-06-25 11:18:42.055 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Rendering with 2 device(s):
2022-06-25 11:18:42.055 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 3080)
2022-06-25 11:18:42.055 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (NVIDIA GeForce RTX 2080 SUPER)
2022-06-25 11:20:14.481 [INFO] :: Finished Rendering
2022-06-25 11:20:14.509 [INFO] :: Total Rendering Time: 1 minutes 41.51 seconds
Going to reboot and try with just the 3080.
Quick reboot (to get consist results). One more render. 2080 Super internal, but not used for the rendering.
2022-06-25 11:33:02.056 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Rendering with 1 device(s):
2022-06-25 11:33:02.056 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 3080)
2022-06-25 11:35:10.650 [INFO] :: Total Rendering Time: 2 minutes 11.45 seconds
Final observations:
1. Internal 2080 Super and External 2070 Super: 2 minutes 26.96 seconds
2. External 3080 used for render, Internal 2080 Super for display only: 2 minutes 11.45 seconds
3. Internal 2080 Super and External 3080 used for render: 1 minutes 41.51 seconds
I expected 1 & 2 configurations to have similar results, but ~20 seconds difference is significant enough that I don't think it's abberant.
3 proves that the 3080 was a good investment, particularly with the additional VRAM. Which can be utilitized when a scene doesn't fit in the 2080.
Hi,
Quick question. Will choosing the 12th generation Intel processor in the "KF" option instead of "K" will it have any influence on the rendering speed in DAZ3d?
Overall, I bought a little bit without further thought the i7 12700KF versions, and now read various conflicting opinions. Fortunately, I still have a few days to exchange.
I made a test on stage from this thread. Turning on the CPU shortened the rendering time only marginally.
Only GPU RTX 3090
2022-07-03 13: 06: 11.186 [INFO] :: Finished Rendering
2022-07-03 13: 06: 11.202 [INFO] :: Total Rendering Time: 1 minutes 47.14 seconds
2022-07-03 13: 06: 54.671 Iray [INFO] - IRAY: RENDER :: 1.0 IRAY rend info: Device statistics:
2022-07-03 13: 06: 54.671 Iray [INFO] - IRAY: RENDER :: 1.0 IRAY rend info: CUDA device 0 (NVIDIA GeForce RTX 3090): 1800 iterations, 0.941s init, 105.076s render
GPU + CPU RTX 3090 / i7 12700KF
2022-07-03 13: 04: 00.888 [INFO] :: Finished Rendering
2022-07-03 13: 04: 00.908 [INFO] :: Total Rendering Time: 1 minutes 46.73 seconds
2022-07-03 13: 04: 19.579 Iray [INFO] - IRAY: RENDER :: 1.0 IRAY rend info: Device statistics:
2022-07-03 13: 04: 19.579 Iray [INFO] - IRAY: RENDER :: 1.0 IRAY rend info: CUDA device 0 (NVIDIA GeForce RTX 3090): 1686 iterations, 0.999s init, 104.134s render
2022-07-03 13: 04: 19.579 Iray [INFO] - IRAY: RENDER :: 1.0 IRAY rend info: CPU: 114 iterations, 0.772s init, 103.805s render
I repeated the same test on my second platform: Ryzen 9 3900X + Nvidia RTX3070. With a total time of about 3 minutes, the difference was less than 3 seconds.
Only GPU RTX 3070
2022-07-03 13:09:26.357 [INFO] :: Finished Rendering
2022-07-03 13:09:26.394 [INFO] :: Total Rendering Time: 2 minutes 58.22 seconds
2022-07-03 13:17:01.481 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2022-07-03 13:17:01.481 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 3070): 1800 iterations, 0.975s init, 174.968s render
GPU + CPU RTX 3070 / Ryzen 9 3900X
2022-07-03 12:58:00.574 [INFO] :: Finished Rendering
2022-07-03 12:58:00.616 [INFO] :: Total Rendering Time: 2 minutes 55.57 seconds
2022-07-03 12:59:02.373 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2022-07-03 12:59:02.374 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (NVIDIA GeForce RTX 3070): 1627 iterations, 1.920s init, 171.289s render
2022-07-03 12:59:02.374 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CPU: 173 iterations, 1.274s init, 170.849s render
Regards.
Short answer: no.
Longer answer: Not really, but having the non-KF version (with iGPU) does open up some potentially very useful alternative multi-GPU rendering system configurations. For example, driving your computer's display(s) with the motherboard's display/hdmi ports so that so that one of your graphics cards isn't more memory constrained (for rendering) than the others. These days, unless I'm doing something like playing a game where I specifically need/want the graphics horsepower of a dedicated GPU for what I'm seeing onscreen, I keep my display(s) connected to the motherboard iGPU ports only out of convenience.
ETA: Having the just K iGPU can also be very useful for troubleshooting situations.
And potential problems.
I would say the benefits outweigh any negatives by far. A GPU is more likely to fail than a CPU, in which case having the igpu is the difference between a non functioning PC versus a working PC. Not everybody has spare gpus. This point alone is worth it.
Intel in particular can leverage their igpu in other tasks. Adobe software can have performance gains with the igpu, even if you have a good GPU in the system.
Sure, maybe you can have an issue on occasion with a igpu...just disable it and be done. You still have that igpu for emergency trouble shooting.