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
Very good points outrider. But I think most hobbyists like myself would very rarely get into the 5 hour+ rendering scene. I support that statement from what I see in the galleries, and not the super-dooper ones, the other 95%.
I define "hobbyist" as the majority of DS users rather than those who use DS for a living. There are certainly people within this category who do spend a great deal of time and effort to produce absolutely wonderful images for no personal profit. However I reckon very few, can I say normal people who, judging by our gallery, are involved in anything like renders that take hours and hours to render. I'd say 90% of them are one figure portraits with a background thrown in. Even if one was to pump up the settings to max, there's no way they'd take multiple hours to render. For those people Ampere wouldn't make much sense dollar-wise when render times are initially low to start with, unless they're on a pig of a GPU, in which case upgrading, if you have the bucks, might make some sense.
The new driver 451.77 SLOWED DOWN 4.10.123 PRO results (which were not the greatest) but increased performance on 4.11.0.366 BETA significantly.
My thoughts are the versions are driver dependent. I see a significant decrease in the time it takes for Daz to load, open in either version.
System Configuration
System/Motherboard: ASUS STRIX Z2070E
CPU: IntelCore i7-7700k @ 4.2 GHz
GPU1: GTX GeForce 1080 Ti @ SPEED/stock
GPU2: EVGA GeForce 1080 Ti @ SPEED/stock
System Memory: G.Skills Ripjaws 64GB @ 2133MHz/stock
OS Drive: Samsung SSD PRO 2TB
Asset Drive: WD Passport Ultra (External) 3TB
Operating System: Windows 10 Pro v 1511 Build 10586.446
Nvidia Drivers Version: 451.77
Daz Studio Version: 4.11.0.366 Beta
Benchmark Results
SLI: DISABLED
Optix Prime Acceleration: ON
Benchmark Results
DAZ STATS
2020-08-15 05:39:34.385 Finished Rendering
2020-08-15 05:39:34.401 Total Rendering Time: 4 minutes 34.22 seconds
IRAY_STATS
2020-08-15 05:39:40.397 Iray INFO - module:category(IRAY:RENDER): 1.0 IRAY rend info : Device statistics:
2020-08-15 05:39:40.397 Iray INFO - module:category(IRAY:RENDER): 1.0 IRAY rend info : CUDA device 0 (GeForce GTX 1080 Ti): 895 iterations, 3.292s init, 268.420s render
2020-08-15 05:39:40.398 Iray INFO - module:category(IRAY:RENDER): 1.0 IRAY rend info : CUDA device 1 (GeForce GTX 1080 Ti): 905 iterations, 3.269s init, 268.105s render
Iteration Rate: (1800 / 268.2625) = 6.709845 iterations per second
Loading Time: ((0 * 3600 + 4 * 60+ 34.22) - 318.7005) = 5.98 seconds
The EVGA booklet says 10/900 series. and it seems to be this one But the sli link is covering the tag for the GTX. Looking through the bag of tidbits from the tech, I remembered, shortly after this PC was built back, in Jan 2017, when the cards were in high demand and scarce, one of the GPU's went bad and had to be returned and replaced. Have no clue which card is performing better but they both did better this new updated driver. At this point, what version of Daz are you using as pro and beta on and what driver on your 1080 rig?
Edit: In the Nvidia panel they read identical which is why I thought they were. So what do you think of the difference in the 411. beta test results?
Well, now your times line up with mine, I get around 4:30 with drivers 441 and up.
So since you have a PC with two 1080tis like me, I am really curious about something. You see I was getting bench times in the 4 minute range in the past. And after a bunch of testing, I believe it was the drivers. With driver 431.36 I was hitting that 4 minute mark across multiple versions of Daz Studio. Unfortunately, the newest version of Daz Studio requires a higher driver number, so I cannot use driver 431.36 on the latest version of Daz. But I can use the latest driver on the older versions of Daz Studio that I have, and each one sees the same drop in speed to around 4:30.
Since you have Daz 4.11 still, you have the option of using driver 431.36 with it. So if you wanted to, you could still test this. That's up to you though.
So yes, I do think drivers can impact render speeds as well. Though I don't recall a driver causing this kind of drop before.
As for your GPU models, the EVGA 1080ti Black was a basic model, that is, it didn't offer any overclocking over the stock Founders Edition. But looking at the Black, it doesn't look quite the same, you can see the heat sink very clearly on a 1080ti Black. The EVGA with a blower fan covers the heat sink completely so you don't see it. And the text on the side is different.
Its kind of a moot point though, since the blower 1080ti and the Black both have the same clock speeds anyway. The Black might have a better cooler, maybe.
And either way you pretty much matched my time by using a new driver. That really surprised me, if your cards are indeed clocked over 400 MHz less than mine, how on earth are they rendering as fast?
The big question is if it will stay that way with 4.12. 4.12 introduced Iray RTX, which overhauls Iray a lot. This allowed the RTX GPUs to get a huge boost because it took full advantage of their new ray tracing cores. So your 2080ti machine should see a huge benefit. But for non RTX GPUs, like the 1080ti, that could be different. It didn't seem to effect my speed too much. At least until driver 441 and up.
System Configuration
System/Motherboard: MSI MEG X570 ACE
CPU: AMD R9 3950X @ Stock with PBO +200
GPU: 2x EVGA GeForce RTX 2080 SUPER XC HYBRID GAMING, 08G-P4-3188-KR @ Stock speed
System Memory: Corsair Vengeance RGB Pro 64 GB @ 3600 MHz CAS18
OS Drive: 1TB Sabrent Rocket NVMe 4.0 SB-ROCKET-NVMe4-1TB
Asset Drive: Same (1TB Sabrent Rocket NVMe 4.0 SB-ROCKET-NVMe4-1TB)
Operating System: Windows 10 Pro version 2004 Build 19041.450
Nvidia Drivers Version: Version 452.06
Daz Studio Version: 4.12.1.118
Benchmark Results - Two 2080 Super only No CPU rendering
2020-08-27 12:33:57.035 Finished Rendering
2020-08-27 12:33:57.071 Total Rendering Time: 2 minutes 44.14 seconds
2020-08-27 12:34:15.366 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2020-08-27 12:34:15.366 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (GeForce RTX 2080 SUPER): 897 iterations, 2.576s init, 158.812s render
2020-08-27 12:34:15.370 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (GeForce RTX 2080 SUPER): 903 iterations, 2.702s init, 157.804s render
Iteration Rate: (1800 / 164.14 seconds) = 10.96357656230966 iterations per second
Loading Time: ((164.14) - 158.812) = 5.328 seconds
duplicate post.
System Configuration
Same as previosu post
Benchmark Results - 3950X and 2x RTX 2080 Super
2020-08-27 13:26:28.294 Finished Rendering
2020-08-27 13:26:28.327 Total Rendering Time: 1 minutes 20.90 seconds
2020-08-27 13:27:16.884 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2020-08-27 13:27:16.884 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (GeForce RTX 2080 SUPER): 330 iterations, 2.365s init, 75.880s render
2020-08-27 13:27:16.888 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 1 (GeForce RTX 2080 SUPER): 332 iterations, 2.670s init, 74.822s render
Iteration Rate: (1800/80.9) = 22.24969097651422 iterations per second
Loading Time: ((80.9) - 75.880) = 5.02 seconds
@shaneseymourstudio
I recently made an attempt to clean my system and to try out a Quadro NVLink bridge, remembered this request. Note that in retrospect, I don't think 4 GPUs in a consumer system is a particularly good idea even if it all works, for several reasons, and with the rumored power budget of Ampere, I don't think it'll even be even possible anymore without more gymnastics.
You can see the 200mm riser ribbon cable running over the PSU and the vertical case mount for the 4th GPU. Because they're all blowers, they don't mind at all being so close together.
Again, though, if I had known then what I know now, I wouldn't have done it this was at all, and when the 4th GPU is in place, there is less than a millimeter of space between it and the NVLink, which sticks out sideways quite a bit; You cannot imagine the dread that welled up in me as I was trying to put it back in the bracket and realized how close it was going to be :) Consumer cases just aren't made for this, and that's the least of the problems.
Hi @chrislb, would you have any data for just the 3950X alone? I am highly suspecting that a TR was simply not worth it, but for the PCIe lanes. It seems like your render times were cut in half adding the 3950X, implying that the times using only the 3950X would be similar to the times using only the 2080s.
I knew my TR 1950 was old, but damn. I would happily buy a 3950X and render on it alone, at half the speed, if it meant I'd never have to worry about VRAM ever again...
I'd say that the Phanteks Enthoo Pro 2 or Phanteks Enthoo 719 would be a better case if you are running 4 GPUs. Plenty of room for the hardware, they support dual power supplies, and you and put quite a few fans in for air cooling or sevral large radiators for air cooling.
I'll run the test on the 3950X alone this weekend at the stock with PBO+200 and also with an all core overclock to see if there is any noticable difference. The only advantages I can see ot the 1950X is that it has more PCIe lanes, it supports more than 128GB of RAM, and it has more memory bandwidth if that's ever an issue for what you use the system for. I'm guessing the 3950X will be 25-35% faster than a 1950X based on other multi threaded benchmarks.
Thanks, @chrislb, that's why I'm so interested in your setup. In my tests, a single 2080ti decimated my 1950 by a factor of 6 I think, and so I'm looking for the discrepancy. I would expect your 3950X to be faster, even a lot faster, but not as much as your timing numbers show.
And you're spot on about the PCIs lanes, that was the main line of reasoning I followed in decided on the TR, but it doesn't look like even a 1600W power supply is going to be able to support 4 3090s, so I'm going to run just two per system, and wouldn't need the additional lanes anyway. And since the only thing I needed a faster processor for was cloth sim in Blender, which I now do largely in MD, I'm thinking that I really don't need the added expense of the TR and its components at all.
And wow, both of those cases look good, but it's not just the airflow, it's the screwy things that start to happen in a system pushed past the configurations in which things were well tested, and rather depending upon the theory of how things should work. From my experiences, I don't think Gigabyte has ever put 4 of the cards I have in the motherboard that I have to see if everything works. I experienced everything from disappearing RAM to disappearing SATA ports to GPUs working but no video signal to firmware upgrades that cause more problems than they solve. But just 2 GPUs is not so uncommon and I'd rather just go with that :) I want to render, not drop a crap load of money for a system I still have to spend weeks debugging.
In any case, thanks for running the test, it'll give me more data so I can make a better decision.
System Configuration
System/Motherboard: MSI MEG X570 ACE
CPU: AMD R9 3950X @ Stock with PBO +200
GPU: 2x EVGA GeForce RTX 2080 SUPER XC HYBRID GAMING, 08G-P4-3188-KR @ Stock speed
System Memory: Corsair Vengeance RGB Pro 64 GB @ 3600 MHz CAS18
OS Drive: 1TB Sabrent Rocket NVMe 4.0 SB-ROCKET-NVMe4-1TB
Asset Drive: Same (1TB Sabrent Rocket NVMe 4.0 SB-ROCKET-NVMe4-1TB)
Operating System: Windows 10 Pro version 2004 Build 19041.450
Nvidia Drivers Version: Version 452.06
Daz Studio Version: 4.12.1.118
Benchmark Results - Only 3950X CPU rendering
2020-08-28 04:50:23.851 Finished Rendering
2020-08-28 04:50:23.888 Total Rendering Time: 16 minutes 8.26 seconds
2020-08-28 07:21:22.140 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2020-08-28 07:21:22.140 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CPU: 768 iterations, 1.538s init, 963.274s render
Iteration Rate: (761 / 963.274 seconds) = 0.7900 iterations per second
Loading Time: ((968.26) - 963.274) = 4.986 seconds
System Configuration
System/Motherboard: MSI MEG X570 ACE
CPU: AMD R9 3950X @ 4.3 GHz all core 1.375 volts power limits raised
GPU: 2x EVGA GeForce RTX 2080 SUPER XC HYBRID GAMING, 08G-P4-3188-KR @ Stock speed
System Memory: Corsair Vengeance RGB Pro 64 GB @ 3600 MHz CAS18
OS Drive: 1TB Sabrent Rocket NVMe 4.0 SB-ROCKET-NVMe4-1TB
Asset Drive: Same (1TB Sabrent Rocket NVMe 4.0 SB-ROCKET-NVMe4-1TB)
Operating System: Windows 10 Pro version 2004 Build 19041.450
Nvidia Drivers Version: Version 452.06
Daz Studio Version: 4.12.1.118
Benchmark Results - Only 3950X@4.3 GHz CPU rendering
2020-08-28 08:35:25.636 Finished Rendering
2020-08-28 08:35:25.674 Total Rendering Time: 15 minutes 31.32 seconds
2020-08-28 08:35:46.714 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2020-08-28 08:35:46.718 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CPU: 788 iterations, 1.637s init, 926.157s render
Iteration Rate: (788 / 926.157 seconds) = 0.8508 iterations per second
Loading Time: ((931.32) - 926.157) = 5.163 seconds
@TheMysteryIsThePoint Here are the 3950X results. The 4.3 GHz all core overclock probably isn't worth the less than 5% reduction in render time. The CPU hits 88-92C during the render with the 4.3 GHz all core overclock. When run at stock with PBO+200 the CPU is 68-71C during the render.
In other benchmarks, my 3950X at 4.3 GHz with raised power limits draws 270-290 watts from the socket according to various monitoring software. That explains the high temepratures.
@chrislb
Thank you so much for running the benchmark on the 3950X. But if, as I suspected, your 3950X is around three times slower than even just one of your Supers, are you sure your 3950X + 2 x RTX 2080 Super numbers are correct?
I'm sure the 3950X with 2x 2080 Supers and 2x 2080 Super numbers are correct. Now I'm not sure of the 3950X only numbers are correct because the test stopped every time before it hit 1800 iterations. I ran the 3950X at stock speed with PBO+200 twice because it went faster than expected. I had similar results both times, so I assumed the results were correct. I rebooted and ran the 4.3 GHz test. THe results seemed appropiate comapred ot the stock speed test results. The log file had a message about stopping because it reached max convergence ratio or something similar to that. However, I used the same scene file and render settings that I used for the GPU tests and GPU with CPU tests. I'm not sure why it stopped before 1800 iterations when I didn't change anything and had the options set correctly according to the instructions on the first page.
I downloaded the scene file again then ran the test again and these are the results I obtained. Those other results are probably incorrect because the test stopped before 1800 iterations.
System Configuration
System/Motherboard: MSI MEG X570 ACE
CPU: AMD R9 3950X @ Stock with PBO +200
GPU: 2x EVGA GeForce RTX 2080 SUPER XC HYBRID GAMING, 08G-P4-3188-KR @ Stock speed
System Memory: Corsair Vengeance RGB Pro 64 GB @ 3600 MHz CAS18
OS Drive: 1TB Sabrent Rocket NVMe 4.0 SB-ROCKET-NVMe4-1TB
Asset Drive: Same (1TB Sabrent Rocket NVMe 4.0 SB-ROCKET-NVMe4-1TB)
Operating System: Windows 10 Pro version 2004 Build 19041.450
Nvidia Drivers Version: Version 452.06
Daz Studio Version: 4.12.1.118
Optix Prime Acceleration: N/A
Benchmark Results
2020-08-29 07:02:40.495 Finished Rendering
2020-08-29 07:02:40.537 Total Rendering Time: 21 minutes 22.49 seconds
2020-08-29 07:05:06.254 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2020-08-29 07:05:06.254 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CPU: 1800 iterations, 1.540s init, 1278.237s render
Iteration Rate: (1800/1278.237) = 1.4081895 iterations per second
Loading Time: ((1,282.49) - 1278.237) = 4.253 seconds
Then I think my next system will be an R9 instead of a TR. Thanks, that's what I needed to know!
Since I upgraded today both nvidia drivers and beta studio
System/Motherboard: GIGABYTE B450 AORUS PRO WIFI-CF
CPU: AMD Ryzen 5 2600X @ 4.0 ghz
GPU: MSI GTX 1070 Armor OC/Aero OC @ 1648mhz clock 2230mhz memory, MSI RTX 2080 Super Gaming X Trio
System Memory: Corsairw DDR4-3200 8gb x2 gskill DDR4-3200 8gb x2
OS Drive: Samsung 960 EVO 250gb
Asset Drive: Seagate 2TB 7200 rpm
Operating System: Win10 Enterprise 18363
Nvidia Drivers Version: 452.06
Daz Studio Version:
Benchmark Results - RTX 2080 Super Only
Daz Studio Version: 4.12.2.6 64-bit
Optrix Prime Acceleration: NA
DAZ_STATS
2020-09-06 22:39:55.180 Finished Rendering
2020-09-06 22:39:55.213 Total Rendering Time: 5 minutes 20.65 seconds
IRAY_STATS
2020-09-06 22:41:54.507 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : Device statistics:
2020-09-06 22:41:54.507 Iray [INFO] - IRAY:RENDER :: 1.0 IRAY rend info : CUDA device 0 (GeForce RTX 2080 SUPER): 1800 iterations, 3.425s init, 313.687s render
Iteration Rate: 5.73 iterations per second
Loading Time: 6.963 seconds
About the same as last test done I think.
Edit: I wish to contribute my result, but I have no "Daz 3D" folder in my AppData/Roaming(windows 10 Pro) any other way I can find my benchmark score?
Took my RTX 2060 little over 8 mins.
I'm sure its there somewhere, but you can use the help file to find this information. Help>Troubleshooting>View Log File
Assuming you just rendered, scroll all the way down to find the info.
I wouldn't overclock as stated. Rendering isn't Gaming. Rendering pushes a system to a whole other level versus gaming.
Watercooled / Aircooled, you're still pushing your system (sure compenets are getting cooled) but they are still being operated at a high level. I actually burnt out a motherboard some years back. It took a while (2 years of lots of rendering), but it was a surprise.
(One of the reasons I avoid second hand, cause ya just dont know.)
Although occasional quick renders are not especially taxing I would think.
I'll be upgrading my TR next year; rendering in Blender seems more CPU friendly. It's a great resource this thread.
Of course, we're all ears for the first 3080 benchmarks!!! :D
Please, if you manage to get your hands on one, we'd all be very thankful if you benchmarked it in Iray (possibly the Beta, that I guess has a newer version of Iray), thank you in advance!
Well l just checked a few websites and no joy. Only place with any cards for sale is E you know who. Dercided to click on higest price first and saw a card foir pre-order sale with the price of $99,900 with 131 bids. Are people stupid or what (I think it is shill bidding and they drove the price way too high).
Amazon claims to deliver my 3080 tomorrow ... will see ..
Is there any kind of reference scene vor Benchmarking? (Could run that on my old 2080 before)?
Ds does not yet support Ampere, though if the change logs and announcements from daz and nVidia are being read correctly we may hope that will change rather sooner than for previous generations.
No, what's happening is that some people have created bots to go after those scalpers (people who buy stuff on launch just so they can sell them at a higher price on places like Ebay) so that they pledge ridiculous amounts of money, but ultimately never end up paying, punishing the scalpers.
Good lol. I hate scalpers with a passion, they ruin everything. Had to pay double the normal ticket price of the free ozzfest some years back due to their sh-shinninnigans.
See here. To echo @Richard Haseltine above, the most current versions of Daz Studio available right now use Iray version 2020.0.1 - and Iray's release notes show Ampere only officially being supported with Iray version 2020.1.0 - so it may not work at all quite yet. But imo it's still worth a try.
Yeah, I tried on my 3080 even with the latest BETA. Not supported yet :(
IRAY rend warn : CUDA device 0 (GeForce RTX 3080): compute capability 8.6: unsupported by this version of iray, please look for an update with your software provider.
IRAY rend warn : GPU 1 (GeForce RTX 3080) with CUDA compute capability 8.6 cannot be used by iray photoreal.
IRAY rend warn : There is no CUDA-capable GPU available to the iray photoreal renderer.
Ok, so I was wrong. What's really ridiculous about this is that other render engines are able to use these cards, even if their OptiX is not updated specifically for Ampere. I mean...they have benchmarks, the cards work right now. Yet Nvidia's own Iray does not work with its own GPUs the day it launches. Let that sink in. The one and only time it has ever worked at launch was with Turing and that was the beta (and back then I stated we got lucky). It may not be Daz's fault, but it does not matter whose fault it is. I think that is just pathetic. There is really no other word for it. Pathetic.
The silver lining is that unlike with Pascal, Nvidia has already released an Iray SDK with support for the new GPUs. So it is only a matter of time. With Pascal, it was several months after launch before the SDK was even released.
This is one of those times that Daz needs to not play things so close to the vest and be open about when they will get the update out. At least acknowledge it is in the works or something. Iray 2020.1 has existed since July as a beta. Iray 2020.1 Final was released on August 17, exactly one month ago. https://blog.irayrender.com/post/626682753751154688/202010-final In the July blog the dev team stated that Ampere support was part of 2020.1, so this was no secret.
I know it takes time, but hey, the Daz beta is a beta for a reason.