genesis 9/victoria 9 8k maps and lower graphic cards
videoninja719
Posts: 343
Can genesis 9/victoria 9 8k maps be rendered using low end cards like rtx 3050 and 3060?
Post edited by videoninja719 on
This discussion has been closed.
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
As far, as I have understood it from the stream, it is enough to have one fully clothed character in the scene. If you scale down your texture mas to 4k with an image editor, it should work to have more figures in your scene.
MAybe, now that the 4000 series of Nvidea is out, you can get a good deal on a 3090 soon.
But that's a big maybe though
How much VRAM do you have on those cards? This is what will limit you.
RTX 3050 has 8GB vram, RTX 3060 has 12GB.
The 3060 12GB should be fine, but 8GB cards are going to be on the edge...
Short version, sure.
Long version, it's complicated.
Just to put this in a bit of perspective...
On my tesla m40(12GB), under default render settings with no hdri, at 1000x1000, a genesis 8 basic female(no hair, no clothes, render sub-d of 2), shows a vram usage of 1564MB, at render. Without g8f, 1155MB.
Approximately 419MB for the figure with textures.
Without textures, using the ITF Dev kit version, the vram utilization drops to 1356MB.
So, the textures take up about 208MB by themselves.
The default textures are a mix of 4k and 2k files.
Batch converting all the textures, 23 of them, to 8k the vram utilization jumps to 2241MB, an increase of 677MB.
Note that this is with texture compression at defaults of 512 and 1024, meaning the textures are getting compressed.
Adjusting these settings to 8192 or above, the vram utilization jumps to 4835MB, an increase of 2594MB
Just ball parking it, we're probably going to see increases of 1-2GB of vram utilization for the G9 series, over the g8 series.
Time to start saving up for an A100, gonna need that 80GB of vram before long. lol.
For those wondering, that's a $10k gpu.
oh well, probably I'll stick with g8. I've a RTX2070Super with 8GB and an i5 6 core, and 32GB of Ram. Jumping to G9 is tempting me, but changing the videocard (and probably the PSU) it's not on my roadmap. Actually it's quite fast for simple renders (under 1 hour for a scene with G8 and not too full of items, less than 30 minutes if I use a G3 character) and I prefer wait a bit (I tons of items never used on a scene).
Also if you talk about a rtx 3xxx as low end card, I think my PC could have troubles managing scenes with more than a G9 character.
An hour to render on a 2070 Super is a long time. I had one before I updated to 3060 12GB and I rendered interior scenes with two to three G8 figures (with clothing and hair) usually in around 20 minutes.
The 3060 is just marginally faster, but I bought it for the 12GB VRAM as 8GB's was getting restrictive.
Running DS 4.15 on i7-5820K, X99, 64GB, 3060 12GB and W7 Ultimate.
I tend to render big scenes with upwards of 20 characters, done in chunks of 4-5 characters at a time. I take it Genesis 9 is pretty useless to me then?
Also- if a 3050 is a low-end card, what does that make my 2070?
:sigh: I guess even basic Genesis 9 will be out of the question for me. Computer is barely 6 months old and has only a GTX-1660 Super. 6 gbs of vram, and lucky to get even that much because it was the last of the Daz-suitable machines left when I ordered it from Dell. Was thinking maybe in a year or so I could get a better card but now I'm not sure. Have hit serious RL issues and trying to hold on to my sanity, let alone my money. So not a good time for new & shiny, expensive toys.
to repeat myself from this thread - https://www.daz3d.com/forums/discussion/comment/7715486/#Comment_7715486
This is for Genesis 9 base. I dont know if Victoria 9 will have 8K skin maps or not
Well the very high-res textures will probably not be of much use for you, but nothing stops you from using lower-res ones. The rest of the features might be more useful for your needs.
I use Gigapixel AI already to downscale textures to 2048x2048 max as my rule of thumb is, unless doing a closeup, textures do not need to be larger than the largest dimension of your render and since my max render size is 1080p, or 1920 at the widest, I don't need a texture over 2kx2k. Comparing a render done that way with one using 8kx8k textures, I can't see the difference. I not only save rendering memory, I also save disk size as I have a 6TB library full of tons of stuff going back to generation 3 and I either need to take the ages it would to go back and delete a lot I havent used in years or spend the money for an even larger drive.
2070 and 2070 super are better than a 3050. The 2070 is about 6% better than the 3060.
https://gpu.userbenchmark.com/Compare/Nvidia-RTX-3050-vs-Nvidia-RTX-2070/4127vs4029 ;
From what I see genesis 9 is just a tremendous resource bogger. There's no real news apart a denser mesh and bigger textures. We already have HD figures for that. It would be great if, at last, daz makes a genesis 10 line optimized for the gpu and with unimesh to freely exchange male and female outfits .. uh .. wait .. there's genesis 1 already LOL.
p.s. Anyway the unimesh proved itself to be a seriuos pita for the PAs to develop content, that's why genesis 2 came out. So I don't think the unimesh in genesis 9 is this great idea.
The unimesh idea is the one of two things about G9, that I really don't like.
I don't mind the denser mesh, for I think modern PC can handle the geometry quite well. I don't like the lack of edge loops and the missing geometry for navel and modeled nipples
I do not like the unimesh/unisex implementation either, however I am willing to give it a chance with G9 to see how well it works. I hope the store labels clothes, footwear, and other items as male or female for the sake of filtering.
Perhaps, with the denser mesh, it would have to be denser than original Genesis, and denser mesh around the areas different between males and females, and more advanced technology, unimesh would work out better than it did for Genesis. I'm not holding my breath even if it would make it easier for an end user like me.
Gaming Benchmarks don't test the features we use in rendering and can be misleading.
Here's a quick summary of some of the results from the Benchmark thread, in terms of Iterations per second in DAZ Studio:
The 3060 is almost twice as fast as the 2070 with around the same amount of Watts but the 3060 also an extra 4GB of memory, just what you need for those 8K textures!
So... Long time no see.
Let's talk about the mesh. I have next to no idea why G1 was such a problem but I can already see why G9 is a single mesh.
Previous posing figures have been built on a design philosophy that might have been a good idea in the 1990s. (I oversimplify.) That logic still works for a one off figure. That is, say, "anime girl."
The idea of a single mesh to rule all shapes didn't die because G1 was a pita. It's evolved very much since. Today's typical production mesh for games and movies isn't really built like Genesis or and of previous generations.
G9 is built like a modern figure. This means it will work out-of-the-box with more platforms and fewer headaches. Given modern workflows, the previous meshes weren't as flexible. It's going to be easier for the modern artist to work with.
Which makes it more of a pita for my own tech but more useful for my art. I can see old school artists being put off, but I can also see failure to adapt to what is the wider industry standard leading Studio down the path of Poser.
Change is inevitable. This change is probably also necessary as it allows Daz to position its ecosystem in the middle of a lot of production pipelines. There's even potential for artists who adapt to capitalize on the larger market.
As for the end user, it ought to be seamless. Again, the modern workflows make some of the old issues moot. It's unlikely G9 wasn't designed with the Genesis issues in mind.
I, for one, look forward to the opportunity to test drive it. It's possible it will fail, sure. If it doesn't, it's going to dominate.
Not saying you have to buy($$$) in. But don't dismiss it until you take the free version for a spin. No doubt there are going to be free morphs around.
Oversimplifying = Uglification. There are already unisex mannequins and lots of free base models available online, without detailed customization support and the differences between opposite sexes, those models would not be attractive and have no advantages over other competitors. Evolution of species has favoured opposite sexes over unisex, that's something worth learning from the nature mechanisms.
There's another reason I prefer improving the G8 system rather than switching to G9 system, because I've invested a lot into the G8 ecosystem, custom morphs, geografts, face generating plugins, 3D body scan plugins, material generating plugins, rigging plugins, lots of very useful plugins from Zev0, and those bridges for other software which support G8 / G8.1. Because of the ecosystem Genesis 8 characters have grown to be widely used beyond image rendering purpose.
Even making Daz to paid software would be far better than pushing such unexciting G9 for quick money, if Daz really needs more money there are always options to switch to subscriptions for more professional features, as ditching G8 for G9 would be technically moving backwards and become less attractive to MetaHuman or other alternatives, unless G9 comes with more exciting features which I haven't found from the disclosure.
G8 system still has lots of potential and needs improvements. The current official bridges can't even handle geografts, and consider integrating this feature that'd be much desired for Unreal users: https://www.daz3d.com/forums/discussion/590086/showcase-bringing-daz-figures-to-life-in-ue5#latest
Except we were talking about the 3050, which isn't in your chart.
That very question is scary to me. My rig has a GTX 1070 Max Q with i7-8750H CPU, 16GB of ram, running Daz 4.15.0.9. I guess here's no point in trying to load and render Genenis 9 without serious bottlenecks.
I feel as though you misunderstood what I said. As in "oversimplification" was what I did to the technical explanation, and has exactly nothing to do with the coming figure.
Further, it's fairly trivial to say that your statements about unisex meshes are not true. PIXAR developed that kind of tech years ago. They put out a scientific paper on it.
As for the bridges: That's a different thing altogether. As in, it's a thing they give away for free. Daz has long been known for putting the money maker first, all the time, every time. There are other tools that can do those kinds of things, easily. Professional tools. Daz already has a nice niche user base of hobbyists. The changes to G9 seem to be about INDUSTRY standards. And the industry doesn't need the bridges. I'm NOT a pro, and I don't really need the bridges. I know how my tools work. Even in light of the limits in the bridge, I wouldn't have issues with the geograft thing. If I were attempting to port G8 to Unreal, it wouldn't be a problem. What they don't already automate, I could write a script to handle, most likely. I'm pretty sure pros who actually know programming and the like wouldn't have a problem.
I need to underline, here, that you are comparing what the industry does now to Genesis, an 11 year old product. How old is that in Computer? Windows XP was the world dominant OS. It was only that year that Windows 7 took the crown. Top GPUs had a massive 1.5ish GB of VRAM. No one would believe you if you told them what was going to happen starting in about December of 2018. I hadn't turned 40, yet. 11 years is a looooooong time in computer tech.
Also: Most game devs don't need the whole model, all naked and stuff. They use clothing. The stuff under gets deleted. I'm not saying that there aren't... er, use cases... out there. But geografts aren't really an industry issue with most of the seams being hidden and all.
My point in all of this is not to "win." I'm just making a point about how things actually are. Art is art. Tools are tools. V4 is still huge, and that's great for art. If you want to make games on G8, it's not gonna be an issue (if you're willing to learn how to use all the tools). Gaming PC speed will make up for much of the losses. G9 is going to be better suited to the bleeding edge, where superior performance is a must have, not the indie scene. Of course, it shouldn't be harder to use for indie devs, but it won't break games if they skip it.
You not buying in isn't going to break Daz. Everyone always announces they aren't getting on the change wagon with each new generation. Then all the cool stuff is on the new figures and people get over it. So I'm not trying sell to you. I'm just trying to be informative. If I just had an opinion, I'd say nothing, and just watch the drama. But the idea that the geometry is the bottom line has been dead for a long time. Daz's own Multiresolution tech demonstrates that. Genesis 3 showed that by having a lower poly count. If the new mesh is a technical failure it will because Daz dropped the ball, not the concept. I might offer any number of complaints about how they do technical things over there, but I haven't seen them drop the ball on a figure since ever. They haven't pleased most of the people all of the time, sure. But the very reason people know about the technical side drawbacks of Genesis is because a lot of users, I was one of them, liked the cross type nature of Genesis.
They had to explain to people like us why G2 was split up. And we didn't like that. We went on to demonstrate that the fundamental structure of G2,3, and 8 are the same across the sexes, with only minor differences in the chesticle region. My strangely popular morph transfer fiasco was based on that fact at the start. The differences are largely invisble at present. I did write that script. I could prove, with relative ease, that know could tell the difference between G8F and G8M without access to the wireframe.
This is a fundamentally invalid argument on multiple levels. 3D models are not subject to the laws of nature, and those mechanisms are only useful when they apply. You can use a single slider to make default G8F look indistinguishable from default G8M, same with previous generations; the original Genesis was unisex, and do you think that Victoria 5 doesn't look like a woman and Michael 5 like a man?
A base figure is exactly that: a BASE upon which to build morphs, and G9 was made to facilitate morph creation. G9 DOES have detailed customization support. If default G9 isn't attractive to you, there's nothing stopping you from making it look exactly the way you want it to, and in fact it would be easier to do with G9 than previous generations.
Why do you keep assuming that Daz is only releasing G9 to make money? You have your own reasons for wanting to stay with G8? Go ahead and do that, but you not wanting G9 doesn't invalidate the advances that G9 offers. What are you basing the latter part of that quote on? How is it "moving backwards"? Why would G9 be "less attractive to MetaHuman"?
Moving from G8 to G9 does not implicate anything about the bridges, and ironically, geografts are one of the big benefits of a unisex mesh because now you'll be able to put any geograft on any figure.
Also, you keep using the word "attractive", and as always in discussions like this, I need to point out that the things you think are attractive are not facts. I think the most attractive G8 base character is Kala. You can disagree with me as hard as you want, but I'm not WRONG about that. There's a thread ranking the different generations of Michael, and the results are all over the map (except that nobody except Serene Night seems to like Michael 8 very much).
I for one, I am really, REALLY happy that G9 will be like Genesis.
So much nicer to work with * in my humble opinion *
I agree that nature mechanisms do not apply to artificial figures.
Victoria 9 looks sexy but G9 unimesh is too primitive to my like :)
People who use Daz as the only DCC software or as the end output would be very happy about G9 unimesh and probably don't care about bridges.
G8 and all the G8 ecosystems are far more useful for tool chains beyond image rendering purpose, until G9 gets the same level of support of G8.
My opinion is to keep support of both G8 and G9 for different purposes.
@Gordig It is not a matter of technology but of complexity and easy of use. They separated the gender meshes in G2 because it was too difficult for the PAs to handle the unimesh for male/female outfits. This was the explanation by daz. Otherwise there was no reason to abandon the unimesh. After that they kept the gender-meshes from G2 to G81. Now they go back to unimesh with G9 that is even more complex than G1, so it is logical to expect that this will fail and will be abandoned same as G1.
As for my personal opinion, the unimesh fits better a morphing figure as genesis is intended to be. But a dedicated mesh can provide better topology for the details of a specific figure, that's also why geografts are used when the figure can't provide enough details. So separated meshes can be more optimized and produce better details for specific features.
"Primitive" in what way?
Again, what are you basing this statement on? Why would G9 be any harder for bridges to deal with? Also, G9's rigging is more in line with industry standards, so G9 could potentially be much easier to work with in other programs.
That depends on what you mean by "support". There's no reason to believe that G8 will stop working in DS at any point, and the bridges were built around G8, so I wouldn't expect future updates to the bridges to drop G8 support. Will Daz continue releasing new content for G8 after G9 is released? Probably not for long, but that's been true of every previous generation as well.
IS that logical to expect? I would posit that it's more logical to think that they didn't go back to a unimesh for no reason, and that they applied lessons they learned from G1 in that decision. They've explained the decision, and the changes they made to facilitate clothing creation.
Watched the promotion video of Victoria 9, when dragging the G9 unimesh slider there is not enough change to distinguish two opposite sexes.
Don't know the G9 rigging until it comes out. Maybe you're one of the G9 developers and know the G9's rigging is more industry compatible?