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
© 2025 Daz Productions Inc. All Rights Reserved.
Comments
...very intriguing. So I'd have an ongoing "storage account for 5$ a month + whatever 1.10$ hourly charge for rendering comed to for each render job.
As I mentioned this would be great for the final high quality rednering that I could then take to a gallery quality print service. 24 GB of VRAM is pretty hefty as that would equate to (using Osso's calculations) about 90 GB of physical memory (and be much faster). I cannot see one of even my epic scenes going beyond say 4 hours on one of those GPUs.
Storage _starts_ at $5. You'll probably want more for DAZ Studio and whatever else you're installiing and renders.. $5 covers only 50 GB. I finally found the storage rates below that I saw on a video. You also will pay for Splashtop to communicate with Paperspace and that starts at $5. I believe it goes up as well. Splashtop gets around the mouse issues.
https://support.paperspace.com/hc/en-us/articles/360003804333-Storage-Pricing
..so do I need my whole runtime in storage or just the scenes?
That's not a gaming card. The Frontier is more of a low tier Quadro, and it is terrible as a gaming card. The Titan is traditionally the fastest gaming card available, even faster than the x80ti of its generation, and it has a few Quadro features (but not all.) Even the Titan V can game, and it has gaming drivers made specifically for it. AMD has no competition for Titan, just like they have no competition for the 2080 and 2080ti. VRAM doesn't matter if the card can't game.
So the new RTX Titan will certainly be faster at gaming than the 2080ti, and offer the full Turing chip. It will be interesting how they price this thing. Considering the Titan V is $3000, the price of the new Titan could pretty much anything. However something is different this time. Having youtube influencers tease this card could be a sign that this Titan will not be positioned like the Titan V was. It sure is strange, since some of these influencers also criticized Nvidia for the 2080ti launch. Perhaps the price of this Titan will come across like an apology, maybe they will not have the obscene Founder's Edition markup over the MSRP, similar to the 1080ti did.
Just what you need for the render. DAZ Studio, models, morphs, textures, etc.
...but it was is great as a rednering card which is what we are looking for. Nvidia still has not yet addressed this beyond pushing their far more espensive Quadro line .
hi, sorry to ask this, I couldnt read the entire thread, is there any news about whether iray will support RTX cores in near future? from reading, its based on optix prime, nvidia will not update optix prime (this was said by a nvidia employee) and I found no hint whether iray will be ported to optix api instead... rather the opposite, google-hit wise iray feels like abadonware... if RTX will not be supported isn't the 980ti still a better alternative to the 2070 card, unless you render so much that power cost kicks in? 980ti used is ~300 and 2070 is currently 530.... so 2070 is 60%+ more expensive for 25% more speed (speed according to octane bench)
It's here.........the Titan RTX! More memory than I had anticipated. 24GB VRAM, yours for a mere $2500. If that ain't enought, they have a custom NVLINK coupler so you can use two. Ack!
https://www.engadget.com/2018/12/03/nvidia-officially-unveils-its-flagship-titan-rtx-gpu/?yptr=yahoo
Specs seem amazingly close to the new Quadro 6000.....so why buy a Quadro?
Driver support I gather - but yea, they are blurring the lines for sure.
Thanks to Kevin for the info on Paperspace, that really makes me think about future hardware purchases....
This Titan has 24GB of memory! That would be nice for creating larger scenes in Studio The RAM limit is one of the two things I really don't like about using Iray.
Nice! Gonna get me 2 of those! Uh, ok maybe not.
sorry to rain on the fun, NvLink memory pooling only works on Tesla cards. Memory pooling is disabled on Titan V and all consumer (RTX20 era) cards. It's just a glorified SLI link for the rasterized frame buffer on consumer cards. As for Quadro RTX, I have yet to see if NvLink even works outside of the dual-card RTX Quadro 8000 model.
Even if the fingers are on the PCB for the card, does not imply that Nvidia did not laser etch the NvLink off of the GPU die, as they do for other lower end parts. Case and point, the new wave of GTX1060 cards with SLI fingers and no SLI support in the GPU die for it.
(EDIT) Kevin Sanderson is correct, I was wrong in assuming the Titan RTX would be like all the other RTX20 cards and only have RT as the selling point for the price.
That's what I thought too...but from the article: "Should that not do the job, you can get the $80 Titan RTX NVLink Bridge (below), letting you join two Titan RTX cards together with double the VRAM and 100 GB/s of total bandwidth." 'Course that may be incorrect, since we don't have all the info yet.
But at the end of the day, I don't have $5080 in spare change to buy this 48 GB dream rig, and it still would not run VUE Infinite 2015 renders.
I feel you on not having a small country GDP to spend on Iray compute cards, all of the RTX lines is out of my reach. Stuck on a lowly 1050ti and a 1070non-ti, (sad face)
It's on the site: https://www.nvidia.com/en-us/titan/titan-rtx/?nvid=nv-int-ndtnrx-66563
:TITAN RTX NVLINK™ BRIDGE
Double the effective GPU memory capacity to 48 GB and scale performance with up to 100 GB/s in total bandwidth of data transfer utilizing the NVIDIA NVLink™ technology."
Yeah, I wish that we could use IRAY on multiple computers in CPU mode without paying the yearly fee for each one. Would not have to worry so much about large scenes and pricey GPUs.
Yeah, a large disposable income would be nice to have so I could get one of these. Hopefully, the price for a card with 24GB of RAM will become more affordable in the next generation or two.
Hmmm. That i might buy; but as you say, close to the Quadro. I have trouble believing it's real.
....read about it on Tom's this morning. Their quoted price was the same as the Titan-V. 2,999$. Still for 24 GB of VRAM and improved performance, that makes it a bit mor worth the price.
Time to go get a Megabucks Lotto ticket for tonight.
The only reason you call it great is purely because of how much VRAM it has. There are other factors needed to be truly great, and Frontier is only a Vega 64 retuned for professional use. But it is a lot slower than the Vega 64, and which was already slower than the 1080 in most cases, so when you add this up, it is terrible as a gaming card. If the Frontier was faster and a real competitor then things would be different. But it is not. The Frontier is only competing with Quadro. The Titan is as more of a true gaming card than anything else, it is the fastest single card of its generation whenever it launches, even the Titan V fit this description, no matter how much Nvidia pitched it as a science and research card. That is the appeal of the Titan. It is quite literally Nvidia's pony car. It crosses into the professional field, but for really pro applications you want Quadro. Rich gamers buy Titans simply to say they have the fastest gaming rig possible.
Actually the vray guys with chaosgroup have proven that memory pooling DOES WORK on both the 2080 and 2080ti.
https://www.chaosgroup.com/blog/profiling-the-nvidia-rtx-cards
In this testing, what they found is that the reported VRAM is not correct, and they actually do not know how much VRAM they have. However, they were able to render a test scene that exceeded the VRAM of a single card. See this chart, the test on the far right was only possible on 2080ti's in Nvlink mode enabled. Using the memory pooling mode reduced performance slightly, but not much. If you recall this exactly what Tom Peterson said about memory pooling. He said that it was possible, but the software had to support it, and there would be a performance hit. In video games nobody would want this performance hit, and so you will not see this in gaming with how Nvlink currently works. That's also why the people that cover gaming all say the same thing about Nvlink, that the memory does not pool. Not to mention no game uses more VRAM than what the 2080ti offers. That could change in a few years, if gaming cards had more VRAM you bet somebody would be taking advantage of it for their "Ultra" settings.
I have not yet been able to find anybody using Iray who has two Turings and Nvlink and has actually taken the steps that chaosgroup describes in order to enable memory pooling. But it should be possible, vray did not need any special update to use the feature. It just worked.
...however, the payoff with 2 cards is the increased VRAM allowing for larger more complex scenes to remain in GPU memory.
So now comes the thought at 2,500$ for a Titan RTX, or about the same for two 2080 Ti's, plus NVLink bridge For one power consumption would be lower with the single Titan (280w) compared with upwards of 300w for two 2080 Ti's and having 2 more GB of VRAM overhead. True, you wouldn't have double the cores to accelerate the process which would be the one disadvantage..
You will not get 22GB with Nvlink, that much we know. Some data must be duplicated for performance, but we do not know how much. I am going to guess that it is somewhere in the 16-18GB range, which is still a solid improvement. That is purely a guess, but I have been more right than wrong on these guessing games. I did actually say the Titan could ship with more VRAM. I was thinking it would be 16 myself, but 24 is logical. After all, the "CEO Edition" of the Titan V has 32GB, so that showed Nvidia was fully capable and willing to add more VRAM if they wanted to.
24GB is a big jump, and I think this may be a hint that the next generation to come (the one AFTER Turing), might finally see a top tier card with more than 12GB. Of course that is a long ways off, so it is silly to think about it now. But I think the 3080 and 3080ti will see VRAM increases.
The Titan will only be slightly faster than the 2080ti. But it will have a few other advantages. Titans can use TCC mode, which allows them to use all of their VRAM. Not only that, but TCC mode has been shown to improve performance on Windows machines in other render engines. Sometimes dramatically so. I also wonder if the Titan will get a faster Nvlink, which would translate to even better performance with pooling enabled.
Just think, you could Nvlink two RTX Titans and have over 40GB of VRAM for a cool $5000 or so.
...indeed, even with the data duplication hit, that would be more than enough for even my most epic of scenes.
Now if I could just get 6 out of 48 nimbers to match....
I just wish somebody would benchmark some 2080ti's with Nvlink for Iray.
...and Octane.
Afaik DAZ Studio Beta has not received any update that would add NVLink support.
All I can do is confirm what others posted.
- You need two identical RTX 2080 Ti (or RTX 2080) cards and one RTX NVLink Bridge.
Asus support recommended a NVLink Bridge 4 Slot for my mainboard to make use of the 16x slots.
(Afaik this will not matter for rendering but may help with some games.)
This will leave two SLI slots free between cards that require two SLI slots.
Slot 1&2: card one
Slot 3&4: empty
Slot 4&5: card two
Note that some RTX cards need more than two slots.
Be sure to inform yourself first if you need a 3 or 4 slot for your mainboard and card combination...
When the NVLink is detected properly in the Nvidia Control Panel it will be reported as SLI enabled:
- Two cards in NVLink and one addtional card for the display seems currently not supported with Windows 10.
The Nvidia Control Panel may still display all three cards but SLI cannot be enabled.
It showed the message: "The display setup is not optimal. Refer to the SLI Monitor configuration options."
- The render times in DAZ Studio are similar with SLI enabled or off. It was 1 second faster with NVLink /SLI enabled for the "Iray render test 2" scene.
- - -
I made some first tests with an experimental build of OctaneRender™ 2018.1 XB1 (aka Build 5.x) but will not perform more benchmarks at this point.
Any results shared now would be obsolete as soon as new test builds are released...
- - -
Does anyone know if there is a difference between the Asus and the EVGA NVLink Bridge?
All I was able to figure out is that the EVGA version costs 70$ more. (~190$ instead of 120$)
I had to preorder blindly a few months ago and choose Asus because that is the same brand as the cards and mainboard.
- - -
Did you follow the steps provided in the chaosgroup link? They got Nvlink to work without needing a software update.
The RTX 2060 is happening, with its "RTX" moniker, and it seams it might actually have some RT and Tensor cores after all. IMO the 2060 is a very weird card. The CUDA count is 1920, which is exactly the same as the current 1070. So factoring in some IPC gains, plus the fact it does contain at least a few ray tracing elements, the 2060 should be faster than the 1070 in every way. However, it will not come in a 8GB flavor, which is a negative. There are no less than 40 different SKUs of the 2060 in the works, and they offer everything from 3, 4, and 6GB versions, plus some of the even contain different kinds of VRAM. Some have plain old GDDR5 and some are showing GDDR6. There may also be some GDDR5X ones out there.
The 1060 has been one of the most popular Nvidia cards ever released. So it makes sense that they would offer a wide range of variants to meet different user needs. I think it is sad though that they are still offering a 3GB version of the 2060, that just seems silly in 2019. While many games are fine with 4GB, that 3GB is super tight and it will not be enough for some of the latest games even at 1080p. Hardware Unboxed did a nice video that looked at VRAM and found the 1060 3GB was already heavily constrained in some games. So a 3GB 2060 better is a very odd thing.
To make things even more interesting, there appears to be a 1160 in the works as well. Yes, 1160. This would seem to strip out the RT and Tensor cores, and this card might be very interesting. If stripping out the RTX features lowers the price enough, this could be a more compelling card for a lot of people who don't care about the RTX features (which seems to be a lot at this time.)
So it looks like we were all only half right in predicting the x60 would strip out the RTX features. It looks like there will be both! However so far the only indicators of the 1160 are in laptops. This is pretty interesting, and as for myself, I wouldn't mind such a laptop. I am looking to buy a decent gaming laptop one day, if the price is right.
Also, this is not really news, but did you guys know that eBay is offering various discounts on everything they sell? These discounts apply to GPUs, too, even used auctions. I just tried it to confirm. The current coupon is 15% off. So a $565 1080ti will drop to $480, that is a $85 discount with that coupon. The coupon has a cap at $100, so no, you can't go gang busters on a 2080ti, but you can get $100 off one if you shoot for it. These types of coupons that apply to everything might just be the best way to find certain GPUs on sale that might otherwise not happen. The 2080ti has not had any sales of any kind, and in fact its price has increased about $100 since its launch. The current 15% off deal lasts until the 27th I think.
I just so happen to have 3 RTX Titans laying around right now, and a RTX 2080 TI. So I thought I might spin up Daz 11 Beta and throw a RTX Titan into my rig and see the difference. First thing I noticed is that Daz did not use the GPU at all. It will only render on my CPU. So I threw my GTX 1080 TI back in, and kept the RTX Titan in, and Daz still would not use either GPU. Took the titan out, and boom, Daz worked fine on the GTX 1080 TI as it always has. Not sure what is going on here, but it would be nice to run two NVLINK RTX Titans for Daz. Currently using them for deep learning, and they are working great for that.