What is the cheapest nvidia card for iray?
Isazforms
Posts: 210
Hello friends I would ask in his experience which is the cheapest nvidia card you can buy to do a decent job with iray . I do not need a high speed ( at least faster than CPU) . My scenes are simple , even there are two characters and a full loaded scenario . I not ask for the minimum requirements, such as those I read. My question is for those who have experienced less powerful card the GTX 970 n vidia
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
A 740 with 4GB is about $100-ish. It has only 384 cores, but the RAM is enough to hold most instances of two characters and scenes.
It's important that you get the version with 4GB. There are some cheaper ones out there with 2GB. Spend a little more to be sure your scenes will fit into the card.
Even though it only has 384 cores, that will be an improvement over CPU only renders, for time.
The best budget card for Iray is a Nvidia Tesla M2090 about $100 US dollars on ebay. This card comes with 512 cuda cores and 6gb of DDR5 video ram. The downside of this card is that you need to run your video on a ati card because you cannot run a Tesla and a Nvidia desktop video card at the same time. If you have a small ati video card or an AMD/ATI APU combo processor you can run both an ati video card and a Tesla at the same time.
The only issue for that card is there can be significant driver conflicts with GeForce Cards and NVIDIA does not recommend mixing Tesla (or Quardo) cards and GeForce Cards. Since it has no video outputs, that means you should also have a Quadro card to drive your monitors. This is not the inexpensive option and you can usually get a pair of GTX 980Ti cards for the same price as that Tesla and a newer Quadro card. (And have some change left over.)
Also note that if you later decide you want more speed, you can, with most mother boards, add a faster card later and then relegate the GT 740 to drive your monitors, which will free up the faster card to just do renders.
One other option, with the Tesla...if you have an onboard (usually Intel) graphics, they seem to play nice...
Although the Geforce and Tesla are supposed to get along in Linux.
Thank you very much for your answers. It have helped a lot.I was not aware of the tesla cards. Neither knew he could have two cards in one system, handling different things. From what I've seen, there are not many 4gb cards in the midrange and low average. :(
What's your absolute maximum budget?
Best value it the 970; that many consider to be the cheapest, due to it being the best value; you can spend more or less and get better render speeds than a CPU; you also need to consider your scene sizes, which gives an indication of memory you will most likely need most.
I just got a 970 and it renders in 20% of the time that it used to with just CPU.
So a small 2 min render comes in at about 24 seconds.
I'm very pleased with it.
The 4GB GT 740 runs about $100. A GTX 960 (4GB) will run about $200. A GTX 970 (4GB) runs about $350. The GTX 980 (4GB) runs around $530. A GTX 980Ti (6GB) runs about $650 and is only 3% slower than a Titan X (12GB) which runs around $1000. Right now, the best bang for the buck appears to be the 980Ti.
I am going to jump and and say that a used 780 Ti off ebay is the biggest bang for the buck at the moment. 2880 CUDA cores (Kepler) trending @ about 280 USD on Ebay at the moment in the US.
The cheapest I got one was 225 a little while back. I have 12 in my render farm, but I am looking to drop down to 8 (making space for Pascal!)
On the new market... the 980Ti would be my pick, going along with that DAZ_Spooky is saying.
I won't buy a used video card as you don't know what it was used for before, and if it was overclocked and rendering the life expectancy will be short. :) Having said that, if you go this route make sure you get the one with 6GB not 3GB of Video Ram. :)
(The card has either of those configurations.)
I wouldn't discount the 780ti so quickly. It can actually be the perfect card. Some 95% of my renders are single figure against a transparent background. Everything is then comped in Photoshop using elements with alpha. So, even though the 780 has only 3GB RAM, it enough for a single character with hair, clothes, props. In fact, it can be enough for two characters, though here, I'll typically render the characters separately using canvases.
I know there are users who set up elaborate scenes with set, multiple characters, vehicles, special effects, and more, and these can easily use up more than 4GB. But I have to wonder if a lot of pro artists work this way. A key benefit of Iray is that you don't have to render everything at once. If you're looking for flexibility in post, you render separate elements using canvas nodes, then finish in Photoshop or other compositing tool.
Yes, that's a nice card. I thought you meant the 780ti being sold used on eBay which, despite having only 3GB, is still pretty decent for lots of render types. The 6GB version would be terrific, but Santa says I've been very naughty this year, so fuggedaboudit.
I bought nvidia gt 740 just to get out of step at this time. And I discovered that render cpu and gpu have different results in the same scene. It is a lie that you obtain the same image with CPU and GPU. A low cost gpu mixed with cpu is faster than only cpu.
GPU and CPU shopuld give the same end result. However, if one times out and the other stops for covnergence then there may well be a noticeable quality difference - that can be fixed by increasing the time allowed.
Nope, have different results in my settings. Transparencies and reflections. I really check it...
Can you post images to see the results?
A small gif is the only I have for the moment, but you can see the diference in the eyes. Same scene, same lights. You can check it for yourself if you do not believe me.
And it stopped rendering by reaching the convergence limit for both CPU and GPU? That would be worth a bug report.
I believe Richard hit it on the head.
For them to be the same they have to stop on the same condition.
There are three stop conditions. (1) Convergence, (2) Time and (3) Iterations. It stops at whichever occurs first. For short renders it is usually convergence. however for comparison renders, such as this one, it usually ends up being Convergence or Iterations for one and Time for the other.
Both images at 95 % convergence. In hair transparencies there are some differences too. You can do ur test If you want.
They ignored me.
You put in a support ticket?
If you can hold out for a few more months, I would suggest doing so. Both nvidia and AMD are set to release major upgrades in 2016, and the competition between them should benifit us consumers. They are using a new kind of VRAM that is twice as fast as GDDR5 and allows for higher capacity, both of which are vital to iray. It could be a very good year for pc owners all around.
Even if you don't go for a brand new gpu, the new releases will certainly cause price drops for all the current cards available. I have an old 670, and that 2gb is killing me. But I'm going to hold out to see what the new cards will do. I believe that HBM2 based nvidia cards will be a major game changer for iray based applications.