Advice on GPU upgrade, please? :)
Illtm
Posts: 15
Hi guys!
Quick question...
My current setup:
3.2 i5 quad core CPU, 32GB RAM, spin disk.
1 GeForce GTX 980, 1 GeForce 980 GTX ti.
I am thinking of replacing the GTX 980 with a new GTX 1070 ti or perhaps go for a used GTX 1080 ti, as these are becoming a bit more affordable these days. But this comes with a risk involved, of course (potentially used extensively for, say.. mining etc?)
The core of my question is: Would replacing the 980 with one of these two cards provide a significant performance increase? (Edit: I'm leaning towards a new 1070 ti..)
I usually don’t create scenes for making final renders, rather I mainly have a screen for iray preview and use the scene for «realtime» reference for 2D work. I still need good image (photoreal) quality, but would like higher speed at a moderate price range.
Any advice would be greatly appreciated. :)
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
They both would improve performance if you pull the weaker card and install the new card alongside the 980, assuming you can fit your scenes in the 980.
Okay, thank you very much! :)
I just realized that I was unprecise in my original description of one of the cards (it's now corrected) and I apologize for any confusion. Would the change of details your assessment in any way?
Pull ther 980 then. The 980ti and either the 1070ti or 1080ti will be faster. Generally speaking Nvidia cards match the previous generationscare one tier higher. So the 980 ti matches, roughly, the 1080 and the 1070ti is just about a1080 so by pulling the 980 and adding a 1070ti you get roughly 2 980ti's performance and the 1070ti has 8Gb which would let you render larger scenes, albeit on only one card. The same with the 1080ti, except it has 11Gb.
Alright.. that does it - I'm now a lot closer to making an order for the 1070 ti! You just confirmed my (vague) suspicions and then some. Thanks a bunch for your help! :)
You might want to consider getting a RTX 2060 Super instead. Both cards are about the same price and have 8GB of VRAM.
The GTX 1070 ti might be a little faster rendering in Studio 4.11 and older but in 4.12 with full RTX support the RTX 2060 Super will render quite a bit faster.
Oh man.. yet another option to consider? ;) Right.. so it looks like the RTX 2060 has quite a bit lower cuda count than the GTX 1070 ti, but I guess there are more factors involved in its rendering performance then? I've snooped around a bit now, and it looks like the 2060-pricing isn't too bad at all compared to the previous gen. To make matters even worse, the 2070 Super isn't all that more expensive than the 2060 Super, and has an additional 25% cuda cores, so... damn it! This is driving me nuts!! :)
You shouldn't compare CUDA count across generations of cards.
Honestly at this point unless you are looking in the used market for a "cheap" GPU (or an amazing clearence deal), the RTX SUPER GPUS (and the 2080 ti / Titan RTX) should be the only new in box GPU purchase you should be looking at for Iray rendering.
To add to that, besides CUDA core count you have clock speed, memory speed and even architecture improvements that affect the rendering times.
This is exactly why I came to this forum hoping for advice. I've finally decided on taking the jump up to one of the new RTX Supers. Thanks a million for your help, guys! :)
Cheers!
VRAM on a card isn't to important in today's world, geforce 10s can use virtual memory, and iray will use as much as it need's, or upto atleast a terrabyte i guess, a single 1070 on windows 10 really has 16373MB of VRAM when virtual memory is set to system managed, and just sitting idle on your desktop, while rendering a decent IRAY scene, virtual memory use will go upto 50+ GB
you can find out how much VRAM you have by going to PC Settings/System/Display/
scrolling to the bottom and clicking on the Advanced Display Settings text link and then the Display Adaptor Properties
No, Iray will not use virtual memory. Whoever told you that is wrong.
the video cards use virtual memory which is part of server rendering and being able to create hollywood movie's, you normally need a nividia titan or amd vega card for, you can download HWINFO and watch the realtime virtual memory use when rendering with iray, I don't know about other card's, but the 1070 uses virtual memory, while physical RAM does'nt get used to much
you can just buy a PCI-E 4.0 motherboard and drive and save $2000 not having to buy physical RAM
Iray does not use virtual memory. Who ever told you that is wrong.
Other GPU based render engines may make use of main system RAM for rendering, but Iray does not work that way.
GPU rendering is limited to physical VRAM on the card.
CPU rendering uses system RAM
There is no cross over while rendering (except over NVlink connections).
virtual memory is the commited virtual memory when your computer is sitting idle on your desktop and displayed as the shared system memory in the display adaptor properties panel, in the attachment image which gives the 1070 a total of 16GB of VRAM
when you start rendering, and virtual memory is set to system managed, the virtual memory use and size when rendering with IRAY will keep increasing until your render is complete or you run out of drive space, and IRAY crashes
There is no image, but as others have said that is not how Iray works when using the GPU.
When Iray activates a compatible GPU for rendering, it loads all of the required assets for the render job into the GPU's physical VRAM.
If the GPU has enough available physical VRAM to fit the scene, it renders the scene.
If the GPU does not have enough available physical VRAM to fit the scene, it does not render the scene and it is notated in the Daz Studio log file.
On another note. The Iray scene does not increase in size as the render progresses, all of the needed VRAM space for the render is reserved at the start of the job.
If your are seeing your virtual memory increase in size during Iray, that may be that you do not have enough system RAM while preparing the job for transfer over to the GPU. This could also be happening because you have the CPU set as a render device (The CPU Iray renderer can use all of the virtual memory that it needss for the job).
it will render the scene, when there isn't enough physical VRAM available, as virtual memory will be used
it is todo with CUDA and unified memory > https://devblogs.nvidia.com/unified-memory-in-cuda-6/ which normally requires an enterprise card for supercomputing, my guess without going back to all the nvidia conferences, is that is was added to geforce 10 cards
I'm sorry to say that full support for Nvidia's Unified Memory is currently crippled at the OS level on all operating systems except Linux. As things stand, the only enhanced memory benefit to be had with a GeForce 10 series card or newer on current Windows is the ability to selectively swap data between system/virtual memory and GPU VRAM - NOT virtually extend the capacity of data storage directly available to the GPU beyond the physical size of VRAM located on the card itself (what Nvidia calls GPU memory oversubscription.)
Add my name to the list of people suggestion something like an RTX 2060. Based on the numbers I've been seeing here, the move from a GTX 980 to a 1070/Ti doesn't look worth it imo. Last I checked 1080/Ti's were ridiculously overpriced compared to even current RTX cards. A 2060 or higher stands to give you quite a significant boost. Don't think that's the case for the others.
nvidia shouldn't buy dell's, on my system virtual memory going up above 50+ GB while rendering with no other background tasks, works fine on my rrandom junk off ebay PC, along with SLI, which makes rendering with a burn't out AMD FX 8300 @ 4.6ghz 100x faster then a single card with optix disabled in daz, and SLI enabled in the nvidia control panel