PCIe 3.0 Quadro Graphics Card on PCIe 2.0 motherboard - will it work?

Shortcut

Hello DAZ People!

I'm hoping someone can answer this technical question to do with hardware.

I want to upgrade my GPU because Iray renders in DAZ are always reverting to CPU only. I only have 1.5GB memory avaialble on my GPU which is plainly just not enough, and it doesn't even support OptiX acceleration. Without wanting to upgrade the entire computer, I was considering an 8GB Quadro M4000 card, as this is within affordable reach and not the latest and greatest in the Quadro lineup. However, as this uses PCIe 3.0 and my board is only PCIe 2.0, my questions mainly are:

1) What is the likelihood of compatability issues running this card?

2) Will I be able to use the full 8GB of memory on the card for renders?

3) Will I be able to use higher resolution textures without issue?

4) Am I approaching this in the wrong way? Is there another alternative to upgrading my graphics hardware and improving render times?

(My system hardware - ASUS board X58 chipset, Intel 980X hex core CPU, 24GB system RAM, Win7 64bit, Legacy BIOS year 2011)

Many thanks in advance if anyone could share their experience and advice :-)

philiphowe

Hi Shortcut,

 From what I have learned both in using Iray and how it works in general, especially in Daz Studio - the higher your card(s) cuda number, the faster it will render IRAY, but only if you have enough threads, or bandwidth on the motherboard and or processor. The fastest processors, I believe, are single or dual Xeons (not single icore7) running on an x99 motherboard, which apparently has more bandwidth than the newer motherboards like the z270, which are made more for gaming, not rendering high resolution scenes. 

So, if all that makes sense, and really I am not sure since the companies that actually build graphics rigs seem to be pushing the Quadro cards, I think because its what Nvidia wants them to do to say they are 'certified', then the cards to get are the much higher, and less expensive GTX 900 and 1000 series. I plan to have two 1080ti's (3584 cudas each or 7168 total cuda) and will try to get a smaller 2 or 3 GB card just to run the monitors and system stuff, as this is supposed to free up the actual IRAY rendering so the twin cards will run in tadem and, I assume, split the time in half. (or double the render speed.)  Again, the bottleneck will be the processor path, threads, and how many they can run through on the motherboard, as I understand it. I do a lot of animation via DS now and its actually right up there with Maya in terms of quality, but what a pain if they take more than a few minutes per frame and I have to wait a day to get 3-400+ frames, then post work, rerender parts for masks,etc. The 1080ti has a whopping 11 gigs of ram on board so my scenes should finally fit in total and maybe I can get back into Octane,too, with multiple figures moving around. Octane renders considerably faster for outdoor scenes, with some limitations. I prefer Iray for indoor lighting and artistic controlled scenes, but its much slower. At least it will render Iray shaders, while Octane either sees them as black or transparent, so I keep two sets of some scenes and figure clothing. 

Anyway, hope that helped. Oh, if you get two cards, try to get the same ram amount on each as Iray will use the amount of the smaller one, not the larger one. I'm writing to Nvidia, or calling them Monday to see if they can give me some straight answers about why they seem to push the expensive, but less cuda cores of the Quadros, instead of the GTX cards, even when they do speed tests with the Titan. Beats me.

kaotkbliss

A pci 3.0 card will work just fine in a pcie 2.0 slot, it just won't be as fast as being in a pcie 3.0 slot.

nicstt

kaotkbliss said:

A pci 3.0 card will work just fine in a pcie 2.0 slot, it just won't be as fast as being in a pcie 3.0 slot.

Not sure how much difference that will take as the rendering on the card takes place on the card; if there is more than one card, then it might make more of a difference, but it is really down to how much data is shifted between cards; it is presumably just so each knows what is being done by the other?

Also, reducing the number of times Studio updates the render might also speed up renders, albeit (perhaps) by only a little.