I have had a GeForce GTX 750 Ti GGDR5 since near its release because it was cheap and popular and has worked. However, in recent months I have developed more complex work on Poser 11 that required me to upgrade from 16GB of computer memory to 32GB (4x8GB) DDR4 2133s. I have a M-ATX motherboard with PCI-E X16. Intel B150 chipset. i5-6500 CPU.

I think it might be time for a new GFX card.

I don't want to spend high-end money at the same price as the PC itself. Mid-range is fine.

I noticed some software programs allow me to use the card for rendering rather than the CPU and I have been reading that CUDA enabled cards are better for Poser?

So if anyone else there is doing lots of stuff like this with 3D files that go over 16GB of memory use and towards the 32GB and know about cards to push those sort of files around on the screen with decreased latency etc., I would like to know.

Thanks

Poser pro 11 can off load rendering to a Nvidia GPU, any other version of Poser including poser 11 use CPU only. GPU can render much faster for typical materials but can struggle with SSS, there is also a limit on texture size, consider the onboard graphics card memory needs to hold the textures and tends to be less than system RAM. For composition of a scene Poser uses OpenGL, I'm not an expert on OpenGL so take the following as informed speculation based on observation and a bit of Googling - whilst Poser can use multiple CPUs to process the various objects in a scene prior to sending it to OpenGL it appears only one process can update OpenGL at a time - if this is correct the speed of the GPU will have minimal impact on how fast Poser responds to user interaction when editing a scene as its limited by the CPU core speed - this matches the experience I have had adding a GPU. Sorry to be a bit negative. I'm very happy with investing in a GPU but it doesn't solve every problem and sometimes its necessary to switch back to CPU.

Just to expand on GPU memory use, I can render a scene that uses 10GB of RAM on a 8GB 1070, there is some kind of optimisation going on but if you're planning to do 16GB+ renders I'd recommend you do more research on the limitations just in case you hit them.

The 1000 GTX series seem an affordable stable way to go.

I am thinking if Quatro or RX is what I should do.

If by RX you are referring to nVidia RTX series, this doesn't work at present for rendering in poser. If you are referring to the AMD RX cards, these also don't work for rendering with poser. GTX 1080and above seem to be running out of stock, but you should be able to get 1060 or 1070cards reasonably priced. I'd agree with ironsoul, that for general responsiveness, single core clock speed of the processor seems to be the limiting factor. FY! rendering with superfly with identical render settings other than bucket size I have found my gtx1080 card about 20% faster than my 14 core 28 thread 3.1GHz i97940x processor. However it might be possible to get equivalent quality in less time using branched path tracing with the cpu, with you can't use with the gpu

I have these choices of cards basically...

https://www.msi.com/Motherboard/support/B150M-NIGHT-ELF#support-vga

As Nails said the Gen 3 10x0 line is a good starting point. In general the range is ordered in performance so a 1050 is lowest performance up to the 1080. The TI version has additional improvements so a 1070ti sits between a 1070 and a 1080 in the performance ranking. Its not quite that simple as some manufacturers provide overclocked versions. Above the 1080i there are the Titan and Titan X cards.
In addition to price the other major considerations are size (some cards have smaller versions) and power consumption, the technical details for a card should indicate requirements. Superfly is a version of Cycles so you can get an idea of performance by checking benchmarking websites or Youtube PC builder channels (eg Linus) as many of these include a Blender test. I'd suggest two important qualities to look for are how effective the cooling system is (to avoid thermal throttling) and if the card is noisy.

Not sure why I added the second bit on last post, got a bit stuck in a line of thought. One area of discussion when Superfly first came out was whether to go for Gen 3 or higher end Gen 2 9X0 series, it has less memory than the 1070 and 1080 but still a potent card. The link is for a Cycles 2.79 benchmark comparison but gives an idea of the relative performance Blender v2.79. Question is kind of complicated by not knowing when/if the RTX series will be supported.

I am currently thinking about a MSI GeForce GTX 1070 Ti GAMING 8G, 8GB GDDR5

I have a Zalman ZM600-GSII 600W PSU.

I'm running a 1070 + 6700 + 64GB on a 500 W PSU and not had any problems. Looking at the MSI charts the Ti version is only 10W more.

I have ordered the MSI GeForce GTX 1070 Ti GAMING 8G after reading as much as I could on the topic and will let people know if it works.

I have a GTX 1070, 8 GB and I am very happy with it.

So basically I got the card.

I had to get a DVI-D cable to HDMI cable though as my old GTX 750 ti was using a standard DVI cable. I have two monitors and use a HDMI to HDMI.

I uninstalled the old Nvidia Drivers.

I installed the MSI with some struggle. I had to remove a blu-ray player... so my system no longer takes optical discs, but that doesn't seem like a big deal these days.

I loaded up some files I am using that take about 8GB of memory to load and it will go higher. Seems I am moving around the viewport quicker. I did a render using just the card and it seems to have gone through it quite fast. I am rendering 2000 x 1600 @ 150 dpi. Basically doing 3D models for photoshopping.

So far so good. Obviously things are better than the 750 ti and I hope this will get me through these bigger projects.

So I didn't go with the professional graphics cards and am probably happier I didn't have to spend so much. Seems this 1070 Ti is working out so far.

Thanks all.

So Task Manager -> Performace Tab. Go to GPU. Select Compute_0 from the drop-down menus on one of the graphs.

Render Superfly. Set renderer to the GTX 1070 ti. Compute_0 graph is hitting 98% to 99% and is rendering nearly twice as fast as my Intel i5-6500 3.2Ghz

My board can take an i7, so that might change things if I decide to update.

BatmansRenderosity posted at 2:10PM Wed, 12 December 2018 - #4341648

So Task Manager -> Performace Tab. Go to GPU. Select Compute_0 from the drop-down menus on one of the graphs.

Render Superfly. Set renderer to the GTX 1070 ti. Compute_0 graph is hitting 98% to 99% and is rendering nearly twice as fast as my Intel i5-6500 3.2Ghz

My board can take an i7, so that might change things if I decide to update.

Are you using Poser Pro? And is the card the only graphic capability in your PC (no Motherboard onboard chipset)?

Azpir8king posted at 7:59PM Wed, 12 December 2018 - #4341661

Are you using Poser Pro? And is the card the only graphic capability in your PC (no Motherboard onboard chipset)?

Yes and Yes.

If not already done try setting the bucket size to 512 to see if that improves the render time.

I did and rendering using the card it didn't change much in terms of time. Roughly the same.

Ok, guess it's more pyschological with the smaller tile number. I get around a 10% improvement (example 796s @ 256 and 689s @ 512)