Another gpu decision.
droidy001
Posts: 282
in The Commons
Really just thinking out loud. If you would like to chip in please do.
I have outgrown my 1050ti. 4gb is no longer big enough.
I've learnt so much about daz recently and my renders have got so much better (in my opinion anyway). Although I do think I'm still scratching the surface.
I'm finding that I have to not only delete everything not in shot and reduce textures, but also in some cases use the geometry tool to hade part of buildings to reduce poly count.
On a simple scene I can just about get 2 characters in. Anything slightly more complex and I'm down to 1.
Cant really afford a new card but will probably get one anyway.
Choices.
2060 super as a 2080 is out of the question this makes more sense than a 2070. The 2070 is just not £100 better.
Or a used 1070 at around £200. That's more like the price I'd like to pay.
Is there going to be a huge difference in performance?
Waiting a bit longer for a 1070 render over the 2060 is not a huge issue.
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
See this thread. There are lots of directly comparable benchmarks between all of these cards there.
The RTX tech is more than real time ray tracing. It greatly speeds iRay rendering. If you can swing a 2060 super, or better, it is a far better choice.
Yeah. The general rule of thumb (at least with recent Nvidia cards) is that a latest gen nth-tier graphics card is roughly equivalent performance-wise to the previous gen graphics card one tier above it (ie. 1070 = 2060, 1080 = 2070.) ETA: (although looking back at the benchmarking data collected in that thread so far, the performance jump from 10XX to 20XX generally looks to be closer to two tiers rather than one.) In any case, based on my experience with a Titan RTX, the benefits you get from RTX enhanced rendering generally don't manifest significantly unless you are dealing with scenes too large to fit into a lower tier card's framebuffer anyway. So yeah - I'd agree that the 1070's +2GB vram seems like a more useful option in your case.
As for fitting it in your case - I'm a big fan of creativity in computing.
PS: Any chance you could contribute a benchmark run or two to that thread using just your 1050 for rendering? The only 1050 numbers we have so far come from its stripped down 2GB VRAM mobile version - not the full card.
Totally forgot the 2060 SUPER has more ram. In that case I'd say stretch for a 2060 SUPER rather than a 1070. Especially with today's inflated 10XX prices.
I don't mean to derail the thread, but, for the sake of not making another "GPU decision thread," I'd like to pose a question here.
Maybe it might help the OP's decision, IDK, but I hope.
I'm also looking at upgrading my graphics card. I've been looking at the RTX 2070 Super and the RTX 2080 Super.
Both are 8GB, but, clock speed aside, the only noticeable difference to me is in the # of Cuda cores.
The 2070 Super has 2560 Cuda cores while the 2080 Super has 3072 Cuda cores.
At 8GB each, how much of a difference does 500 or so Cuda cores make?
Or, going from the 2060 to the 2070, how much of a difference is there between 2176 Cuda cores in the 2060 and 2560 in the 2070?
I've not studied the benchmarks too much but it would be interesting to know if number of cores scales up evenly. As you say clock speed and memory speed play a part but cuda count seems to be the biggie.
So it has 16% more cores, does that mean at same clock speeds It should be about 16% quicker at renders?
CUDA "cores" operate in parallel so they should scale very directly. There might be a small difference if you're saturatuing the VRAM bandwidth or something of that sort but otherwise it should be close.
The real question is: is the performance difference worth the $200 (US) price difference. The 2070 Super is a very nice card and probably the best bang fo your buck in the price/perfmance matrix. You won't be sorry with either card. I know, not too helpful.
Keep this in mind when looking at CUDA core counts, more is not always better if the card is an older card. CUDA cores are only equal when comparing the same generation cards. The easiest example of this is comparing your 1050ti to a GTX 680 that released a few years prior. The 680 has 1536 CUDA cores, that is exactly twice as many CUDA core as the 1050ti. So at first you might think, whoa that sounds fast. But it isn't. The 680 is from 2012, and when tested the 1050ti is actually as fast or even slightly faster. And it is important to note that the 680 was a flagship product in 2012.
Here is a fun article exploring the 680 VS a 1050ti. https://www.techspot.com/article/1588-geforce-gtx-680-revisit/
So CUDA cores are not the full story. Basically each new generation features more optimized CUDA performance than the previous.
RTX features the fastest CUDA performance yet, but there is also the RT factor. The effect that RT cores have on performance is a bit of a wild card. You still get the CUDA performance, but now you get RT on top of that, its almost like a bonus (depending on how you view the price of the card perhaps). The improvement that RT delivers on top of CUDA can vary wildly, depending on your scenes. Generally it seems that more complex geometry lends itself to seeing better RT benefits. So the more complex your scenes get, the better the benefit of having those RT cores becomes. There is a bench for this as well, which uses dforce strand hair to generate the geometry that can show this benefit.
Unfortunately we cannot enable or disable the RT cores in Iray. This makes directly comparing their impact a little harder than one would like. We can compare it to the previous version of Daz Studio. which has the previous version of Iray before RT core support was released. It might look like a simple way to compare, but Iray was quite drastically overhauled in this update for RTX. However, comparing the Daz 4.11 render times to 4.12 with RT support does give us a ball park idea. So you can compare these results, along with yours.
I have said before the 2060 Super is a great deal, it having 8GB is a big reason for that. The 2060 Super is very close the original 2070 in performance. Then you have the 2070 Super, the 2080, and the 2080 Super.
Yeah...Nvidia really goofed up the naming scheme this generation. The 2070 Super is about like a 2080. The 2080 Super is faster than both. (But not as fast as a 2080ti, which is its own special kind of beast).
If money is a concern, I would strongly suggest looking at the 2060 Super. Its already decently fast and the RT cores can make it even faster depending on your scenes, and it has the 8GB which is practically standard these days in good cards. It will be a lot faster than a 1070.
Went for the 2060 super. Will benchmark later. Thanks to everyone who chipped in with advice and info.
Edit:
Bench done. I would have been more than happy with the improvement of a 1070 over my 1050, but effing wow. The 2060 Super blows it out of the water. around double the speed on the benchmark.
My 2070 renders indoor scenes at roughly the same rate as my 1080ti. The two together...
I'm happy it worked out for you!