Nvidia Titan Z for Iray, A good or a bad Idea ?
JPiat
Posts: 70
Hello;
I trying to figure ou whethert to get a Titan X (Pascal) or a Titan Z for my system. I already own a GTX 970 and a Titan X,
From what I see from the drivers specification, the Titan Z is a 700 serie... but 5760 Cuda and 12 Go...
Did anyone use a Titan Z with DazStudio, any advice ?
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
5760 Cuda cores and 12 Gb?
that's a no brainer
just go for it
Titan Z (released in 2014) vs Titan X (2015), very interesting
Titan Z: 5760 Cuda Cores, 705Mhz / 876Mhz, 12GB
Titan X: 3072 Cuda Cores, 1000Mhz / 1075Mhz, 12GB
From this : http://gpuboss.com/graphics-card/GeForce-GTX-TITAN-Z
If computing is a indicator for IRAY (no evidence for that). I will go for a secondary Titan X Pascal GPU... Much less expensive.
Don't compare CUDA cores from different generations; it isn't a relaible method.
Why not? http://furryball.aaa-studio.eu/support/benchmark.html
I thought the Titan Z was two GPUs, each with only 6GB of RAM, on a single card - the original Titan X was the first in that line with a full 12GB.
I prefer such sites as AndandTech and TomsHardware; there are others.
But like I said it isn't reliable; you need to go and investigate. It is easy to find links that differ from others.
And then there is what Richard said, they share the 12GB.
Is that just one result for the 780 ti? 4 cards? Well it says one result, so statistically significant it is not.
And it's a league table; folks love being at the top. Someone asked something similar, and provided a link showing the 780ti is better; I provided about 3 or 4, and they were legitimate sites; not that in this case I question the sites, only the results from such low numbers. Even the Titan with 20 results is not expecially significant, but certainly more likely to offer an idea.
Take the 980ti and the 780ti using three cards; this is not a fair comparrison, as it compares 6 results with 1. We can not reach a conclusion with such low numbers, but especially when the 780 is only one result. But if you want to reach a conclusion, then the 980ti when used in a x3 combo is better. Personally I don't. You were linking so that suggests to me you do.
So I stand by my original statement, it isn't reliable.
IIRC you are correct.
Also, from what MEC4D has posted in other threads, it seems that when mixing cards, the faster cards match the speed/performance of the slower cards. So the Titan Z would reduce the performance of your Titan X.
Exactly the situation I'm in now. The Titan Z is very tempting. I know the clock speed is slower and the architecture is older, but from everything I've read the issue with Iray is number of CUDA cores and memory, not clock speed or architecture. So, you can get a Z for around $800 these days, same with a Titan X, same with 2 980 ti's. 2 980's give you more CUDA but less memory. I guess the question I need answered, is what can't youu do with 6 gig of VRAM? Right now, I run a 780 ti and a 970. So for both to work on one scene, I'm limited to 3 gig VRAM. 2 980 ti's will double that, and give me about 1200 more CUDAs then I'm using now. Of course, I would have a ridiculous number of CUDA's if I just kept the 780 ti and added a Titan Z, and if the scene size exceeds the 780 I'd still be working with more then 5000 CUDA's. Thoughts?
If one of you guys does upgrade to the monster Z, can I have your old GPU :P
Well, I didn't have the nerve. Thought if it was the thing to do, I'd see more support for it. I went with 2 980 ti's. Just hope the VRAM is enough.
No, no, no.
CUDA cores are only good to compare within the same generation.
I will for instance generally take the same CUDA or a few less on a newer architecture card versus an older one.
No trying to be argumentative or anything, but what hard data do you base that on?
In my case, experience. I have a long-running render scene that takes about 26 minutes with both my 980 TI and my 1080; the 1080 alone does the scene in about 1 hour 3 minutes and the 980 TI in about 57 minutes. So - 6 minutes slower and noticeably less power on the 1080.
Hi there
GTX Titan Z is great card but essentialy this card is SLI GTX Titan Black or GTX780 6GB on one board,due this not sure if I would go with that card as this card is based I think on Kepler,although Kepler cards are still pretty potent cards in rendering as I've used GTX 780 in IRAY and on other renderers
In all cases if you will go like any two cards like GTX 980Ti or GTX Titan Z your scene cannot be higher as 6GB,Titan Z will not give you 12GB worth of VRAM but only 6GB,same as GTX 980Ti,if you will have like in my case GTX1080 and Titan X,you will have not 20GB VRAM for rendering,but only 8GB VRAM for rendering,simple NVIDIA or CUDA rather doesn't support shared memory like OpenCL 2.0
But still think if budget allows two GTX 1080 or GTX1070 will be better choice like running two 980Ti,right now IRAY is not very much optimized and hopefully in coming moths they should optimize this bit further,running like Titan X and GTX1080 and in some renderers GTX1080 is already few times faster
Poser Pro 11 render times:
Titan X has rendered that image at 29.1 minutes(1746 seconds)
GTX1080 has rendered that image in 8.70 minutes (525 seconds)
GTX1080 with Titan X has rendered that image in 4.60 seconds(281 seconds)
Hope this helps
Thanks,Jura
Ah. I didn't think the Z was just 2 Titan 6 gig slapped together, giving Iray only 6 gig. Well, I wanted 2 cards and 2 1080's at this point is more than I wanted to spend. I more or less like using slightly older tech because you get more for your money. I went with 2 980 ti's, which should significantly increase my scene size as I'm basically using 3 gig now, and should increase my speed somewhat. Should tide me over for a year or so, then I'll upgrade to 2 pascals when the next faster thing comes out.
Tough one to link with just a link or two.
Consider that newer generation cards, out-perform older ones, when comparing the same of very similar CUDA cores. It takes a lot to look at the different figures, in time I mean. Compare within generation, then more cores is better; clock speed makes some difference, but that is much more of a problem when trying to decide how much.
The 980ti has 2816 cores; the 1080: 2560. It's faster than the 980ti from the couple of render posts I've seen, and I expect that to improve as time goes on, as drivers are tweaked; presuming they do
. There have been odd reports (iirc) of the 1070 being faster than the 1080; no idea how much truth there is in that, or if it's down to bugs not yet sorted. I believe the 1070 is faster, but can't be certain I am remembering correctly.
But things are still in beta, so I'm still waiting to see; i'm not planning on upgrading, Nvidia charge too much, maybe AMD Ryzon processors will be a viable alternative next year.
The 1080 is slightly faster than the 980ti even though it has less cores, it manages do this becuase of it's higher clock rate. If it could out perform the 980ti at the same clock, that would be something.
Well, I can tell you that my set up right now is no slouch in the time it takes to render, I just run up against scene size limitations. I'm going to be running 2 980 Ti's, should get them tomorrow. Then I'll see if the speed is significantly faster than what I have now. Didn't want to for 2 1080's, or the soon to be released 1080 Ti. Just too much money right now. The prices drop pretty fast when the new ones come out, so I could see upgrading to 2 1080ti's for example in about 1 1/2 years.
Yes there,Titan Z is or are essentially two card slapped together which will give you same as with 980Ti VRAM.regarding the 1080,IRAY in current state is not optimized "enough" for 1080,maybe in coming months we will see better optimization,but I don't think we will see massive jump,have look on Octane which in latest beta supports 10xx series but still performance against the older generation(Maxwell) is still not big as many would think so,its faster but not 2 times faster there
With two cards you will have faster renders what I know and what I tried myself
Hope this helps
Thanks,Jura
Hi there
Right now GTX10xx series are disappointment in rendering,many people expected bigger gains in rendering speeds,I expected bigger gains in render speeds mainly in IRAY,what I didn't expected GTX1080 render speed in Poser 11 where is around 3x faster than my Titan X SC
Yes agree we shouldn't compare older and newer generation cards,but we must compare this as we do renders and we are expecting faster speeds with every new generation cards,IRAY and mainly CUDA 8 in current stage is still in beta and due this we are not seeing faster renders with new generation of Pascal
This I saw too,when GTX1070 is faster than my GTX1080 or my Titan X is faster than GTX1080.right now I would wait there if you are planning to upgrade,as Nvidia will be releasing Ti and AMD VEGA and this should bring prices a bit down for newer generation,Nvidia is still stuck at GDDR5X and AMD is going with HBM2 memory and their new CPU we will see,hopefully price will be in $500 USD mark,as Intel charge too much for their 8 core CPU or 10 core CPU
Interesting times ahead at least
Hope this helps
Thanks,Jura
Talking to a friend today who works with computers next year Nvidia are releasing the GTX1080TI and will supposedly come with either 10 or 12 gig Vram.. Here is an article about it.. http://www.hngn.com/articles/218039/20161121/nvidia-geforce-gtx-1080-ti-vs-titan-x-leaked-specs.htm
And another with supposed comparisons between the 1080 and 1080TI.. http://gpuboss.com/gpus/GeForce-GTX-1080-Ti-vs-GeForce-GTX-1080
And a supposed comparison between the Titan X and 1080TI.. http://gpuboss.com/gpus/GeForce-GTX-TITAN-X-vs-GeForce-GTX-1080-Ti
People keep comparing the 1080 to the 980ti, but that's not the correct comparison. Compare it to the 980. It's a signficant boost. When the 1080Ti comes out, THEN compare it to the 980Ti.
Also remember that those gains also come with double the memory (980 had 4GB), More CUDA cores (980 had 2048), a faster clock (980 had a base clock of 1126 MHz), AND draws about 40% less power (980 uses around 300W at full load). That's a BIG increase in power and efficiency.
Even when comparing to the 980Ti, most are comparing to heavily overclocked versions. The 1080 STILL has 50% more memory, faster clock, and draws a LOT less power, and with all that, still manages about 10% more performance without significant overclocking.
If/When the 1080 Ti comes out, then you'll have something you can properly compare to the 980 Ti. After all, look at the comparison of the 980 to the 980Ti...... even just looking at gaming benchmarks, there was only about a 10-15% jump in performance. And if you compare the 780Ti to the 980? About a 15% jump in performance. The 780Ti drew around 375W at full load, the 980Ti draws about 390W. And these are reference boards at stock clocks.
The 1080 is an incredible card. And a lot of the big improvements are in areas they haven't figured out how to leverage for things like Iray.....like SMP, Ansel, and others. Great for Gaming and VR......no one knows how well Iray could perform with SMP being used creatively......because nVidia hasn't done it yet.
(oh, and when it first came out....the 980 had an MSRP of $550, the 980Ti had an MSRP of $650 at release, and the 780Ti? It was $600 MSRP at release. So the 1080 only cost $50 more at release than the 980Ti. So for $50 you get ALL those added bonuses, and performance gains as well.)
Well... I had not bought a Titan Z, but a second Titan X maxwell and finally a third one on Ebay... Two Titan X maxwell have a classic design, the last is watercooled. It's not a choice but a obligation. I can plug 3 PCI Express card on my (cheap) motherboard. The Titan X on the center slot can't be cooled by the fan : there is no enough space.
Is it fast ? Yes. Mec4d has posted several benchmark for this configuration. The down side is the loading time from Daz3D to the GPUs. 2 or 3 minutes for complex scene, but I have a gamer motherboard, not a professionnal one with X99 Chipset and dedicated path for the PCI Express.
The reason for the comparison really boils down to cost and what's available. I suspect when the 1080 Ti comes out it's be around 800. My guess, it won't be cheap. So I'm really comparing cost effectiveness. I just got 2 980 Ti's for 800, and that should render faster than either 1 1080 or a 1080 Ti, at least based on everything I've read.
Tech sites are guessing that the GTX1080Ti will sit between the current GTX1080 and the Titan X in terms of price.. But one difference between the GTX1080 and the GTX1080Ti is that the GTX1080Ti will use a modified GP102 chip that is in the Titan X where as the GTX1080 uses the GP104 chip.
There's quite a lot of in between room there. I saw a site that said they doubt it will be under 1000. All I can say, is I upgraded my 780 Ti and 970 to 2 980 Ti's, classified ACX with backplates, and I'm getting scenes done in 15 minutes that used to take twice that or more. With fan control temps on both cards stay under 55 c. The Sickleyield benchmark came in at 1 min 55 seconds. I spent $800. That's what I'm talking about when I say cost effectiveness. One of the cards is a year old, the other 4 months. EVGA warranties are by serial #, so they're both covered. For what I do, it was a good move.
That is an interesting observation; 2-3 minutes load time on 3 Titans? I seem to recall Mec4D also mentioning that when she upgrade to higher VRAM; she noticed a slow-down in the initial loading time just prior to render.
I am beginning to wonder about VRAM. It sounds like there is a threshold beyond which it actually hurts more than it helps.
With only 3GB in my 780 Ti's, my loading time for the largest scenes I've ever worked on was never more than ~20 seconds. Perhaps you are rendering without iRay on in your viewport?
-P
If I render a simple scene, IE Genesis + Outfit, I wait 20 seconds, no more. But most of my scene are very complex, With my cheap motherboard, and 3 GPUs, the PCI Express speed is only x1. So I have to wait until the texture and the geometrie had been loaded into the GPU
Leaked (or made-up) information about the 1080ti - I'll pass; it will be even less reliable than what Nvidia releases, and that has been demonstrated on more than one occasion to be flawed.
I waited before deciding on the 10 series; I've decided that waiting is in order; I'm also considering Xeon or the new Ryzen.
... A card with insufficient memory, is an expensive way of both using electric, and denying myself the cash for something more effective.
It isn't a comparison between the same cards from different generations, that would be true; it is however a correct comparison, and comparing the 1080 to 980ti, is useful way of obtaining data to see how effective the performance increases are; time will tell how effective they are - or perhaps more accurately (considering Nvidia throttle using drivers), how effective the improvement is allowed to be.