jasonbogert Posted April 11, 2014 Share Posted April 11, 2014 Hi everyone. So I was just wondering if I could get some insight from people who have had real world exerpiences with certain GPU's with applications such as maya/zbrush/sculptris/photoshop and maybe some other applications used for game development. I am debating whether or not I should get a single, GTX titan black for my needs or go with a GTX 780 + a quadro k4000. I already have the GTX 780 at the moment but can still return it. Now I know that quadro cards are made more specifically for workstation and 3d kinda work but there are limited reviews out there in relation to the titan black and how well it can do in these types of applications vs the quadro k4000. I know that the purpose of the titan black is to be both a gaming and an entry level workstation card due to its double precision point performance and other things. So any thoughts from people who have used any of these cards and their performance? Thanks in advance. Link to comment Share on other sites More sharing options...
RyderSK Posted April 11, 2014 Share Posted April 11, 2014 Zbrush is yet to be accelerated by GPU (at the moment it's not), Maya's Viewport 2.0 handles direct11 acceleration much better than Max natively and even in OpenGL holds its own with regular cards (as illustrated by benchmark on this forum section by Dimitris), Photoshop get's almost eaqual acceleration from all GPU's across market, it's not like you need much performance there though, while Sculptris is openGL accelerated, it's much more taxing the CPU, in same way Zbrush does. Plus, you already have GTX780. All in all out, you can't get any better. You shouldn't be pondering about exchange, upgrade or whatever. Feel the need to throw some money into trash :- ) ? "double precision point performance" doesn't make it entry level workstation card, but entry level general computing (GP-GPU) card, separate from all tasks you listed and will ever do. Link to comment Share on other sites More sharing options...
jasonbogert Posted April 11, 2014 Author Share Posted April 11, 2014 Okay thank you so much. Ya that was the one thing I was trying to figure out but I couldn't seem to find an answer to was the point of the double precision performance and if it applied specifically to what I was trying to do. I have an i7 4770k cpu so hopefully that will be good for my needs. Link to comment Share on other sites More sharing options...
Dimitris Tolios Posted April 11, 2014 Share Posted April 11, 2014 (edited) I don't want to repeat things over and over. I know that the purpose of the titan black is to be both a gaming and an entry level workstation card due to its double precision point performance and other things. Well, there are workstations and workstations. As far as Quadro and Firepro cards being the "workstation cards" for CG artists, the Titan doesn't have that kind of "workstation" built-into it, as - as said above - DP/FP64 is mostly used to make calcs in situations that require extreme precision (more than 17 decimal points I think) and viewport work simply doesn't require that. Usually this is useful for very precise scientific models, or when looking up for holes in coding while compiling, end even then, some ppl " accuse " coders of being lazy and using FP64 precision as a default, without it really being needed. Most of the FP64 intensive code is run through the CPU of the end-user anyways, even if the coder used a GPU to accelerate the compiler. The Titan was the consumer card with the most VRam, so too many idealized it thinking "more VRam, has to be better across the board", which is ofc wrong: the original titan could be matched in performance by an overclocked 780, and easily outperformed by a 780Ti, both of which had "crippled" DP/FP64 performance, ontop of half the VRam! Just keep in mind that 6GB versions of both 780 and 780Ti are in the works/about to be shipped by EVGA and probably other manufacturers will follow Titan's niche is shrinking even more, simply because DP/FP64 calcs that are not very relevant to CG work or any viewport acceleration. Most OpenCL / CUDA calculations performed for renderings etc, are also unaffected by lowering DP performance. Edited April 11, 2014 by dtolios Link to comment Share on other sites More sharing options...
jasonbogert Posted April 11, 2014 Author Share Posted April 11, 2014 ya I heard about that. I should be able to upgrade to the 6bg version of the 780 then through evga. Thanks for validating what Juraj said. I was def having a lot of trouble figuring out what was really important for viewport performance and other things like that. I just want these applications to run as smooth as I can afford for what I need to do. Link to comment Share on other sites More sharing options...
Dimitris Tolios Posted April 11, 2014 Share Posted April 11, 2014 The "vanilla" 780 was of great value if 3GB could "contain you", and there were VERY VERY VERY few scenarios that 3GB are not enough for viewport acceleration. We are talking multiple monitors and fasts paced gaming. Even 4K is not a real challenge for 3GB. So all the 6GB (or 4GB etc) cards are left with, is GPGPU and progressive renderings, something nearly none of them was designed for. Developers adapt to the hardware for GPGPU, not the other way around, and so far this was the case for most GPU accelerated applications: custom code is expected to be written. Depending on pricing, the 6GB 780 will still be the better choice, as you are called to pay a pretty hefty premium for those extra SMX clusters in the 780Ti and Titan Black. If 1x 780 is "not enough" as far as raw power goes, probably neither Titan or 780Ti will be. In my opinion it is better to start thinking about scaling up to more than one 780, than going for a much more expensive card that is 10% or so faster. 6GB being a Titan exclusive, made it an valid excuse for some, but FP64 still isn't in the CG graphics field. Link to comment Share on other sites More sharing options...
jasonbogert Posted April 11, 2014 Author Share Posted April 11, 2014 Well would SLI'ing two gtx 780's do anything for things like maya? I heard from other forums that 3d modeling programs such as this don't really benefit from two cards. When asked this specific question on tomshardware.com people were saying that maya and such could only reference the cores on a single GPU. Unless anything has changed since 2013 when these were posted. Link to comment Share on other sites More sharing options...
Dimitris Tolios Posted April 11, 2014 Share Posted April 11, 2014 Well would SLI'ing two gtx 780's do anything for things like maya? I heard from other forums that 3d modeling programs such as this don't really benefit from two cards. When asked this specific question on tomshardware.com people were saying that maya and such could only reference the cores on a single GPU. Unless anything has changed since 2013 when these were posted. Sorry for the miss-understanding, I was talking exclusively for GPGPU and progressive rendering when suggesting a 2nd 780. Correctly said, for 3D viewports you will see no benefit. If a 780 is "slow" on any 3D viewport, it is almost guaranteed that it is a GeForce driver deficiency, and not lack of raw compute power. And all GeForce cards share the same drivers, so... Link to comment Share on other sites More sharing options...
jasonbogert Posted April 11, 2014 Author Share Posted April 11, 2014 oh okay. thanks for clearing that up! Link to comment Share on other sites More sharing options...
komyali Posted April 11, 2014 Share Posted April 11, 2014 hehehe don't kill me I already posted on another thread xD just do this and tell us what happened http://www.eevblog.com/forum/chat/hacking-nvidia-cards-into-their-professional-counterparts/ Link to comment Share on other sites More sharing options...
Dimitris Tolios Posted April 12, 2014 Share Posted April 12, 2014 hehehe don't kill me I already posted on another thread xD just do this and tell us what happened http://www.eevblog.com/forum/chat/hacking-nvidia-cards-into-their-professional-counterparts/ Despite "resoldering" new product IDs and installing Quadro BIOSs, most models don't really gain Quadro functionality. Something is "wrong". It is believed that certain architecture features are after all disabled in GTX cards and the Quadro driver cannot be "fulled" in allowing access to all the potential. Link to comment Share on other sites More sharing options...
terryosser Posted April 12, 2014 Share Posted April 12, 2014 Good forum and cool thread. Hey Dimitris could you tell me who the titan black it really aimed at then? i too need a highend card for game development but i don't want to buy a Quadro. i was thinking of buying a titan black but after reading what you have wrote i might wait for a 780 6gb. Also do you know if a titan black be good at running the unreal 4 editor? as in building lightmapping etc? Link to comment Share on other sites More sharing options...
Dimitris Tolios Posted April 12, 2014 Share Posted April 12, 2014 (edited) The anecdotal story of how "a friend of mine" got a GTX Titan 1st gen (14/15 SMX units vs. 15/15 the Black has). Why did "he" get it? "He" was happy with "his" 670 4GB (was 4 for VRay RT tbh), till "He" got a 1440p monitor...then gaming started being slower at full res, but not that bad. But "He" still started dreaming for a 2nd, going SLI (for gaming) and ofc getting some progressive rendering benefits. But at the same time, "He" was building a watercooled rig in a AIR540 to transplant "his" 3930K/X79 system from the otherwise fine CM 692 Adv. 670 4GB at that point were around $260-300 used, with waterblocks for it at around $75-100 a piece new. So "he" was looking at a minimum total investment of $650+ to get 2x GPUs under water. Tried to find a used 780 that was brand brand new into the market, could not find anything below $500, ended up finding a Titan for $650 and pulled the trigger. For $750 give or take, there it was: a GTX Titan 6GB under a full cover XSPC block. Less absolute performance in both gaming and progressive rendering than the 2x 670, but a single card that would give benefits all around. It also pulled some 250,000 points or more in F@H (yes, me thinks that likes DP) and... But Many ppl just go for the "blink". "He" was impressed by the blink too, "he" doesn't judge. Just rationalizes. UE4: it is DirectX 11...I don't see why a 760/770 would not be "ok"...furthermore a 780 or Titan. Edited April 12, 2014 by dtolios Link to comment Share on other sites More sharing options...
terryosser Posted April 12, 2014 Share Posted April 12, 2014 Thanks for ur response, im mainly using maya, 3dmax, photoshop, unreal4 and zbrush and i have 3 monitors...one high res monitor 2500x1500 pixel i think it is. so basically your saying a titan black just isn,t worth the money for these programs? Link to comment Share on other sites More sharing options...
Dimitris Tolios Posted April 12, 2014 Share Posted April 12, 2014 (edited) I have the same, 2560*1440p 27", and a secondary 1920*1080p 24".I believe the GTX Titan will have very small to no difference over a 780 3GB, with the 780 6GB remaining a cheaper-than-titan option with 90% or more of the potential should you wanted to go for VRay RT GPU, Octane, Mari etc. A 780Ti will be as fast as the Titan Black, and cheaper, but again I think for a CG workstation the even cheaper "vanilla" 780 is more than enough. To be honest for the most part and kinda low poly models for games, a 760 will also be "ok". The performance returns vs. cash spent is diminishing. The "jump" from a 760 to a 780 will be much more obvious than the jump from a 780 to a Titan or 780Ti. The 780 "buys" you into the big boy performance in the $500 range - far from cheap, but more promising than a K2000 easily. A 290 or 290X are also good choices if you don't care about CUDA (nVidia proprietary) GPGPU. Edited April 12, 2014 by dtolios Link to comment Share on other sites More sharing options...
terryosser Posted April 12, 2014 Share Posted April 12, 2014 (edited) I cant do ati because of unreal 4. running udk with my last ati card was a nightmare. Plus i want to use phys and cuda. I also model highpoly stuff so viewport rendering is very important. Im fed up with splitting scenes. I forgot about mari, sometime in the future i would love to start using mari. Thanks for ur help man. Edited April 12, 2014 by terryosser Link to comment Share on other sites More sharing options...
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now