Question about Graphic Cards
R25S
Posts: 595
With an eye on purchasing a new computer I have a littel question with a huge impact.
What Grafic Card would be the better on (iray)... should I buy NVIDIA TITAN X or NVIDIA QUADRO M6000?
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
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
The difference is about £3000.
Otherwise the cards are almost identical.
(If you are planning to use the Iray server in the future then it might make a difference as some of the server's features are for Quadro only. Though by that time they may have the server working with all nVidia cards.)
my understanding is the Quadros are much more versatile for professional application but Titan cards around the same price point outperform them with CUDA renders in Iray. My understanding is not always "spock" ready so take with a grain of salt.
No; I don`t thing that I`m gonna use a iray server.
But at this moment I have the ability to choose between an used Intel Xeon E5-2699 dual (18 Core, 2,3GHz Turbo 45MB) and 2 NVIDIA Quadro M6000 (and space for a third one) for nearly 9000€
or a brand new Alienware Area-51 with Intel® Core i7-5960X (8 Core, 4GHz Turbo, 20 MB), with 3 NVIDIA GeForce® GTX™ TITAN X for around 8000€.
And if I choose one of them I want to be shure to buy the best for the money.
I have seen similar reports to StrarDragon, but have nothing concrete to back up what we believe.
I would however consider waiting for Pascal cards later in the year.
That`s the point... wich one of them brings faster render time... and will it be significant faster or is it a waste of money because the online cost for the Titan X is between 1100€ and 1300€ the cost for the Quadro M6000 is between 3000€ and 8000€ has significant difference...
The Titans have better resale value...and will sell faster.
Everyone wants to sell Quadros at close to full price so they are slow to sell...because everyone wants to buy them at 'off lease' discount pricing. Titans on the other hand...they just sell.
That said, the basic difference between them is the 'precision' of the compute calculations. Iray does not demand the 'precision' and doesn't use it, any way, so on a Quadro, Iray is running as if it were on a Titan...at several times the cost. Vray (Iray's 'professional' brother), on the other hand DOES demand the precision...
This is a really helpful answer. So goinig for the Titan would be the right descission.
That Quadro system sounds like a bargain though! 72 threads!!!
Yes; it has more Cores but it has only 2,3 GHz while the other one has 4GHZ and it has only 2 Cards whil the Alienweare has 3 Cards on the other hand you can add a third card on the Workstation (because their a one space left). But on the other Hand the Workstation has no guarantee while the Alienweare has full guarantee.
I would suggest the i7-5960X is the best option. So many functions are still single threaded. You will find that 2.3 is slow compared to 4GHz.
Nice to have the 72 logical processors for a few functions e.g. like rendering. But for me that is less than .00001% of my time and generally I do that overnight anyway. If you do a lot of rendering then you may find it of value. The E5-2699 is more a server CPU than a workstation CPU (of the E5 family the 2687 is the workstation variant). I have tried forcing E5-2690's to run at full turbo speeds (turbo and overclocking) but even with the water cooling and 12 fans I could not find a stable arrangement (heat was not the problem), they just like to plod along at their steady 2.6.
I am not sure how 3 Titan's will load down the i7. A pair of E5-2690's run a continuous 25% to 30% just to feed a pair of TitanX's when rendering GPU only, so the i7 might find it hard to keep up with three. Two Titan's might be better than three for the i7.
It sounds like you won't use the benifits that the M6000 can provide. Not unless you are prepared to spend as much again on professional software and even then you may not notice the difference between the M6000 and TitanX.
I think you will find that a lot of money often does not buy as much speed as one would expect, but then there is always the bragging rights :)
The main thing I`m looking for is fast rendering.
By now I use a i7-4700HQ with 2.4GHZ and NVIDIA GEFORCE GTX 880M.
It works fine for simple renders but when it comes to a bit more pretentious renders it will end up very slow.
For an example; the render I addes currently to the Gallery http://www.daz3d.com/gallery/#images/104449 needed more than 14 hours... and it is not very pretentious... so I`m looking for something that will do the job a lot faster. Rendering somethin that uses lamh and maybe some glasses and liquides needs a few days.
Definitely, definitely, definitely get the 5960X system. Three titans will be fine with the i7. I've run two Xs, a titan black, and a titan z at the same time from a 4790K on a Z97 classified. Everything scales almost perfectly based on CUDA cores.
The benefits of the M6000 are driver and application based. It renders slower than the Titan X and provides much less value for rendering.
Honestly though, 8 grand is still high. You could build the same rig for 2 grand less if you put it together yourself.
Thats good advise.
But I`m a bit afried of putting a sytem toghter by my self.
I did this a few times and ending up with a lot of trouble. Most of the trouble came from internal problems like the Soundcard want to use the same port as the TV/Videocard wants to use or the Graphic-Card... and so on; wich end up that some Program or Games will freeze the whole PC shortly after starting it and pushing the Reset-Button is the only way to start again. (you know, doing a hot-reset is not the best way to restart a PC). And any two weeks I can`t even start the Computer because it freezes shortly after the start (or while starting a program) or shut down by itself directly after start or during rendering.
This is what happens to all of my first 5 PC`s that I put together by my self. (And I didn`t use low-budget parts. Each of this PC`s cost me around 3000 bucks).
So I start buyed pre-build PC and Laptops and that problems are gone.
One of the main problems of the cost is that Computer Parts are more expensive here in Germany. For example; the new Alienware Area 51 Desktop (Basic Version) cost 1600 $ in the USA, here in Germany it will cost 1800 Euro wich is 2036 $.
The Alienware Desktop, wich I speak of above, that cost around 8000 Euro would cost only 5700 $ in the USA (wich is only 5038 Euro)... you see; the same PC cost nearly the double...
Option A if you are planning on rendering with CPU only like wtih 3delight.
Option B if you want to render with GPU.
I find that the CPU contributes very little to Iray GPU rendering and I have mine off so I can do other things while rendering. In the case of Iray GPU rendering three titans willl definitely render faster than 2 Quadros since they are basically the same spec.
I have two Titan X in my setup. In the future I may get a 3rd or add the new Pascal card coming out which may have double the CUDA cores of the Titan X.
Quatro is just the titan with the floating-point precision unlocked. It has twice the floating-point precision, as it is intended for "accurate rendering", while the Titan is setup for fast-rendering, for games. However, the "code" does not depend on those values for anything other than "on-screen rendering".
What is the point of twice as long floating-point precision? {0.000000000000001} 16x places V.S. {0.0000000000000000000000000000001} 32x places
- It displays less "clipping" of near surfaces. (Surfaces microns near one another have better z-sorting.)
- Joints have less "cracks", where seams of two intersecting planes touch. (Think of hair-line cracks in the corners, where the "sky" might show through a wall interior.)
- AA and bumps and ray-traced beams are more "true" to origin. (Well, it also extends rendering-time ten-fold too.)
Honestly, the only difference is the drivers. Older cards could be turned into Quatro, but now there is code in the chip that stops that from happening. Even though nothing is any difference inside of the chips.
Some examples...
http://www.tomshardware.com/news/Nvidia-GTX-690-Quadro-K5000,21656.html
http://hackaday.com/2013/03/18/hack-removes-firmware-crippling-from-nvidia-graphics-card/
Also, there was a software mod that once allowed PhysX to be used, if you had an nVidia card present. If nVidia detects an AMD card, it self-disables PhysX from working, even though it is actually able to work.
https://forums.geforce.com/default/topic/415601/nvidia-disables-physx-when-ati-card-is-present-in-your-system-wtf-33-/