Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Nvidia Quadro K5000 for mac

Jeff Kilgroe said:
There is no one single card to rule them all, unfortunately. For a new PC build with enough PCIe v3 slots and if budget were not a consideration, I would go with dual overclocked GTX680 4GB cards for use in Premiere and Resolve. If I needed more robust OpenGL support and spent more time in Maya, but also Premiere and Resolve as well, I would go with the K5000 + K20X combo.
Click to expand...

Jeff, these 2 options you mention (dual GTX680s versus K5000+K20X) differ HUGELY in price. AFAIK, you can get a GTX680 4GB card for around $800. $1600 for dual, whereas the K5000 alone costs $2200 and the retail for a K20X is above $3000. Can you really compare a $1600 setup versus a $5200 setup? There must be a trade-off there somewhere, right?
 
Rienis if you have to hack the GTX680 to get it working right then some of us cant do it.
I work for a corporation that doesn t allow that option.
For personal use, not a problem.
 
Eric Santiago said:
Rienis if you have to hack the GTX680 to get it working right then some of us cant do it.
I work for a corporation that doesn t allow that option.
For personal use, not a problem.
Click to expand...

In what way you have to hack them and why? I understand that you may need firmware hacking in order to get them working in a Mac Pro, but a PC?
 
Reinis Traidas said:
In what way you have to hack them and why? I understand that you may need firmware hacking in order to get them working in a Mac Pro, but a PC?
Click to expand...

it would be on a Mac Pro and we have a policy on hardware.
Im sure some corps have similar.
Im only referring as to why some will pay the high cost of GPU power.
 
Eric Santiago said:
it would be on a Mac Pro and we have a policy on hardware.
Im sure some corps have similar.
Im only referring as to why some will pay the high cost of GPU power.
Click to expand...

Right. I'm actually interested in the fact that from what I gather - the hardware in a GTX680 is about the same as the K5000, so my question is - what is there in the Quadro that justifies the extra $1200 of its cost?
 
Reinis Traidas said:
Right. I'm actually interested in the fact that from what I gather - the hardware in a GTX680 is about the same as the K5000, so my question is - what is there in the Quadro that justifies the extra $1200 of its cost?
Click to expand...


Its been like that for years now.

Same issues with Quadro 4000, 4800, etc.....
 
Quadro cards are specified for different purposes vs. the GTX series of cards. They cost more for certain reasons and it's not directly based on performance. The GeForce cards offer superior performance for many common tasks and yes, there are trade-offs to be made either way. Quadro cards have features like ECC memory, better warranties, superior tuning for OpenGL and greater precision as their primary advantages. If you're just looking for raw CUDA performance for accelerating Resolve or Premiere, the Quadro isn't your best option. If you need the extra precision and boost the Quadro cards give within OpenGL and the more fluid performance for apps like Pro/E, SolidWorks, Maya, XSI, etc.. Then the Quadro is your best option.

In terms of hardware, the cards will typically use the same GPUs at their core, but the Quadros have a more robust memory interface, going back to the ECC RAM as an example. They are typically clocked slower and built to run within certain power profiles we see in a lot of the mainstream workstations. Drivers and firmware are tuned differently to perform in a different way, mostly for OpenGL. GPU selection really comes down to using the right tool for the job. In the Mac world, the tool selection is usually quite limited as far as GPUs go.
 
Jeff Kilgroe said:
Quadro cards are specified for different purposes vs. the GTX series of cards. They cost more for certain reasons and it's not directly based on performance. The GeForce cards offer superior performance for many common tasks and yes, there are trade-offs to be made either way. Quadro cards have features like ECC memory, better warranties, superior tuning for OpenGL and greater precision as their primary advantages. If you're just looking for raw CUDA performance for accelerating Resolve or Premiere, the Quadro isn't your best option. If you need the extra precision and boost the Quadro cards give within OpenGL and the more fluid performance for apps like Pro/E, SolidWorks, Maya, XSI, etc.. Then the Quadro is your best option.

In terms of hardware, the cards will typically use the same GPUs at their core, but the Quadros have a more robust memory interface, going back to the ECC RAM as an example. They are typically clocked slower and built to run within certain power profiles we see in a lot of the mainstream workstations. Drivers and firmware are tuned differently to perform in a different way, mostly for OpenGL. GPU selection really comes down to using the right tool for the job. In the Mac world, the tool selection is usually quite limited as far as GPUs go.
Click to expand...

Thanks for this. Like many others, I am eager to find out what Apple has in store for the Mac Pro (if anything). I'm hoping for enough PCI v3 slots to have dual Nvidia cards, as well as space for a Red Rocket and a Decklink card. Thunderbolt would be great for storage.
 
If you cant wait get a CUBIX XPANDER.
We run Avid and Resolve along with FCPX and Premiere Pro on the same Mac Pro with Quadro 4000s as GPU plus a ROCKET and the DeckLink to boot.
 
I just tried my new Quadro 4000 for Mac on the Windows 7 side. I tried a render in Maya 2013 and actually, to be honest, I didn't even saw a difference than what my previous GeForce GT 120 did.. In fact, while just browsing around I don't see any difference (Though on the Mac side, open pages on internet seemed faster since I installed the Quadro 4000..)

So yea, any reason you can see why it doesn't seems to have a difference? I actually downloaded the lastest drivers on the Windows side, but nothing seems to change on the performance side.

EDIT: Alright, I just tried both cards. I made the EXACT SAME RENDER with the Quadro 4000 and the GeForce GT 120, and both did the render in 25 seconds.. I guess there is a problem here, I can't beleive I paid $800+ for nothing better.
 
Last edited:
Um, what do you mean by render, specifically? These cards don't accelerate rendering in Maya, they accelerate your interface in terms of OpenGL and the number of polygons you can push around. The Quadro 4000 should be significantly better than a GT120 if it's running right. Once you click "render", it's up to your CPUs from there on out. Some rendering software can use OpenGL or CUDA for acceleration, but that's not very widespread at the moment. There's the RenderGL extension API for incorporating OpenGL acceleration into Raytrace and GI rendering systems... For the most part, rendering is very CPU dependent and usually handled via render farms or multiple CPU nodes. Much cheaper to farm it out onto lots of commodity CPUs than trying to do it on one box with expensive GPUs/ or compute cards.
 
Open Maya, create a polygon sphere, up the faces to 10000, then apply poly smooth set to 4.

Lets see how the GT120 handle that baby :)

Trust me, it wont.

Been there too many times.
 
Jeff Kilgroe said:
Um, what do you mean by render, specifically? These cards don't accelerate rendering in Maya, they accelerate your interface in terms of OpenGL and the number of polygons you can push around. The Quadro 4000 should be significantly better than a GT120 if it's running right. Once you click "render", it's up to your CPUs from there on out. Some rendering software can use OpenGL or CUDA for acceleration, but that's not very widespread at the moment. There's the RenderGL extension API for incorporating OpenGL acceleration into Raytrace and GI rendering systems... For the most part, rendering is very CPU dependent and usually handled via render farms or multiple CPU nodes. Much cheaper to farm it out onto lots of commodity CPUs than trying to do it on one box with expensive GPUs/ or compute cards.
Click to expand...

Yea, I'll probably make a fool of me, but I thought the Graphic Card helped the performance of the render time (when rendering a scene). As I got some answer from CGsociety, the graphic card is more for live render as generaing high poly scenes. Sorry about that noob talk here, I guess I still have a lot to learn.

But other than for the 3D, why is these Graphic Cards seems to help a lot when rendering in AE/Premiere etc? I guess I should keep the card anyway since I'd like to accelerate my work in these softwares too.

Also, talking about graphic cards, I got an issue when booting in Windows 7 (BootCamp) when my 2 cards are at the same time in the computer (Quadro 4000 and GeForce GT 120). I made a topic about it, don't know if you have knowledge/can help here, jeff? http://www.reduser.net/forum/showth...enter-with-Quadro-4000-For-Mac-GeForce-GT-120


And then, talking about render farm/GPU etc, which would be the best way to lower my renders time? Upgrading my actuals GPU or hetting another comp for render pharm?

thanks a lot for your help btw, Jeff!
 
Olivier Brisson said:
Ahah, yea, so graphic cards are more for "live" work as far as I understand.
Click to expand...


Trust me its the live work that you need a hand with.

Try dealing with a few particle sets in real time along with some heavy mesh data.

Rendering you can walk away and just relax.

Watching your screen refresh every 15 seconds...forget about it :P
 
Olivier Brisson said:
Yea, I guess it's a good deal here, though, I still really need to improve my render time so I'll have to do something for this.
Click to expand...


Whats your specs and where is the bottle neck?

Maybe being so deep you never know whats causing the slowdown.

Some cases its your I/O, network, etc...

Ive only been using Renderman for the last few years.

I gave up on the Maya Renderer years ago.
 
I'd actually love to use Renderman, but I'm at my last year school and I need to keep it like school work.

What mean "I/O" exactly?

btw here's my specs:


2 x 2.26 GHz Quad-Core Intel Xeon

RAM 20 GO 1066 MHz DDR3


Version du système : Mac OS X 10.6.8 (10K549)
Version du noyau : Darwin 10.8.0
Volume de démarrage : Macintosh HD
Mode de démarrage : Normal
Mémoire virtuelle sécurisée : non activée
Noyau et extensions 64 bits : Non


Nom du modèle : Mac Pro
Identifiant du modèle : MacPro4,1
Nom du processeur : Quad-Core Intel Xeon
Vitesse du processeur : 2,26 GHz
Nombre de processeurs : 2
Nombre total de cœurs : 8
Cache de niveau 2 (par cœur) : 256 Ko
Cache de niveau 3 (par processeur) : 8 Mo
Mémoire : 20 Go
Vitesse d’interconnexion du processeur : 5.86 GT/s
Version de la ROM de démarrage : MP41.0081.B07
Version SMC (système) : 1.39f5
Version SMC (support pour processeur) : 1.39f5
 
Rendering differs from one application and rendering system to another. It is also a very broadly used term, "rendering", meaning the computer is crunching all the data to assemble your image or video or other media you're creating. Generally speaking, 3D rendering software is highly dependent on CPUs. It's easy to task out to multiple CPUs, cores, individual systems (as nodes on a "render farm"), etc.. GPU or graphics card acceleration has been tried a lot over the years, but it never really takes off in this field. When it comes to complex data and calculations for raytracing, global illumination, subsurface scattering and all the good stuff, the larger CPUs still perform the best here. Not only that, but they're dirt cheap in terms of the amount of computing power you get in contrast to other solutions like GPU or compute cards. We can put two of the fastest 8-core Xeons, giving us 16 cores / 32 threads, into a single U of rack space. In a "standard" full-height 42U rack, with a good power management system installed, we can get 36 U of CPUs or a total of 1152 concurrent threads on which to run render tasks.

Some rendering in other types of applications, like video compositing, encoding, etc.. is more often accelerated by GPU hardware these days. Adobe Premiere uses nVidia GPUs (and a few select ATI GPUs) to accelerate many real-time operations while editing as well as several GPU-specific FX for editing and in final rendering. So a good GPU can assist the CPUs in these situations.

You have a Mac Pro 4,1 -- a 2009 Mac Pro. Dual quad-core at 2.26GHz with 20GB RAM.... It's a good computer, but it's still a generation behind the current Mac Pro, which is also a generation behind what Intel is offering and other companies like HP, Dell, etc.. are shipping. There are some ways to speed it up -- such as more RAM, maybe, depending on your work load. Faster drives/ storage (this is where I/O or Input/ Output speeds come into play). If you are transacting data over a LAN while rendering, as in bringing in your data or sending rendered frames back out on the network, this can really slow down the process.

Rendering is a time consuming process. As you do more CG work, you will get a feel for how fast certain things should render on your hardware. You will often have to take these estimates into consideration when starting a project to know how much detail or realism you can deliver within your project's deadline.

Generally speaking, the Quadro 4000 card will help give a boost to Adobe Premiere and some limited acceleration to Adobe After Effects within it's Open-GL accelerated raytrace engine, which is used mostly for title generation and 3D object placement. It will also accelerate the polygon performance or interactive speed within your Maya GUI. While modeling and animating, you can move more detailed meshes around on the Quadro 4000 card than you can with that GT120. For the "rendering" part, the Quadro 4000 will kick in within the Adobe apps to accelerate a few things, but most still happens on the CPUs. For Maya, or similar 3D rendering systems, it almost always comes down to your CPUs. If you want to render faster, add more CPUs, more systems, faster CPUs, faster systems, etc..

You're talking about a 25 second render in one of your posts above. Sounds fine to me... but I don't know what it is you're rendering. By comparison, the last pro animation job I worked on was last summer, we were averaging 18 to 20 minutes per frame render times and were constantly fighting a number of factors to keep it at 20 minutes per frame or less. Otherwise we would have never met our deadline.
 
You must log in or register to reply here.
Back
Top