Nvidia Takes Fermi to Entry-Level Professionals

how do they compare to similary priced AMD offerings ?

Yes! Please let's get some benchmarks between these and AMD workstation graphics cards, that would be oh so helpful in getting my IT dept. to get the card I need for Solidworks. THG to the rescue?

I wouldn't expect the comparrison to go in Nvidia's favor. Otherwise this advertisement article would have made sure to point that out.

So we see a Quadro version of the GTS 430 before a retail one. The specs just give a big "meh" especially since competing Firepro products have been out for a while now.

I agree this article is an advertisement. That said, in the professional graphics arena, drivers are king. I would not be surprised to see AMD parts with much better theoretical performance get beaten in the real world by these nvidia parts. It may be purely for historical reasons, but give credit where credit is due: nvidia generally has good driver optimizations for many applications, and these optimized drivers often best AMD's offerings.

I would also like to see a solidworks shoot out. If you dont have a test maybe something like this one: http://www.solidmuse.com/solidwork [...] model.html
Help us poor people stuck with IT built rigs who think any run of the mill card can handle solidworks!!! Let alone the p4 im still forced to use... Damn IT dept.

Yeah, at least I managed to salvage a rig with a dual-core. Boo to the bad small-business IT deptartments!

Tom's hasn't done a Workstation Chart since 2007, if their search engine is working right. Back then, the FX 570 were beating out top of the line FireGl cards costing 10 times as much in Solidworks benchmarks. As Anna and everyone else on the Solidworks forums says, it's all about CPU. Get the fastest processor you can get, which is Core i7, fast memory, and a good hard drive. Don't blow more than 500 on the video card, and 200 is really enough unless you do a lot of rendering and presentation stuff. I'm using a Dell M6400 notebook. About two years old, Core 2 extreme worth every penny, and FX2700m graphics card. My co-worker opted for the quad core and the 3700m, and his is considerably slower than mine in everything solidworks. He is a modeling and sim guy so he got the right combo for what he normally does with it. I know a lot has changed, and AMD has done some catching up, but Nvidia still leads in drivers.

Alot of popular software is CUDA only in this space, so no comparrison can be made.

Regardless, ATI stream blows. Go look at folding at home, Nvidia crushes AMD when it comes to parallel performance.

I bet you work for nvidia.

ATI drivers have alwasy sucked,and I don't work for Nvidia.

How about some benchmarks for Autodesk Inventor 2011 instead of AutoCAD. AutoCAD is for drawing pretty lines. Inventor is for solid modeling, which needs a higher end machine much more than AutoCAD.

Of course there still remains the problem that except for UG (as far as I know) no solid modelers are multithreaded for the core modeling and assembly environments. This makes the software very CPU limited. So I'd like to see benchmarks with an i5-680 and workstation versus gaming cards. Inventor uses Direct3D, but I've seen mixed results. For example with an E8400 these cards provide identical performance: 8800GTS 640MB, HD5850 1GB,m HD5870 2GB. Although with the 5850 and 5870 in CF, there was a performance improvement. My benchmarking assembly has 7000 parts.

And i bet every ATI fan boy bashing on nvidia works for AMD right!?!

NOP, I don't work for AMD I just have a £700 waste of space in the form of a Dell XPS 1710 with a crappy nvidia GPU inside it.

I know I am noob or just only play games but what makes these better than, lets a gtx 480?

Drivers/small hardware change. Still have not found a GREAT article or anything that explains it, but in general, it is the drivers, and a small hardware change that some people can imitate by "soft-modding" a gaming GPU. All in all, gaming GPUS do not do well in applications that workstation GPUs excel at, and vice versa. Different beasts. high rpm-low torque vs. low rpm-high torque (possible analogy?)

I thought AMD does not support CUDA? So how can you compare -- unless we are talking OpenCL.

Well considering how much cheaper AMDs offerings are, they would not compare well. You are looking at comparing the entry-level Quaddro 2000 to a mid-range part from AMD. Considering AMDs most expensive Professional card is 1/3rd the price of nVidia's, in most cases when comparing dollar per dollar AMD will win by a notable margin despite any driver or optimization issues.
However, when weighing the cards to use for a workstation you should always do research. The best card will flip flop depending on application. For instance if you are going to use Autodesk Maya, you will be using an AMD. AMDs just do significantly better here and you would be wasting money using nVidia. For more entry-level applications, chances are you will want the lower watt card over the cheaper card. You would just hope the better card is also the cheaper card.

haha which would you get the GTS450 or the 5570, nuf said

Are the amd fanboys not aware this is a tech site, and this is news that every other tech site is covering. So jealous of Nvidia that your always worrying about some slight to ati(amd) whatever you call them. Its pathetic drama.
http://www.slashgear.com/nvidia-qu [...] -05106033/
http://www.anandtech.com/show/3961 [...] o-2000-600

Nvidia never released a GT200-based Quadro if I'm not mistaken (after all, they couldn't scale the architecture at all). I'm surprised they only get 1.5x performance with a jump from G92 to Fermi.

EDIT: Nevermind, I guess they did (FX4800 and FX5800).

Sounds great. I think they should make more cards like the Quadro 600 that can work in hotter and more cramped environments.

Since the last 2 releases Nvidia is outperforming AMD on price and performance even on Autodesk Maya 2009 and 2010 and most professional applications, not a fan appreciation but pure facts. I did my research a year ago and went to Nvidia for an entry level Quadro 580. At that time Nvidia Quadro 580 was better on performance than the AMD near price option and offered 512 MB of memory, twice what AMD did. As a reference the new Quadro 600 is slightly better than the 580 in Maya (12.20 Vs 10.82) in Nvidia's own current Performance numbers. But I also read every other review and Nvidia was almost always better on every price. Is not obvious but there are specific scenarios and apps that work better on AMD professional cards so if you are planning to buy a professional card do a specific research and ask real users for the specific combination of software and hardware you want to get.

And finally on real use for almost a year I can say the Q580 works really well and stable with Maya on Windows 7 and XP too. The new Maya version "2011" has a new and improved Rendering engine and this year more powerful graphic card options offer real advantages over less advanced and cheaper cards. But on this latest version I haven't read any performance review, just users comments.

I worked at a few architectural firms over the years and they all use nvidia workstation cards. The computer guys said they just trust nvidia. I got a firepro in my laptop myself.

If these cards perform well I can see a few of my friends being all over them at that pricepoint.

if AMD and nVidia made the exact same card, the only difference being the drivers. the nVidia one would beat the AMD one. AMD drivers plain suck, and if you use linux it's even worse.

I don't understand this. The architecture of both cards are the same and the gaming cards have over twice the amount of cuda cores for about a $100 less. I'd like a more definitive answer as to what makes these cards better for business.

Can someone benchmark the GTX 460 against the similarly priced Quadro 600 on workstation performance? It's 336 cuda cores v 96. Does a little bit of tweaking make that big of a difference or is there another performance point I'm missing?

Professional 3D applications are optimized (the drivers) to move very heavy geometries at relatively low fps with basic shaders and gaming cards are optimized for shader power, fast fps and now with an increasing geometry complexity, but not with a vital precision. Even they use the same hardware, the Pro apps use the resources for different task. The goal in pro 3D apps is geometry first as if you are drawing a very complex building or industrial model you need to see the mesh (geometry) with great precision and detail and you only need basic shader resources for a simple solid rendering in the viewport. Professionals don't need to see beautiful and complex shaders on screen at 30-60 fps as when the models are finished they are exported for rendering purposes to more specialized software/hardware rendering engines and machines. Usually a studio of engineers know very little on how to create complex shaders for rendering purposes. They know about structures, forces or angles. And the same happens with architects, they need color only to differentiate one structure to another. When the high precision mesh is ready it goes to a different pipeline where the model is "cleaned" and prepared for rendering, if at all needed. This is accomplished usually by a different studio that has the shaders, textures, lighting and the rendering knowledge.

In the digital content creation DCC industry the modelers, programers, technical directors and even fx professionals usually wont see their work in all their glory when working. They will perform renderings but again not using the full rendering power as color is not even close to final in this levels. It is just when the model is ready when the lighting, texturing and shading process begins. But it wont end there. When the model is shaded and textured another specialist adjust the lighting of different objects separated and creates the illusion of light as a whole. So objects or models are rendered individually or in groups as layers, like in Photoshop. But is the composer in the post production level who is responsible for the final color and appearance of every frame. He receives the images in layers from the animation sequences and he/she composes and adjust these layers into one single picture. Even as the composer needs the shader power in the graphic card is for very high color precision of pre-rendered pixels and not for fast moving 3D objects. In this scenarios a professional needs 12 bit depth for every color channel plus alfa. Hence a 12 bit professional monitor with high color accuracy reproduction.

So, is the game industry mostly the one that needs superfast pixels and moving geometry at 30-60 fps. But in a game you wont care or even see bad pixels and geometry rendered with relative low precision as long as you are able to see your opponent and shoot first. Software drivers are given for gaming cards but in the professional world the drivers are so specialized they use different versions for specific software. This is certification.

One last thing is that these different Pro industries have a relative low volume demand and different specialized needs compared to the millions the gaming industry sells in a more unified world. But Pro studios and some professionals are willing to pay more for the particular features they need, usually very high precision, 3D stereo glasses, very specialized rendering features and a lot of on board memory, 2-4 gbs for very heavy geometry. And all this with Reliability so overclocked hardware is not a good idea if not extremely well implemented.

As everything is changing and evolving new 3D professional software is starting to use the great shading power from GPUs for high quality final renderings as is the case with Octane Render, using Nvidia's gaming hardware or MachStudio Pro, exclusive with AMD PRO GPUs. Still in its infancy and every single frame still needs seconds or minutes to render.

I hope I helped a bit to clear the difference between gaming and the darker professional side.

I don,t see continuos leadership, it's kind off jumping one over onother
http://hothardware.com/Reviews/ATI [...] rd/?page=5
An yes, new generation of Quadros is faster, but an half year was faster FirePro. I am using Solidworks, and in Solidworks forums I see users more like ATI cards...

I am using AutoCAD 2009 I will try this card.