Imagine the possibilities of having all the raw GPU power you can handle. Then imagine a bit less, then just a tad more. That’s where NVIDIA’s Quadro 2000 sits, right in the middle. The Quadro 2000 is NVIDIA’s latest mid-range graphics card pushing memory up to 1GB from the previous Quadro FX 1800 and updates the CAD crunching computation of their CUDA technology. It was just last year that NVIDIA launched the Quadro 2000. We talked about it of course, but now we’ve been able to test it out. You’ll see no benchmarks here though. Foul things. We’ll tell you what works, what doesn’t work and ultimately what you need to look for in a GPU for 3D modeling.
Quadro 2000 Review: At a Glance
The Quadro 2000, released October 2010, is NVIDIA’s newest mid-range release of their Fermi-based cards. The size is reduced from previous cards, making it easier to install without tearing power lines out of your motherboard. It has two DisplayPort and one Dual Link DVI-I connection and supports up to two monitors. The maximum resolution is a healthy 2560×1600 and the NVIDIA Control Panel allows you to customize and adjust global 3D settings and program specific 3D settings. You get a slight increase of the memory bandwidth over the FX 1800, but the biggest gain is with the increase to 1GB of memory over the 768MB with the FX 1800. You’ll find the price for a Quadro 2000 currently in the $350-$450 (USD) with Buy.com being one of the cheapest, reputable places to purchase it. (Quadro 2000 at Buy.com)
Quadro 2000 Review: Bottom line
I’m going to cut to the chase on this review. You can look up all sorts of specs for the Quadro 2000, read all sorts of reviews and get someones opinion about cores and clock speeds. You can see some of those details down below. What I want to tell you is how it work and what you’ll notice if you decide to use an NVIDIA GPU and in particular, the Quadro 2000.
Perhaps the best quality of a piece of hardware installed inside a chunky computer case is not knowing it’s even there. For the most part, that’s the case with the Quadro 2000. Hidden away, quietly buzzing and serving up a smooth display throughout your day. However, there are times where you will notice what this graphics card is doing and what it’s capable of.
Dual monitor connection
The Quadro 2000 has the same set-up video ports as the FX 1800 (two DisplayPort connections and one DVI-I connection), so no new cables or adapters are needed if you’re upgrading. If you have an older GPU and two monitors, you’ll need at least one DisplayPort cable. I’m using two DisplayPort cables. There has been some reports of problem, that I’ve experienced as well, with some cords working and others not. If you are having problems getting two monitors running, try using a different cable before tossing out a $400 GPU.
Dual Monitor set-up
The most frustrating event in my whole time using the Quadro 2000 was getting the monitors set-up for dual screen mode. I’ve never really been satisfied with the options on the NVIDIA Control Panel for this. As simple as it seems, it downright difficult to set my monitor on the left to be the main display. Switching cords didn’t work, so I ended up physically moving the monitors around. No big deal, but also a pain.
*Update* – Shadley comments below that it is possible to switch monitors around. However, it’s not through the Nvidia control panel. You can drag and drop your monitors from one side to the other in the Windows, Display, Screen Resolution settings.
Quadro 2000 Driver
Most likely you’ll receive the GPU driver software with the card. If not, as in my case with the evaluation unit I received, you’ll have to download the Quadro 2000 driver and in many cases you’ll need to update to the driver your software supports. You can get any of the latest NVIDIA drivers here.
Realview Graphics on SolidWorks 2010
If you look up SolidWorks Graphics Card Drivers compatibility for the Quadro 2000, the drivers are supported for SolidWorks 2011. For SolidWorks 2010, you’ll get the message “Requires SolidWorks 2010 SP01 or later. For RealView support with SolidWorks 2010 SP04 and earlier, sw2010Quadro.msi must be installed.” It’s a quick installation and then you’re all set to toggle RealView mode in SolidWorks.
This is one test I’ll share, because I don’t think I’ve ever actually seen anyone mentions it and the assumptions can be misleading, especially if you only use one 3D modeling program. Below you’ll see 3 screenshots of 3 different programs: Autodesk Inventor 2011, MoI3D and SolidWorks 2010. In each, I tested the ability to zoom in tight and select the edge of a solid.
With Inventor, I was able to zoom in to the maximum extent and still pick the edge of the solid. The rotation at this level remains smooth, but panning the model is not possible. Overall, the graphics don’t flicker and there is no visible clipping of the geometry.
With MoI, if you zoom in too tight, you pass through the model. So, zooming in as close as possible, I found that the edges of the geometry can still be easily selected, but the selection area is larger on each side of the edge. The zooming and panning are slightly more jittery than in Inventor, but no clipping or flicker occur.
With SolidWorks, selecting an edge zoomed in as tight as possible was an adventure. And by adventure, I mean you had to hunt around for the edge before it actually highlighted. At this zoom level, there was a lot flicker and it was impossible to rotate or pan the model without the model jumping off the screen.
This was all a surprise to me. Typically you would test out a program, see the results and determine that the GPU performance is good or bad. Just this simple test, shows that there’s much more to graphical performance than what the GPU offers. It’s also how the software uses the GPU.
2D and 3D Benchmark
Ok, I’m not totally anti-benchmark. I’ve run quite a few card cookers on the Quadro 2000 and have compared it against an FX 1800 and an FX 3800 I also use. I don’t much care for reading through benchmarks on other sites, so I won’t bore you here. Benchmarks are completely relative to the entire computer system, so it’s best you run a benchmark of your own, granted you can get your hands on the hardware you want. For 2D and 3D benchmark I prefer to use Passmark’s Performance Test. It’s fast, easy to use and provides a better overall assessment of your system than the SPECview benchmark.
Quadro 2000 Details
NVIDIA’s Quadro FX family of GPU’s, while still available, is being overshadowed by the new Quadro Fermi family of video cards. With these cards, promises of increased performance come with increases in the number of CUDA processing cores packed into the GPU. Of course, doubling CUDA cores doesn’t exactly mean doubling performance. You simply have more simple algorithms being calculated at the same time. Errors can occur from all this information being passed around which can be protected with ECC (“Error Correction Codes”) Memory. However, ECC Memory is only available on the higher end Quadro cards like the 5000 and 6000. The Quadro 2000 does not come with ECC Memory.
Quadro FX 1800 vs Quadro 2000
The form factor of the Quadro 2000 is smaller than the Quadro FX 1800. Like the 1800, there is no auxiliary power line needed, whereas the higher end 4000, 5000 and 6000 will require an additional 6-pin PCI Express power connector to function.
As you see in the table below, the Quadro 2000, while generally quite the same, gets an upgrade on some features over the FX 1800.
This is just one experience. You can read other reviews, see list upon list of benchmarks and data, but if you really want to know about the performance or results of a GPU, ask an engineer who’s putting it to work everyday, or better yet, test some of your own models out to see the difference. Some think GPU will be around for a long time to come, some think that GPU will be made obsolete by the CPU. Regardless of either, we use, and in most cases need, a graphics card that can push a large amount of pixels, handle large set of 3D data and speed up renderings for the programs which rely on operations dedicated to GPU intensive calculations. The Quadro 2000 handles all of this well, and from my experience, is an improvement over the FX 1800. Where clipping and black screens occurred with the FX 1800, the Quadro 2000 keeps the graphics smooth. For a $400-$500 graphics card, it’s a shame they didn’t include ECC Memory, but the argument is that if you need that, you really need to upgrade to the Quadro 5000 or 6000 anyway. For work involving concept modeling and mid-size assemblies, the Quadro 2000 is completely capable. For the work I do, on large assemblies and occasional rendering, I prefer the FX 3800 (or 4000.) It’s quite a bit more power hungry, but also provides much smoother performance overall. FX 3800′s can be bought for around $700 with the Quadro 4000 running just slightly higher.
What’s your experience? There are certainly a lot of options. In the end it’s going to come down to what you use your computer for. What’s worked for you?
Disclosure: NVIDIA provided the video card for the testing along with multiple emails about the status of the review which has taken forever to complete.