NVIDIA Quadro 6000
Fresh off the SIGGRAPH presses in sunny Los Angeles, NVIDIA has announced a full refresh of their Quadro professional graphics solutions. Dropping the previously used “FX” title from the Quadro nomenclature, NVIDIA has unleashed the Fermi-powered Quadro 4000, Quadro 5000, and Quadro 6000 GPUs.
Fermi has been a “better late than never” kind of chip. While ATI poured out DirectX 11 offerings for desktop graphics solutions, NVIDIA wouldn’t respond with Fermi-powered GeForce cards for a further six months. Soon afterward, ATI refreshed their entire lineup of workstation GPUs, offering five new DirectX 11-compliant FirePro cards. Again, ATI went uncontested as rumors spun as to what kind of Fermi madness NVIDIA would ultimately unleash with a refresh of Quadro. Today is finally that day, and the new Quadro lineup is impressive and ambitious. For Quadro customers, the results are well worth the wait.
Icrontic was fortunate to receive the Quadro 6000 for review, and I’ve been putting it through the paces on our testbench for the last few days. This GPU is an absolute monster, perhaps even in a class of its own. I’m getting ahead of myself, however. Let’s start from the beginning and take a closer look at what’s on the inside of the NVIDIA Quadro 6000.
The NVIDIA Quadro 6000
The ultra high-end NVIDIA Quadro 6000 is powered by the GF100 GPU. It is a two-slot solution that requires either a single 8-pin PCIe power connector, or two 6-pin PCIe connectors. It boasts an enormous 6GB of memory and has an impressive array of features. It is introduced as the next-gen answer to the Quadro FX 5800, which makes it the most powerful GPU that NVIDIA offers. Full product specifications are as follows:
- 547MHz core clock
- 448 CUDA Cores
- 6GB 800MHz GDDR5
- 384-bit Memory Bus
- 144GBps Memory Bandwidth
- 2×6-pin (or 1×8-pin) PCIe power connector
- 225W max power consumption
- 2 x DisplayPort, 1 x DL-DVI Outputs
- 3-Pin Stereoscopic (3D) Support (Optional)
- Serial Digital Interface (Optional)
- SLI Multi-OS Support
- NVIDIA 3D Vision Support
- Supports DirectX 11 and OpenGL 4.0
- USD$4,999 MSRP
This GPU breaks new technological ground with its massive amounts of memory. At six gigabytes of GDDR5, the Quadro 6000 holds the title of the GPU with the most amount of memory ever. It really makes me feel cheeky when I remember how proud I was of my Geforce 8800GTX’s 768MB of memory just four years ago. As if the sheer amount of memory wasn’t enough to impress you, the 384-bit memory bus and 144GBps memory bandwidth should. With memory specs like this, there is no question that the Quadro 6000 means business.
The Quadro 5000 and 6000 GPUs are also the first to feature ECC, or Error-Correcting Code memory. ECC is implemented to check data accuracy and integrity as it passes through. Having ECC on these boards ensures better reliability and security when dealing with large data sets.
NVIDIA is directing this launch as a radical shift in how their GPUs tackle creative tasks. Their new focus is what they call “Computational Visualization”. The idea is that the modern GPU is capable of so much more beyond just rendering graphics, and computation of complex systems thanks to CUDA and raw processing power is making this happen. Adobe Premiere Pro CS5 and the Mercury Playback Engine is a good example of this new focus that can be used today on Quadro GPUs. Project complexity is growing, simulations are becoming more intensive, and the new line of Quadro GPUs are ready to take on this bold new environment.
An example of computational visualization is how RTT is using modern Quadro GPUs to test prototypes of new cars. A designer can have a high resolution, full detail model of the car in software and run wind tunnel simulation tests on the body. The GPU can handle the simulation of the air passing over the car body, which is displayed by moving arrows. The model can be inspected in real-time as the simulation is in progress. Designers can instantly modify their designs without having to render out—or worse yet, build an expensive physical prototype.
The Fermi-powered Quadro 6000 marks a bold new direction for NVIDIA and their workstation products. Though we can talk all day about what these boards are designed to do and how NVIDIA sees them fitting into the big picture, it’s time to come back to reality. How does this GPU actually perform?
The Evaluation
For purposes of this evaluation, the Quadro 6000 is going up against the ATI FirePro V8800, V8750, and V8700. Though it could technically be considered an ultra-ultra high end part, the Quadro 6000 does not have a direct competitor, as the recently announced Quadro 5000 is more directly aligned to compete with the FirePro V8800. The V8800 is ATI’s most powerful offering currently, and the closest competitor to the Quadro 6000’s spec level. For further analysis, the FirePro V8750 and V8700 have been included in the test results, which rounds off the three most powerful workstation GPUs that ATI produces.
All of today’s combatants have been through Cinebench R10 and Cinebench R11 to test OpenGL performance. Cinebench is a benchmark tool made by Maxon based on their Cinema 4D software. These results test the real time OpenGL playback performance of these GPUs.
SPECviewperf is a staple of any workstation GPU benchmark suite. Viewperf is a collection of DCC tests engineered specifically to test performance in industry standard software packages such as Maya, Lightwave, and Solidworks. As of a few months ago, Viewperf 11 was released. This much needed update retooled the tests to be more intensive on today’s more powerful hardware. Viewperf 11 ensures much more accurate benchmark results.
Test Results
(Click for larger versions)
To say I was surprised by the OpenGL results is a gross understatement. With a Fermi core and a metric ton of memory, I expected the Quadro 6000 to breeze by its FirePro competition. It is quite interesting that all three offerings from ATI brought in better framerates than the Quadro did in Cinebench R11. These results aren’t to say that the Quadro 6000 has bad OpenGL performance. Rather, the FirePro GPUs just seem to be much better in the real-time OpenGL rendering segment. Though OpenGL rendering is not typically a central focal point of workstation GPU usage, it certainly is an interesting detail in this evaluation.
I was concerned that the Cinebench results might be off, that they weren’t accurately portraying the Quadro 6000’s actual performance. For consistency’s sake, I put the FirePro V8800 and the Quadro 6000 through a full benchmark in 3DMark Vantage to test their performance in DirectX 10. The results were unchanged—the FirePro V8800 scored 16,631 3DMarks averaging 51 FPS while the Quadro 6000 scored 14,186 3DMarks averaging 42 FPS. The Quadro 6000 scores well, but can’t keep up with the V8800’s impressive real-time speeds.
Update: 08/12/2010: We originally published the results with vsync on. The apology and update has been posted here. The results below are with vsync ON:
The Viewperf results are a different story, and those scores tell a striking tale of dominance on behalf of the Quadro 6000. It is hardly a fair comparison as it is clear that the massively expanded memory capacity and bandwidth are kicking some serious tail in all of the Viewperf applications. The Quadro 6000 managed to score higher than the FirePro V8800 in all of the tests, and in roughly half of the tests on average, it managed to double the performance of the V8800. That is quite the impressive increase. Professionals who work with incredibly intensive projects should take note.
The Quadro 6000 does run on the hot side. While testing, the GPU averaged a temperature of 88 degrees Celsius. Under full load, the GPU ramped up to 94 degrees Celsius, at which point the blower kicked into top gear and promptly brought the temps back down to a more manageable 86 degrees. The blower, while effective, does get on the loud side when it really gets going. In fact, while playing Alien Swarm and Grand Theft Auto IV, both games managed to push temps high enough that the fan became audible over my microphone. I’m sure the arid temperatures of Southern California are not helping the situation, but when the GPU idles at 80-84 degrees Celsius, that’s a spicy meatball.
The Final Word
The Quadro 6000 is an incredible step forward for NVIDIA. The board combines great technologies with sheer power and magnitude. With such monstrous memory capacity, it would take a seriously impressive project to bring this GPU to its knees. The Quadro 6000 will stand tall to any challenge thrown at it. This GPU is in a league of its own.
Real-time performance was not as impressive as I had hoped, but it isn’t a deal breaker either. If you’re really concerned about having great gaming performance on a workstation GPU, there are better alternatives. Only the obsessive framerate users would have such a concern, as the Quadro 6000’s real-time performance is more than good enough despite its back seat position to ATI’s solutions. Drivers for Fermi Quadro boards are also in their infancy, however, and performance in this area is a something that definitely has potential to increase with future releases as the drivers mature.
If you’re looking for unbridled GPU power and have the cash to go the distance, then the NVIDIA Quadro 6000 is an outstanding workstation solution that will chew through the DCC polygons. There are few alternatives that will deliver in the way that the Quadro 6000 does. Fermi lives in the workstation world, and it is a hungry, vicious beast. Welcome back, NVIDIA.
Heat and power issues aside, this is the fastest workstation card alive. We’re proud to award it our highest honor, the Icrontic Golden Fedora.
Articles RSS