by Ilya Gavrichenkov
06/04/2008 | 07:15 PM
Our site writes about the new graphics solutions on a regular basis. We examine the graphics hardware in detail and test them all in the latest and most impressive 3D games, because the results of these tests are what most computer enthusiasts are expecting from us. However, we shouldn’t forget that gaming graphics accelerators are not the only types of cards available in this market. In fact, all graphics accelerators can be split into three big groups depending on their application. Besides widely spread gaming cards, the other two groups include integrated graphics and professional cards. Graphics solutions built into core logic sets are primarily intended for office use. They can’t boast high 3D performance, but are inexpensive and cope perfectly fine with the graphics interfaces of the operating system and majority of other applications. Professional graphics accelerators include elite graphics cards that speed up the interface of computer-aided design and 3D modeling applications. Solutions like that boast certain unique features that are unavailable in graphics cards from other categories.
Nevertheless, the main designers of discrete GPUs – AMD and Nvidia – do not develop special processors for the professional graphics cards. Instead, they adapt the same chips that are used for gaming accelerators. As a result, a lot of users out there do not really see any dramatic difference between professional and gaming graphics cards, which often leads to inadequate use of both of them.
Although many differences between the cards of these two types start with their PCBs, the optimizations for different applications are implemented primarily in the drivers. The gaming graphics card drivers are optimized for scenes with relatively few polygons, where bright photo-realistic image is achieved with thoroughly drawn textures and various effects, including pixel and vertex shaders. Professional applications do not strive for photo-realistic visualization. Therefore, they use completely different GPU features: they require high performance during work with numerous polygons with changing positions in wireframe mode or with primitive shading. Texturing speed becomes secondary in this case, and the shaders are hardly used at all. So, the professional graphics card drivers are optimized for maximizing the geometric GPU performance. Since the above described driver optimization is necessary for successful work in 3D modeling and CAD applications, the gaming graphics cards perform relatively slow in them, which puzzles the users most of the time. Professional graphics cards, in their turn, aren’t fast enough in 3D games.
Besides, we have to point out one more thing. Although transferring computational workload from the CPU to the GPU has become pretty popular lately, professional graphics accelerators cannot claim taking part in the final rendering process. So, you have to understand that the professional graphics accelerators have actual effect on performance in CAD/CAM and DCC systems only in viewports, but not during final rendering when processor speed matters most. The only exception to this rule is Nvidia’s brand name Gelato renderer that knows to use the GPU resources. However, its practical value is quite insignificant because the rendered is not widely spread yet.
To back up these statements with some real facts and to see what the current situation in the professional graphics card market looks like, we undertook another test session involving mid-range professional graphics solutions from AMD and Nvidia. Especially, since it’s been a while since our last roundup of professional graphics accelerators for CAD and CAM applications and both manufacturers have already released new product generations. Besides, responding to readers’ numerous questions we have also added the performance comparison against one of the top gaming cards.
So, let’s take a closer look at our today’s testing participants.
ATI FireGL V3600 is the youngest professional graphics accelerator in the current lineup that AMD offers for CAD and CAM applications. However, we can hardly call it a budget solution, because it retails for a little over $200, which is the price of a gaming Radeon HD 3870.
At the same time, ATI FireGL V3600 is not based on the newest GPU, but on the 65nm RV630GL that ahs already stood the test of time and that belongs to inexpensive solutions from the “gaming” prospective. The graphics core frequency is also not that impressive: only 600MHz. It means that ATI FireGL V3600 is very similar to Radeon HD 2600 Pro without the introduced professional optimizations. However, the professional card is equipped with 256MB of DDr2 memory with 128-bit bus that works at 1GHz frequency. So, this professional graphics accelerator compatible with DirextX 10, Shader Model 4.0 and OpenGL 2.1 features 120 shader processors, provides about 2.4GPixel/s peak fillrate and has memory with 16GB/s bandwidth.
The exterior looks of ATI FireGL V3600 turned out exactly as we have expected. It uses the same PCB as Radeon HD 2600 Pro. The only distinguishing feature is a small cooler that is made of copper in this case. This cooling system cools only the GPU: the memory chips on both sides of the graphics card PCB do not have any cooling of their own.
The card is compatible with PCI Express x16 interface version 2.0 and is equipped with two Dual-Link DVI outs, so it supports monitors with up to 2560x1600 resolution.
The manufacturer positions its next professional graphics accelerator, ATI FireGL V5600, as a more advanced solution than FireGL V3600. It costs about $500. Nevertheless, this card is based on the same RV630GL chip compatible with DirectX 10 and OpenGL 2.1.
So, the GPU used in ATI FireGL V5600 can boast only 120 shader and 8 texturing processors. This graphics card performs faster (ATI FireGL V5600 offers 3.2GPixel/s fillrate) due to higher core frequency of 800MHz. it also has more graphics memory: 512MB. While its memory bus is equal to the same 128bit, the 35GB/s bandwidth is achieved by clocking the memory to 2.2GHz, which was possible thanks to the latest GDDR4 SDRAM chips.
In other words, ATI FireGL V5600 looks very much like Radeon HD 2600 XT equipped with GDDR4 memory. Besides the similar technical characteristics, these two different graphics accelerators also boast similar PCB layouts. In fact, the only noticeable visual difference is the absence of a TV-Out on a professional card. Even the cooler is also borrowed from the gaming counterpart. It is a single-slot active solution made of copper and aluminum that is cooled with a small centrifugal fan.
ATI FireGL V5600 card is a little bigger than ATI FireGL V3600. Nevertheless, the graphics card receives enough power via the PCI Express x16 2.0 bus and there are no additional power connectors on the PCB. However, there are two Crossfire connectors, which do not really have any real practical value: professional graphics card drivers from AMD do not support multi-GPU modes.
However, to be fair we have to say that nevertheless it is possible to install several FireGL cards into the same system and each of the cards will work independently with its own monitors. Just like ATI FireGL V3600, ATI FireGL V5600 features two Dual-Link DVI Outs supporting LCD panels with resolutions up to 2560x1600.
Mainstream professional graphics accelerators from AMD are currently represented by two solutions: ATI FireGL V7600 and ATI FireGL V7700. Unfortunately, AMD didn’t provide us with their newer solution, V7700 that is why we had to test only the FireGL V7600 card. This card currently sells for $900.
ATI FireGL V7600 is based on a relatively old R600GL GPU manufactured with 80nm process. Although the manufacturer formally puts ATI FireGL V7600 and ATI FireGL V7700 into the same generation and price range, ATI FireGL V7700 looks a little bit more attractive, as it is based on a 55nm RV670 chip. However, both 7XXX graphics accelerators perform at close speeds in professional applications.
ATI FireGL V7600 core with 240 shader and 16 pixel processors works at 600MHz. So, in terms of its major technical parameters and theoretical fillrate of 9.6GPixel/s ATI FireGL V7600 is comparable with Radeon HD 2900 GT. Just like the gaming analogue, this professional solution features 512MB of GDDR3 SDRAM working at 1.6GHz frequency. So, the memory bandwidth on ATI FireGL V7600 with a 256-bit bus equals 51GB/s.
Looks like GPU-Z utility faces some difficulty in detecting the specs
of our ATI FireGL V7600…
Since this graphics card uses a relatively old chip, ATI FireGL V7600 is incompatible with PCI Express x16 2.0 and the card supports only the first version of this protocol providing half the bandwidth. However, professional graphics cards very rarely need to transfer large textures and do not support Crossfire, so incompatibility with the new bus standard is not that big of a deal.
The exterior looks of ATI FireGL V7600 is very similar to that of its gaming analogue. It has a common form-factor for graphics cards of this class (242mm long) and a massive dual-slot cooler using some copper parts, two 6mm heatpipes and a powerful centrifugal fan. This cooler is covered with a plastic casing and exhausts the air outside the system case. The card requires additional power supply, so it has an 8-pin power connector, which can also use a 6-pin plug.
On the card rear panel there are two Dual-Link DVI outs and a connector for stereoscopic glasses, which can only be found on professional graphics accelerators these days. I would like to add that ATI FireGL V7600 is bundled with a component adapter for HDMI devices.
High-performance professional graphics cards in the AMD camp are represented by two solutions: ATI FireGL V8600 and ATI FireGL V8650. These two cards are practically identical, the only difference between them is in the amount of onboard graphics memory. At the time of the review we managed to get our hands on the less expensive ATI FireGL V8600 model equipped with 1GB of graphics memory. But even its price seems outrageous: it retails for $1600. The modification with 2GB of memory, ATI FireGL V8650, will cost you additional $800. However, CAD and CAM applications very rarely require that much video memory, so most users will be ok with ATI FireGL V8600.
This graphics card from the upper price range is based on the same R600GL chip as ATI FireGL V7600. However, in this case it runs at higher frequency of 688MHz. So, the theoretical filtrate provided by ATI FireGL V8600 reaches 11GPixel/s. ATI FireGL V8600 has one more distinguishing feature. Although this graphics card is based on the same GPU as ATI FireGL V7600, it has more streaming processors: 320. Together with higher core frequency it affects the performance. The memory on ATI FireGL V8600 also works faster (they used GDDR4 SDRAM). Its frequency is 1.73GHz and the bus is 512bit wide. As a result, ATI FireGL V8600 boasts over 110GB/s memory bus bandwidth.
AMD chose a pretty advanced GPU from R600 series for its high-performance CAD and CAM graphics cards, however, once the newer graphics processors, such as RV670, appeared, it doesn’t seem that advanced any ore. However, despite this fact, ATI FireGL V8600, like ATI FireGL V7600, is fully compatible with DirectX 10, Shader Model 4.0 and OpenGL 2.1. The only thing about ATI FireGL V8600 that doesn’t meet the today’s standards is the absence of PCI express x16 2.0 support.
ATI FireGL V8600 exceeds ATI FireGL V7600 from all aspects, but doesn’t have any gaming analogues. It is similar to Radeon HD 2900 XT, but not more than that. This professional graphics card features a lot of onboard video memory running at higher frequency. And it means that AMD had to develop unique PCB layout specifically for ATI FireGL V8600 graphics card.
ATI FireGL V8600 looks really impressive. It is a full-size PCI Express x16 graphics card 285mm long. There is a metal plate on the bottom side of the PCB that ensures robustness of the card and is used to lock the card securely in special workstation cases. Of course, you can remove this pate and fit ATI FireGL V8600 into some regular cases, but there are very few cases it will fit in, so you will have to put up with this fact.
The cooling system is also highly original. It looks like a stretched cooler that we saw on ATI FireGL V7600: the heatsink ribs are also made of copper and you can see two heatpipes inside, there is a centrifugal fan and the whole thing is covered with a similar casing. However, ATI FireGL V8600 cooler features a unique base: it is a massive aluminum plate that covers the entire card and features a pin-heatsink above the voltage regulator circuitry. As a result, ATI FireGL V8600 can be called the heaviest graphics card, because it is dressed up in “plated” metal armor.
The size of this cooling system suggests that this card requires pretty serious power supply. True, as there are two power connectors on the PCB: a 6-pin one and an 8-pin one. You should use both of them at all times. Moreover, the 8-pin connector is compatible with six-pin plugs. AMD recommends using at least 650W power supply units in systems with ATI FireGL V8600.
Note that the card also has a connector for a daughter board with Genlock and Framelock functionality. However, as far as we know, AMD doesn’t offer boards like that at this time. ATI FireGL V8600 also has the same useless Crossfire connectors.
Like all other representatives of the FireGL family, ATI FireGL V8600 supports up to two monitors with 2560x1600 maximum resolution that can be connected to two Dual-Link DVI Outs. The card is bundled with DVI → D-Sub and DVI → Component HDTV adapters. There is also a connector for stereoscopic glasses on the card’s rear panel.
AMD drivers for ATI FireGL graphics card family looks very much like the software most gamers are familiar with. It can be managed in exactly the same way: through the Catalyst Control Center.
I would like to point out a few peculiarities of the ATI FireGL driver. First, it doesn’t offer you to choose any preset profiles optimized for different professional applications. The reason for that is the support of AutoDetect technology that automatically selects the necessary profile depending on the type of application you are running.
The second peculiarity is the support of 10-bit color instead of the standard 8-bit. This expansion of color space is used primarily in medical applications that lack gray gradients for tomography and x-ray visualizations. For these particular cases AMD offers 10-bit color representation and 1024 gray gradients instead of the standard 256.
Although contemporary professional graphics accelerators from Nvidia start from Quadro FX 370 we decided not to review this card today. For the price of $120 it offers very low performance and cannot compete successfully against the youngest competitor’s solutions. That is why the slowest professional graphics accelerator from Nvidia participating in our today’s roundup will be Quadro FX 570. It sells for only $60 more - $180 – but its GPU works at higher frequency and its memory bus is twice as wide.
Quadro FX 570 (just like Quadro FX 370) is based on G84GL chip manufactured with 80nm technological process. We can’t say that it belongs to the most progressive Nvidia GPU generation, because they already have 65nm chips at their disposal. However, this GPU supports DirectX 10, Shader Model 4.0 and OpenGL 2.1, so it is still quite up-to-date.
On the gaming side, GeForce 860 graphics cards use the same chip, however, there is no complete analogue to Quadro FX 570 in the gaming camp. The thing is that the GPU of this professional graphics accelerator works at a surprisingly low frequency of only 460MHz (the shader domain frequency equals 920MHz). Besides, it only has half the shaders – 16. And since the GPU has 8 rasterizers, Quadro FX 570 theoretical fillrate makes 3.7GPixel/s.
Since technical specifications of Quadro FX 570 solution are not like those of any Nvidia’s gamin cards, we were not surprised to discover that its features a unique PCB layout. The card looks like a budget solution: it is small, has very simple layout, has no additional power supply connector, a small chip cooler with an aluminum heatsink.
Quadro FX 570 features 256MB of graphics memory. It has the same amount of memory as the professional budget solution from AMD has. Nvidia card uses the same DDR2 SDRAM chips and 128bit memory bus, however, its memory works at lower frequency of 800MHz. As a result, the memory bus bandwidth of this Nvidia solution equals only 12.8GB/s.
The card uses PCI Express x16 v2.0 bus and is equipped with two Dual-Link DVI Outs for monitors with 2560x1600 maximum resolution.
Nvidia Quadro FX 1700 professional graphics card belongs to mainstream segment and costs $420. However, we were very surprised to find out that it looks exactly like the budget Quadro FX 570 described above. Of course, it is also based on the same G84GL graphics processor.
The frequencies of the Quadro FX 1700 solution are even more puzzling: they are exactly the same as those of Quadro FX 570. The chip works at 460MHz and the memory – at 800MHz. So what are the differences between these two solutions from different market segments? There are not that many of them actually. First, Quadro FX 1700 has twice as much DDR2 SDRAM onboard: 512MB. However, despite larger memory capacity the memory bus is still the same 128-bit one. And second, the GPU has the full set of 32 shader processors.
The third difference can be noticed with a bare eye. Quadro FX 1700 has not just two Dual-Link DVI connectors, but also a component out for HDTV devices.
By the way, just like its younger sister, Quadro FX 1700 is pretty economical. It doesn’t require additional power and is OK with a small active cooler. According to Nvidia, its peak power consumption doesn’t exceed 42W.
The newest professional graphics accelerator from Nvidia is Quadro FX 3700. The manufacturer positions it as a high-end solution that can be found in retail for about $1100. However, Nvidia also has the so-called Ultra High-End professional graphics cards, such as Quadro FX 4600 and Quadro FX 5600. We will not discuss these two solutions today, because they are priced much higher than the top competitor’s offerings, so it doesn’t make much sense to compare them. Especially, since you can check out our article called Nvidia Quadro FX 5600 and Nvidia Quadro FX 4600: New Leaders in Professional Graphics devoted solely to high-speed professional solutions from Nvidia.
Since they launched Quadro FX 3700 in very beginning of this year, it is not at all surprising that this graphics card uses the latest G92GL GPU manufactured with the most advanced 65nm process. As you probably know, these Nvidia GPUs are also used in their latest generation gaming solutions from GeForce 9800 family. However, Quadro FX 3700 looks more like GeForce 8800 GT, because it only has 112 shader processors. However, it is not that bad for a professional graphics card. The 500MHz chip and 1250MHz shader domain frequency is a much bigger issue. As a result, this professional card provides about 8GPixel/s fillrate and yields in this aspect not only to the gaming counterparts, but also to the competitor solutions.
As for the memory, Quadro FX 3700 uses 512MB of GDDR3 SDRAM working at 1600MHz. Since the G92GL chip uses a 256-bit bus, the memory bandwidth on this professional solution makes 51GB/s.
You can see that Quadro FX 3700 and GeForce 8800 GT are related from the way these two cards look. In fact, the professional solution uses the same PCB as the gaming card. There is a single-slot cooler covering the entire front side of the PCB that dissipates the heat from the chip and the memory. It is made of copper and aluminum, features a metal casing and a small centrifugal fan creating air circulation beneath it. However, there is nothing original about this cooling system: it was also borrowed from the gaming GeForce 8800 GT, where this cooler proved to be highly efficient and pretty quiet.
The card has a 6-pin connector for additional power supply, which must be used as Quadro FX 3700 consumes up to 78W of power. Besides, there is an SLI connector. I would like to stress that unlike professional AMD solutions, Quadro FX drivers support multi-GPU rendering. Moreover, Nvidia even announced a dual-chip professional graphics accelerator not that long ago, which was based on two G92Gl chips – Quadro FX 4700 X2.
Quadro FX 3700 is compatible with the latest version 2.0 of PCI Express x16 interface and features two Dual-Link DVI connectors for monitors with up to 2560x1600 resolution. Besides, the card has an additional connector for 3D glasses.
Drivers for professional Nvidia Quadro FX graphics cards remind us of Forceware that we all know very well from the gaming solutions. The control panel looks the same and offers almost the same settings. The only difference is that the offered settings profiles belong to the professional applications and not to 3D games.
Just like Catalyst, Nvidia drivers can automatically switch between different profiles depending on the running application. However, Forceware software allows editing these profiles manually, while catalyst doesn’t.
I would like to point out that Nvidia offers not only basic drivers for its professional solutions. You can also download two additional utilities compatible with Quadro FX from the company web-site: MAXtreme D3Dand Powerdraft. These are specialized drivers for Autodesk 3ds max and Autodesk AutoCAD. These special drivers help boost the professional graphics cards performance in the above listed CAD applications.
Before we move on to the actual benchmark results of our today’s testing participants, we have to sum up their technical specifications side by side for a more illustrative comparison. It will help us interpret the test results adequately.
We tested our professional graphics cards in a workstation based on the today’s fastest quad-core Intel Core 2 Extreme QX9770 processor working at 3.2GHz. Our testbed also used an Intel X38 Express based mainboard and 4GB of DDr3 SDRAM.
All tests were performed in 64bit Windows XP Professional. Unfortunately, we couldn’t complete the tests in a more contemporary Windows Vista because some CAD/CAM applications are not completely compatible with it.
As we promised, we also included the results of a top single-chip gaming graphics accelerator – Nvidia GeForce 9800 GTX. It is based on G92 GPU with 128 shader processors and is equipped with 512MB of memory. This card worked at its standard frequencies: 600MHz for the chip and 2000MHz for the memory.
So, we ended up using the following equipment to put together our test platforms:
The cards were tested in 1920x1200 resolution with Vsync disabled, unless stated otherwise.
First of all we decided to check the results in the main professional graphics card test - SPECviewperf benchmark, which has already become a set industry standard. Although SPECViewperf is a synthetic test, it shows very well the performance of graphics accelerators in OpenGL, which is determined by hardware peculiarities as well as driver optimization quality. This benchmark transmits to the graphics card driver a pre-formed succession of OpenGL commands typical of different professional applications. The built-in scripts emulate the user’s work in viewports of the following applications (the names of the corresponding benchmarks are given in brackets): 3ds max (3dsmax-04), CATIA (catia-02), EnSight (ensight-03), Maya (maya-02), Pro/ENGINEER (proe-04), SolidWorks (sw-02), UGS Teamcenter Visualization Mockup (tcvis-01) and UGS NX (ugnx-01).
The obtained results are pretty diverse. You can see that there is no universal professional solution. FireGL cards show best results in 3ds max, EnSight, Maya, Solidworks, VisMockup and UGS NX, while in CATIA and Pro/ENGINEER the winner’s laurels go to Quadro FX family. As for the top gaming solution, GeForce 9800 GTX, its performance is below any possible criticism, since its OpenGL driver is optimize for gaming and not for professional needs.
I have to say that SPECviewperf offers one more interesting feature. It allows measuring the performance drop with enabled anti-aliasing (FSAA). The diagrams below show how the results of standard SPECviewperf tests change in different FSAA modes:
I would like to say that sometimes anti-aliasing helps Nvidia cards outperform the pricing and positioning analogues from FireGL family.
However, do not blindly follow the obtained results: the practical value of SPECviewperf numbers should be taken with certain allowances. This test does show the peak performance of the OpenGL driver in real applications. However, you should bear in mind the fact, that this benchmark’s scripts use far not the latest versions of CAD and CAM systems. Although SPECviewprf benchmarking suit is updated regularly, it still uses almost 5-year-old model of applications behavior. Moreover, do not forget about the ever growing tendency to transfer professional applications from OpenGL to DirectX that can currently offer more options, especially when it comes to shader processing. For example, Microsoft API is already moving into 3ds max and AutoCAD, which is not taken into account by the benchmark. Therefore, SPECviewperf tests cannot really replace real professional applications that we are going to talk about next.
The peculiarity of 3d max 9 tests is that it uses not OpenGL but DirectX. They recommend this particular API for the latest versions of this popular 3D modeling suite. Besides, I would like to say that Nvidia offers a special driver for 3ds max application: MAXtreme D3D, which speeds up the performance in viewports even more.
Quadro FX graphics cards look better in general performance tests of contemporary 3ds max suite. However, if we take a closer look at the detailed performance check, we will be able to single out a few very curious tendencies. Check out the diagrams below illustrating the results of specific operations:
As we see, Quadro FX cards are not always better in 3ds max. For example, they lose shader performance tests to AMD solutions.
Moreover, the real 3ds max tests show that the gaming graphics accelerator is not that dramatically slow any more. Yes, GeForce 9800 GTX loses to high-performance professional solutions, but can compete pretty successfully against the cards from the 500-dollar range.
Another application that prefers DirectX to OpenGL is AutoCAD. I would like to remind you that Nvidia software developers are theoretically offering their special PowerDraft driver for this suite as well. However, we couldn’t use it again, because it currently works only with a 2008 version of the suite and only in 32-bit Windows XP OS. In other words, a special Quadro FX driver for AutoCAD is a formal advantage of Nvidia’s professional graphics card. In reality, far not all the users will be able to really take advantage of it.
Professional Nvidia Quadro FX cards yielded to FireGL solutions in this test. But the most unexpected result was demonstrated by the gaming GeForce 9800 GTX that outperformed 1000-dollar professional accelerators here. Let’s take a closer look at the individual test results to find out why:
I believe no comments are necessary here. Without the PowerDraft driver, Quadro FX series cannot compete against GeForce 9800 GTX. However, the top ATI FireGL solutions, do not give in to the gaming card.
Professional ATI FireGL cards are far ahead their competitors in Maya 2008. It has already turned into a tradition: ATI (and now AMD) solutions used to and still do outperform Quadro FX in this test, despite the changing GPU generations. Moreover, the performance advantage has become so dramatic that Quadro FX 3700 yields even to ATI FireGL V5600 selling for half the price of the Quadro.
We see very curious result in SolidWorks 2008. The highest performance belongs to the gaming GeForce 9800 GTX that left everyone behind, including much more expensive professional solutions.
Besides the real applications, we also resorted to a special CINEBENCH R10 test. It shows the systems performance in Cinema 4D suite that is extremely popular among Mac users and fans.
According to this benchmark, professional graphics solutions from AMD win this round in all price ranges.
Although professional graphics accelerators are hardly ever used for gaming, we decided to add the results of a popular FutureMark 3DMark06 gaming benchmark. Unlike CAD and 3D modeling applications, this benchmark can test the GPU computational capabilities used for shader processing best of all.
This is where we see that the gaming workload is very different from the load in professional applications. The results of 3Dmark06 are completely different, as it doesn’t care about how fast the triangles are displayed, but needs to process shaders, which are barely used in professional graphics.
Here we are going to discuss the measured power consumption level of complete systems (without the monitor) equipped with different professional graphics accelerators. The measurements were taken in idle mode and under workload - in 3dsmax-04 test from the SPECViewperf 10 suite.
The results are quite logical. The systems equipped with high-performance professional graphics accelerators based on 80nm chips require most power. That is why the top ATI FireGL solutions will hardly be the best choice for power-conscious users. Quadro FX 3700 demonstrated pretty attractive balance between performance and power consumption. However, it is the newest card among all our today’s testing participants: it is based on a 65nm chip.
The power consumption of ATI FireGL V5600 and V3600 turned out higher than that of Quadro FX 1700 and FX 570, although these AMD products use GPUs manufactured with more advanced production process. However the youngest Nvidia graphics cards are also slower than their rivals, since their clock speeds are lower to begin with.
Or today’s tests of contemporary mid-range graphics accelerators once again confirmed that there remains parity between AMD and Nvidia solutions. In fact, it means that there are no evident favorites and outsiders. In certain applications the winner may be different. For example, ATI FireGL solutions perform best in Maya and AutoCAD, while Nvidia takes over the leadership in SolidWorks and 3ds max.
So, we cannot really recommend any particular professional graphic accelerator at this time. These cards are pretty expensive and choosing the right solution for your needs is very serious and responsible task. Especially since true professionals rarely use different application suites at the same time. Therefore, the choice of the optimal graphics accelerator for a workstation should first of all depend on the way this workstation is used. All the data you might need to make the right choice is given in this article.
In addition to the conclusions regarding the performance made above, I would like to say that we do not see any parity from the power consumption standpoint. The top AMD solutions use 80nm chips, so they turn out considerably less economical than Quadro FX cards. Therefore, it would be really interesting to check out the new ATI FireGL V7700 graphics card based on the brand new RV670, which we are going to do in the nearest future.
The last thing we can say basing on the obtained results, is that gaming graphics cards can also demonstrate good performance in contemporary CAD and CAM systems. The situation has changed: the top gaming graphics accelerator we took as n example this time – GeForce 9800 GTX – was completely defeated by its professional opponents only in one synthetic SPECviewperf testing suite working with an OpenGL driver. In real applications the systems equipped with this card performed very well. Moreover, in some tasks such as SolidWorks 2008 and AutoCAD, it worked even faster than 100-dollare professional solutions.