by Ilya Gavrichenkov
03/20/2006 | 03:02 AM
Those who have been visiting our site for quite a while now have already understood that we primarily focus on the hardware components that would make up a high-performance desktop platform. The reason for that is certainly your, our readers’, interest to this type of hardware in the first place. However the news from the market sometimes force us to deviate a little bit from the main course and pay due attention to less typical but not any less interesting solutions. For example, we started devoting more time to products that are more suitable for professional users rather than hardware enthusiasts. For instance, not so long ago we posted a comparative testing report on the dual-core Intel Xeon and AMD Opteron processors (for details please see our article called Workstation Processors Duel: AMD Opteron against Intel Xeon). The response to this article was so great that we decided not to stop there and continue delivering more information on the expensive high-end hardware for professional applications.
Today we are going to pay special attention to professional graphics accelerators. This test session is extremely interesting because the main competitors in this market are the two well-known names – ATI and Nvidia. They are offering the professional users the specially adapted modifications of their gaming solutions. It is a really interesting question if the situation with professional ATI and Nvidia graphics cards in 3D modeling applications will resemble what we are used to seeing in the gaming world.
Here it is important to keep in mind that the professional graphics cards market is a very tasty morsel for the manufacturers. The thing is that the graphics cards for professional use built from the same component base as the gaming graphics solutions are being sold at a much higher price than the gaming product. The price difference may sometimes be close to 4-5 times, which is no joke already. This mere fact inspires ATI and Nvidia to take active part in development of corresponding professional solutions and promote them into the market alongside with the gaming Radeon and GeForce cards. Creative Company has recently turned into an unwilling stimulant for this development process having announced the reorganization of their 3DLabs division that used to offer the most high-performance solutions for 3D modeling applications.
In our today’s review we are going to introduce to you a new flagship product from the professional 3D modeling graphics card family by ATI. I have to point out right away that it is very unlikely to surprise a sophisticated reader from the architectural standpoint, because it takes quite a bit of time before the professional solution follows the gaming product line-up on the same GPU, so there appears quite a bit of information around about the chip architecture. This time is spent on software (BIOS and drivers) debugging and modification. However, it would be totally unfair towards those guys who work in 3D modeling applications if we didn’t pay due attention to the new professional graphics solution. So, here we go!
ATI FireGL V7100 has been ATI’s top professional graphics solution until today. Note that since this graphics card is over 1 year old already it hasn’t been a threat for the latest NVIDIA Quadro competitors lately. No wonder since FireGL V7100 is based on R423 architecture, which corresponds to the power of Radeon X800 XT.
So, ATI really needed to upgrade its top of the line professional graphics solution with something newer. Moreover, the developer has long had the opportunity to do so: new R520 and R580 GPU architecture has long been used for the gaming graphics solutions and has proven very successful. However, until recently, ATI didn’t dare move this architecture into the professional field, which requires a much higher level of reliability, while the typical working conditions for the solutions of this kind can usually be much more severe. Moreover, it takes much more from the software developers when it comes to launching a professional graphics accelerator. The work on the BIOS and drivers optimized for the most widely spread 3D modeling suites may take much more time and effort, which is one of the reasons why it takes the professional graphics solutions much longer than their gaming brothers based on the same architecture to finally make it to the market.
As a result, the second generation of FireGL V cards for the PCI Express x16 bus based on the 90nm R520 chip is coming to the market only today. The graphics cards for professionals based on R520 architecture will be available in two major modifications called FireGL V7350 and FireGL V7300 respectively. These graphics cards boast similar features and the difference lies only in the amount of onboard graphics memory. While ATI FireGL V7300 is equipped with 512MB of GDDR3 video memory, the ATI FireGL V7350 carries 1GB of memory onboard. As a result, ATI FireGL V7350 can be considered the industry’s first mass graphics accelerator equipped with so much video memory. In fact, the need for so much memory is arguable even for a professional graphics solution. There won’t be too many users today who would be interested in getting themselves one of the new ATI FireGL V7350 cards: there are still very few tasks that can really make use of all this memory for texture storing. So it looks like ATI FireGL V7300 will still become a more popular product in this family. Especially since its recommended price is about $400 lower than that of the top model.
All in all, the specifications of ATI FireGL V7350 and FireGL V7300 are very similar to those of ATI Radeon X1800 XT. So if you really want to refresh your memory and read more about the architecture and functionality of the new graphics accelerators, you should take a look at our previous article called ATI RADEON X1800 XT and XL Performance: Crushing NVIDIA's 7800? As for us, we believe it would make sense to repeat just a few key things here about the new ATI FireGL V7350 and FireGL V7300: these cards can lay 16 textures per clock cycle, feature 8 shader pipelines supporting Shader Model 3.0 and are fully compliant with DirectX 9 and OpenGL 2.0 API.
Also you shouldn’t forget that professional graphics cards also feature the Avivo video engine that ensures high quality video playback and allows H.264 hardware decoding supported by the new generation DVD. It is important to note that the Avivo version implemented on the mew professional graphics cards should also provide future support for Genlock and Framelock (professional technologies for output synchronization to an external source and cluster visualization). The daughter card for implementation of these technologies should be available in H2 2006.
It is important to mention that ATI doesn’t yet claim that the new FireGL V7350 and FireGL V7300 will support configurations with more than one graphics cards installed into a single workstation. At the same time, the competitive Nvidia solutions do offer SLI support in the professional field, too. Maybe the reason here is the fact that ATI doesn’t yet have any chipset solutions for the workstation platforms, so there are really no platforms that would allow implementing CrossFire technology with FireGL accelerators. At the same time it is essential to understand that there is no 3D modeling software that would be optimized for multi-GPU platforms at this time, and as for the full-screen work modes, these applications do not normally use them.
Now let’s take a closer look at our today’s main hero – the top-of-the-line ATI FireGL V7350 professional graphics card with 1GB of onboard memory. I have already said a few times today that this graphics card has a lot in common with the ATI Radeon X1800 XT. This is also true for the exterior appearance and design of the card:
The front side of the FireGL V7350 card differs from the corresponding gaming solution only by the decorative picture on the cooling system.
The reverse side of the PCB is very much different, this is where your eye catches all the major distinguishing features at once: ATI FireGL V7350 has additional memory chips here that are absent by the gaming cards with only 512MB of memory.
Of course, I would like to say a few words about the cooling system of the ATI FireGL V7350. It requires an additional slot and is built with a 60mm turbine that throws warm air outside the case right away. Note that this cooling system serves to cool down the GPU and the memory chips on the front side of the PCB. As for the memory chips on the reverse side of the PCB, they do not use any cooling at all.
The similarity between ATI FireGL V7350 and ATI Radeon X1800 XT is also proven by the component base used for these cards. Both of them are equipped with the same GDDR3 memory chips – Samsung K4J52324QC-BJ12. And even though the graphics processor is marked as FireGL, the way it looks suggests very clearly that it is a close relative of the ATI R520 GPU.
As for the working frequencies of the ATI FireGL V7350, they do not correspond to those of the Radeon X1800 XT. The professional card works at lower clock speeds: 600MHz on the chip and 1300MHz on the memory. Here I would like to stress that professional graphics cards do not have the dynamic frequency change mechanism when they switch from 2D to 3D mode. Therefore, the numbers listed above are true for any work mode of the ATI FireGL V7350 graphics card. So, the performance of ATI FireGL V7350 in gaming applications will be slightly lower than that of Radeon X1800 XT (its working frequencies are 625MHz for the chip and 1500MHz for the memory), because the increase in the amount of onboard memory will in no case ensure any performance growth in this type of applications.
As for the performance in professional applications, which is of primary interest to us today, of course, a lot depends on the drivers here. The developers have special drivers ready for their new ATI FireGL V7350 and FireGL V7300 graphics cards. These drivers are optimized for work in specific professional 3D modeling applications. The list of these applications is constantly available on ATI’s web-site .
Together with the ATI FireGL V7350 graphics card we also received the driver version 8.223.
Although this driver offers suspiciously few setting options, especially compared with the drivers for the competitive solutions from Nvidia, it contains the necessary optimizations for all major applications used by the today’s professional users.
ATI positions the new professional FireGL V7350 graphics accelerator as a primary rival to Nvidia’s Quadro FX 4500 based on the same G70 architecture as the GeForce 7800 GTX gaming graphics solutions. Moreover, Nvidia Quadro FX 4500 boasts the same design as Nvidia GeForce 7800 GTX 512MB, including the cooling system with four heatpipes and 80mm fan. You can read more about the Nvidia Quadro FX 4500 design and features in our article called Performance Monsters from AMD: Dual-Core CPUs in a Dual-Processor Workstation .
If we compare the specifications of the new top professional graphics product from ATI against the competitor from Nvidia, the picture will be pretty ambiguous:
ATI FireGL V7350
NVIDIA Quadro FX 4500
PCI Express x16
PCI Express x16
GPU production technology
Memory bus width
Amount of onboard
2 x DVI
2 x DVI
Dual-link DVI support
$1800 (average retail)
Judging by the table above it would be pretty hard to claim the superiority of ATI FireGL V7350 over Nvidia Quadro FX 4500. In fact, the only characteristic of the ATI solution that is better over Nvidia Quadro is the amount of onboard graphics memory. From the architectural standpoint Nvidia Quadro FX 4500 seems to be a more suitable solution for the professional application field. ATI FireGL V7350 yields to the rival in the lower number of texturing units in the first place.
I would like to point out that the workload laid on the GPU by professional 3D modeling applications is very much different from the gaming workload. The thing is that the gaming industry decided to give up active use of multi-texturing in favor of shader effects. As a result, the developers of contemporary graphics accelerators have also adjusted their approach to hardware designs. They moved away from simple increase in the number of texturing units towards massive improvement of the GPU computational power required for faster and more efficient shader processing. ATI R580 chip currently used in Radeon X1900 graphics card family is a vivid example of this concept implementation. This processor features only 16 texturing units thus yielding to G70 with 24 units of the kind. However, this 1.5 times difference doesn’t really affect the gaming performance, because R580 boasts huge computational power, which we have already told you about in our previous articles, such as The Fast and Furious: ATI Radeon X1900 XTX Review . However the use of the same architecture in professional applications can have really bad outcome for the developers. The thing is that contemporary 3D modeling applications use very limited number of shaders. The image is output onto the screen in a good old manner: by laying textures one over another. Therefore, shader performance is of secondary importance for professional accelerators, while the geometry and texturing speeds become the top priority parameters. In other words, if in games ATI Radeon X1900 XT turns out overall faster than Nvidia GeForce 7800 GTX, then in professional applications the situation may turn out completely different.
This is why they didn’t base their professional solutions on the existing R580 chip, which is in mass production right now. The higher shader performance it boasts is of simply no use in this case. This is exactly why ATI decided to use the older R520 GPU for its professional graphics cards. This chip is free from any excessive extras and at the same time boasts identical geometrical performance and texturing speed.
The primary goal of this test session was to compare the performance of the top-of-the-line professional solutions from Nvidia and ATI. We tested these graphics cards on a dual-processor platform built with AMD Opteron processors and an NVIDIA nForce Professional based mainboard. Our test platform was configured as follows:
We used the following drivers:
Vsync was disabled everywhere, if possible.
First of all we decided to test the performance of our top professional testing participants in SPECviewperf 8.1 benchmark, which serves as a sort of reference when it comes to graphics cards performance in 3D modeling tasks. This benchmark shows the performance during rotation of typical models created in various professional applications.
Note that this benchmark doesn’t allow us to evaluate the performance of our testing participants in each application separately. Since the developers do not optimize their drivers specifically for SPECviewperf 8.1, this benchmark can be regarded as a great tool for “raw” performance analysis of the graphics card and the driver in OpenGL. Here are the obtained results:
According to the results, Nvidia Quadro FX 4500 retains the leadership in most cases. The reasons for that have already been described above: the “raw” performance of this GPU is higher due to more texturing units. Nevertheless, the new ATI FireGL V7350 is up to the mark in most cases. As we can see it is fatally behind only in three tests: maya-01 (Maya), sw-01 (SolidWorks) and ugs-04 (Unigraphics). However, there are a few models where ATI FireGL V7350 card turns out an indisputable winner. They are 3dsmax-03 (3ds max) and light-01 (Lightscape).
Besides the SPECviewperf tests, we have also tested our professional graphics accelerators in some widely spread 3D modeling suites with the SPECapc test scripts.
All in all, ATI FireGL V7350 can boast higher speed than Nvidia Quadro FX 4500.
However, these results demonstrate only the performance in OpenGL mode. As is known, 3ds max can also work in DirectX. And those professional graphics accelerators that are based on the initial gaming GPU designs are usually faster when working via DirectX. And our results prove this point, actually.
As for the ATI FireGL V7350 performance, it manages to defeat Nvidia Quadro FX 4500 here. The ATI newcomer is faster in all test modes and its average performance rate is about 12% higher than the average performance of the rival. The performance advantage is the biggest when we work in wireframe mode.
Maya is another professional suite for 3D modeling, which has become a de-facto standard when it comes to computer animation projects.
In Maya ATI FireGL V7350 cannot boast the high results we have just observed in 3ds max. NVIDIA Quadro FX 4500 retains its leadership here.
SolidWorks 2005 is an integrated 3D modeling environment for creating mechanical CAD designs.
The results of ATI FireGL V7350 are again pretty disappointing. The older solution from Nvidia based on G70 architecture works over 50% faster here.
AutoCAD 2006 doesn’t need an introduction, I assume. This is a popular software tool for computer-aided designing and modeling.
However, the performance of our newcomer turned out quite low here as well. It is probably the raw ATI driver that led to ATI FireGL V7350 turning a few times slower than the Nvidia Quadro FX 4500. Hopefully, the upcoming driver releases will eliminate performance issues in AutoCAD suite.
Besides the SPECapc family of applications we have also resorted to a special CINEBENCH 2003 test. It shows the systems’ performance in a popular Cinema 4D application, which is especially beloved by Mac users.
Note that here both professional graphics accelerators run about neck and neck.
Although professional graphics accelerators are hardly ever used for games, we decided to still run the FutureMark 3DMark06 test. The thing is that unlike 3D modeling applications, this benchmark allows to better estimate the computational power of the GPU used for shader processing.
This is where the excellent shader processing speed of the new ATI FireGL V7350 solution reveals itself. Unfortunately, most of the contemporary 3D modeling applications and tools cannot take advantage of this ATI’s strength, because they deal primarily with triangles processing and texturing, where ATI FireGL V7350 appeared to be not very fast.
It is evident that the new professional ATI graphics solution based on the 90nm R520 GPU has lifted the performance level of the ATI’s professional family to a completely new level. However, despite this fact we cannot yet claim that ATI FireGL V7350 is the today’s fastest professional graphics solution. So far ATI FireGL V7350 is just a worthy competitor to Nvidia Quadro FX 4500, but not an indisputable leader. In some tasks Ati wins the race, in another tasks – Nvidia leads. So, if you are looking for a professional graphics solution, you should first of all determine the type of applications you will be working with, because the price of these two products is about the same at this time, so cannot be a crucial factor.
For example, our test session showed that ATI FireGL V7350 boasts high speed in 3ds max, this application is a definite trump of the newcomer from ATI. At the same time, the performance in Maya is not that great, and the results in AutoCAD also leave much to be desired. This can be partially explained by far not the best driver optimization. So, in some cases the performance of ATI FireGL V7350 can be significantly improved with the new driver updates. However, I wouldn’t really hope that new drivers can change things dramatically, especially since Nvidia is also working on new software and moreover, is going to launch a new next-generation professional Quadro graphics card fairly soon.
Summing up the results of our today’s test session I have to admit that the new professional FireGL graphics card from ATI didn’t manage to win a convincing victory over the old rival – Quadro FX 4500. From the features prospective, no superiority can be pronounced either. It basically means that there won’t be overwhelming demand for the new ATI FireGL V7350 solution. This solution will be of interest either for those users who work in professional applications where ATI FireGL V7350 shows the best results, or for those users who need 1GB of onboard graphics memory for work with large chunks of texturing data.