by Sergey Lepilov
04/05/2010 | 05:31 PM
Half a year since AMD's release of the RV870 Cypress processor and the ATI Radeon HD 5800 series based on it, Nvidia has finally responded with the new graphics architecture called Fermi and two graphics cards that embody it: GeForce GTX 480 and GTX 470. Many users have been eagerly anticipating this release and now, basing on various reviews and tests, begin to make up their minds as to what graphics card would be best for playing modern DirectX 11 games. There is still some time left for making a shopping decision as the Fermi cards are going to hit the shops by mid-April only. We had not managed to get hold of the card prior to the announcement of the Fermi architecture, so this first review and test session are published with a few days delay. Let's hope the wait was worth it!
The technical specifications of the Nvidia GeForce GTX 480 graphics card are listed in the next table in comparison with its market opponent as well as with its predecessor GeForce GTX 285.
Click to enlarge
There is a lot of information about the Fermi architecture on the Web already. You can also read through the official documentation. So, let's move on straight to business.
I've got an OEM version of the Nvidia GeForce GTX 480 - without packaging or accessories. The new card won't look really new to anyone who is familiar with Nvidia's GeForce GTX 260 and 285 series products.
The five nickel-plated copper heat pipes sticking up from the GPU are the only thing that may catch your eye as unusual.
Otherwise, there is nothing extraordinary about the appearance of the reference GeForce GTX 480. The PCB is 267 millimeters long. It is shorter than the ATI Radeon HD 5870 and thus easier to install into system cases.
The graphics card has a version 2.0 PCI Express interface, two dual-link DVI-I connectors and one HDMI. There is a vent grid in the card's mounting bracket to exhaust the hot air out of the computer case.
There is a gap at the other butt-end of the card but it does not open a way for the air to get to the blower as might be expected. 6- and 8-pin power connectors can be seen on the PCB. There are also two MIO connectors for building SLI configurations with two or three GeForce GTX 480 cards.
Take note that there is another vent grid near those connectors. Leaving the card through it, the hot air remains inside the system case. This is no good considering the high power consumption and, consequently, heat dissipation of the GeForce GTX 480 (up to 250 watts according to its specs). The engineers who developed this cooling system must have just had no choice.
The cooler's plastic casing is secured with locks that can be easily opened.
The heatsink can also be removed easily so that we could take a closer look at the card's PCB.
All of the graphics memory chips can be found on the face side of the PCB. The power section of the GeForce GTX 480 includes a 6-phase GPU voltage regulator based on a CHL8266 controller and a dual-phase regulator for the graphics memory.
The GPU die incorporates a tremendous 3 billion transistors. It is covered by a heat-spreading cap with the chip's marking.
Judging by that marking, this is a revision 3 chip (GF100-375-A3) manufactured on the 4th week of 2010. The GPU incorporates 480 universal shader processors, 60 texture-mapping units and 48 raster operators. The GPU's geometrical domain is clocked at 700 MHz whereas its shader domain is clocked 1401 MHz. The frequencies are dropped to 51 and 101 MHz, respectively, in 2D applications. The rest of the characteristics have been shown in the specifications table above.
The reference Nvidia GeForce GTX 480 is equipped with 12 GDDR5 memory chips for a total of 1.5 gigabytes. They are all located on the face side of the PCB. These are Samsung's K4G10325FE-HC04 chips.
The specifications suggest that these memory chips have an access time of 0.4 nanoseconds and a rated frequency of 5000 MHz. However, the memory frequency of the GeForce GTX 480 is only 3696 MHz, which leaves ample room for overclocking. To save power and dissipate less heat the card drops its memory frequency to 270 MHz in 2D applications. The graphics memory bus is 384 bits wide, providing an impressive peak bandwidth of 177.4 GBps.
The latest version of the GPU-Z tool can tell everything about the GeForce GTX 480 with almost no errors.
Now let's check out the card's cooling system and its efficiency. The key component of the reference cooler is the GPU heatsink.
It consists of five copper heat pipes, 6 millimeters in diameter, which are part of the cooler's base (the so-called direct touch technology). The pipes go through aluminum fins, about 0.35 millimeters thick and placed 1.5 millimeters apart. It must be noted that the heatsink's dissipation area is not really large. The pipes in the base are placed 1.5 millimeters apart from each other, too. The heatsink is all nickel-plated.
There is a thick layer of gray-colored thermal grease between the GPU and the cooler's base.
The amount and quality of thermal grease is highly important for coolers with direct-touch technology to show their best performance. More so than for coolers with a classic base. Running a little ahead, I can tell you that when I replaced the default thermal grease with a thin layer of Arctic Cooling MX-3, the peak GPU temperature lowered by 3°C. The temperature did not change in 2D applications, though.
The second component of the reference cooler installed on the GeForce GTX 480 is the metallic plate with a blower.
The plate has contact with the graphics memory chips and power circuit elements via thermal pads. The speed of the blower (whose maximum output power is 21 watts, by the way) is adjusted by the card automatically depending on the temperature. Interestingly, the blower's speed grows up smoothly but when the load is removed from the GPU, it plummets down suddenly. After the noise produced by the GeForce GTX 480 cooler at near-maximum speeds, you get an impression that the blower turns off altogether but that's not true. In 2D mode, when the card's clock rates are lowered, the blower is working at 44-46% of its full capacity. I will discuss the noise factor later on. Right now, let's check out the efficiency of the GeForce GTX 480's reference cooler.
The graphics card was loaded by the resource-consuming Firefly Forest test from the semi-synthetic 3DMark 2006 benchmark running at 2560x1600 with 16x anisotropic filtering. The GPU temperature and the blower's speed (in percent from its maximum) were monitored with MSI Afterburner version 1.5.1 which does not yet fully support the GeForce GTX 480 series. The ambient temperature was 25°C during this test. The graphics card was installed into a closed system case whose configuration is listed in the Testbed and Methods section. The card's default thermal interface was used.
So, let's see how hot the GeForce GTX 480 is when its blower works in the automatic mode as well as at the maximum speed.
The new graphics card is obviously very hot. Running the 3DMark 2006 test, its GPU quickly grew as hot as 95°C but then, thanks to the blower accelerating to 70-78% (about 3600 RPM), dropped to 91-92°C and did not change thereafter. If the blower's speed is manually set at its maximum speed (about 4780 RPM), the GPU temperature is no higher than 68°C. The performance of the heatsink depends greatly on the blower's speed, which indicates that its heat dissipation area is not large enough.
I also checked out the GeForce GTX 480's reference cooler with FurMark version 1.8.0 (with the EXE file renamed) which was running in full-screen mode at 2560x1600 with 16x anisotropic filtering enabled in the GeForce driver. When the blower was working in automatic mode, the picture was almost the same as in 3DMark 2006, except that the peak temperature was 98°C, but when the blower automatically accelerated to 4150 RPM, the temperature lowered to 91-92°C. And here are the results at the maximum speed of the blower:
The resulting GPU temperature is as high as 86°C. As you can see, the new graphics card is very hot and its cooling system is very noisy in 3D mode. This shouldn't come as a surprise, though. As a matter of fact, top-end products from both Nvidia and AMD/ATI have never been cool or quiet. Besides, there will surely appear alternative cooling solutions that can often be up to 30°C better than the reference samples while being incomparably quieter (I can recall the Arctic Cooling Accelero Xtreme GTX 280 or Thermalright's solutions). The only problem is that it does not feel good to spend $500 for a graphics card that requires you to change its default cooler with a better one (possibly losing your warranty). Well, perhaps we should wait for GeForce GTX 480 with alternative coolers to come out.
I measured the overclocking potential of my GeForce GTX 480 with EVGA Precision version 1.9.2.
Considering the temperature, I did not expect anything impressive from the new GPU at overclocking. Indeed, I could only increase its frequency by 45 MHz without losing stability or provoking image artifacts. The resulting frequency was 745 MHz (+6.4%). As opposed to the GPU, the 0.4-nanosecond memory chips did much better, being perfectly stable at 4780 MHz (+29.3%).
I am not absolutely sure that the memory test from the latest version of OCCT works correctly with the GeForce GTX 480, but it takes almost all of the graphics card's 1.5 gigabytes of memory to run.
The higher memory frequency did not affect the temperature of the graphics card's PCB and GPU.
Winding up this description, I want to give you a link to the graphics card's BIOS and remind you that the recommended price of the Nvidia GeForce GTX 480 is $499. The card is going to start selling worldwide on the 12th of April.
Every test was performed in a closed system case with the following configuration:
To avoid being limited by the CPU's performance, the 45nm quad-core CPU was overclocked to 4.0 GHz (at a frequency multiplier of 21, voltage of 1.3725 V, and with Load-Line Calibration enabled) in some games included into this test program.
The computer's system memory was working at 1.53 GHz with 7-7-7-14_1T timings at a voltage of 1.64 V. The rest of the BIOS settings pertaining to CPU and memory overclocking were left at their defaults (“Auto”).
The Nvidia GeForce GTX 480 graphics card will be compared with a Leadtek WinFast GTX 285 1024 MB and with an XFX GeForce GTX 295 2x896 MB in this test session.
AMD's camp will be represented by a Radeon HD 5870 1024 MB and by a dual-processor Radeon HD 5970 2x1024 MB.
Now let's move on to the software and benchmarking tools used in the tests. I started the test session on the 23rd of March and carried it out under the Windows 7 Ultimate x64 operating system with all critical updates installed. I used the following drivers:
The graphics cards were tested at two resolutions: 1920x1080 and 2560x1600. I believe that it wouldn't make sense to test the premium-class products at lower resolutions as it wouldn't bring any valuable results, the cards being often limited by the speed of the platform.
The tests were performed in two image quality modes: “High Quality + AF16x” (maximum textures with 16x anisotropic filtering) and “High Quality + AF16x + AA 4(8)x” with 16x anisotropic filtering and full-screen antialiasing (MSAA) of 4x or 8x level (if the average frame rate remained high enough for comfortable play). I enabled anisotropic filtering and full-screen antialiasing in the games' settings or configuration files. If such options were missing, I enabled these settings in the Control Panels of the Catalyst and GeForce drivers. Vertical sync was always off in the drivers' control panels.
All games were updated with the latest patches available at the time of tests. So, the complete list of test applications includes two popular semi-synthetic benchmarking suites, one technical demo and 21 games, including the latest titles. Here is the complete list of tests with the settings (the games are sorted in the order of their release):
If the game allowed recording the bottom fps readings, they were also added to the charts. I ran each game test or benchmark twice and took the best result for the diagrams, but only if the difference between them didn't exceed 1%. If it did exceed 1%, I ran the tests at least one more time to achieve repeatability of results.
The results of AMD's Radeon HD 5970 and Radeon HD 5870 cards are marked in red in the diagrams. The GeForce GTX 480 is Nvidia's traditional green. The GeForce GTX 295 and GTX 285 are teal. I did not benchmark the cards at overclocked frequencies as we are going to do this in a separate review.
Now, let's see what performance we can get from Nvidia's new card.
In the first semi-synthetic benchmark the GeForce GTX 480 is just a little faster than its main opponent Radeon HD 5870, both being ahead of the dual-processor GeForce GTX 295. The Radeon HD 5970 beats every other runner in the high quality mode at 2560x1600, leaving the GeForce GTX 480 far behind.
3DMark Vantage shows us a different picture at the Performance settings where the GeForce GTX 480 scores almost 2000 points more than the Radeon HD 5870. When the load is higher, these cards deliver similar performance.
The GeForce GTX 285 and GTX 295 do not support DirectX 11, so I compared them with the GeForce GTX 480 by testing all of them in DirectX 10 mode with disabled tessellation:
The GeForce GTX 480 is impressive in the Heaven demo, showing its very best to the potential buyer. The new card delivers superb performance, being far ahead of the two other top-end cards from Nvidia in the hardest test mode.
Now, let's see how good the GeForce GTX 480 is in DirectX 11 mode with turned-on tessellation (the results of the cards in DirectX 10 mode are formatted in italics).
As you can see, the GeForce GTX 480 easily competes with the dual-processor Radeon HD 5970 and leaves its market opponent Radeon HD 5870 far behind. If tessellation becomes a popular feature in games, the GeForce GTX 480 will be far more appealing than AMD's cards. But that's what's going to be (or not to be) in the near future. Let's return to the present and check the graphics cards out in today's games.
In World in Conflict the GeForce GTX 480 proves to be faster than the Radeon HD 5870 and successfully competes with the dual-processor GeForce GTX 295, being superior to the latter in terms of bottom speed. The dual-processor Radeon HD 5970 is unrivalled in terms of average performance, but its bottom speed is not any better than that of the other tested cards.
Many gamers expected the Fermi to finally conquer the resource-consuming Crysis as a single-processor graphics card, but it doesn't. The GeForce GTX 480 is but a little faster than the Radeon HD 5870 (something that has not been expected by most gamers, either). On the other hand, the GTX 480 delivers a higher bottom speed in the hardest test mode and easily beats the GeForce GTX 295.
Unreal Tournament 3 seems to rank the graphics cards up according to their price positioning. Thus, the GeForce GTX 480 is faster than the Radeon HD 5870 but slower than either of the dual-processor solutions.
The best that the GeForce GTX 480 can do in Lost Planet is to beat the Radeon HD 5870 in three out of the four test modes and equal the GeForce GTX 295. The dual-processor Radeon HD 5970 is still the fastest card in this test session.
Nvidia's new card performs confidently in Far Cry 2 where it is always ahead of the Radeon HD 5870 and even challenges the Radeon HD 5970 in the highest-quality mode with FSAA enabled.
Call of Duty 5: World at War has modest system requirements and runs at a playable speed on any of the tested cards even at 2560x1600 with 8x MSAA. The GeForce GTX 480 and the Radeon HD 5870 are roughly comparable in this game.
Despite the fact that BattleForge: Lost Souls is a Radeon-optimized application, Nvidia's GF100-based graphics card is only inferior to the Radeon HD 5870 in the FSAA-less mode. When full-screen antialiasing is turned on, the GeForce GTX 480 goes ahead.
The GeForce series cards have no chance in this game, which is optimized for the Radeon architecture, too.
The GeForce GTX 480 beats both the Radeon HD 5870 and the dual-processor Radeon HD 5970 in this flight simulator. In the high-quality mode the GTX 480 is twice as fast as the GTX 285!
Like BattleForge and Stormrise, Call of Juarez: Bound in Blood runs better on AMD's solutions. However, this does not prevent the GeForce GTX 480 from beating the Radeon HD 5870. It must be noted that Call of Juarez is not a resource-consuming application. A single GeForce GTX 285 or a Radeon HD 4870/HD 5770 would suffice for playing this game at the highest resolution.
It is quite a surprise to see AMD's cards win in Wolfenstein because Nvidia's OpenGL support used to be superior. But as you can see, the GeForce GTX 480 cannot match the Radeon HD 5870 in this game.
Nvidia's new product has a mighty weapon in Batman: Arkham Asylum in the form of hardware support for PhysX effects. For AMD's graphics cards to perform successfully in this game, you have to either install a GeForce as a second graphics card or turn off those PhysX effects altogether. In the latter case the game's visuals degenerate greatly, but the former solution is costlier and more troublesome as you have to make the Catalyst and GeForce drivers run together on a single computer. We must note that the GeForce GTX 480 is just brilliant in Batman: Arkham Asylum, outperforming the dual-processor GeForce GTX 295.
The GeForce 480 leaves no chance to its opponent in Resident Evil 5, either. Moreover, the new GF100-based card can successfully compete with the dual-processor GeForce GTX 295 and Radeon HD 5970.
I have to remind you that the GeForce GTX 285 and GTX 295 cards do not support DirectX 11. Therefore I benchmarked the GeForce GTX 480 in DirectX 10 mode in order to make a correct comparison.
The GeForce GTX 480 cannot impress us with its performance in this game. It is not much faster than the GTX 285 and is no better than the GTX 295. The new card is far from brilliant in the DirectX 11 mode, too:
The Radeon HD 5870 and GeForce GTX 480 cannot get a decisive win in Borderlands. The Nvidia card is faster in the FSAA-less mode whereas the Radeon is ahead at 1920x1080 when we turn on 8x MSAA.
If you are keen on wiping out thousands of zombies in Left 4 Dead 2, you may want to choose an AMD card since they are faster than Nvidia's solutions. On the other hand, the game's system requirements are not high and it can run smoothly on any of the tested cards.
The GeForce GTX 2xx series cards do not support DirectX 11, so I will first test the GeForce GTX 480 in DirectX 9 mode for the comparison's sake.
Now let's see what we have in DirectX 11 mode where the Radeon series cards join in (the DirectX 9 results of the GTX 2xx series are shown in italics).
The GeForce GTX 480 is a little bit faster than the Radeon HD 5870 in this game.
CrossFireX technology does not yet work in the new flight simulator Wings of Prey, so the Radeon HD 5970 proves to be the slowest card in this test. SLI technology, on the contrary, works well, making the GeForce GTX 295 an unrivalled leader. The GeForce GTX 480 is a little faster than the Radeon HD 5870 across all of the test modes.
Warhammer 40 000: Dawn of War II is one of the most CPU-dependent games in this review, and the graphics cards only differ in the highest-quality mode at 2560x1600 where the GeForce GTX 480 is a little faster than the Radeon HD 5870 in terms of average frame rate but inferior to the latter in terms of bottom speed. By the way, the bottom speed of the GeForce GTX 480 is the lowest one among all the tested graphics cards in this game. This must be due to insufficient optimizations of the GeForce driver and may be corrected in near future.
I benchmarked the graphics cards in Metro 2033 at the beginning of the Chaser level using the scripted scene where the protagonist with two friends is driving in a trolley and cannot move about. Thus, the results of the test are highly repeatable. The frame rate is measured with FRAPS for 160 seconds after the loading of the map. First, I used the DirectX 10 renderer:
The GeForce GTX 480 is somewhat faster than the Radeon HD 5870, both cards being considerably slower than the dual-processor Radeon HD 5970. While CrossFireX did not work in Wings of Prey, Metro 2033 proved to be poorly optimized for SLI technology. In the next game, Just Cause 2, SLI did not work altogether.
I also benchmarked the GeForce GTX 480 in DirectX 11 mode in comparison with an ATI Radeon HD 5850 1 GB overclocked from its default 725/4000 MHz to 950/4800 MHz (I performed this test after the main test program and had already returned the other Radeons). It was the same Chaser scene at the highest graphics quality settings. Here are the results:
The overclocked Radeon HD 5850 is not far behind the GeForce GTX 480 at 1920x1080 but begins to display a slideshow at 2560x1600 whereas the GeForce GTX 480 is three times as fast as its opponent then. However, the new card is still unable to deliver a playable frame rate at the highest resolution. Moreover, I noticed the image to get somewhat blurry on the GeForce when running the test demo and checked this out with screenshots (on another level as I could not make identical screenshots on the Chaser map). You can see the difference yourself (the screenshots were captured at 1920x1080).
ATI Radeon HD 5850
NVIDIA GeForce GTX 480
It is easy to see that the Radeon produces a better-looking picture. All the textures are sharp, without fuzziness or anything. To remind you, I selected the same graphics quality settings for both graphics cards: DirectX 11, Very High, 16x AF, AAA, Advanced DOF and Tessellation. Besides, I selected the High Quality texture filtering mode in the Catalyst and GeForce/ION driver (the Quality mode is selected by default). Perhaps this reduction of quality is the trick the GeForce GTX 480 resorts to in order to deliver a higher speed in Metro 2033? The game developer answered to our question promptly. Here is what Oles Shishkovtsov, 4A Games Chief Technical Officer said:
No, the observed difference in quality is not due to the performance of the graphics cards. Indeed, graphics cards from Nvidia and ATI render the scene differently in some antialiasing modes. We are trying to find a way to correct this issue. Again, this has nothing to do with speed.
Hopefully, people from 4A Games will find a solution.
The GeForce GTX 480 is inferior to the Radeon HD 5870 in this recently released game.
The first pair of summary diagrams show the performance growth delivered by the GeForce GTX 480 relative to the fastest single-processor graphics card of Nvidia's previous generation, the GeForce GTX 285. The latter's performance is the reference point and the GeForce GTX 480's performance is calculated in percent to it (DX10 and DX9 renderers, respectively, were used in S.T.A.L.K.E.R.: Call of Pripyat and Colin McRae: DiRT 2).
Everything is clear enough. The GeForce GTX 480 enjoys an average advantage of 30-37%, depending on the resolution, over the previous-generation flagship GeForce GTX 285. When full-screen antialiasing is turned on, the difference is 46-48%. I must note, however, that the Radeon 5 series brought about bigger performance benefits in comparison with the previous-generation card from AMD. The biggest performance growth can be observed in such games as Tom Clancy's H.A.W.X., Far Cry 2 and Metro 2033. It is in the undemanding applications like Call of Duty 5: World at War or Wolfenstein that the two cards differ less.
The next diagrams show the performance of the dual-processor GeForce GTX 295 in comparison with the new single-processor flagship GeForce GTX 480. The GTX 295 is the reference point (DX10 and DX9 renderers, respectively, were used in S.T.A.L.K.E.R.: Call of Pripyat and Colin McRae: DiRT 2).
So, the GeForce GTX 480 is faster in some games while the GeForce GTX 295, in others. There is no definite leader. I can recall the Radeon HD 5870 being nearly always faster than the previous-generation dual-processor Radeon HD 4870 X2. Anyway, the GeForce GTX 480 enjoys a large advantage in the newest games Metro 2033 and Just Cause 2 because SLI technology does not yet work correctly in them.
Now, let's see how the new GeForce GTX 480 compares with the Radeon HD 5870. The performance of the Radeon HD 5870 is taken as the reference point.
Again, we've got no definite winner as both cards win some tests and lose others. The GeForce GTX 480 proves to be ahead in twelve games, namely World in Conflict, Crysis, Unreal Tournament 3, Lost Planet: Colonies, Far Cry 2, Tom Clancy's H.A.W.X., Call of Juarez: Bound in Blood (!), Batman: Arkham Asylum (the Radeon working without a PPU), Resident Evil 5, Borderlands, Colin McRae: DiRT 2 and Metro 2033. The Radeon HD 5870, on its part, is faster in five games: BattleForge, Stormrise, S.T.A.L.K.E.R.: Call of Pripyat, Left 4 Dead 2 and Just Cause 2. So, I can say it quite definitely that the GeForce GTX 480 does not enjoy an overwhelming advantage over the Radeon HD 5870 as many gamers anticipated.
And the last pair of diagrams show how slower the GeForce GTX 480 is in comparison with today's fastest graphics card Radeon HD 5970.
The GeForce GTX 480 is ahead of the dual-processor Radeon HD 5970 in Wings of Prey (because CrossFireX technology does not work there) and Batman: Arkham Asylum (because the Radeon does not have PhysX support). The cards are equals in Tom Clancy's H.A.W.X. and Resident Evil 5. In the rest of the games the Radeon HD 5970 is faster than the GeForce GTX 480.
I measured the power consumption of systems with different graphics cards using a specially modified power supply. To create maximum load I launched FurMark 1.8.0 in stability check mode at 2560x1600 singly and together with Linpack x64 (LinX 0.6.4, 4096 MB, 7 threads). These two programs load heavily the graphics card and CPU, respectively, so we can determine the peak power draw of the whole system and see what power supply will suffice for it. You can see the results in the diagram:
Easy to see, the GeForce GTX 480 configuration needs about 130 watts more than the Radeon HD 5870 configuration under FurMark as well as under FurMark and Linpack x64 combined. Moreover, the GeForce GTX 480 configuration proves to require more juice than the configuration with the dual-processor Radeon HD 5970! That's a real hungry graphics card! On the other hand, the GeForce GTX 295 still remains the highest-consumption card according to our tests. When it comes to running 2D applications, the GeForce GTX 480 needs 26 watts more than the Radeon HD 5870.
Now let's compare the temperature of the graphics cards in the automatic mode of their coolers. Each card was running 15 cycles of the Firefly Forest test from the semi-synthetic 3DMark 2006 benchmark at 2560x1600 with 16x anisotropic filtering. The tests were performed in a closed system case at an ambient temperature of 25°C. Here are the results:
The Radeon HD 5870 is much better than the GeForce GTX 480 in terms of GPU temperature in both idle and loaded modes.
We also measured the level of noise using an electronic noise-level meter CENTER-321 in a closed room about 20 sq. meters large. The noise-level meter was set on a tripod at a distance of 15 centimeters from the graphics card which was installed on an open testbed. The mainboard with the graphics card was placed at an edge of the desk on a foam-rubber tray:
The bottom limit of my noise-level meter is 29.8 dBA whereas the subjectively comfortable (not low, but comfortable) level of noise when measured from that distance is about 37 dBA. The speed of the graphics cards' coolers was being adjusted by means of a controller that changed the supply voltage in steps of 0.5 V.
The Radeon HD 5970 and HD 5870 proved to be within 0.1 dBA in this test, so they are represented by the same graph in the diagram. I could not measure the noise of the GeForce GTX 295 on my testbed because I would have had to dismantle the whole cooler to reach the blower's connector.
Here are the results:
The first thing that must be noted is that none of the reference coolers is quiet. Second, we can note that the graph of the GeForce GTX 480 cooler goes lower than the graph of the Radeon HD 5870, meaning that the new card's cooler is quieter than the reference cooler of the Radeon HD 5870. However, this is only true when the coolers are working at the same speed, which is not the case in practice. In the automatic mode the speed of the GeForce GTX 480's cooler varies from 2100 to 3600 RPM whereas the fan of the Radeon HD 5870's cooler works at a speed of 1270 to 2040 RPM (see the temperature diagram above). Considering the high heat dissipation of the GF100 processor, Nvidia's engineers had to increase the top temperature threshold in the graphics card's BIOS and set a higher fan speed. As a result, the new card is noisy and hot. Alas, the GeForce GTX 480 is going to be inferior to the Radeon HD 5870 in this respect until new revisions of its core.
What do we have half a year since the release of the Radeon HD 5870? Taking all the tests together (which is not exactly correct), the new GeForce GTX 480 is an average 5 to 15% faster than the Radeon HD 5870 depending on the specific test, quality mode and resolution. In the individual games we can see both impressive victories and losses. Therefore, it is better to discuss each particular game as I have done in the performance summary section. And I guess it is quite right to claim that the GeForce GTX 480 becomes the fastest single-processor graphics card in the world.
This achievement is not without a downside, though. Apart from its high power consumption, price, noise, and temperature, I also mean the time that has passed since the release of the Radeon HD 5870. AMD's top-end graphics cards have been selling freely for months and have as many as six official driver versions already whereas Nvidia's GeForce GTX 480 still has no official driver. Its beta version is still being tested. Such a long delay might have been justified only by an overwhelming advantage (50% and more) in performance over the opponent, but Nvidia cannot claim it.
On the other hand, the problem of high noise and temperature can be solved by installing an alternative cooler many of which have already been announced by the leading graphics card makers. Nvidia also offers such benefits as PhysX and CUDA and, supposedly, higher performance in DirectX 11 games that use tessellation. I have no doubt that the GeForce driver will be optimized further, too. So, this review is just the first look at the new card as we will explore its capabilities in more detail later.
I guess there is something unquestionably positive about Nvidia's newly released product. Tough competition will bring the prices down. This is true for graphics cards like for any other hi-tech devices. When there is only one indisputable leader, it may be good for that leader, but not for end users like all of us. That's why the GeForce GTX 480 is really good news for every gamer, even those who stick to the AMD brand.