The First DX11 NVIDIA GPU - GeForce GTX 480
The high performance video card landscape normally runs on an ever evolving plane, with both NVIDIA and ATI trying to out gun each other every few months. It's a classic battle of one up that came to griding halt on September 22, 2009 (6 months and 4 days for those of you counting) when ATI launched Radeon HD 5870, the first DirectX 11 video card to market. Performance over previous generation cards was fantastic and showed that a powerful graphics card could also use very little power when in an idle state. Another feather in the cap of the Radeon HD 5870 was a new feature called Eyefinity, which has quickly become the best way to enjoy your favorite games!
For the last 14 months, the highest performing graphics card you could get from NVIDIA was the GeForce GTX 295 and it's performance was very similar to the much less expensive Radeon 5870. Over the last 6 months it's been slim pickings on GeForce GTX 200 series inventory, so it is with much anticipation that NVIDIA is truly, finally, for real this time, taking the wrapper off their next generation graphics card code named Fermi, the GeForce GTX 480. Yes, we can finally show you the full performance numbers!
The GeForce GTX 480 is the new flagship product in NVIDIA's video card arsenal and is designed for maximum gaming performance. The GeForce GTX 480 launches with an estimated price point of $499 USD and is a DirectX 11 ready graphics card that competes directly with the ATI Radeon HD 5870 in the single GPU market.
Based on the Fermi third-generation Streaming Multiprocessor (SM) architecture, GeForce GTX 480 has 480 CUDA cores, doubling the shader horsepower of GT200. The reference card that we will be benchmarking has a GPU clock of 700MHz, with a Stream Processor clock of 1401MHz. The GeForce GTX 480 has a 1536MB frame buffer consisting of Samsung GDDR5 memory clocked at 924MHz, for an effective data rate of 3696MHz. The memory is connected via a 384-bit memory interface and the GPU has 60 texture units and 48 ROPs. This configuration offers a peak texture fillrate of a 42GTexels/s and over 177GB/s of memory bandwidth. Compared to the previous generation, the geometry pipeline is significantly revamped, with vastly improved performance in geometry shading, stream out, and culling. The number of ROP (Raster Operations) units per ROP partition is doubled (to 48 in total). Fillrate is greatly improved, enabling multiple displays to be driven with ease. 8xMSAA performance optimized through enhanced ROP compression. Additional ROP units allow better balance and performance even for uncompressed portions of the scene. The texture units have also been redesigned for improved efficiency and higher real-world performance.
NVIDIA is also releasing the GeForce GTX 470 graphics card today and like the name suggests it will be a little less expensive at $349 and offer less performance than the flagship GeForce GTX 480 graphics card. The GeForce GTX 470 features 448 CUDA cores, 56 texture units, and 40 ROP units. NVIDIA claims that a GTX 470 strikes a perfect balance between price, performance, and power.
The architecture of this new GPU has already been covered, so we will let you reference that article here if you'd like to read it. The rasterization pipeline has come a long way, but as games aspire to film quality, graphics is moving toward advanced algorithms that require the GPU to perform general computation along with programmable shading. G80 was the first NVIDIA GPU to include compute features. GF100 benefits from what we learned on G80 in order to significantly improve compute features for gaming.
GeForce GTX 480 and 470 leverages Fermi’s revolutionary compute architecture for gaming applications. In graphics, threads operate independently, with a predetermined pipeline, and exhibit good memory access locality. Compute threads on the other hand often communicate with each other, work in no predetermined fashion, and often read and write to different parts of memory.
If you like to compare specifications, here is a great chart that NVIDIA provided that shows differences between the GeForce GTX 480 and the GTC 470 video cards. Notice that the only similarities between the two are in the physical and thermal categories!
That being said, let's take a closer look at the NVIDIA GeForce GTX 480 reference card that NVIDIA provided us with for testing.
A Closer Look At The GeForce GTX 480
The GeForce GTX 480 graphics card that we have on the test bench today is a dual-slot single GPU video card that measures in at 10.5" in length. The front is pretty plain, but keep in mind that this is just a reference card.
Flipping the GeForce GTX 480 over we don't find too many interesting things, but we can make out small holes. These allow for a small amount of extra airflow into the video cards cooling fan, which helps improve cooling performance. We have seen this for a couple years now on several NVIDIA graphics cards and it appears to work
The NVIDIA GeForce GTX 480 GDDR5 graphics card has a pair of dual-link DVI-I outputs along with a mini-HDMI output header. This is the first time we have come across this connector and we didn't have an adapter to try it out. We asked NVIDIA if their board partners would be including an adapter and they said that they expect AICs to include either a cable or dongle. Both the Dual-link DVI and HDMI outputs can be used to send high-definition video to an HDTV via single cable (audio too, if running HDMI). A regular sized HDMI header was not used since it couldn't fit next to the pair of DVI outputs.
The NVIDIA GeForce GTX 480 video card requires a 600 Watt or greater power supply to power the GeForce GTX 480 as it has a max board power (TDP) of 250 Watts. NVIDIA also suggests that your power supply have a minimum of 42 Amps on the +12V rail. It also requires that the power supply has one 6-pin PCI Express power connector and an 8-pin PCI Express power connector for proper connection. It should be noted that the NVIDIA minimum system power requirement based is based on a PC configured with an Intel Core i7 3.2GHz CPU. The GeForce GTX 470 needs a minimum 550W or greater system power supply (with a minimum 12V current rating of 38A) in case you are wondering what the little brother for this card requires.
The GeForce GTX 480 and 470 graphics cards do support SLI and have a pair of SLI bridges located along the top edge of the graphics card. The NVIDIA GTX 400 series support two, three, and quad SLI configurations. Since the GeForce GTX 400 series run rather hot it was noted to us that SLI requires optimal cooling and airflow. Poor ventilation will result in high fan speeds.
For SLI use NVIDIA strongly recommends that you test in a well ventilated case (rather
than an open bench). If three PCI-E slots are available, the cards should be installed in
the inner most and outer most slots. This ensures optimal cooling for the GPUs. NVIDIA said that 4-way SLI (Quad SLI) support for GTX480 will be supported in specific motherboards (same as GTX 285).
The next thing that needs some explaining on the GeForce GTX 480 graphic card is the cooling solution. This card runs hot. It runs so hot that NVIDIA even sent out a 'guide' on the power than these cards use. That guide starts out stating this:
"We’ve received some questions about the power consumption of GTX 480. We acknowledge that GTX 480 consumes more power than any other GPU we have produced. However, it’s important to view its power in the context of what GTX 480 delivers..." - From "A Perspective On Power" by NVIDIA
NVIDIA flat out says that the GF100 core runs hot and the five heat-pipe heatsink with the large 'radiator' that hangs off the side of the video card is a testament to this! It does look great though as I love the black nickle plating!
Taking The GeForce GTX 480 Apart
Since this is the first time I have actually seen a GeForce GTX 480 production card in person I had to take it apart and see what parts are being used. Okay, I really wanted to just see the GF100 GPU core and see what brand GDDR5 memory ICs were being used, but since I had to take it all the way apart I might as well walk you through the whole process. The front of the GeForce GTX 480 graphics card consists of a glossy black plastic shroud, the five heatpipe cooler and large radiating surface and the squirrel-cage fan.
The first step in taking apart the GeForce GTX 480 is to remove the fan shroud, which is easy to do as it is a tooless process. I really like this design as you can easily clean out the fins of the heat sink without having to remove any screws! You can just squeeze the plastic around the clips and they release and you can use compressed air to blow the dust out of the fins. Notice that once the fan shroud is removed you can see how ginormous the metal plate is on the heat sink!
Next you can remove the four larger Philips head spring screws as they attach the heatsink to the video card. With the heatsink removed you can see the core for the very first time! The design of this cooler is great as once again you can change out the thermal paste in a matter of seconds.
NVIDIA wasn't stingy on the thermal paste as they had plenty on the black nickle heatsink. Notice that the heat pipes are direct touch and the heatsink weighs in at 350g or just under 11oz. This is a very nice heatsink design and you can tell that this cooler design wasn't cheap to manufacture.
Our cooler was dated March 11, 2010 and we received the video card on March 19th, 2010! You can see that NVIDIA got us these video cards pretty quick considering they were flown in from China, went through customs, delivered to NVIDIA and packed up and sent out to each review site. We were told roughly 150 cards were sampled for the launch reviews, so if you want to read performance reviews on NVIDIA GeForce GTX 480 video cards you should have an issue finding one. We were told 29 sites in the US were sent cards and only six of those sites got two for SLI testing and that five got a GeForce GTX 470 for testing. Legit Reviews has just shy of one million monthly readers and qualified for just a single GeForce GTX 480 graphics card. Spread the word of this review to your friends and help LR get more traffic, so we can bring you reviews on better hardware!
The next step is removing all 15 of the small screws that hold on the
aluminum frame! Once all these screws are removed you can pull off the aluminum frame that helps cool the boards components and doubles as the fan mount. Notice the thermal pads on the frame to help cool down the voltage regulators and the GDDR5 memory ICs.
After the aluminum frame is removed you can take off the three Philips screws that hold the squirrel-cage fan down. Flipping it over we can see it is a model made by Delta Electronics. The fan has part number BFB0721HF on it and Delta has been known to produce quality cooling solutions that can move some air. This specific model is part of the BFB series and is 50mm across.
Now that the card if stripped down and pretty much bare we can take a look at the a good look at the heart of the GeForce GTX 480 and that would be the GF100 core shown above. The core is labeled as GF100-375-A3, which we are guessing would be an A3 stepping of the GF100 silicon. This GF100 GPU has 512 cores, but only 480 are enabled due to various reasons. We've heard rumors of heat issue, bad yields, and that NVIDIA or its board partners might be releasing 'enthusiast' boards with all the cores enabled down the road with even better improved cooling solutions.
Here is the picture that our readers always request to give you an idea of the core size next to something most people are familiar with.
Located around the GF100 GPU core are twelve Samsung GDDR5 memory chips for the cards memory. I wasn't able to get a good picture of the etching, but the exact model number is "K4G10325FE-HC04". That means these are 0.40ns parts that are speed rated at 5Gbps according to Samsung. It should also be pointed out that only solid-state capacitors are used on the GeForce GTX 480 video cards and you can see a few of them in the photo above.
Here you can see the NVIDIA GeForce GTX 480 graphics cards on top and the ATI Radeon HD 5870 graphics card on the bottom to get an idea of how the cards look. As you can see both cards use 10.5" Printed Circuit Boards (PCB) and have a layout that is shockingly similar if you ask me. You can tell these two companies have traded engineers over the years!
GeForce GTX 480 Test Settings
All testing was done on a fresh install of Windows 7 Ultimate 64-bit with all the latest updates installed. All benchmarks were completed on the desktop with no other software programs running. The Corsair DOMINATOR memory modules were run in triple-channel mode at 1866MHz with 8-8-8-24 timings. The ATI Radeon HD 5000 series cards were all tested using CATALYST 10.3 drivers and the NVIDIA GeForce GTX 285 used Forceware 197.25, while the GeForce GTX 480 used Forceware 197.17. The ASUS P6T Deluxe V2 motherboard was run using BIOS 0610 with the processor running stock settings and Turbo enabled.
Windows 7 Drivers Used:
Intel Chipset Inf Update Program V188.8.131.524
SoundMAX 2000B Audio Driver V184.108.40.20685 for Windows 64bit Windows 7.(WHQL)
Marvell Yukon Gigabit Ethernet Driver V220.127.116.11 for 32/64bit Windows 7.(WHQL)
Here is the Intel LGA 1366 Test platform:
|Intel Test Platform|
|Intel Core i7 975|
ASUS P6T Deluxe V2
6GB Corsair DDR3 1866MHz
Western Digital VelociRaptor
Cooler Master 1000W
None (Open Bench)
Windows 7 Ultimate 64-Bit
Video Cards Tested:
- Sapphire Radeon HD 5970 1GB - 735MHz Core / 1010MHz Memory
- MSI R5870 Lighting 1GB - 900MHz Core / 1200MHz Memory
- ATI Radeon HD 5870 1GB - 850MHz Core / 1200MHz Memory
- Sapphire Radeon HD 5850 Toxic 1GB - 765MHz Core / 1125MHz Memory
- ATI Radeon HD 5850 1GB - 700MHz Core / 1000MHz Memory
- NVIDIA GeForce GTX 480 - 700MHz Core/1401MHz Shader/1848MHz Memory
- PNY GeForce GTX 285 - 648MHz Core/1476MHz Shader/ 2484MHz Memory
NVIDIA GeForce GTX 480 GPU-Z 0.4.0 Details:
Batman: Arkham Asylum
Batman: Arkham Asylum is an action-adventure stealth video game based on DC Comics' Batman for PlayStation 3, Xbox 360 and Microsoft Windows. It was developed by Rocksteady Studios and published by Eidos Interactive in conjunction with Warner Bros.
For our testing we set everything as high as it would go, except for Physx and NVIDIA Multi Sample Anti-Aliasing.
Benchmark Results: In the popular game title Batman: Arkham Asylum we found that the NVIDIA GeForce GTX 480 had around a 4% performance increase over the MSI R5870 Lighting and reference designed Radeon HD 5870 graphics cards at 1920x1200. Not a very large difference at all. The performance improvements at 1280x1024 are virtually negligible.
Resident Evil 5
Resident Evil 5 is a survival horror video game developed and published by Capcom. Resident Evil 5 features similar gameplay to Resident Evil 4, with context-sensitive controls and dynamic cut scenes also making a return. The player can control Chris Redfield or Sheva Alomar in a similar fashion to Leon S. Kennedy in Resident Evil 4, with the same over-the-shoulder perspective.
All in game settings were set to their maximum. We used 8xAA for all of our performance tests in Resident Evil 5.
Benchmark Results: The NVIDIA GeForce GTX 480 starts to stretch it legs in Resident Evil 5 and shows us solid 12% gain in performance over the MSI R5870 Lighting at both 1920x1200 and 1280x1024 resolutions.
Tom Clancy's H.A.W.X.
Tom Clancy's H.A.W.X. (High Altitude Warfare eXperimental squadron) is an aerial warfare video game developed by Ubisoft Romania and published by Ubisoft for Microsoft Windows. It was released in United States on March 6, 2009 and features Microsoft DirectX 10.1 game play.
The game has a built-in benchmark that is very tough on video cards!
For this game VSync was turned off, but
Antialiasing was turned on and set to 8x for better image quality.
All of the DirectX 10 options were set to high including Ambient occlusion (SSAO) on both the NVIDIA and ATI graphics cards. The game was patched with update v1.2, which was the most current patch.
Benchmark Results: When we ran Tom Clancy's H.A.W.X. in DirectX 10 mode with AA enabled, we saw the NVIDIA GeForce GTX 480 had a 9% performance improvement over the MSI R5870 Lighting at 1920x1200. In lower resolutions (1280x1024), however, the GTX 480 gained almost 18% in performance over both 5870's.
3DMark Vantage is the new industry standard PC gaming performance benchmark from Futuremark, newly designed for Windows Vista and DirectX10. It includes two new graphics tests, two new CPU tests, several new feature tests, and support for the latest hardware. 3DMark Vantage is based on a completely new rendering engine, developed specifically to take full advantage of DirectX10, the new graphics API from Microsoft.
The Extreme settings were used for testing, so a resolution of 1920x1200 was used.
Benchmark Results: Overall number for the GeForce GTX 480 were almost identical to the Radeon HD 5870 cards we tested. Statistically, these cards are pulling virtually the same numbers (less than 2% difference). As expected, the NVIDIA GeForce GTX 480 did show a 17-19% advantage over the Radeon 5850 cards that we tested.
Unigine 'Heaven' DX11
The 'Heaven' benchmark that uses the Unigine easily shows off the full potential of DirectX 11 graphics cards. It reveals the enchanting magic of floating islands with a tiny village hidden in the cloudy skies. With the interactive mode emerging, experience of exploring the intricate world is within reach. Through its advanced renderer, Unigine is one of the first to set precedence in showcasing the art assets with tessellation, bringing compelling visual finesse, utilizing the technology to the full extent and exhibiting the possibilities of enriching 3D gaming. The distinguishing feature of the benchmark is a hardware tessellation that is a scalable technology aimed for automatic subdivision of polygons into smaller and finer pieces so that developers can gain a more detailed look of their games almost free of charge in terms of performance. Thanks to this procedure, the elaboration of the rendered image finally approaches the boundary of veridical visual perception: the virtual reality transcends conjured by your hand.
For this benchmark VSync was turned off and we tested with 0x AA and 4x AF to check out system performance. We also ran the benchmark at 1920x1200 and 1280x1024 to see how the benchmark ran at some different monitor resolutions.
Benchmark Results: The Unigine Heaven Benchmark showed that the NVIDIA GTX 480 card had a 28% performance increase over the MSI R5870 Lighting in 1280x1024. The GTX 480 was still well ahead of the closest 5870 in 1920x1200 resolutions as it stayed ahead of the MSI Lighting by almost 16%!
We also ran the new Heaven 2.0 benchmark that just came out on a couple of the cards as we were in a time crunch. For this benchmark VSync was turned off and we tested with 8x AA and 16x AF to check out system performance. We also ran the benchmark at 1920x1200 and 1280x1024 to see how the benchmark ran at some different monitor resolutions. It should be noted that we ran the new extreme tessellation mode on this benchmark. These are the toughest settings that you can run on this benchmark, so it should really put the hurt on any graphics card.
Benchmark Results: The Unigine Heaven 2.0 Benchmark is tessellation heavy and it showed that the NVIDIA GTX 480 video card had a 25% performance increase over the dual-GPU powered Sapphire Radeon HD 5970 that beat it in the original Heaven 1.0 benchmark. This goes to show that when it comes to extreme tessellation settings that the GTX 480 has some serious potential and if games get into heavy tessellation the architecture of the GeForce GTX 480 will dominate the competition.
S.T.A.L.K.E.R.: Call of Pripyat
The events of S.T.A.L.K.E.R.: Call of Pripyat unfold shortly after the end of S.T.A.L.K.E.R.: Shadow of Chernobyl following the ending in which Strelok destroys the C-Consciousness. Having discovered the open path to the Zone's center, the government decides to stage a large-scale operation to take control of the Chernobyl nuclear plant.
S.T.A.L.K.E.R.: Call of Pripyat utilizes the XRAY 1.6 Engine, allowing advanced modern graphical features through the use of DirectX 11 to be fully integrated; one outstanding feature being the inclusion of real-time GPU tessellation. Regions and maps feature photo realistic scenes of the region it is made to represent. There is also extensive support for older versions of DirectX, meaning that Call of Pripyat is also compatible with older DirectX 8, 9, 10 and 10.1 graphics cards.
The game S.T.A.L.K.E.R.: CoP has no internal benchmarking tools built into the game engine, but they do have a standalone benchmark available that we used for our testing purposes. The screen capture above shows the main window of the benchmark with our settings. Notice we are running Enhanced Full Dynamic Lighting "DX11" as our renderer.
Under the advanced settings we enabled tessellation and 4x MSAA. We didn't enable ambient occlusion as we wanted to use these test settings for mainstream cards down the road and these settings should be tough enough to stress the Radeon HD 5000 series cards.
The benchmark gives you results for four different lighting scenarios and
the easiest on graphics cards are the 'day' settings, so we will start
there. As you can see, all of the cards were able to easily average above 45FPS, at a resolution of 1920x1200. The NVIDIA GeForce GTX 480 outpaced the rest of the test cards in this benchmark. The GTX 480 was ~10% faster than the MSI R5870 Lighting in 1280x1024. The performance numbers go up from there when you examine how the GTX 480 does in higher resolutions. In 1920x1200, the difference is a whooping 18% over both 5870s.
Since video card temperatures and the heat generated by next-generation cards have become an area of concern among enthusiasts and gamers, we want to take a closer look at how the NVIDIA GeForce GTX 480 graphics card does at idle and under a full load.
NVIDIA GeForce GTX 480 Video Card Idle Temperature:
At idle on an open test bench the GeForce GTX 480 had some pretty good temperature numbers
for being such a hot card. We were seeing right around 51C at idle, which isn't as bad as we expected to be honest. As you can see from the screen shot above, the idle state
of the GeForce GTX 480 drops the GPU frequency down to just 50MHz and the
memory clock down to 67.5MHz to help conserve power and lower
temperatures. The fan spins at 1750RPM at 44% rotation speed when at idle.
NVIDIA GeForce GTX 480 Video Card Load Temperature:
We fired up FurMark and ran the stability at 640x480, which was enough to put the GPU at 100% load in order to get the highest load temperature possible. This application also charted the temperature results so you can see how the temperature rises and levels off, which is very nice. The GeForce GTX 480 maxed out at 96C. The fan on the GeForce GTX 480 video card was left on auto during temperature testing and the fan speed is dynamic. This means that it will increase/decrease depending on the GPU temp and GPU load. You can see a slight dip in temperatures at the start of the test as the GPU fan was still spinning up to a higher speed to keep up with all the heat. The card peaked at 96C then the fan speed increase was able to cool the card down and keep the temperature level at around 92C. The CPU Fan peaked at around 88%, which meant the Delta cooling fan was spinning at ~4500RPM. When we hit the space bar to stop the rendering the temperature dropped, but since a 2D application was open the card wouldn't go back to a complete idle state. Since the fan then slowed down again and the temperature started to rise again.
This picture was taken during the temperature testing shown above and it shows the GeForce GTX 480 screaming along at 67dB at 85% fan speed, which is ~4200RPM. This is not a silent video card by any means.
GeForce GTX 480 Overclocking
A performance analysis of the GeForce GTX 480 video card wouldn't be complete without some overclocking results, so we got in touch with our friends over at EVGA and they sent the latest build of their Precision software for us to try out on our reference card.
Using the EVGA Precision software utility for the GeForce GTX 480 graphics card is a little different though as the shader clock slider is gone. We contacted NVIDIA about this and they had this to say on the issue:
"When you move the processor clock slider on GTX 480 skin on the latest GF100 Precision tool the core clock goes up by proportional amount, since it’s ½ shader clock on this architecture. This is by design, GF100’s NV clock domain is controlled solely by the processor/shader domain so EVGA figured it would be easier on the user to remove the NV domain to avoid confusion. Old skins are still necessary to support all adjustable domains in the G8X – Gt21X products." NVIDIA PR
This makes sense, so you have one less slider to worry about! The bad news is that those that run F@H won't be able to adjust just the shaders as the core and shader clocked are locked together and can't be un-linked.
I cranked up the fan to 100% and found the max that my reference card could overclock to was 800MHz on the core and 1025MHz on the GDDR5 memory. If I pushed the core or memory any higher it would run some benchmarks, but would crash on the longer ones like 3DMark Vantage. The idle and load temperatures were actually better than default since the fan was at 100%. The noise level was higher than you'd like to use 24/7, but for overclocking it is good enough.
NVIDIA GeForce GTX 480 Graphics Card at 700MHz/1401MHz/924MHz:
NVIDIA GeForce GTX 480 Graphics Card at 800MHz/1600MHz/1025MHz:
Not a bad jump in performance! Let's see what it looks like against other cards with the latest drivers on the market.
The jump from 700MHz to 800MHz on the core clock really looked to help performance as the 14% clock frequency increase resulted in a 13% performance gain. Not a bad overclock, but I really wanted to get a little more out of the card. Even with this overclock the card was still idling in the 45C range since the fan was at 100%. It I had to take a guess it would be that I was voltage limited and not temperature limited. I'll explore this more in the future.
For testing power consumption, we took our test system and plugged
it into a Kill-A-Watt power meter. For idle numbers, we allowed the system
to idle on the desktop for 15 minutes and took the reading. For load
numbers we measured the peak wattage used by the system while running
the OpenGL benchmark FurMark 1.7.0 at 1920x1200 resolution.
Power Consumption Results: The NVIDIA GeForce GTX 480 used 20% more power than the MSI R5870 Lighting at idle. The difference is even more apparent when you look at the numbers under load where the GTX 480 uses 27% more power than the MSI R5870 Lighting. The ATI Radeon HD 5850 power consumption is totally eclipsed by the GTX 480 by 60% under load. The NVIDIA GeForce GTX 480 is clearly a power-hungry beast.
GeForce GTX F@H Performance
For those new to F@H, Folding@home is a distributed computing project, that very simply stated, studies protein folding and misfolding.
In the last few years Folding@home on graphics cards has gained popularity as it offers great performance per watt. By utilizing the power of the GPU, Stanford has been able to leverage ever evolving graphics technology and put it to good use trying to understand and possibly solve the problem of protiens misfolding.
Only recently has the pendulum swung back to the favor of folding on the CPU for ppd (points per day) crown with the release of the -bigadv client.
Since the release of the NVIDIA GPU client, their graphics cards have consistently been at the top of points per day, meaning the highest production video card. Nvidia has been touting Fermi as THE next big thing for GPGPU and projects like Folding@home. Many have been giddy with anticipation since Nvidia first broke the news of Fermi way back on September 30th 2009.
With the release of the GeForce GTX 470 and GeForce GTX 480, Stanford will be dropping a GPU3 client to work on these new cards as the GPU2 client does not work. As of 2am this morning we are being told "coming soon" for the full release of this client, but the early rumor is some time in April. Please do not ask for our copy as it does not download work or return results to Stanford. This new client works on both GeForce GTX 400 series as well as the GeForce GTX 200 line so we can get a comparison between the two generations. Note that this is a Cuda build and does not work on ATI hardware, so a direct comparison to the Radeon HD 5000 series will have to wait.
|GTX 480 Lambda
||GTX 480 Villin
|| GTX 285 Lambda
|| GTX 285 Villin
| Seconds per step
| GPU Idle Temp
| GPU Folding Temp
| System Load
| CPU Load
For our test we were sent three different work units, one of which (Spectrin) we weren't able to get our system to work correctly with on either graphics card. The results pretty much speak for themselves, a 50%-70% speed increase with about a 40% increase in power consumption. So indeed we are gaining performance per watt and actually giving the CPU a little more to keep the new GeForce GTX 480 fed with data. For those of you planning on running one or two of these beasts 24/7 will need to work hard to keep them cool. At 96C under a folding load the GeForce GTX 480 will go a long way to keep your room warm on those cold winter nights. With spring and summer on the way for the North though it could be one very high electric bill!
As far as expected points per day go, I will quote Nvidia "a GTX280 does about 7700 points per day (PPD), so the PPD for the GTX480 would be based on the speed up. So, if the GTX 480 is doing 70% better, then it would get 1.7 x 7700 = 13,090 PPD."
Dual LCD Display Testing
For years I have used a dual monitor setup in the office as it makes life easier and you honestly do become more productive when doing the daily tasks. When I first got the NVIDIA GeForce GTX 480 reference card I stuck it into the office system as I was curious how it would perform in my main system. Yes, I actually spent a few days with this card leaving it powered on 24/7 for a little longer than a 3-day stint to see what it would be like in a system like any of our readers would own. The card worked great in Adobe Photoshop CS4 and all the other GPU accelerated applications I use on a daily basis, but sadly it doesn't have a working F@H GPU client available for it. The day after I initially installed the GeForce GTX 480 graphics card in my system I noticed that my office was much hotter than any room in my house and I knew the graphics card was to blame.
I fired up EVGA Precision and instantly two things jumped out at me. The first is that the GPU was idling around 89-90C and the second is that the core clock of the GPU looked to be pulsing at a regular cycle.
After a little more a little more investigation I discovered that the GeForce GTX 480 video card was sitting at 90C in an idle state since I had two monitors installed on my system. I talked with some of the NVIDIA engineers about this 'issue' I was having and found that it wasn't really an issue per say as they do it to prevent screen flickering. This is what NVIDIA said in response to our questions:
"We are currently keeping memory clock high to avoid some screen flicker when changing power states, so for now we are running higher idle power in dual-screen setups. Not sure when/if this will be changed. Also note we're trading off temps for acoustic quality at idle. We could ratchet down the temp, but need to turn up the fan to do so. Our fan control is set to not start increasing fan until we're up near the 80's, so the higher temp is actually by design to keep the acoustics lower." - NVIDIA PR
Regardless what the reasons are behind this, running a two monitor setup will cause your system to literally bake. Just for fun I compared the GeForce GTX 285 video card that has been in the PC for the past year to the GeForce GTX 480 to see what the differences were when it comes to heat and energy efficiency in a multi-monitor setup. I have a pair of Samsung SyncMaster 22233RZ monitors that I used for testing.
|GTX 480 1 LCD
||GTX 480 2 LCD
|| GTX 285 1 LCD
|| GTX 285 2 LCD
| Mem Clock
| Shader Clock
| Idle Temp
| Idle Power
| Fan Speed
I'm not sure about you guys, but the results here really shocked me. I always knew that running two monitors over one increased power consumption, but I didn't expect it to jump up this high. On the previous generation GeForce GTX 285 graphics card running to cards increased the system power consumption by ~30W and increased the temperature ~10C. You can see that the idle state of the graphics card is also difference as the clock frequencies and voltages have to remain higher for the two LCD panels. Running two LCD panels on a GeForce GTX 285 has been fine over the past year and I never really had any complaints.
All that changed with the GeForce GTX 480 as you notice some dramatic changes right away. For starters adding a second monitor to the GeForce GTX 480 increased the systems power consumption by 80W, which is a 43% increase in the systems idle power consumption! If that wasn't bad enough the card would idle at 90C, which is 33C or ~58% higher than with one monitor! For comparison sake the GeForce GTX 285's GPU temperature went up ~21%.
|GTX 480 1 LCD
||GTX 480 2 LCD
|| Radeon 5870 1 LCD
|| Radeon 5870 2 LCD
| Mem Clock
| Shader Clock
| Idle Temp
| Idle Power
| Fan Speed
If you are going to buy a new video card today you are likely looking at the GeForce GTX 480 and Radeon HD 5870, so here are those two display performance results if you are interested in it. The Radeon HD 5870 uses 41 Watts more power at idle for the second monitor, which is slightly more than the 33 Watt increase that I saw on the GeForce GTX 285 video card. If you are looking to run a pair of monitors 24/7 on your PC the difference between the Radeon HD 5870 and the GeForce GTX 480 is night and day. The Radeon HD 5870 will run nearly 40% cooler and use 75 Watts less power doing so. AMD is the clear winner on this battle.
All this testing was done inside the Corsair 800D chassis, which has great airflow and is a true enthusiast grade case. If you only plan on running one monitor this might seem silly to you, but for those that run two monitors this could be an issue that you were not aware of as I don't see too many review sites testing things like this. We only got the GeForce GTX 480 graphics card 7 days before it launched, so it's tough to test all aspects of a video card in such a short period of time.
Final Thoughts and Conclusions
The NVIDIA GeForce GTX 480 graphics card that I have spent the past week using has left me with mixed feelings. Looking at the benchmark numbers alone the card does very well and beats the ATI Radeon HD 5870 in the majority of the benchmarks that I ran in our testing. The downside to the NVIDIA GeForce GTX 480 video card is of course the excessive heat, noise and power draw that comes along with it. NVIDIA had numerous issues bringing the GF100 to the market and we know for a fact that the core has 512 CUDA cores, but they disabled 32 of them with a BIOS patch. We were told this was done to improve yields and also keep the cards thermal profile in check. If those cores were enabled that would give the GeForce GTX 480 nearly 7% more compute power and it would have really pulled away in the performance charts. If only NVIDIA could have pulled that off.
In my personal system (Corsair 800D Chassis) with two monitors the GeForce GTX 480 graphics card would idle at 90C and if it was a sunny day and my office was warmer it would idle at 92. I fired up the new DX11 game title Aliens Versus Predator and with GPU-Z in the background I saw the temperature reach 99C while gaming for around 30 minutes. At this temperature the fan is spinning at 70dB and it honestly was not an enjoyable gaming experience. I asked NVIDIA if the card was built to run at temperatures this high and they claim that the GeForce GTX 400 series was built to operate at high temperatures and reminded me that 105C was the peak temperature for the GeForce GTX 480 video card. While benchmarking the GeForce GTX 480 graphics card on the open test bench I found the outside of the heatsink to reach 50C on the fan side and 59C on the exhaust side, so this card without a doubt will put out some heat.
If you want to run out and buy a GeForce GTX 480/470 graphics card now that they have officially launched here on Friday, March 26th at 7:01PM EDT don't get your hopes up. NVIDIA informed us that widespread e-tail availability of both GeForce GTX 480 and GTX 470 will happen the week of April 12, 2010. NVIDIA is currently building tens of thousands of units for initial availability, and they claim they will have enough produced to ensure that their partners will have ample volume for what is likely the most anticipated GPU launch ever.
Heat and power aside, the GeForce GTX 480 graphics card is impressive as they took the single GPU performance crown back from ATI and that was no easy task. They also improved the GPGPU performance over the previous generation core by a fair amount. Seeing F@H performance jump up nearly 70% was amazing, but at the same time the power and heat increases took the fun out of that performance increase. Having waited this long for Fermi to come out, I personally expected a little different user experience and many of you might have as well. I've been told by some add-in board partners that some custom designed cards are in the works and those should make things very interesting. The GeForce GTX 480 video card appears to be a diamond in the rough and in the weeks to come we can't wait to see the retail cards and see what driver enhancements NVIDIA will make for this new series! It should only get better from this point on!
I look forward to hearing your feedback in the forums!
Legit Bottom Line: The NVIDIA GeForce GTX 480 was known to be hot and fast before it came out and that is exactly what it turned out as being.