nVidia's GTX580 Hits the Streets and Leaves an Impact
11/9/2010 by: Sean Kalinich
Last year around this same time the air was full of rumors about nVidia GPUs. Did they have anything at all to challenge the Evergreen lineup that had been thrown down by AMD? We heard that TSMC [Taiwan Semiconductor Manufacturing Company] was having issues trying to make the giant GF100 die on their troublesome 40nm process. We saw a rush of articles about the mock-up card used by NV at their GTC 2009 keynote and so much more. Finally in the opening weeks of 2010 we saw the GF100. It was fast, it was big and it was also a power hungry GPU and puts out a ton of heat.
To put it simply, the GTX480 with its GF100 was not what nVidia wanted and not what we all expected. However, not too long after we were slightly disappointed [but in many ways still impressed] with the performance of the GF100 nVidia dropped a revamped GPU on us, this was the GF104 and showed up in the fast and nimble GTX460. This inexpensive GPU also featured a streamlined design that helped to reduce heat and power consumption while maintaining performance.
The new GPU under the hood of the GTX580 is something of a new design. While the Shader Model configuration of each shader processor found on the GTX580 is the same as we saw on the GTX480 - there are some improvements to the way it works. One of the first is the inclusion of support for 16-bit Floating Point texture filtering. nVidia has also dropped in support for a new tile formats that show an improvement in Z-cull efficiency. As a final 'improvement' the GeForce GTX580 simply has more shader, texture and CUDA cores and a much faster clock speed to boot.
nVidia GF110 die is hidden below this heatsink
For those of you planning on dropping in a window AC unit into your PC case I would hold off for now. nVidia did some tweaking to the overall design of the GPU which has allowed them to drop in 512 CUDA cores [whole 16SM clusters] while pulling less power and generating less heat that its warm blooded older brother the GTX480. They did this [according to nVidia] by reengineering the Fermi Die at the transistor level.
Again according to nVidia [you would have to tear down a GPU to really know for certain] they have adjusted the type of transistor and are using low leakage [pronounce somewhat slower] transistors where the instructions do not need full speed and, of course, fasters ones where the data throughput is more critical. This type of reconfiguration makes the GPU much more power efficient in that you are reducing the amount of power leakage across the board. This is combined with an improved cooling system [although it looks the same on the surface] means that you are also going to be generating less heat [we will talk more about this later].
But even these items are not all to the new GTX580. As you have heard us say before; AMD is great for bringing things to the market, but they often do not have the resources to pursue them. One of these items is DirectX 11. We have to give full credit to AMD for having the first full family of DX11 GPUs; they beat nVidia to the punch by a few months. However, AMD did not have the same type of funds that nVidia has and have not been able to developers offset the cost of game production in the same way that nVidia does and has. As such we are seeing nVidia pull ahead in DX11 implementation. This is appearing in the GTX580 by a new distributed tessellation system. nVidia uses a 16 PolyMorph Raster Engines in the GTX580. It pushes the workload out across all of these to not only improve performance but also to improve efficiency. As one AAA lead engine designer told us: "With DirectX 11 hardware, Tessellation is finally usable. When AMD, nVidia and Microsoft asked us what do we need to use Tessellation, we listed five mandatory features. It took AMD the same amount of time to make Tessellation unit usable as nVidia, which did not had Tessellation at all... and don't get me started on N-Patches."
Hawx2 Tessellation Wire frame image
The final specs of the GTX580 come out something like this:
GPU Clock Speed 772MHz
512 Cores at 1,544MHz
6 x 64-bit memory controllers [nVidia lists this as 384-bit]
1.5GB GDDR5 memory at 1002MHz QDR [4.008 billion transfers per second]
192.4GB/s memory bandwidth
The interesting bit about this design is that we received press releases from board vendors and they cited clock speeds as high as 850MHz for the GPU and 1.7GHz for those 512 Cores, i.e. the same level of overclockability as experienced on the smaller GF104 GPU i.e. GeForce GTX 460.
The GTX 580
nVidia GeForce GTX 580 follows the same style of design as the GeForce 4xx series, at least on the surface. Inside the silicon is another story. We find the typical dual DVI ports and the single Mini-HDMI 1.4a port. Above this is a grill for ventilation/exhaust of the hot air generated by the GPU and memory.
For power you get a single 6-pin PCIe power port along with an 8-pin one. It is just over 10.5 inches long and to be painfully honest not much to talk about when it is just sitting on your desk.
The Test System Build and Comments
Processor: Intel Core i7 975 Extreme
Mainboard Asus P6X58D Premium [Supplied by ASUS]
Memory: 6GB Kingston KHX12800D3T1K3/6GX [Supplied by Kingston]
Hard Disk: Patriot 128GB Torqx SSD [Supplied by Patriot]
Graphics Cards: GeForce GTX580 [Supplied by nVidia]
Graphics Cards: Asus EAH5870 V2 S.T.A.L.K.E.R. Call of Pripyat Edition [Supplied by ASUS]
MSi N465GTX Twin Frozr II [Supplied by MSi]
eVGA GTX480 Hydro Copper [Supplied by eVGA]
ASUS EAH5850TOP DirectCU [Supplied by ASUS]
AMD Radeon HD 5870 [Supplied by AMD]
AMD Radeon HD 5970 [Supplied by AMD]
Zotac GTX 285 AMP! Edition 1GB [flashed to stock BIOS] [Supplied by Zotac]
Heat Generation and Power
As we mentioned before the cooling has been beefed up on the GTX580. nVidia also improved the power regulation on the 580 too. These two combine to make sure that you are getting a more efficient GPU than the GTX480. Then again the GTX480 was something of a pig when it came to power consumption and heat generation.
As with most GeForce GPUs the GTX580 has hardware to monitor the temperatures generated by the GPU and to throttle it down to prevent damage to it. Starting with the GTX580 nVidia has thrown in some real time voltage monitoring hardware to help prevent the possibility of damage due to the card drawing too much voltage. This is not one part but one part for each of the two 12-Volt connections. These two circuits are monitored by the drivers and can dynamically adjust the voltage if an app tries to draw too much.
To see what types of temperatures we could get we ran Furmark and found that we hit a max temp of 87C during a 1920x1200 run @ x16 MSAA. This is actually a little down from our previous temperature mark of 89C with the GeForce GTX480 on the same run. But, let’s be honest it is not that much off the mark.
***Note, we do not remove the cooling for photographing the PCB until after all testing is done. This prevents any discrepancies in heat generation readings. All add-in graphics cards are tested as they arrive from the factory.
Unfortunately our time with the GTX580 was a little on the short side. We were not able to get too deep into the overclocking side of things before we had to starting putting the article together. We are currently working on pushing the GTX580 in addition to looking much deeper into the Tessellation performance that nVidia is boasting of. We will get that up to you as quickly as we are able to.
***Overclocking is always going to be a hit and miss thing. The speeds that we are able to reach do not indicate what all cards are capable of. We can show you potential only, your speeds will vary based on the systems you use and the individual cards purchased.
Synthetic tests are nice for many reasons; they offer an easy to use repeatable standard for testing. Most are readily available for free on the internet and can be used by anyone. But unfortunately they rarely tell the full story. Still they are an important part of the story and need to be included in any testing. Our typical synthetic tests are Futuremark’ s 3DMark Vantage, Furmark, Stalker Call of Pripyat, Direct Compute, and Unigine’s Heaven benchmark for both DX10 and DX11. With these three tests we can give you a good idea of base performance of the GPU in question.
3DMark Vantage is one of the benchmarks that everyone knows. It is used commonly for bragging rights. If you have the highest score, you win. The suite of tests covers the gamut of DX9, DX10, AI and Physics processing that you would normally see in gaming. We will run both the performance test as well as pushing the cards to complete the Extreme test run as well.
Well that is impressive; the GTX580 manages to outperform even the dual GPU 5970 in 3DMark Vantage and it can do it without PhysX enabled. The actual 3DMark Scores are close, but look at the GPU scores here. The GTX580 wins by over 4,000 points which could be a good sign for some of our later tests.
We see a very similar result with the Extreme test run although the margin this time is only 2,000 GPU points.
Unigine Heaven [DX 10 and DX11]
Unigine’s Heaven bench was one of the first tests that offered a good measure of DX11 performance. Although we now have DX11 cards from both AMD and nVidia we are still going to maintain the DX10 runs just to see how these GPUs handle the more bulky DX10 Code.
In our DX11 run the GTX580 pulls out ahead again. This time the margin for victory is much smaller than we saw with 3DMark. Here there is a gap of less than 1FPS and 24 points between the GTX580 and the HD5970.
The DX10 run tells a different tale. The GTX580 is quite a bit behind the HD5970 in both FPS and points. We are wondering if the DX10 ‘bug’ we have been seeing in the 4xx series GPUs is showing up in the 5xx now.
Stalker Call Of Pripyat Bench [DX10 and DX11]
The Stalker Call of Pripyat test uses the stalker rendering engine and images. It is capable of emulating the most common effects from DX11 and DX10 in a run of different scenes. We ran our GPUs through both and have recorded the average frame rate below. We sorted this bench by the ‘SunShafts’ test as it appeared to be the toughest on the GPUs in question.
Ok, wow. The GTX580 does VERY well with the COP bench and DX11. It does lag a little behind in the Sun Shafts test [as does everyone else] but in the other three it is more than comparable and beats the HD5970 in two of them [Rain and Night]
And the ugly DX10 bug crawls back through the cracks. Just looking at the performance here it is a little embarrassing. The GTC580 comes in slower than the HD5850 in terms of DX10 performance with this bench. Of course this is a benchmark that was at least partially subsidised by AMD, so that could be part of it…
If we take the DX11 and 3DMark Vantage scores all by themselves it looks like nVidia has this one in the bag. After all, the GTX580 scores quite well on the DX11 tests; especially with Tessellation running. However, when you put DX10 into the mix [with the exception of 3DMark Vantage] things get a little different. The GTX580 and all of the other nVidia GPUs drop behind by quite a large margin. We are not entirely sure that this is bench specific as this has appeared across multiple driver revisions from nVidia. In all reality it could be an indication of a problem with DX10 code. We are trying to find out the source of this with nVidia and will let you know once we do find anything out.
As we mentioned above we use both synthetic and real-world testing to give you the full picture of any product. For GPUs the best real world test is in-game testing. Time Demos and benchmarking tools with render engines cannot come close to showing what a graphics card can really do once it is in you system. As such we have chosen five games two DX9, two DX10 and two DX11. These games are Metro 2033, Just Cause 2, Bioshock 2, Battlefield Bad Company 2, FarCry2, and Modern Warfare 2. We played a specific level of each of these and used FRAPS 3.0.3 to measure the ability of our test subjects to render each game. All of the charts below are sorted based on the minimum frames per second each card was able to achieve. We have also pulled out the DX10 based GTX 285 from all of the DX11 games.
Modern Warfare 2 [DX9]
The final sequel is Call of Duty Modern Warfare 2. This game brings the stress and chaos of modern combat to life. It is a very short but very intense game, in it you take on multiple roles as you move through different locations around the globe and engage in all forms of combat. For testing I ran from after you are selected until the liberation of the school. Settings are shown below.
I know what you are thinking; why are we still using games like Modern Warfare 2... well because it is a popular game and it represents a piece of the DX9 puzzle. However, with the debut of the latest CoD: Black Ops, we will be refreshing the test suite. Here the GTX580 came in 5 FPS faster than the HD5870 Stalker Edition from Asus [while the Stalker was overclocked]. It is still 23 FPS behind the HD5970, but then again at speeds of over 100 FPS MINIMUM I doubt you would be able to tell them apart from watching them on the monitor.
Battlefield Bad Company 2 [DX11]
Another game in the BF franchise this one takes a stab at a more modern form of warfare. You do start off toward the end of WWII but soon are in the future [present?] for the single player campaign. The game features some decent AI [more mob style than anything] and has some great weapons. Our test run is the entire first mission. Settings are shown below.
In Bad Company 2 we find the GTX580 nipping at the heels of the Dual-GPU HD5970 with a gap of about 4FPS. Despite the ‘loss’ this is actually a very impressive showing; especially considering the gap between the GTX580 and the GTX480 here [24 FPS].
FarCry 2 is the third of the sequels we used for testing. It is a large sandbox game that does not truly have levels like most first person shooters. For testing I played from after you wake up until you free the hostage. Settings are shown below.
FarCry2 is a game that is way too close to call. The GTX580 comes in only 2FPS faster than the GTX480. That margin is well within normal variances in testing. Plus, let’s face it at the speeds they are all running it would be hard to tell even the top three apart.
Bioshock 2 [DX10]
Another sequel shows up in our testing, this one pits you in an undersea utopia gone horribly wrong. The engine is based off of the Unreal engine and offers us some DX10 surfaces to help increase the rendering load. For testing I ran the level from the beginning of the game until you find Dr Tenenbaum. Settings are shown below.
Now this is interesting. Bioshock2 shows the GTX580 WAY out in front. The gap is over 20FPS minimum and even though you would be hard pressed to see it, it does represent a very significant performance lead.
Metro 2033 [DX11]
Metro 2033 is another in the post-apocalyptic genre. This one takes place in the year 2033 20 years after a terrible event took place that almost wiped out mankind. The survivors took refuge in metro stations in Moscow. Now you must venture out to prevent the death of the few remaining people left. T The game itself plays a lot like Stalker; it is slow and filled with mutated horrors [*note 4A games, was founded by developers that worked on the X-Ray Engine used in STALKER.]. For our testing we played for 30 minutes starting with your exit from the station. Settings are shown below.
As we have mentioned in the past Metro 2033 is hard on the GPU. However, the GTX580 managed to get us very close to our 32FPS minimum mark. Much closer than any other single GPU we have tested. The game was noticeably smoother especially during the combat sequences.
Just Cause 2 [DX9]
This one is plain cheesy, but is a decent one for third person action. It is also a DX9 console port. The game put you as a CIA operative looking to regain some lost data cards and destroying everything in your path. The graphics are ok, but the AI and general story are terrible.
For Just Cause 2 the performance shown by the GTX580 was less than we hoped for. The GTX580 was even with the overclocked Stalker edition GPU but behind on the average FPS. It was also not much faster than the GTX480.
The GTX580 showed us some good performance increases over its predecessor the GTX480. In some cases the difference was almost 100%. However in most it was 35-30%. Still in a couple of games we saw almost no difference in performance. This was even taking the three runs we do into consideration. All of that aside the GTX580’s gaming performance is like most other GPUs, there are going to be games it does very well in and those that it does not. I imagine that with time nVidia will tweak the drivers and squeeze out some additional FPS to make up for those games that are not where they should be.
The GTX580 will slide into the top spot that the GTX480 used to occupy and hit the streets for the tidy sum of $499.99. But wait there is actually more to this. The GTX580 will be completely replacing the GTX480. That means that the GTX480 will be going away after the launch and ramp up of the GTX580. GF100 is becoming a GPGPU-only product, selling as Quadro and Tesla boards.
What this means to consumer market? You are seeing one of the first times that nVidia has replaced a GPU mid-stream like this. Well ok, not really it is important to remember that the GTX480 was supposed to launch in early November 2009. So what we are seeing here is a fairly normal timing, but an abbreviated schedule. When we first looked at the GTX480 we felt that its pricing was a little outside the day-to-day performance that we saw. We can tell you that we do not feel that with the GTX580. While the $499.99 price is steep, it is not beyond market expectations and the performance is certainly much more in tune with this price than what we saw from the GTX480.
When we first saw the specs on the Fermi GPU we were very excited about its potential. However one delay after another combined with issues at TSMC and then rumors that the GTX480 would not have access to all of the CUDA Cores it was designed with began to diminish our enthusiasm. Then when we finally did see the GTX480 we found it to be a very hot running card that needed a lot of power to just to operate. We were not terribly impressed with the raw performance of the GTX480, but we did start to see some of the original potential. In our opinion nVidia was starting to get back some of their image quality. The colors were more vivid, the textures seemed cleaner and the games we had played over and over felt a little more ‘alive’. Still the GTX480 was too hot, too power hungry and did not perform as well as it should have.
Now we see the GTX580 on the scene, it does perform like the GTX480 should have. It can run fast without setting off fire alarms or doubling your power bill. It is exactly what the GTX480 should have been last year. While we are not completely finished with our testing, so far we have been impressed.
Given that we heard the news from AMD's partners about the delayed launch of Radeon HD 6900, it just may as well be that the GTX580 ends the 2010 as the fastest single GPU graphics card on the planet.
GeForce GTX 580, GeForce GTX 480, GTX 480, GTX 580, GTX480, GTX580, nVidia, NVDA, GF110, GF100, GF104, Gaming, CUDA, PhysX, 3D Vision, 3D, DVI, HDMI 1.4a, DX11, DX10, DirectX,
© 2009 - 2011 Bright Side Of News*, All rights reserved.