ASUS GTX 580 Reviewccokeman - November 29, 2010
» Discuss this article (19)
For all intents and purposes, the ASUS GTX 580 looks to be a reference design card, but the shroud is equipped with ASUS branding so you know where this card came from and can proudly show off your loyalties. The ASUS GTX 580 comes to market with the clock speed of the fixed function units set to 782MHz on the core ,or 10MHz higher than the reference design clock speed, while the 1.5GB of GDDR5 memory is at the standard 2004MHz. This gives you the potential for a light performance increase right out of the box. The GTX 580 measures 10.5 inches long from end to end, meaning it is on the larger end of the spectrum, but most of today's enthusiast-style chassis should present no problems. When you look at the front of the card, the differences between the GTX 580 and GTX 480 are quite easily seen. Gone are the protruding heat pipes and massive heat sink that dominated the external view of the GTX 480. That large of a cooling solution is not required with the GTX 580 since NVIDIA went back to the drawing board and completely redesigned the architecture down to the transistor level for improved power efficiency and thermal performance. The rear of the card is pretty mundane with the surface mount components and the screws used to hold the shroud and heat sinks onto the PCB. ASUS uses a system of covers on the Dual Link DVI connectors and the PCIe slot connector to make sure there is no damage to these connection points once it leaves the factory. Just one added precaution.
The ASUS GTX 580 uses the reference connectivity solution in the form of two Dual Link DVI ports and a single Mini HDMI port. With this configuration, you are still limited to two displays, so if you are looking at a surround setup, or better yet a 3D surround setup with the addition of NVIDIA's 3D Vision system, you will need a second graphics card. At least this gives you the graphics horsepower to get the best visual experience with the frame rates demanded by gamers. The rear end of the card contains a depression near the fan opening to bring in as much cool air to the GPU when used in an SLI multi GPU setup. This should allow the top card to run cooler for some worry free gaming.
The dual bridge connections on the ASUS GTX 580 means that it can support up to a three-card SLI setup at resolutions up to 7680x1600. By going with more than one card, you can use an NVIDIA 3D Surround multi monitor setup for a truly immersive gaming experience with a trio of 120Hz monitors. The ASUS GTX 580 has a TDP of 244 watts and you will need a power supply that includes both a 6 and 8-pin connector if you do not want to use adapters. This TDP is 56 watts less than the power hungry GTX 480. NVIDIA recommends a 600 watt or greater power supply when using the GTX 580. The power supply requirements obviously scale up when you use more than one card. You can look up the listing of NVIDIA certified power supplies over at SLI Zone under the Certified Products tab.
To verify that this truly is a reference design card, you have to get under the hood and take a look. NVIDIA took what it learned from the GTX 480 and improved the thermal design characteristics. By fixing this problem, it gained some flexibility in the size and type of cooling solution that could be used. For this build, the card uses a large Vapor Chamber heat sink that is smaller and more efficient than the massive cooler used on the GTX 480. To keep the voltage regulation circuits and memory modules cool, an aluminum plate is used as a full cover heat sink. This plate is held to the PCB by screws and keeps the PCB from flexing as well as functioning as a heat sink. The diagram below provides an illustrated view as to how a "Vapor Chamber" works. Just imagine it as a large flat heat pipe.
Looking at the heatsink assembly once it is removed, you see that it is a substantial cooler. The large fin array runs the length of the Vapor Chamber and directs heat out through the vent on the I/O bracket. On top of the heat sink are a series of strips of insulation that keep the air driven by the fan going through the fins of the array to maximize cooling efficiency. The Vapor Chamber has a raised platform that makes contact with the integrated heat spreader of the GF 110 core so that is the only point of contact.
The GF 110 GPU was completely redesigned down to the transistor level with "lower leakage transistors on less timing sensitive processing paths and higher speed transistors on more critical processing paths," as NVIDIA puts it. Through this redesign, the clock speeds were able to be increased as well as lowering the power consumption — a win on both counts. With the GF 110 GPU used on the ASUS GTX 580, we get the full capabilities of the Fermi architecture. You get four graphics processing clusters, 16 streaming multiprocessors (instead of 15), 512 CUDA cores (instead of 480), 16 Polymorph engines, 64 texture units, 48 ROP units. The clock speeds have been bumped from 772MHz on the fixed function units and 1544MHz on the CUDA cores to 782MHz and 1564MHz, respectively, on this ASUS card. The ASUS GTX 580 uses the reference specification 1536MB of GDDR5 memory that runs through a 384-bit interface with 768K of shared L2 cache. The memory used on this card is manufactured by Samsung and is rated to run at 1250MHz.
Let's see if the additional 10MHz on the core makes a difference in performance. I'm sure the Voltage Tweaking will, but how much is the question.