NVIDIA, ASUS GTX 560 Ti Reviewccokeman - January 25, 2011
Category: Video Cards
Price: NVIDIA $249, ASUS $269
» Discuss this article (55)
When introduced last year in July, the GTX 460 was NVIDIA's card that hit the right performance points for the right price and really shook up the mid-range segment by running the latest DX 11 games at frame rates that approached (if not surpassed) the performance of AMD's flagship single-GPU cards, the HD 5870 and HD 5850. This for a much more appealing price point, offering the gamer on a budget the ability to step up to the plate and get big league performance for minor league pricing. The GTX 460 fell into what NVIDIA calls the "Hunter" class of video cards that don't have all the firepower, but use their speed to help close the performance gap. When the GTX 460 dropped, the board partners quickly had a multitude of hopped up cards with better cooling and non-reference PCB designs that took the performance to another level. You had ECS with a massive cooler from Arctic Cooling, Palit with its Sonic version, MSI with the Hawk that featured an overbuilt beast of a board with additional voltage regulation capabilities for the enthusiast, to EVGA with its FTW edition card with an 850MHz core clock right out of the box. These cards helped boost the stature of the GTX 460 as a worthy successor to the 8800GT in terms of price-to-performance. But I digress — let's fast forward a scant six months later and we have what should be the replacement for the GTX 460, the GTX 560 Ti. The Ti stands for Titanium, which is lighter and stronger than steel, referencing that you get more for less. The size of the card comes in at a mere nine inches, which looks to deliver brutal performance for its price point.
The GTX 560 Ti features a revamped GF114 core that has been redesigned down to the transistor level to make the core more efficient so that the amount of hardware on board could be increased without creating a hot-running beast. This allowed NVIDIA to bump the CUDA core count up to 384 while also increasing the clock speeds. The GTX 460 that the GTX 560 Ti is replacing was an overclocking beast, so NVIDIA made sure this line has the same capabilities, providing better power regulation and better cooling on the reference design. Let's take a look at what NVIDIA and its partners have to offer with the introduction of the GTX 560 Ti, the latest rendition of its "Hunter" class card.
The first card I will be looking at is the reference card from NVIDIA. Since this one comes without any retail packaging, we'll get right to the card. The GTX 560 Ti measures a smallish nine inches in length and features a central fan blowing through a heat-pipe based cooling solution. The shroud has been designed with the center portion in a dish shape to allow plenty of airflow into the card even when running two of the GTX 560 Ti's in an SLI configuration. The fan blades sit up higher than the shroud to increase the airflow into the card without increasing the noise signature. The back side has little of interest, but it looks like a series of Hi-C caps are in use around the bottom of the GPU socket for the four phase power circuit.
Connectivity comes in the form of two Dual Link DVI ports and a single mini HDMI port that offers Bitstreaming support for Dolby True HD and DTS-HD Master Audio. To run a surround setup, you will still need to use a pair of cards since the GTX 560 Ti only supports two outputs being used at one time, so the third monitor must be connected to a second card. NVIDIA's stance is, if you are going to run a three monitor setup, the company wants you to have the best possible experience and that takes more than a single card to deliver great performance at a resolution of 5760x1080. The power requirements for the GTX 560 TI are two 6-pin PCIe power connections. The TDP on the reference card is 170 watts with a 500 watt power supply recommended.
The shroud comes off to show the heat pipe-based cooling solution employed by NVIDIA for the GTX 560 Ti. This cooling solution is a two part system with the cooling assembly for the GPU and a plate-style system to provide cooling for the VRM circuit components and the 1GB of memory. The whole assembly comes off when you remove the 15 screws that hold it onto the PCB. The GPU heat sink can then be removed once the plate is off the card. The cooling plate has a small set of fins cast into it above the VRM circuits for added cooling from the airflow passing over them. With the overbuilt four phase power design, cooling was significantly upgraded as well to take advantage of design. This seems to help with the clock speeds this card was capable of reaching. The power monitoring hardware that was introduced with the GTX 580 has been incorporated into the GTX 560 Ti and, as before, NVIDIA states that this will not affect overclocking, but is meant to prevent damage to the power subsystems from the loads imparted by programs such as Furmark and OCCT.
The cooling solution has a copper plate with three heat pipes running through it that transfer the heat from the GPU to the fin arrays. The three heat pipes are 6mm in diameter and spread out to a pair of aluminum fin arrays. In the center is still a large solid aluminum heat sink that is the base for the cooling fan from AVC. About the only information I can glean from a few quick searches is that the info on the back of the fan is all I have — the fan is 80mm in diameter, 15mm thick, runs on 12V, and uses ball-type bearings. When pushed, the fan was not loud at its 75% maximum indicated speed.
The GF114 GPU is the latest in NVIDIA's Fermi arsenal. As such, it has benefited from all the latest tweaks all the way down to the transistor level, much like the work done to its more powerful cousins, the GTX 580 and 570. These increased power efficiency, allowing NVIDIA the ability to increase the clock speeds while at the same time reducing leakage. While the GTX 560 is the successor to the wildly popular GTX 460, it also caries with it the same streaming multiprocessor configuration used in the GTX 460's GF104 core. The GF114 is built using a 40nm process and has two GPC (Graphics Processing Clusters), eight Streaming Multiprocessors, 384 CUDA cores, 64 Texture units, and 32 ROP units. Each SM is equipped with a polymorph engine for a total of eight with a total of two raster engines. On board is 1GB of 5Gbps rated GDDR5 memory from Samsung that runs through four 64-bit (256-bit) memory controllers. The clock speeds delivered on this reference card are 822MHz on the Fixed Function units, 1644MHz on the CUDA cores, and 1000MHz (4000MHz QDR).
Sure the reference card is nice and will certainly be popular, but let's see what an AIB partner has to offer as a point of difference from the reference design. Usually this includes a hopped up PCB and better cooling to increase performance even further than what is available from the OEM. Let's find out.