Product Listing

NVIDIA GeForce GTX 780 Ti - More Cores Unlocked!

By Wong Chung Wee - 7 Nov 2013

NVIDIA GeForce GTX 780 Ti - More Cores Unlocked!

NVIDIA GeForce GTX 780 Ti - More Cores Unlocked!

The NVIDIA GeForce GTX 780 Ti card was announced in Montreal under a shroud of mystery last month. At NVIDIA’s Montreal event, the card made an appearance but it was whisked away before any of the card’s details, especially its specifications, were made known to public. In addition to its flash-in-the-pan appearance, the price of the card, at US$699, was announced subsequently. We understand this card is most likely seen as NVIDIA’s attempt to steal some thunder from AMD’s recent launch of their new Radeon R9 290 series Hawaii GPU-based cards. The NVIDIA GeForce GTX 780 Ti is clearly targeted at the top-end R9 290X edition.

The NVIDIA GeForce 780 Ti sports a fully unlocked GK110 core, with 2880 CUDA cores. It has a base clock of 876MHz, and a boost clock of 928MHz. Its 3GB of GDDR5 video memory is rated at 7000MHz, with a 384-bit memory interface. (Image Source: NVIDIA)

Based on its appearance, the GTX 780 Ti appears identical to the GTX Titan, GTX 780 and GTX 770 cards. In fact, many would have thought it's nothing more than an overclocked GeForce GTX 780. However, that's far from reality when looking underneath its massive cooler.

Under its hood, the GTX 780 Ti uses the same GK110 core that's used on the GTX 770, GTX 780 and the Titan. However, the GTX 780 Ti has 15 Streaming Multiprocessors (SMX), which is one more than GTX Titan's count of 14! On the GTX 780, its core only features 12 operational SMX units. Just when we thought the GeForce GTX Titan has unlocked all the potential of the GK110, NVIDIA surprised us once more. 

As a result, the GTX 780 Ti has a total of 2880 CUDA cores for single precision processing, and a total of 960 CUDA cores for double precision operations. However, do note that the double precision cores on the GTX 780 Ti operate at 1/24 their full speed, to the tune of 210 GFLOPS or 0.21 TFLOPS. On the other hand, despite GTX Titan's lower number of 896 double precision CUDA cores, they operate at full speed to churn 1.3 TFLOPS. This is the major difference that separates the Titan apart from the GTX 780 Ti and a clear signal that the Titan has always been a solution for workstation and engineering tasks (but also doubled up nicely for gaming) as we've personally found from the GeForce GTX Titan PlayTest event we held with NVIDIA.

Summary of Single vs. Double Precision Processing  Throughput
  GeForce GTX 780 Ti GeForce GTX Titan GeForce GTX 780
CUDA cores (single precision) 2880 2688 2304
CUDA cores (double precision) 960 896 NIL
Double precision processing speed 1/24 Full Speed NIL
Double precision processing throughput 0.21 TFLOPS 1.3 TFLOPS NIL

Back to the GTX 780 Ti, it has a total of 240 texture mapping units, and 48 raster operators. While the GTX 780 Ti has the same of of ROPs as the Titan and GTX 780, its increased number of TMUs will put the card in good stead during 3D scene rendering. As such, we are already expecting the card to perform better.

 

New Feature - A Balance of Power

The GTX 780 Ti has support for GPU Boost 2.0 that primarily makes use of the GPU core's temperature threshold to  dynamically adjust core clock speeds for better performance. It also supports the new fan controller that was first featured in the GTX 780 card. However, the newer GTX 780 Ti has a nifty feature that balances power over its three power sources or rails. Like any modern graphics card, it draws power from the Molex power connectors as well as from the PCIe interface. During overclocking, the user runs the risk of maxing out power draw at any of the three rails. Once this occurs, the card will throttle itself, and will result in a lower overclocked state. However, with this new power balancing feature, the GTX 780 Ti is able draw upon power from the other rails to supplement its needs level up the lagging power sources, so that all of them would be drawing the same maximum power level. As such, the GTX 780 Ti is touted to have better overclocking capabilities than the other GeForce GTX cards.

In terms of video connectivity options, the GTX 780 Ti has two Dual-Link DVI ports, a HDMI port and a DisplayPort port. The card supports G-Sync, which is a new technology that synchronizes the monitor’s refresh rates to the GPU’s draw rates. The end result, really is that monitors will now have a variable refresh rate, ranging from as low as 30Hz to as high as 144Hz. However, the monitor needs to feature a G-Sync add-on board that will significantly increase its costs.

The card features two Dual-Link DVI ports, a HDMI port and a DisplayPort port for support of up to four displays. (Image Source: NVIDIA)

To recap, the launch price of the card is US$699. In the wake of its announcement, NVIDIA also announced price cuts to the incumbent cards, consisting of the GeForce GTX 780 and GTX 770. Their new prices are US$499 and US$329 respectively. However, there will be no price reduction for the GTX Titan as the company wants to preserve the supercomputing pedigree of the card. Although the Titan sports less CUDA cores than the GTX 780 Ti, it has a much more higher double precision compute capabilities than the former. It also sports twice as much video memory buffer at 6GB. At launch, the GTX 780 Ti is positioned as a "high-end enthusiast grade graphics card", and it is expected to take the spot as NVIDIA's top consumer graphics card. Let us put the card through its paces to see if the statement rings true!

Here's a look at how the NVIDIA GeForce GTX 780 Ti card compares against the competing cards:

NVIDIA GeForce GTX 780 Ti 3GB GDDR5 and competitive SKUs compared
  NVIDIA GeForce GTX 780 Ti NVIDIA GeForce GTX Titan NVIDIA GeForce GTX 780 AMD Radeon R9 290X AMD Radeon HD 7970 GHz Edition 3GB DDR5
  NVIDIA GeForce GTX 780 Ti NVIDIA GeForce GTX Titan NVIDIA GeForce GTX 780 AMD Radeon R9 290X AMD Radeon HD 7970 GHz Edition 3GB DDR5
Core Code
  • GK110
  • GK110
  • GK110
  • Hawaii
  • Tahiti XT
GPU Transistor Count
  • 7.1 billion
  • 7.1 billion
  • 7.1 billion
  • 6.2 billion
  • 4300 million
Manufacturing Process
  • 28nm
  • 28nm
  • 28nm
  • 28nm
  • 28nm
Core Clock
  • 876MHz
  • 836MHz
  • 863MHz
  • 1000MHz
  • 1050MHz
Stream Processors
  • 2880
  • 2688
  • 2304
  • 2816
  • 2048 Stream processing units
Stream Processor Clock
  • 876MHz
  • 836MHz
  • 863MHz
  • 1000MHz
  • 1050MHz
Texture Mapping Units (TMUs)
  • 240
  • 224
  • 192
  • 176
  • 128
Raster Operator units (ROP)
  • 48
  • 48
  • 48
  • 64
  • 32
Memory Clock (DDR)
  • 7000MHz
  • 6008MHz
  • 6008MHz
  • 5000MHz
  • 6000MHz DDR (GDDR5)
Memory Bus width
  • 384-bit
  • 384-bit
  • 384-bit
  • 512-bit
  • 384-bit
Memory Bandwidth
  • 336 GB/s
  • 288.4 GB/s
  • 288.4 GB/s
  • 320GB/s
  • 288GB/s
PCI Express Interface
  • PCI Express 3.0
  • PCI Express 3.0
  • PCI Express 3.0
  • PCIe v3.0 x16
  • PCIe ver 3.0 x16
Power Connectors
  • 1 x 6-pin, 1 x 8-pin
  • 1 x 6-pin, 1 x 8-pin
  • 1 x 6-pin, 1 x 8-pin
  • 2 x 8-pin power cables
  • 1 x 6-pin, 1 x 8-pin
Multi GPU Technology
  • SLI
  • SLI
  • SLI
  • Improved AMD CrossFire
  • CrossFireX
DVI Outputs
  • 2
  • 2
  • 2
  • 2
  • 1 x Dual-Link
HDMI Outputs
  • 1
  • 1
  • 1
  • 1
  • 1
DisplayPort Outputs
  • 1
  • 1
  • 1
  • 2
  • 2 (version 1.2 HBR2)
HDCP Output Support
  • Yes
  • Yes
  • Yes
  • Yes
  • Yes

 

Test Setup

These are the specifications of our graphics testbed:

  • Intel Core i7-3960X (3.3GHz)
  • ASUS P9X79 Pro (Intel X79 chipset) Motherboard
  • 4 x 2GB DDR3-1600 G.Skill Ripjaws Memory
  • Seagate 7200.10 200GB SATA hard drive (OS)
  • Western Digital Caviar Black 7200 RPM 1TB SATA hard drive (Benchmarks + Games)
  • Windows 7 Ultimate 64-bit

Below is the list of cards we'll be testing. For the reference GTX 780 card, we had to clock down the ASUS GeForce GTX 780 DirectCU II 3GB GDDR5 to the default operating values of the intended reference card. For the AMD Radeon HD 7970, the downclocked Sapphire Radeon HD 7970 OC 3GB GDDR5 was used. We also took the chance to pit the top-end AMD R9 290X "Hawaii" card against the best ones from NVIDIA, in order to have a gauge of its performance standings.

  • NVIDIA GeForce GTX 780 Ti 3GB GDDR5 (ForceWare 331.70)
  • NVIDIA GeForce GTX Titan 6GB GDDR5 (ForceWare 331.65)
  • NVIDIA GeForce GTX 780 3GB GDDR5 (ForceWare 331.65)
  • AMD Radeon R9 290X 4GB GDDR5 (AMD Catalyst 13.11 Beta 7)
  • AMD Radeon HD 7970 GHz Edition 3GB GDDR5 (AMD Catalyst 13.9)

Note 1: In temperature and power consumption comparisons, the results used were from the data gathered from the actual reference cards. Please refer to our reviews for the NVIDIA GeForce GTX 780 and AMD Radeon HD 7970 GHz Edition.

Note 2: In our overclocking testing, we overclocked the ASUS GeForce GTX 780 DirectCU II 3GB GDDR5 to simulate an overclocked reference GTX 780 card.

Benchmarks

Here's the full list of benchmarks that we'll be using for our assessment:-

  • Futuremark 3DMark 2013
  • Crysis 3
  • Unigine 4.0 "Heaven"
  • Hitman: Absolution
  • Far Cry 3

For our temperature and power consumption tests, 3DMark 2011 was used.

 

Join HWZ's Telegram channel here and catch all the latest tech news!
9.0
  • Performance 9
  • Features 9
  • Value 8
The Good
Great overall performance
Power balancing feature for overclockers
Quiet operation
The Bad
Pricey
Our articles may contain affiliate links. If you buy through these links, we may earn a small commission.