GTX 780 Ti in 2016: Benchmark vs. GTX 1080, 1070, 1060, RX 480, More

By Published December 17, 2016 at 2:47 pm

The second card in our “revisit” series – sort of semi-re-reviews – is the GTX 780 Ti from November of 2013, which originally shipped for $700. This was the flagship of the Kepler architecture, followed later by Maxwell architecture on GTX 900 series GPUs, and then the modern Pascal. The 780 Ti was in competition with AMD’s R9 200 series and (a bit later) R9 300 series cards, and was accompanied by the expected 780, 770, and 760 video cards.

Our last revisit looked at the GTX 770 2GB card, and our next one plans to look at an AMD R9 200-series card. For today, we’re revisiting the GTX 780 Ti 3GB card for an analysis of its performance in 2016, as pitted against the modern GTX 1080, 1070, 1060, 1050 Ti, and RX 480, 470, and others.

GTX 780 Ti Specs

In case anyone’s forgotten the specs (it’s been a while), we’ve got a recap table of the GTX 780 Ti specifications below. This card came out just shortly before we really got heavy into GPU reviews, so we unfortunately have no original review to compare against – but it had about a 2-year period on our bench.

  GTX 780 Ti 
GPU GK-110
Fab Process 28nm
Texture Filter Rate
TjMax 95C
Transistor Count 7.1B
ROPs 48
TMUs 240
CUDA Cores 2880 
Base Clock (GPU) 875MHz 
Boost Clock (GPU) 928MHz 
GDDR5 Memory /
Memory Interface
3GB / 384-bit
Memory Bandwidth (GPU) 336GB/s
Mem Speed  7Gbps 
Power  1x6-pin
TDP  250W 
Output 1xDVI-D
MSRP $700, later $600

The GTX 780 Ti operates on nVidia’s Kepler architecture, using the fully enabled GK110 die (including all 15 SMXs). Kepler had a vastly different core and SM architecture from today’s Pascal, and as such, the core count is not comparable to the 10-series on a strict numbers basis. Depending on whether you’re looking at Maxwell or Pascal, core-to-core, there is as much as a 40% increase in performance per Watt versus the old Kepler architecture.

Precision has also changed, with modern Pascal GPUs focusing primarily on FP32 (short of going GP100). The 780 Ti and its GK110 GPU support FP64 on a 1:24 ratio (1x FP64 enabled core, effectively, for every 24x FP32 cores).

The point of mentioning this core architecture change is to reinforce that we’re focusing solely on gaming performance for today. This does not take production workloads into account, which may still be advantaged on the 780 Ti over some lower-end, equally-matched (in gaming) GPUs of the modern era. The 780 Ti runs more cores, and while that may not be the biggest identifier of gaming performance (especially cross-generation), it is potentially a point of advantage in some production applications.

The 780 Ti GPU hosts 2880 CUDA cores with a reference Boost 2.0 clock of 928MHz. Modern datapath optimization, clock-gating, and compression tech (like advancements in color compression) mean modern architectures are vastly more efficient. This key word, “efficient,” generally can be taken to mean greater performance per watt of energy expended. By way of example, greater color compression in memory means power savings on every bit transacted across the bus, significantly reducing overall power consumption versus effective computational power. Compression and optimizations in the pipe, like instruction-level interrupts, also mean faster and more intelligent processing and queuing of commands. Prefetching routines also help, and have also been improved since Kepler.

Other specs for the GTX 780 Ti include a 7Gbps memory clock – a bit slower than most modern devices – with 3GB of GDDR5 memory. This is on a 384-bit memory interface.

The TDP of the reference card is 250W, about 70W higher than today’s significantly more powerful GTX 1080.

A lot of the power savings are in lower voltage process. Moving to FinFET from planar gets you down to about 0.8v at the low-end, rather than 1.0v. GDDR5X on the GTX 1080 also reduces its voltage requirement to the memory, and compressing colors in memory reduces the power consumed per bit transacted.

Comparing against Pascal is sort of unfair. Kepler just can’t match up to today’s GPUs – but that’s how this industry works. The point of this comparison, of course, is to help determine which upgrades make the most sense for owners of the GTX 780 Ti cards.

We’re working with a reference GTX 780 Ti. That’s the only one we’ve got, so that means we’re operating at the specs defined in the above table. This is not a pre-overclocked partner card.

Game Test Methodology

We tested using our GPU test bench, detailed in the table below. Our thanks to supporting hardware vendors for supplying some of the test components.

Battlefield 1 and Gears of War use the AMD 16.10.1 drivers, which have less than a 1% performance change from 16.10.2 in BF1. Driver package 375.57 was used for Battlefield 1 and Gears 4 testing on nVidia hardware, with 375.57 used on the GTX 1050 & GTX 1050 Ti (and GTX 770/780 Ti in the retest). NVidia's 372.54 drivers were used for game (FPS) testing on the GTX 1080 and 1060. The 368.69 drivers were used for other devices. Game settings were manually controlled for the DUT. All games were run at presets defined in their respective charts.

Windows 10-64 Anniversary Edition was used for testing.

Each game was tested for 30 seconds in an identical scenario, then repeated three times for parity.

Average FPS, 1% low, and 0.1% low times are measured. We do not measure maximum or minimum FPS results as we consider these numbers to be pure outliers. Instead, we take an average of the lowest 1% of results (1% low) to show real-world, noticeable dips; we then take an average of the lowest 0.1% of results for severe spikes.

GN Test Bench 2015 Name Courtesy Of Cost
Video Card This is what we're testing! - -
CPU Intel i7-5930K CPU 3.8GHz iBUYPOWER  
Memory Corsair Dominator 32GB 3200MHz Corsair $210
Motherboard EVGA X99 Classified GamersNexus $365
Power Supply NZXT 1200W HALE90 V2 NZXT $300
SSD HyperX Savage SSD Kingston Tech. $130
Case Top Deck Tech Station GamersNexus $250
CPU Cooler NZXT Kraken X41 CLC NZXT $110

For Dx12 and Vulkan API testing, we use the PresentMon onPresent variable, then use our own Python script to extract the data.

Specific Test Methodologies for Games

Video Cards Tested

Continue to Page 2 for the FPS results & conclusion.

Prev Next »

Last modified on December 17, 2016 at 2:47 pm
Steve Burke

Steve started GamersNexus back when it was just a cool name, and now it's grown into an expansive website with an overwhelming amount of features. He recalls his first difficult decision with GN's direction: "I didn't know whether or not I wanted 'Gamers' to have a possessive apostrophe -- I mean, grammatically it should, but I didn't like it in the name. It was ugly. I also had people who were typing apostrophes into the address bar - sigh. It made sense to just leave it as 'Gamers.'"

First world problems, Steve. First world problems.

We moderate comments on a ~24~48 hour cycle. There will be some delay after submitting a comment.


  VigLink badge