The GTX 1080 Ti Performance Review vs. the TITAN XP & the GTX 1080

On August 3, 2016 NVIDIA launched their Pascal flagship TITAN X at $1200.  With 3584 CUDA cores and 12GB of VRAM, it is significantly faster than the (now) $499 GTX 1080, and much faster than AMD’s aging flagship, the Fury X.   Now it is 7 months later on March 9, and NVIDIA has just launched their new GeForce gaming flagship for $699, the GTX 1080 Ti, which is supposed to be faster than the TITAN X!

BTR received a GTX 1080 Ti from NVIDIA about a week ago, and we have put it through its paces with our updated 25-game benchmark suite against the TITAN X, and versus the GTX 1080.  This evaluation will determine the world’s fastest video card using the top three fastest single-GPU cards, but we did not include AMD’s top card as it falls well below GTX 1080 performance.

We are testing our three competing cards on a clean installation of Windows 10 64-bit Home Edition using resolutions of 1920×1080, 2560×1440, 3440×1440 and at 3840×2160.  As befits testing top video cards, we use Intel’s enthusiast Z7170 platform with Core i7-6700K’s all 4 cores locked and turboed to 4.6GHz, and 16GB of Kingston’s 3333MHz DDR4.

We have already covered NVIDIA’s Editor’s Day 2017 launch of the GTX 1080 Ti which updated GameWorks for DX12 and brought us a new software benching tool for VR, but here is a brief summary of the specifications and features of the new gaming flagship, the GTX 1080 Ti.

Key Features of the GTX 1080 Ti

The next generation of games will not only look better and run faster on the Pascal GPUs, they’ll also be more immersive thanks to virtual reality headsets like the Oculus Rift and the HTC Vive.  NVIDIA has developed a number of advancements for virtual reality called VRWorks for reducing latency, improving image quality, and bringing a whole range of new content to VR.  VR framerates need to be locked to a minimum of 90 fps for fluidity – and they need to be rendered twice, once for each eye so using a very fast video card is essential.  NVIDIA also uses Multi-Res and Lens Match Shading for higher performance in VR games and applications.

The TITAN X versus the GTX 1080 Ti

To produce the GTX 1080 Ti variant of the TITAN X, its 12GB of GDDR5X was cut to 11GB so the memory interface has also dropped, from 384-bit to 352-bit.  In addition, the Ti’s ROP count is down from 96 to 88 and 256K of the Titan X Pascal’s L2 cache is gone.

To balance the effects of the drop in memory bandwidth from using 12GBs down to 11GB, NVIDIA uses a higher speed GDDR5X which means that GTX 1080 Ti actually has slightly more bandwidth than that of the TITAN X.

The base clock speed of the GeForce GTX 1080 Ti is 1480MHz and the typical Boost Clock speed is 1582MHz, slightly above the TITAN X boost of 1530MHz. The GeForce GTX 1080 Ti has also been outfitted with next generation GDDR5X memory modules running at 11 Gbps – the TITAN X memory is 10Gbps.

Check out the GTX 1080 Ti overview and block diagram.

Here are some more specifications for the GTX 1080 Ti:

From comparing the specifications of both cards, we can predict that the TITAN X will be up to about 5% slower than the GTX 1080 Ti. The price of GTX 1080 Ti is set to start at $699 and will also include a choice of a bundled copy of For Honor or of Ghost Recon Wildlands.

Since the GTX 1080 Ti is a new addition to NVIDIA’s product lineup, the original GTX 1080 is moving to occupy the $499 spot by dropping $100 off of its launch price.  The GeForce GTX 1080 Ti will be released on March 10th, tomorrow. It will be a hard launch, with immediate availability and will be sold on nvidia.com as well as by their board partners.  Board partners will offer GeForce GTX 1080 Ti cards based on their own custom designs in April along with overclocked GeForce GTX 1080 boards with 11 Gbps G5X memory and overclocked GTX 1060 boards with 9 Gbps G5 memory.

Here is NVIDIA’s line up with the GTX 1080 Ti now the gaming flagship.   The TITAN X will remain at $1200 and will mostly remain popular among the scientific application crowd for Deep Learning calculations.

GeForce GTX 1080 Ti
GeForce GTX 1080 11 Gbps
GeForce GTX 1080
GeForce GTX 1070
GeForce GTX 1060 9 Gbps
GeForce GTX 1060
GeForce GTX 1050 Ti
GeForce GTX 1050

How does the GTX 1080 Ti compare with AMD’s flagship, the Fury X?

It doesn’t.  The GTX 1080 Ti and the TITAN X are simply in a class or two above the Fury X as the GTX 1080 dominates also.  In fact, NVIDIA’s 4th card, the GTX 1070, is also faster than the Fury X.  Because of this performance disparity, we did not test the Fury X in this evaluation.  We are looking forward to testing AMD’s upcoming Vega video cards which should be significantly faster than the Fury X.

However, before we do performance testing, let’s take a closer look at the GTX 1080 Ti and check out overclocking and noise.

  • Joe Calidude

    Great Review! Much appreciated!

  • kevmolio

    What if you get a Titan XP for $500 from the fan boys running to the 1080ti for 15 more FPS :)? As long as I hit 4k 60FPS, I don’t need much more than that :/ I’ll snag the cheapest card, thanks.

  • Me

    “If you currently game on any other video card, you will do yourself a favor by upgrading. ” Better said
    “If you currently game on any other video card than GTX1080 or TitanXP, you will do yourself a favor by upgrading”. I see no point in to pay another 700 USD at not even one year since 1080 launch for 30% performance improvment.

  • femonk

    hi,
    the article says ‘The only reason a gamer might choose a TITAN X instead would be if he also works with Deep Learning algorithms.’.
    Can anyone tell me, why the TITAN X would be better for Deep Learing than the GTX 1080 Ti even if the specs are almost identical? Thanks a lot.

    • apoppin

      Yes. The TITAN X has 12GB of vRAM vs 11GB and it has an instruction set for deep learning inference.

      From the TITAN X launch article:
      http://www.babeltechreviews.com/pascal-titan-x-vs-gtx-1080-first-benchmarks-revealed/
      “TITAN X features Peak Single Precision performance – 11 TFLOPs FP32 (32-bit floating point) and also has a new instruction set for deep learning inference, 44 TOPS INT8. It features 12GB of non-ECC GDDR5X memory for scientific applications and also for extreme resolutions in gaming.”