Posted: Jon Peddie and Robert Dow 05.04.17
28% more performance for 35% more money
The Nvidia GeForce GTX1080 Ti is the world’s most powerful AIB, when run on an AMD Ryzen-based platform it gets the highest scores, but the greatest gains from generation to generation are on Intel—see charts in conclusion.
In our 7 March edition of Tech Watch , we wrote about the Nvidia GTX 1080 Ti introduction, and commented that The GTX1080Ti comes close to Nvidia’s TitanX for 42% of the price. At the time Nvidia announced the GTX 1080 Ti, the company said it would be 35% faster than GTX 1080 and 40% more expensive at $699. In our testing, we got slightly different numbers, and a surprise from AMD.
We ran a series of tests of the Nvidia GeForce GTX 1080 Ti with the 1080Ti in an Alienware Aurora (4GHz i7-6700k) and a CyberpowerPC (3.6 GHz Ryzen), using four games (DeusEx MD, Fallout 4, Sniper Elite 4, and Tomb Raider) and two benchmarks (FireStrike Ultra and TimeSpy) and found that the GTX1080 Ti on average delivered 5.1% higher scores on the AMD platform (4.0% on benchmarks, 5.5% on games, Fallout 4 being the exception). No doubt, different combinations of an Intel CPU, and/or choice of games (and even settings—we ran everything at max on a 4k screen) might skew the scores to favor an Intel-based PC, we leave that to others to demonstrate. This is what we found. Note: We are comparing against last generation (Skylake) CPU, not the latest i7-7700K. We speculate it might be only 5 – 10% faster, but it is faster.
Regardless of the CPU, the GTX 1080 Ti outclasses a standard 1080, an AMD Radeon RX 480, or even two 480s, the 1080 Ti is clearly the highest performance gaming AIB on the planet (until we test the Titan Xp).
Depending upon test, and platform, the Nvidia GTX 1080 Ti out performs the GTX 1080 by 28%, and yet costs 35% more. The GTX 1080 Ti is a completely designed and manufactured Nvidia AIB. The company sells the finished AIBs to its partners who rebrand and box them, and then sell them to distributors, retail outlets, and consumers directly worldwide.
Compared to AMD’s RX 480.
Not exactly an “apples-to-apples” comparison, we took a look at how the AMD RX 480 stacked up against the GTX 1080 Ti, and if AMD’s claims that using two RX 480s would be comparable in performance for less cost. It wasn’t.
However, when other factors are considered such as power and price, and we employ the Pmark, the story changes. The specifications used in the Pmark calculation are shown in the following table.
If a consumer is concerned about price, and/or power, AMD’s RX480 becomes more attractive.
And the surprising result of all this testing is how well both the AMD RX 480(s) and the Nvidia GTX 1080 Ti run on an AMD platform.
What do we think?
The price performance ratio doesn’t scale in moving from a GTX 1080 to a GTX 1080 Ti. However, if performance is the most important factor then the difference isn’t so great as to influence a purchase.
Nvidia is pushing Moore’s law to its limits with the GTX 1080 Ti, and we suspect the Titan Xp may push it even further.
If you examine the change from generation-to-generation of the GTX 1080 series, some remarkable results appear. We compared the test scores and price, and also the power ratio (the product of transistors times clock, which is a function of Moore’s law), and normalized it all to an AMD RX 480.
The GTX 1080 Ti’s efficiency in games however, improves considerably on an Intel platform, which shows the inter-relationship between CPU and GPU in (certain) games.
And, just as remarkable, is the tremendous falloff in generation-to-generation efficiency when using synthetic benchmarks to measure performance. This suggests aggressive driver tuning for games, something Nvidia is very good at, and probably hardware efficiencies in the GPU targeted at game performance. This also points out that the synthetic benchmarks are not providing realistic workloads relative to actual game play.
And now we await the opportunity to apply these test concepts to the Titan Xp.–JP