AMD Radeon R9 Fury X Video Card Review

We review AMD's new Fiji GPU comprising the new AMD Radeon R9 Fury X video card with stacked chip technology High Bandwidth Memory. We take this video card through its paces, make comparisons and find out what it can do for us in real world gameplay. Is this $649 video card competitive? Is it truly geared for 4K gaming as AMD says?

 

The Radeon R9 300 Series

Before we begin with what exactly Fiji is, and the Fury video card series lineup, let's start by going back just a bit and talking about the AMD Radeon 300 series, which also launched "alongside" Fiji. The AMD Radeon 300 series is a refresh (or rebrand according to many) of products based on very familiar GPUs and feature sets. We recently evaluated the high-end GPU of this lot, the AMD Radeon R9 390X via the MSI Radeon R9 390X GAMING 8G video card.
The AMD Radeon R9 390 and 390X are very familiar products because these are inherently re-brands of the AMD Radeon R9 290 and 290X with a higher clock speeds and 8GB of VRAM standard. It is important to note these video cards and note AMD's design goals with these video cards, before we start talking about the AMD Radeon Fury series, which supersedes the Radeon R9 390 and R9 390X series.
AMD made it clear that the AMD Radeon R9 390 and 390X are video cards that are quote: "Designed for great 4K gaming experiences." In fact, as you travel around the AMD reviewers guide and press presentation documents you will find many references to the fact that the AMD Radeon R9 390 series is designed for a 4K gaming experience. This is not a suggestion, but rather a solid bullet point that AMD PR is not afraid to loudly claim.
In fact, here is a very poignant quote directly from the reviewers guide that needs to be brought to light so you understand where AMD sees both the Radeon R9 390 series and the Radeon R9 Fury series. This quote pertains to the Radeon R9 390 series:

"The Radeon™ R9 390X and R9 390, are best used in 4K gaming applications. 4K is the present and future of gaming, and as such AMD is dedicated to providing the premiere 4K gaming experience. Some gamers prefer 1440p however. For them, we have the Radeon™ R9 380. When it comes to playing all the latest games at 1440p, the best choice all around, is the R9 380."

That's right, AMD sees the AMD Radeon R9 390 series as the cards best used in 4K, and doesn't even recommend the lower and more common and popular resolutions of 1440p or 1080p for the R9 390 series. The AMD Radeon R9 390 and 390X now have 8GB of VRAM standard to help facilitate better 4K performance.
Why are we dragging out the design goals for the R9 390 series in a Radeon R9 Fury X review? Because AMDs design goals for the AMD Radeon Fury series are also for a 4K experience, but there are certainly differences between the cards. It is interesting to take note of AMD's firm stance on 4K, and firm stance that lesser resolutions like 1440p and 1080p are not the goals with Fury X, even if those are the more common resolutions gamers use.

---

AMD's new Fiji GPU

The AMD Radeon Fury series is based on a new GPU called Fiji. In the grand scale of GCN this would be above a Tonga GPU in architecture (R9 285). This means it is updated when compared to Hawaii, Radeon R9 290X and 390X (390X would be GCN 1.1 along with Hawaii). The best way we can show you the changes between GPUs is in this chart below that we have created, along with a spec comparison chart. Please click these for bigger versions and study the differences.

You will find that in this newer GCN version with Fiji that all the things Tonga supports are in Fiji, along with some new things. That means updated geometry processors for better tessellation, the updated delta color compression in Tonga, and several other things. On the whole, Fiji is the most advanced GPU ever built by AMD, containing the latest feature set possible.
The complete list of new architecture changes in Fiji are:
• New memory interface design based on HBM technology
• Updated geometry processors with improved tessellation performance
• Lossless color compression for frame buffer reads and writes
• New 16-bit floating point and integer instructions
• New data parallel processing instructions (sharing between SIMD lanes)
• Improved compute task scheduling
• L2 cache size increased to 2 MB
• Updated display scaler with improved quality
• Updated video decode engine with support for HEVC

LINK