Jul 12, 2022
As a result of the stacking, 1 GB of HBM takes up a massive 94% less onboard surface area than 1 GB of regular GDDR5 memory does. It is now time for the Fury X. After much anticipation, and we can demonstrate how well AMD's newest high-end GPU operates on the company's spiffy new liquid-cooled graphics card. This comes after a great deal of marketing around the product. We have conducted tests in various games using our well-known frame-time-based metrics, and we now have a complete set of data to share with you. Let's get to it.
Over the last several weeks, practically everything about the new Radeon GPU that most people would want to know has become public information; the only thing that hasn't been revealed is how well it performs. Let's take a minute to recap this information in case you've been living under a rock for the last few moments. Two brand-new fundamental technologies lie at the foundation of the Radeon R9 Fury X: the Fiji graphics processor and a brand-new category of RAM called High Bandwidth Memory (HBM).
It has been over two years since AMD Radeon released a new high-end graphics processing unit (GPU), and the Fiji GPU is the giant chip in a series of products based on the GCN architecture that dates back to 2011. Even the Xbox One and the PlayStation 4 are built on GCN, but Fiji is an upgraded version of that technology constructed on a far grander scale.
I find that most conversations on a graphics card's physical elements are dull compared to the topics around the GPU architecture. Though, the Fury X card because this card is unique in comparison to the standard fare in several rather significant aspects. The card is unique in that it is just 7.7 inches long and has an umbilical cord protruding from its abdomen in the direction of an external water cooler. This is where the peculiarities begin. Due to the stringent constraints, AMD has put on this product; you can anticipate all Fury X cards to have the same unique configuration. Board manufacturers won't have the freedom to adjust clock speeds or provide bespoke cooling solutions for the Fury X.
Instead, the vanilla Radeon R9 Fury, scheduled to be released in the middle of July at prices beginning at $550, will be the territory of custom graphics cards. Compared to the Fury X, the R9 Fury's graphics processing unit (GPU) resources will have some reductions made, and customized boards and cooling will be usual. AMD has informed us that we may anticipate specific variants of the Fury with liquid cooling and others with more traditional air cooling.
The Catalyst Control Center now has individual sliders for the GPU's clock speed, power limit, temperature, and maximum fan speed. This change was made to make the overclocking process easier. Users may direct the PowerTune algorithm from AMD to achieve their system's optimal balance of acoustics and performance.
Our review unit of the Fury X does have one significant limitation, even though it has a lot of positive qualities. The card makes a consistently high-pitched whine whenever it is turned on, regardless of whether it is actively being used. It is not the typical burble of pump noise, the whoosh of a fan, or the erratic chatter of coil whine; instead, it is a constant screech similar to what an outdated CRT monitor may produce. Although the noise does not reach the threshold at which our sound level meter registers it, it is not difficult to pick up on.
Fraps is a piece of software that can record the amount of time it takes to render each frame of animation and collect the majority of the data you will see on the following pages. When collecting data from setups with a single GPU, we sometimes use a program known as FCAT. However, this step is not required in most cases to get accurate results. On the other hand, we have filtered the findings of Fraps by utilizing a moving average of three frames. It is expected that this filter will consider the impact of Direct3D's three-frame submission queue.
To navigate between the different scenarios, use the buttons located up top. The frame timings that you will see are taken from one of the three separate test runs that we carried out for each card. You will note that personal computer graphics cards do not always create smoothly flowing progressions of subsequent animation frames, contrary to what the phrase "frames per second" indicates. Instead, the frame time distribution is a hairy and capricious beast that may vary significantly in appearance and behavior.