Product Listing

AMD Radeon RX 480 review: A budget card with not-so-budget performance

By Koh Wanzi - 2 Jul 2016

Introduction

Note: This article was first published on June 29.

Performance for the masses

AMD Radeon RX 480

We’ve said it before, but we’ll say it again. These are some exciting times to be a PC gamer. Following the launch of the Pascal-based NVIDIA GeForce GTX 1070 and 1080 – both of which offered unprecedented levels of performance at their respective price points – AMD announced the Radeon RX 480 at Computex 2016 at a staggeringly value-oriented price of just US$199 for the 4GB version.

More interestingly, AMD made the tantalizing claim that two Radeon RX 480 cards in CrossFire would perform better than a single GeForce GTX 1080. That meant that you could theoretically surpass NVIDIA’s US$699 flagship for less than US$500, which added points in the value department once again (if CrossFire works out well everywhere).

Since then, AMD has completed its Polaris product stack with the announcement of the Radeon RX 460 and 470 at E3, both of which target gamers seeking more value oriented options that can still tackle some level of VR gaming and high performance non-VR performance respectively.

That actually sets quite an interesting tone for AMD overall. Until Vega – AMD’s next-generation performance silicon – launches, it looks like the red camp will not be competing directly with NVIDIA’s high-end Pascal cards. Instead, AMD has set its sights on appealing to the masses, and given the price of the RX 480 and the performance it offers, that may just turn out to be a rather sound decision.

 

A fourth-generation GCN architecture

AMD Radeon RX 480

The Radeon RX 480 is based on Polaris, which is actually a fourth-generation Graphics Core Next (GCN) architecture. While Fiji added features like support for High Bandwidth Memory (HBM), Polaris appears more focused on optimizations like improved geometry processing, enhanced shader efficiency, and more flexible asynchronous compute capabilities.

These efficiency gains have allowed AMD to squeeze extra performance from the GCN architecture, which debuted in its first iteration in 2012. For instance, the enhanced geometry engine now features a new indexed cache that reduces data movement and frees up internal bandwidth resources.

And like Pascal, Polaris also utilizes a more efficient memory and delta color compression engines. Delta color compression, or DCC for short, is a method of increasing the effective memory bandwidth by taking advantage of the fact that whole blocks of pixels often store very similar data. Normally, a GPU would treat individual pixels as if they contained unique and vastly disparate values, but because DCC processes blocks of pixels and stores only a single, precise value (the rest are stored as a delta), it is able to reduce the required bandwidth.

DCC has been in use since GCN 1.2 (or the AMD Radeon R9 285 and later cards), and Polaris now features an updated memory controller that supports up to 8Gbps GDDR5 memory. The overall effect is to extend the viability of GDDR5 memory, a standard that is all too likely going to be phased out in favor of GDDR5X and HBM2 memory in future cards.

AMD Polaris DCC

And since this is still GCN we’re talking about here, Polaris also has asynchronous compute capabilities, which is the ability to process compute and graphics workloads simultaneously. This includes recently added features like Quick Response Queue, which allows greater flexibility in handling both types of tasks. While GCN might prioritize graphics over compute tasks, Quick Response Queue now enables the GPU to devote more resources to compute tasks if they are more urgent, without stopping the graphics task entirely

Alternatively, compute tasks are also able to preempt graphics workloads, and the GPU is able to divert resources entirely to the former where necessary.

The graph area signifies how much of the GPU's resources are being directed to either compute or graphics tasks over time.

 

Making VR affordable

AMD Radeon RX 480 block diagram

The Radeon RX 480 features 36 Compute Units (CUs), for a total of 2,304 stream processors and 144 texture mapping units (TMUs). The number of CUs actually puts it somewhere between the Radeon R9 380 and 390X, but at a much lower price than the latter. Our review unit is the 8GB model that's priced at US$239, which is slightly more expensive than its 4GB counterpart. Here’s a quick snapshot of the card’s specifications:

A quick look at the card's specifications.

It is clocked at a 1,266MHz base clock, and boasts up to 5.8 TFLOPS of single-precision floating point performance, which isn’t that far behind the 6.46 and 8.9 TFLOPS the NVIDIA GeForce GTX 1070 and 1080 offer respectively, especially when you consider their prices.

The reference design has a fairly short PCB, although the blower-style fan extends beyond that. But because of the vents on the top of the card, it looks like the fan might actually exhaust some heat into the case, and not just out the back of the chassis. The cooling shroud is constructed of plastic, and the dotted grid pattern is reminiscent of the design we saw on cards like the Radeon R9 Fury X.

AMD Radeon RX 480

Round the back, the Radeon RX 480 sports three DisplayPort 1.3/1.4 ports and an HDMI 2.0b connector. These are essentially the latest display standards, so the RX 480 supports things like HDR video and higher resolution displays with higher refresh rates, for instance standard definition 4K video at 120Hz. AMD announced its plans to enable more vibrant and higher fidelity multimedia content earlier this year, and you can read about that in more detail here.

AMD Radeon RX 480 display connectors

This is also AMD’s first GPU to be based on the new and more efficient 14nm FinFET process, so we can expect power savings as well (more on that in the results section). When it comes to power consumption, the card has a TDP of 150 watts, which means it still requires a 6-pin PCIe connector. That’s pretty low in and of itself, but we can’t help think that it’s actually not that impressive – despite AMD’s emphasis on how efficient Polaris is – because the NVIDIA GeForce GTX 1070 has the same TDP and is so much more powerful. The GeForce GTX 1080 also isn’t that much more power hungry at 180 watts, so AMD isn’t really breaking any new ground in the power consumption department at least.

AMD Radeon RX 480 power

That aside, the card forms the crux of AMD’s pitch to make VR affordable to a wider audience. The company noted that only the top 16 percent of PC users purchase graphics cards that support VR games and applications, and the overwhelming majority make do with cards in the US$100 to US$300 price range. Normally, those cards wouldn’t be powerful enough to handle the demands of VR gaming – AMD puts the proportion of VR-capable PCs at just 1 percent of the 1.43 billion PCs globally – but that’s exactly what the company wants to change with the RX 480.

In a word, AMD says the card will bring about more affordable and compact VR machines. At a brief demonstration during the PC Gaming Show at E3, CEO Lisa Su even showed off an Alienware VR backpack machine powered by a Radeon RX 480, teasing a possible new wave of small and robust VR-ready systems.

And now for the card’s performance. It’s definitely not in the same class as the GeForce GTX 1070 and 1080, but for its price, there’s really little to complain about.  

Join HWZ's Telegram channel here and catch all the latest tech news!
8.0
  • Performance 8
  • Features 7.5
  • Value 9
The Good
Delivers great bang for your buck
Performance is impressive for its price
Fairly low power consumption
The Bad
Lacks new features that are readily perceived by end users
Not much overclocking headroom
Same old reference design
Our articles may contain affiliate links. If you buy through these links, we may earn a small commission.