AMD is in the spotlight right now, mainly thanks to its range of Ryzen processors. They may not be thrashing Intel’s best at outright performance, but they certainly are on price. Read our Ryzen 7 review for more. The Radeon RX Vega is the name of AMD’s next-generation graphics cards, and we have high hopes they will take the fight to Nvidia, whose consumer range tops out with the awesome GeForce GTX 1080 Ti.
If you are viewing a mobile version of this page, click here to see the latest version of this story.
What is AMD Vega?
Vega is the name of the new Radeon graphics cards. The full name is AMD Radeon Vega and the cards are set to compete with Nvidia’s best.
In a similar way to what it did with Ryzen, AMD has made huge changes to the processor’s architecture and has already shown a few demos and figures which hint at big performance increases. See also: Best graphics cards
The Vega brand will apply to both consumer gaming cards as well as workstation cards.
AMD has now unveiled the first of the latter type, the Radeon Vega Frontier Edition.
When will AMD Vega be released?
AMD said it “expects to ship” RX Vega cards in the second quarter of the year, and the rumours were correct that this would end up being late June. However, there's still no sign of the consumer cards: only one version of the workstation card has gone on sale.
That's the air-cooled model, which was released on 28 June. The liquid-cooled version will go on sale in Q3 - there's no exact date yet.
How much will AMD Vega cost in the UK?
The air-cooled Frontier Edition (below) costs US$999 from Newegg (around £999), and it's $1500 (£1500) for the liquid-cooled version. These, as we've said, aren't gaming cards. Really, they're equivalent to Nvidia's Titan Xp.
While you could buy a Titan Xp or Frontier Edition and put it in your home PC, it's overkill and you're better off buying a cheaper consumer card, and we don't know just yet any details about these Vega cards.
The current flagship Radeon RX 580 is surprisingly cheap, but it’s also way behind Nvidia’s best for performance.
Vega should address that, and we hope AMD can keep the prices down. A GTX 1080 Ti will set you back the best part of £700 ($700) which makes it the most expensive component in most PCs.
What are AMD Vega’s specifications?
We don’t know the exact specifications of the consumer cards which will launch yet, but AMD has revealed enough to whet gamers’ appetites.
The Vega Frontier Edition has the following specs:
- 14nm GFX9 GPU
- 64 NCUs
- 4096 stream processors
- 16GB HBM2
- 2048-bit memory bus
- 480GB/s bandwidth
- PCIe Gen 3 x16
- 225W TDP
Here’s a summary of new features to expect from Vega:
- Next-generation Compute Unit (NCU)
- New geometry engine is twice as fast
- High-bandwidth cache controller (HBCC)
- High-bandwidth Memory 2 (HBM2)
- Primitive Shader
- Draw Stream Binning Rasterizer
- More power efficient
Let’s briefly explain those.
First, the NCU is – in essence – what Nvidia did with its new 10-series graphics cards last year. This is far more scalable than AMD’s previous GPUs and what it means is that, if there’s not much work to do, only a portion of the chip needs to be used. This saves power and generates less heat.
When there’s lots of work required, such as playing a game at high quality settings and high resolution, the full power of the chip can be used.
The HBCC is part and parcel of this efficiency. The former will use the available memory more efficiently, and developers will be able to make full use of all memory to make games run faster. In one demo, HBCC was turned off and then on: results showed that average frame rates were increased by 50 percent, and minimum framerates by 100 percent.
The new programmable geometry pipeline is said to be 100 percent quicker, which means it can program twice as many polygons per clock cycle as the previous generation of Radeon cards. This is obviously another key improvement that should directly impact frame rates.
Two clever features called Primitive Shader and Draw Stream Binning Rasterizer save on unnecessary calculations by figuring out which pixels won’t be seen in a frame, and ensure those aren’t given processing time.
Spied on an engineering sample of Vega were an eight-pin PCIe power connector and a six-pin. This backs up the lower power draw and better efficiency of Vega compared to the two eight-pin connectors required by the Radeon R9 Fury X.
HBM2 is arguably one of the most interesting new features. Most graphics cards, including Nvidia’s, use GDDR5 memory. This has the drawback of a relatively narrow bus, which is offset by high memory frequencies.
The bus width and memory clock speeds have been steadily increasing over the years, but the improvements aren’t really enough to avoid memory being a bottleneck.
AMD came up with HBM in 2015, but while this had a big bus width, modules were limited to 1GB, and just 4GB maximum. Now, HBM2 offers not only double the bandwidth but also modules up to 8GB in size.
Plus, rather than being limited to the memory on the graphics card, Vega will allow the PC’s main RAM and even fast SSDs to be used for storing data.
These features are great, but game developers need to optimise their games to work with the new technology. And this is why AMD has already done deals with Bethesda and other companies to make games which use AMD’s Vulkan API. Vulkan is an alternative to DirectX 12, but it’s early days for both APIs.
As with Ryzen, there’s plenty to get excited about when it comes to Vega. But until AMD officially announces models and prices, we won’t know how they will stack up against Nvidia’s offerings.
But we’ll be updating this article regularly as more information – including rumours – arrives.