AMD Vega has been somewhat of a mixed product. The incredibly hyped, yet often delayed graphics architecture from AMD was finally released, but it’s release wasn’t without its issues last year. From stock issues thanks to low supply and demand from ethereum miners to a confusing way of packaging the product with monitors and game discounts, AMD Vega had a lot to live up to. So, was the hype worth the wait?
- MSRP: $499
- GPU: Vega 10
- GPU Cores: 4096
- Base Clock: 1247 MHz
- Boost Clock: 1546 MHz
- Texture Units: 256
- ROP Unites: 64
- Memory: 8GB HMB2
- Memory Clock: 1890 MHz
- Memory Interface: 2048-bit HMB2
- Memory Bandwidth: 484 GB/s
- TDP: 295 Watts
- Peak Compute: 12.6 TFLOPs
- Transistor Count: 12.5B
- Process tech: 14nm
The entire Vega architecture has been hyped since it was first unveiled at CES 2017. From things such as the high bandwidth cache controller to the new rendering pipeline, Vega had a lot to live up to. But at the end of the day, many people are looking at the Vega GPU line as an upgrade from Polaris, the AMD architecture which came before Vega and is currently being used in the current generation of consoles. So how much of an impact does upgrading from a Polaris card, such as a Rx 580 to the Vega 64 make, and is it as much of a jump as its next nearest-priced competitor, the Nvidia GTX 1080?
First things first - a little bit of caution: the Vega 64 is a power hungry card. You really need a beefy power supply if you are thinking of upgrading from an older rig that didn’t require much power. I have an all-AMD rig, which is pretty representative of an average AMD rig of the last few years running the Rx 580. The PSU in it is - well, was - only a year old, and ran everything just fine. However, I realized rather quickling into benchmarking that something was amiss as synthetics were causing the PC to completely reset. One such experience was using the Final Fantasy XIV Stormblood Benchmarking tool and getting it to go through at 1080p just fine, however it would continually crash the minute I ran it at 1440p. However, to make matters even murkier, the benchmark ran just fine at 4K on this rig. However, when I shut down the PC for the evening, it has refused to turn back on. So, if you’re looking to upgrade to the Vega, make sure your have a higher wattage PSU than 500W (AMD actually recommends a 750W unit). I felt it was good to test it this way, however, because most people when upgrading GPUs don’t take into account the power supply. So - a word of caution to those who are looking to upgrade: check your PSU! That being all said - here are the specs for the PC which all the tests were finally performed on for this review: i7-6700K @ 4.0 GHz, Corsair H60 cooler, Gigabyte GA-z170-HD3 motherboard, 16 GB GDDR4 RAM @ 3200MHz, 256GB Samsung SSD, 2TB Seagate HDD @ 7200rpm.
I tested some of the more popular MMOs on the market today, as well as some multiplayer games that really push GPU. And the Vega came away with some pretty respectable numbers.
The stable 100 frames per second in World of Warcraft: Legion is nice - and if you’re experiencing framerate drops in 4K, I would check to ensure you have anti-aliasing disabled. At 4K, it’s pretty much overkill to have a lot of AA enabled and it is a drain on performance. For reference, when using FXAA High at 4K, that average levels out to about 80 frames per second.
Ghost Recon: Wildlands is another title that surprised me. It’s an Nvidia title with a metric ton of graphics being thrown at the card in the benchmark, yet at full ultra the game ran respectively well. Well, at 1080p. However, tweak a few settings in 4K and you could get console level performance with absolutely breathtaking PC visuals.
The real shocker was PUBG. Well, not really a shocker, the game has suffered from optimization issues - especially with AMD - since its early access launch. However, with the 1.0 and most current drivers, I expected a little better at 1440p, and at 4K.
But what if you’re going from a Polaris card to the Vega? How much of an increase will you see? Well, it depends on the title and resolution.
There is a massive gulf between the Rx 580 and the Rx Vega 64 on this graph. The most interesting result, however was the Overwatch test at 1080p on the 580 versus the Vega. The 580 outstrips the vega by 22 frames per second - 160 to 138fps respectively. I’m not sure if the 580 was just having a good day, or if it knew it would be ripped from the system for the Vega right afterwards - however, it’s surprising to see. Otherwise, aside from Black Desert Online and World of Warcraft, there is a sizeable gulf between the Vega and the 580.
That gulf is even more pronounced with the jump to 1440p - especially in ames like Ghost Recon and Battlefield 1. However, things even out with PUBG and Final Fantasy XIV, with the Vega only holding the slight edge.
The real shocker to me is the Elder Scrolls Online performance on the 580. This is a game that would not keep a stable framerate no matter what I did on the Rx 580. Vega 64, though, does hadle the game well at 1440p, but not as well as you’d hope for a $500 card - especially seeing the way the GTX 1080 outpaces the Vega in that game - more on that later. Also, the Rx 580 takes its spot as last in the pack in Overwatch, with the Vega outsriping by only about 5 frames per second.
This card really seems to settle into its own at 1440p, though. Other games I tested while working with the Vega were The Witcher 3: Wild Hunt, Destiny 2 and more and each time I found the experience at 1440p absolutely incredible. The Witcher did require a few tweaks to get an above 60 framerate (notably to foliage detail - thanks transparencies!) but otherwise I felt that 1440p really shines with the Vega 64. However, this card has also been marketed as being to do 4K, so we’d be remiss if we didn’t test that.
Simply put: the Vega 64 Reference card is not a good 4K card, at least in my experience. Other than World of Warcraft, which surprisingly kept at the framerate limit for all of the tests, Battlefield 1 and Overwatch, this is a card that was brought to its knees at 4K. Now, in terms of the 4K landscape with consoles now pushing 4K Graphics* these numbers are respectable. Indeed, as mentioned previously, it you want visuals and don’t mind the 30 frames per second lock, Ghost Recon: Wildlands is definitely playable with the Vega. In fact, Vega - and the GTX 1080 - give you the ideal Xbox One X PUBG experience at 30 frames per second.
If you’re a 580 user, the Vega does offer more opportunities for higher framerates at 4K as well - as these are all with the settings set to max. You can definitely tweak settings to get more respectable framerates with the Vega as well - I was able to hit 45-50 frames per second average in PUBG just by turning a few settings down to high/medium.
However, this brings up a important point: this was a card and architecture that was hyped for years - literal years - as being AMD’s next big thing. Indeed, when it was shown running games at conventions (I played DOOM running on a Ryzen/Vega Rig at CES last year in 4K running on Vulkan) it was always running at 4K. This gave gamers the expectation that Vega would be a full-tilt 4K gaming card - and in my experience it just isn’t. Sure, you can game at 4K, but it would be compromised. And the rig used for testing isn’t a slouch either. This card is at its best when running at 1440p, which for many users is the target right now.
The other major issue with Vega is the performance compared to the next direct competitor: the GTX 1080. Simliarly priced, the 1080 is now over 18 months old, yet still outstipes the Rx Vega at almost every turn. Games such as Battlefield 1 have an edge thanks to AMD optimizations, and the same can be said for the games preferring the 1080 thanks to Gameworks implementation. However, for a card to be hyped and touted for so long to come up short consistently against your aging competitor - a competitor that is similarly priced, yet uses less power and runs games better - it doesn’t say much for the Vega as a whole. It also leaves questions about how AMD can compete with Volta, especially with Navi being in the distance and the company doesn’t even have an answer yet to the GTX 1080 ti.
Isolated from the competition, the Rx Vega seems like a great deal - it’s a sizeable upgrade from the previous AMD tech and it does provide great gaming experiences at higher resolutions than standard HD. However, once you add the Pascal architecture into the equation, the Vega does leave some to be desired.
Once more note - the Radeon card does run a little hot in some games - though that is the same experience I’ve had with the GTX 1080 mini I reviewed a few months back. This doesn’t leave a lot of headroom to overclock the reference card, and as a result I decided against doing so for this review.
The hottest the card got was while benchmarking Black Desert Online - in my opinion the most graphically demanding traditional MMO on this list. However, when the coldest temp you see is still above 80c, I start to worry about overclocking headroom. Setting a custom fan curve didn’t really help alleviate these temps either. Now with the card hitting a sweet spot at 1440p, you might see some room to overclock there, but at the end of the day, if you grab the reference card, I would simply make due with the stock speeds to be safe.
Radeon Vega 64 is a card that I personally have looked forward to for years - I held off grabbing the Pascal card upgrade until I knew for certain what Vega would be. However, I’m left feeling a bit underwhelmed in the end. Yes, 1440p is a great target for the Vega - but the architecture was hyped for so much more. Consumers, such as myself, expected more. At least the expectation was that the card would hold serve with the GTX 1080 - and even there the Vega 64 falls flat.
However, for many fans of AMD, this card is a sizeable upgrade - especially going from the Polaris architecture. The jump from the Rx 580 to the Rx Vega 64 is certainly enough to warrant looking at an upgrade. However, the power consumption, temperatures and middling framerates at 4K do leave a lot to be desired. But, if you’re in the market for an AMD card, the Vega isn’t a bad choice overall - and games running Vulkan will definitely see a tremendous benefit. In the end, though, the Rx Vega 64 still doesn’t do enough to overcome the shortfalls it sees when compared with its most direct competitor - the 18-month old GTX 1080.
- Discernible upgrade from previous AMD architecture
- Great 1440p card
- Runs relatively quiet under load
- Doesn’t do enough to push it over the Pascal equivalent
- Extremely power hungry
- Runs hot making Overclocking worrisome
The product discussed in this article was provided by the manufacturer for the purposes of review.