Putting It All Together, Big Graphs
We kind of breezed through the 32 games tested and didn't go into the margins as much as we often do throughout the benchmark pages. With so many games we didn't want to bore some readers with a mass of percentages. Instead, we're about to hit you with all those percentages now as we look at performance on a per game basis in a single graph. Be warned: we have four comparisons here with 128 individual data points so grab your reading glasses and let's get into it.
Previously we found Vega 56 to be 2% faster than the MSI GTX 1070 Gaming X on average across 25 games. We've since downgraded slightly to the GTX 1070 Founder Edition which is a few percent slower and we've also removed Crysis 3 and replaced it with eight other games. So this has changed the margin slightly and now Vega 56 is 5% faster, though only about 3% faster than the MSI model.
Anyway as you can see the big wins for AMD here is again Dirt 4 and Dawn of War III. AMD does well in many of the newly released titles, particularly those that use modern APIs so that's well worth noting. Still troubling performance in massively and we mean massively popular titles such as GTA V and Battlegrounds is an issue and AMD need to work on this. We can probably turn the other way for games such as Assassin's Creed and Dishonored though. Anyway, overall a strong showing for Vega 56 and once pricing settles down and we get custom board partner models it should be a worthy alternative to the GTX 1070.
Moving on we have a much less exciting comparison to make with the air-cooled Vega 64 model -- what a stinker. Overall it was 5% slower than the GTX 1080 Founder Edition and of course it runs much hotter, louder and burns more power than a Skylake-X processor on steroids. Not much else needs to be said here. The reference card just isn't cutting it. We do feel like the board partner versions will be significantly better, especially for this model, but that's a story for another day.
Liquid cooling Vega 64 and allowing it to have those higher clock speeds while avoiding throttling really helps this complex GPU. The liquid cooled version was 4% faster than the GTX 1080 Founders Edition overall but in 10 of the games tested it was faster by a 10% margin or greater. Of course the liquid cooled edition is priced $100 above the GTX 1080 when looking at the standalone MSRP, so it's hardly "hot" in terms of price vs. performance.
Compared to the Aorus Xtreme GTX 1080 Ti, the liquid cooled Vega 64 was 18% slower and once pricing corrects it should only be 14% cheaper, so another sketchy result for Vega.
At this point we've established a pretty good baseline of Vega's performance. Most other reviewers have concluded by saying that Vega 56 is slightly faster than the GTX 1070 and Vega 64 is slightly slower than the GTX 1080 while the liquid cooled version wins more than it loses but struggles massively against the GTX 1080 Ti.
That's pretty much the situation from a frame rate standpoint and we didn't need a 32 game benchmark for that conclusion, but it was interesting to see where Vega did well, and not so well.
It was also interesting to see where the air-cooled Vega 64 offered gains over Vega 56, to see which games have very little separation between the two GPUs, and to see where the Vega 64 dominated the GTX 1080 -- there is some hope for Vega yet.
Setting power consumption, pricing and even performance aside for a moment, I do believe one of Vega's biggest issues is AMD's own reference card design. This is yet to be proven, but if history has shown us anything it's that AMD's reference cards suck and they rob their own products of performance.
The liquid cooled model might very well be a best case scenario for Vega and that's kind of a good thing. If board partners can adequately cool Vega 64 using air coolers then we might see a significant improvement in performance for that model. That would also leave room for Vega 56 to improve and I can easily see that model beating the GTX 1070 by a comfortable 10% margin while remaining cool and quiet.
- Core i7-7700K - Amazon, Newegg
- GeForce GTX 1080 - Amazon, Newegg
- GeForce GTX 1070 - Amazon, Newegg
- GeForce GTX 1060 - Amazon, Newegg
- Radeon RX Vega 64 Liquid Cooled - Amazon, Newegg
- Radeon RX Vega 64 - Amazon, Newegg
- Radeon RX Vega 56 - Amazon
On the subject of pricing, there is currently a supply issue with Vega that is reminescent of what we saw with the RX 580, RX 480 and even the GTX 1060, but this was always likely to happen and even more so with cryptocurrency mining as big as it is currently. I'm confident that Vega 56 and 64 models should start hitting the standalone MSRP once we see custom board partner cards arrive later next month. Of course, if there's still a serious supply issue due to miners then that might not be the case.
For now we'd recommend to avoid the reference cards, sit tight and wait for the much better custom versions and give supply a chance to catch up with demand before paying an inflated price. Overall, I'm still not 100% sure how I feel about Vega yet. There are some promising signs that it can be competitive, but I would like to reserve ultimate judgment until we have a few board partner models in hand.
Pros: Vega 64 is slightly slower than the GTX 1080, while the liquid cooled version wins more than it loses against the GTX 1080.
Cons: Poor availability and thus pricing. AMD's air-cooled RX Vega 64 reference card is ~5% slower than the GTX 1080 FE despite being hotter, louder and using more power (though we have high hopes for custom cards).