AMD’s Vega for gamers is finally here. The Radeon RX Vega 64 and RX Vega 56 mark AMD’s return to the high end of the graphics-card world. After a long, long stretch wherein 2015’s Radeon R9 Fury X and R9 Fury were asked to hold the fort against Nvidia’s Pascal onslaught, the company is relieving them with what promises to be competition for Nvidia’s long-dominant GeForce GTX 1080 and GeForce GTX 1070.
The Vega 10 GPU that’s riding in on those cards is a massive and massively complex piece of silicon. It packs 12.5 billion transistors into a 486 mm² die fabricated on GlobalFoundries’ 14nm LPP FinFET process. (For comparison, the Nvidia GP102 chip aboard the GTX 1080 Ti and friends is a similarly massive 12 billion transistors on a 471 mm² die.) The card’s compute resources are organized into 64 “Next-gen Compute Units.” Each of these hosts 64 stream processors for a total of 4096. The full Vega 10 chip has 256 texture units and 64 ROPs, too.
While the basic organization of Vega 10 may sound similar to Fiji before it, the similarities largely end at those broad outlines. I would love to explore Vega’s many changes and capabilities in more depth, but when you have two days and change to review two brand-new graphics cards in the wake of testing and writing for a CPU review, stuff has to be left on the cutting-room floor, and a deep dive on Vega’s new talents is one of them. We’ve known broadly what the new bits of Vega would be since January, however, so my architecture introduction is as good a place as any to start if you need to catch up. I’ll be trying to add more information to this article as time goes on, but AMD should have a white paper available soon with full architectural details if you want to know much, much more.
The Radeon RX Vega 64 and RX Vega 56
The implementations of the Vega 10 GPU that will be available to consumers were revealed at SIGGRAPH a couple weeks ago, but to recap, AMD will be selling the fully-enabled Vega 10 GPU aboard the Radeon RX Vega 64 in both air- and liquid-cooled varieties. The Radeon Vega 56, on the other hand, loses eight NCUs to the world’s tiniest chainsaw, and it’ll only be available as an air-cooled card in its reference form. Curiously, AMD left all 64 ROPs intact on both the RX Vega 64 and RX Vega 56, meaning the cuts to the 56 may hurt less than they otherwise might.
|GTX 980 Ti
|Titan X (Maxwell)
|GTX 1080 Ti
|R9 Fury X
|Radeon RX Vega 64
|Radeon RX Vega 64
|Radeon RX Vega 56
Both cards also ship with 8GB of HBM2 memory on board. That memory communicates with the Vega 10 chip across a 2048-bit bus. AMD doesn’t disclose as much, but on the RX Vega 64 cards, that HBM2 runs at an effective rate of 1890 MT/s, and on the RX Vega 56, it runs at about 1600 MT/s.
|Asus R9 290X
|Radeon R9 295 X2
|Radeon R9 Fury X
|GeForce GTX 780 Ti
|Gigabyte GTX 980
|GeForce GTX 980 Ti
|GeForce Titan X
|GeForce GTX 1070
|GeForce GTX 1080
|GeForce GTX 1080 Ti
|GeForce Titan X (Pascal)
|Radeon RX Vega 64 (air)
|Radeon RX Vega 64 (liquid)
|Radeon RX Vega 56
We didn’t have time (yet) ahead of publication to run our fancy Beyond3D test suite, so here are some potential peak rates for the RX Vega family. Compared to the GTX 1080, the RX Vega 64s trail slightly in pixel fill and peak rasterization rates, but they bring a prodigious array of shader and texturing power to the table, along with much higher memory bandwidth. The RX Vega 56 similarly trails and trounces the GTX 1070 in these theoretical peak rates. Our tests will tease out whether those theoretical victories translate into real performance shortly.
AMD’s reference design for both the air-cooled RX Vega 64 and RX Vega 56 uses the same black shroud with an axial fan exhausting air directly out of the rear of the card. While this design may bear some similarities to the Radeon RX 480 before it, they’re ony skin-deep. The shroud on this card is mostly metal and features a full metal backplate, as well. The Radeon logo on the side of the card lights up with red LEDs when the card is on.
One neat little touch from the R9 Fury X returns on board the RX Vegas: the “GPU Tach.” This line of red LEDs (or blue, if you flip a DIP switch) will show you your GPU’s occupancy, presuming you can see it from your desk or other site of sitting. The LEDs can also be turned off.
The Radeon RX Vega 56 will list for $399, or $20 more than the GTX 1070’s original $379 suggested price. The RX Vega 64 air-cooled card will sticker for $499, or the same as the GTX 1080’s most recent suggested price. Both cards are also available as limited editions and as parts of “Radeon Packs,” which we described in more detail in our RX Vega reveal. We’ve been talking about RX Vega for ages, so now it’s time to shut up and share performance numbers.
Our testing methods
Most of the numbers you’ll see on the following pages were captured with OCAT, a software utility that uses data from the Event Timers for Windows API to tell us when critical events happen in the graphics pipeline. We run each test run at least three times and take the median of those runs where applicable to arrive at a final result.
As ever, we did our best to deliver clean benchmark numbers. Our test systems were configured like so:
|Gigabyte Aorus GA-Z270X-Gaming 8
|16GB (2 DIMMs)
|G.Skill Trident Z
(run at DDR4-3200)
|Samsung 960 EVO 500GB
Kingston HyperX 480GB
2x Corsair Neutron XT 480GB
|Seasonic Prime Platinum 1000W
|Windows 10 Pro with Creators Update
|Radeon RX 580
|Radeon Software 17.7.2
|Radeon RX Vega 56
|Radeon Software beta
|Radeon RX Vega 64
|EVGA GeForce GTX 1070 SC2
|GeForce GTX 1080 Founders Edition
Thanks to Intel, Corsair, Kingston, and Gigabyte for helping to outfit our test rigs with some of the finest hardware available. AMD, Nvidia, and EVGA supplied the graphics cards for testing, as well. Behold our Gigabyte Aorus Z270X-Gaming 8 motherboard in all its glory:
Unless otherwise specified, image quality settings for the graphics cards were left at the control panel defaults. Vertical refresh sync (vsync) was disabled for all tests. We tested each graphics card at a resolution of 2560×1440 and 144 Hz, unless otherwise noted.
The tests and methods we employ are generally publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.
Doom‘s Vulkan renderer is a staple of our graphics card reviews and a favorite of AMD tech demos. Let’s see if the RX Vegas have made a deal with the devil for the performance lead at maximum settings and a resolution of 2560×1440.
Doom‘s Vulkan renderer is always a bright spot for Radeons, and the Vega cards don’t disappoint out of the gate. The RX Vega 56 outperforms even the GTX 1080, and the RX Vega 64 is 11% faster still in average frame rates. Both Radeons and the GeForce GTX 1080 deliver impeccable 99th-percentile frame times, as well. The GTX 1070’s 16.1-ms 99th-percentile result is still good, but in a fast-paced game like Doom, you’ll notice the frame-rate drops that lead to such a figure.
Our “time-spent-beyond-X” graphs can be a bit tricky to interpret, so bear with us for just a moment before you go rocketing off to the conclusion. We set a number of crucial thresholds (or bins) in our data-processing tools—50 ms, 33.3 ms, 16.7 ms, 8.3 ms, and 6.94 ms—and determine how long the graphics card spent on frames that took longer than those times to render. Any time over the limit ends up aggregated in the graphs above. Those thresholds correspond to instantaneous frame rates of 20 FPS, 30 FPS, 60 FPS, 120 FPS, and 144 FPS, and “time spent beyond X” means time spent beneath those respective frame rates. We usually talk about these results as a proportion of the one-minute test runs we use to collect our data.
If even a handful of milliseconds make it into our 50-ms bucket, we know that the system is struggling to run a game smoothly, and it’s likely that the end user will notice severe roughness in their gameplay experience. Too much time spent on frames that take more than 33.3 ms to render means that a system running with traditional vsync on will start running into equally ugly hitches and stutters. Ideally, we want to see a system spend as little time as possible past 16.7 ms rendering frames, and too much time spent past 8.3 ms or 6.94 ms is starting to become an important consideration for gamers with high-refresh-rate monitors and powerful graphics cards.
With a fast-running game like Doom, it makes the most sense to start our analysis at the 8.3-ms mark. Recall that any time spent past this point means the frame rate will drop below 120 FPS. Here, the RX Vegas prove their mettle. The RX Vega 64 spends just about three-and-a-half seconds of our one-minute test run below 120 FPS, while the RX Vega 56 spends five seconds working on tough frames that similarly dip its instantaneous frame rate. The GTX 1080 delivers a similarly respectable result, but the GTX 1070 spends a whopping one-quarter of our test run working hard.
Hitman (DirectX 12)
The RX Vega 64 doesn’t repeat its commanding performance in Doom here, but a tie with the GTX 1080 is still a fine result. The RX Vega 56 is slightly faster and slightly smoother than the GTX 1070, as well.
The Vegas and GeForces deliver practically perfect performances at 50 ms, 33.3 ms, and 16.7 ms, so we can say they all offer fine gaming experiences. Looking at the high time-spent-beyond-8.3-ms bar, though, the RX Vega and GTX 1080 are as virtually tied as it gets in these matters. In a win for the Radeon camp, the GTX 1070 spends more than two seconds longer than the RX Vega 64 does on these tough frames.
Rise of the Tomb Raider (DirectX 12)
Rise of the Tomb Raider‘s DX12 mode results in another tight matchup for our contestants. The RX Vega 64 can’t quite catch the GTX 1080, but the Vega 56 delivers a slightly more fluid experience than the GTX 1070 while turning in virtually the same 99th-percentile frame time.
Those 99th-percentile frame times don’t tell the whole story for the RX Vega 56 and the GTX 1070, though. For that, we have to turn to our time spent beyond 16.7-ms threshold. Here, the Vega 56 spends nearly a second less on tough frames compared to the GTX 1070. Both the GTX 1080 and the RX Vega 64 spend only imperceptible amounts of time on tricky frames, as well, although the GTX 1080 maintains its slight edge.
The Witcher 3
Chalk up another good showing for the Vegas here. The RX Vega 64 only slightly lags the GTX 1080 in both average frame rate and 99th-percentile frame time, while the RX Vega 56 and the GTX 1070 are as dead-even as it gets.
Even better, all of the Vega cards and the GeForces spend just a handful of milliseconds past the crucial 16.7-ms mark. Save for the RX 580, all of these cards will deliver a smooth, high-fidelity Witcher 3 experience at 2560×1440.
Deus Ex: Mankind Divided (DX11)
The RX Vegas post a fine showing in DXMD. The RX Vega 64 trades punches with the GTX 1080, while the RX Vega 56 narrowly beats out the GTX 1070 in both its average frame rate and 99th-percentile frame time.
All of our cards spend plenty of time working on frames past 8.3 ms, so we’ll stick with the standard 16.7-ms mark here. Some minor hiccups at 50 ms and 33 ms aside, the RX Vega 64 shares company with the GTX 1080 in delivering only a handful of frames that took more than 16.7 ms to render. The RX Vega 56 also provides a slightly smoother experience than the GTX 1070 here. Even so, I’m surprised Radeons continue to have those noticeable stutters in DXMD, since they’ve been present for nearly a year now. Perhaps future driver updates will help smooth out the Vegas’ performance for good here.
Gears of War 4
Gears of War 4 is one of Microsoft’s first-party DirectX 12-only efforts, and it has plenty of PC-friendly graphics settings to make even the most powerful graphics card sweat. We dialed in the game’s Ultra preset and went to work.
I was hoping Gears of War 4 would provide even footing for our graphics cards to strut their stuff. Things didn’t quite play out that way. The RX Vega 56 trails the GeForce GTX 1070, and the RX Vega 64 just matches the lesser GP104 card. Seems Nvidia’s DX12 driver update earlier this year was no joke.
None of the cards save the RX 580 spend a noticeable amount of time past the 16.7-ms mark on tough frames, so it makes more sense to observe what happens past 8.3 ms. Here, none of the cards are standouts, but the GTX 1080 does spend three seconds less working hard than the RX Vega 64 and GTX 1070 do. The RX Vega 56 spends about six seconds longer than its fully-enabled counterpart on those tough frames, an unusually large gap given the cards’ relatively similar clocks and resource provisions. We may need to investigate further.
Grand Theft Auto V
For a change of pace, I cued up our usual GTA V test run at 4K using maximum settings for everything save MSAA.
I had hoped moving to 4K with GTA V would help show the virtues of the Vega cards a bit better versus the competition, but that turned out not to be the case. The GTX 1080 is the only card of this bunch that delivers what I would deem a playable 4K experience on a traditional 4K monitor. What may be interesting to some who are flirting with 4K is that both RX Vegas deliver 99th-percentile frame times that will keep a typical 4K FreeSync monitor happy. FreeSync doesn’t help GTA V feel as fluid as it does on the GTX 1080, but the smooth motion both cards delivered with FreeSync enabled in my informal testing was a far sight better than the alternative.
Although I’ve been ruined by the GTX 1080 Ti for 4K gaming, the GTX 1080 still delivers a surprisingly playable 4K, 60-FPS experience with GTA V. It spends just over a second working on tough frames that would drop instantaneous frame rates below 60 FPS. Nothing else in this company even comes close, but FreeSync does help make the RX Vegas more appealing if you’re mulling the jump to 4K. None of the Vegas (or GeForces) spend even a millisecond past 33.3 ms, which augurs well for a fine 4K FreeSync experience.
Watch Dogs 2
For some reason, Watch Dogs 2 simply does not play well on the Vega cards. The GTX 1080 is way out in front of the pack here, and the RX Vega 64 can’t beat the GTX 1070, let alone the GTX 1080. The Radeons exhibit slightly more competitive 99th-percentile frame times here, but the gameplay experiences they deliver still feel less smooth and fluid than that of the GeForces.
Given the average frame rate ballpark above, the most relevant “time spent beyond” metric is our 16.7 ms threshold. Both Vega cards spend over twice as long working on tough frames here as do the GeForces, and the RX 580 is thoroughly outclassed.
Six TDPs are better than one
A long-running pursuit among Radeon fans has been to find just how far they can undervolt their cards with minimal loss of performance. Whether deservedly or not, AMD cards have developed a reputation of being pushed to the ragged edge of the voltage-and-frequency-scaling curve in pursuit of the highest possible performance. We observed as much with the move from the RX 480 to the RX 580, whose slightly higher performance than its forebear required 48W more power for the privilege.
Perhaps because of that reputation, AMD is giving builders the option to easily explore more efficient points on that curve. The company is specifying six separate TDPs for each RX Vega card. Three options for this configurable TDP will be available in Radeon Wattman at any given time: “Power Saver,” “Balanced,” and “Turbo.” “Balanced” is the default mode, and the one we used in our main gaming tests. A two-position BIOS switch on the card will drop the base TDPs of each of these settings even further. (Yes, there is some overlap.) While the most devoted undervolters will still likely want to resort to manual tweaking, AMD is making the practice both officially sanctioned and more accessible to folks who just want to toy around with increasing performance-per-watt without a major time investment.
I decided to test both the power consumption and noise levels of these cards while also exploring the performance implications of the three TDPs available from Wattman in the BIOS switch’s default position. To do so, I ran through our Hitman test run and logged performance at each setting.
In Hitman, the RX Vega 64’s “turbo” mode offers no additional performance. “Power save,” however, has a negligible impact on both average frame rates and 99th-percentile frame times. The RX Vega 56 gets a bit of additional headroom from “turbo,” and its performance slightly decreases in “power saver” mode, as we might expect. Of course, we’re most interested in the changes in power consumption each mode offers. To measure this result, I used my trusty Watts Up power meter while looking at a static but complex scene in our Hitman test run.
So that’s something. The RX Vega 64 maintains most of its performance while shedding a whole 86W from our test system’s power draw. Just goes to show how far AMD is pushing the Vega 64 to extract the last 1% of its performance potential, we suppose. Let’s see how these numbers stack up against the rest of our test cards.
Across the board, the numbers aren’t good for the Radeons. The RX 580 is drawing as much power under our Hitman load as the GeForce GTX 1080, and the numbers only go up from there. The RX Vega 56’s default “Balanced” mode makes our test system ask 91W more from the wall than the GTX 1070, and the RX Vega 64’s default “Balanced” mode makes our system yank an astonishing 136W more than the GTX 1080 does. The Vega architecture seems to offer a major increase in performance per watt compared to current Polaris chips, but in absolute terms, AMD’s latest is still far behind Pascal GPUs in efficiency. At least in the case of the Vega 64, “Power saver” mode might be a prudent thing to try for folks willing to give up a tiny sliver of performance for a rather big slice of system power draw saved.
Before we issue a verdict on the RX Vega 56 and RX Vega 64, it’s time once again to sum up our results in our famous value scatter plots. To produce these charts, we take the geometric mean of the average FPS figures for each game we tested. We also convert the geometric mean of the 99th-percentile frame times we collected for each card into a 99th-percentile FPS figure so that our higher-is-better system works. Where practical, we used retail or e-tail pricing for each card tested.
If you’ve been following our Vega coverage, you’ve already known how these graphs would shake out for some time now. The only question was how close the race would be. The answer: pretty close. In our testing so far, the RX Vega 64 trails the GTX 1080 by about 11.5% in our all-important 99th-percentile frame time metric, but its performance potential in our tests (as measured by average FPS) only lags the GeForce by about 6%.
Those gaps are a bit of a letdown after the long wait for Vega, but they don’t seem insurmountable. AMD has proven it can lower 99th-percentile frame times with driver updates in the past. Vega marks the biggest change to AMD’s GCN tech in years, and the company may need some more time to fully tune its software for the best performance from the product. Vega is also is launching into a gaming market that has been dominated by the GTX 1070 and GTX 1080 for over a year. Game developers may find some new tweaks and optimizations are necessary to get the best performance from the Vega GPU, as well.
Another reason for the performance deficit between the RX Vega 64 and the GTX 1080 in our initial standings is, I think, a lot simpler. The RX Vega 64 reference card seems to be running on the ragged edge of its voltage-and-frequency-scaling curve, and its factory fan profile doesn’t allow the GPU to run at its peak clock speeds for extended periods, if at all. Our card was plenty happy to overclock with its blower cranked and a bunch of fans blowing on it, but the eyebrow-raising power draw and physically painful noise levels that ensued showed why AMD isn’t pushing Vega over the shoulder of the voltage-and-frequency-scaling curve and into its ear.
Even at stock speeds, power consumption is the bane of the Vega GPU. Our system power draw with the RX Vega 64 installed exceeded that of even the GeForce GTX 1080 Ti, at 408W for the RX Vega 64 and about 350W for the GTX 1080 Ti, and it peaked at over 500W for the Vega chip once we informally explored overclocking. For a more apples-to-apples comparison, installing the GTX 1080 Founders Edition caused our system to draw about 272W. Those extra watts mean more expensive power supplies, more robust cooling fans, potentially higher noise levels, and a need for better climate control in one’s gaming den. The Power Saver Wattman profile goes a long way toward taming the RX Vega 64 for next to no performance cost, but there is no denying that performance-per-watt remains a challenge for AMD’s architects.
I’m curious, then, how the RX Vega 64 performs in its liquid-cooled guise. AMD itself showed that liquid cooling can be good for power-dense and power-hungry graphics cards with the R9 Fury X. Keeping hot-tempered GPUs way cool can help reduce wasted energy and could potentially open up more performance headroom. This time around, though, a liquid-cooled Vega 64 is a $200 upcharge over the base RX Vega 64 as part of a Radeon Pack. $699 happens to be the same price as the sticker for Nvidia’s GTX 1080 Ti Founders Edition, and given the huge delta in performance between those two products, I can’t imagine any but the most ardent AMD fans will pick the liquid-cooled Vega. Even with the enticement of packed-in games, accessible FreeSync monitors and heavy discounts on other hardware, leaving that much performance on the table is rough going. Perhaps we’ll see affordable aftermarket liquid-cooling solutions that can help bridge the gap.
The RX Vega 56 is a happier story for AMD. As was the case with the R9 Fury versus the R9 Fury X, losing eight of the full Vega 10 chip’s compute units to the world’s tiniest chainsaw just doesn’t hurt the Vega 56 that much. Our indices of 99th-percentile frame times and average frames per second put the Vega 56 dead-on with a hot-clocked GTX 1070, and only 10% behind the Vega 64 in our FPS index. The 56 does draw much more power than a GTX 1070 in our test system, but not in the eye-popping way of the Vega 64.
The RX Vega 56 looks especially nice in light of the FreeSync monitor selection these days. One can get a 144-Hz IPS gaming display with a 30-Hz-to-144-Hz FreeSync range for just $550 now, compared to about $800 for a comparable G-Sync display. That $300 could go a long way toward more powerful components elsewhere in a system, or it could stay in one’s pocket. It’s not a stretch any longer to say that a variable-refresh-rate display is the way to game, and the RX Vega duo completes a puzzle for FreeSync that’s been unfinished for a long time. I could go on and on about how revelatory VRR tech still is for the gaming experience, and I’m happy to see it become potentially more accessible again for high-end gaming. Assuming you can tolerate the higher heat and noise levels of the Vega 56 compared to the Pascal competition, its FreeSync support makes it a strong contender for the entry-level high-end graphics card to get.