The Radeon VII was AMD's surprise unveiling at CES 2019.
Its new high-end graphics card for gamers, this is the world's first graphics processor (GPU) to use a 7nm manufacturing process.
(See our Radeon VII preview from CES.) It certainly won't be the last, but until AMD releases its next-generation cards later in 2019, currently known under the code-name "Navi," the $699 Radeon VII is its flagship card, and its chief rival to Nvidia's GeForce RTX 2080.
AMD's last entrant into the high-end card stakes was the mid-2017 Radeon RX Vega 64.
The Radeon VII is a clear improvement on that card for high-end 4K play, and solid in that niche today, but several emerging technologies and its pricing put it in a slippery position versus the RTX competition.
We ultimately find the Radeon VII to be a fair alternative to the GeForce RTX 2080, although the GeForce RTX 2080 is a better overall performer if you're gaming at 1080p and 1440p resolutions.
Similar Products
2019: The State of High-End Graphics
Although competitive with the GeForce GTX 1080, the Radeon RX Vega 64 was more than a year late to the party, and it had much higher power demands.
There are parallels between the Radeon VII and the Radeon RX Vega 64.
For one, the Radeon VII is also late to the game, if less so; the competing GeForce RTX 2080 hit the streets in September 2018.
In addition, the Radeon VII draws significantly more power than the GeForce RTX 2080.
View All 11 Photos in Gallery
The difference this time is that the Radeon VII has a few practical real-world advantages, which was hard to say about the Radeon RX Vega 64.
The Radeon VII has a massive 16GB video-memory buffer, or twice as much as the 8GB that the GeForce RTX 2080 carries; even the elite GeForce RTX 2080 Ti$1,199.00 at Nvidia has only 11GB.
The Radeon VII also more than doubles the memory bandwidth of the GeForce RTX 2080, a metric that could pay dividends in ultra-high-resolution scenarios.
AMD's pricing of the Radeon VII is the same as Nvidia's recommended starting price for the GeForce RTX 2080 ($699), meaning it's do-or-die; the Radeon VII has to measure up in almost every respect to top it.
We won't be spoiling anything by saying both cards perform closely in today's AAA titles at a 4K resolution.
Therefore, the "value" of either card (a term we use lightly, as this amount of money can buy you an entire gaming PC) goes beyond sheer performance.
The 16GB of memory on the Radeon VII is designed for future applications, such as tomorrow's 4K games and 8K content creation, while the GeForce RTX 2080 has hardware-accelerated ray tracing and AI-driven scaling.
(For a further rundown of the differences, see our companion piece, AMD Radeon VII vs.
Nvidia RTX 2080: Which High-End Gaming Card to Buy?)
The Radeon VII Deep Dive
The roman numerals "VII" in the Radeon VII's name simultaneously allude to its second-generation Vega architecture (Vega 20), and its 7nm fabrication process.
AMD gave few specifics regarding the architectural improvements, citing increased bandwidth for render output units (ROPs), overall reductions in latency, and larger accumulators for floating point and integer operations.
In other words, "Vega II" is an evolution of the original Vega architecture.
The 7nm fabrication process is more significant for the Radeon VII than its architectural improvements.
The smaller process allows the GPU die to be smaller; it's just 331mm2, whereas the Radeon RX Vega 64 came in at 495mm2.
The GeForce RTX 2080 has a seemingly huge 545mm2 die, although much of that is allocated to its proprietary Tensor and ray tracing (RT) cores, of which the Radeon VII has none.
(I'll talk more about future gaming considerations later in this article.)
Double the Video Memory
The smaller GPU die explains how AMD was able to cram 16GB of high-bandwidth memory (HBM2) onto the Radeon VII, or double the amount from the Radeon RX Vega 64.
It's normal for a performance gaming machine to have 16GB of main system memory, but video memory? Professional GPUs have included even more, but as far as gaming-focused GPUs go, the Radeon VII has the most we've seen to date.
AMD bills the Radeon VII as a gaming card that can also be used for content creation.
Its 16GB of video memory, AMD says, is ideal for ultra-high 8K resolutions.
There's some objective truth to that: Video-memory demands scale with resolution.
While 8K may not be popular just yet, it's coming down the pipeline just as 4K once did.
It was all but impossible to walk across the show floor at CES 2019 and not hear or see about it.
(Look for a bit of content-creation benchmarking after the gaming-tests section in this review.)
The Radeon VII's 16GB of video memory is also an asset for gaming, as our benchmarks will show.
I've logged recent AAA titles, like Rise of the Tomb Raider, using about 8GB of video memory while running at a 4K resolution.
It would appear that 8GB of video memory is sufficient, but having more than that certainly won't hurt your prospects for future gaming.
The rearview mirror tells all: Just five short years ago, the 3GB of video memory in the GeForce GTX 780 Ti was considered extravagant surplus.
Today, that amount barely cuts it for 1080p gaming.
Radeon VII Versus Radeon RX Vega 64
Speaking of the past, here's how the Radeon VII stacks up to its predecessor, the Radeon RX Vega 64...
The Radeon VII makes some concessions in the number of compute units and stream processors, but it has the same number of ROPs as the Radeon RX Vega 64.
The shrink to a 7nm process allows the Radeon VII to fit more transistors in a smaller area, which is exactly what we want to see in the computer world.
What we don't want to see in this world, though, are higher prices.
AMD took a page out of Nvidia's book by pricing the flagship Radeon VII much higher than its predecessor.
At least AMD won't fall victim to competing with itself, as the GeForce RTX 2080 has been on the market for some time.
The Radeon VII will likely get most of its performance benefits from its higher GPU clocks.
Its 1,750MHz boost clock is about 13 percent higher than the 1,546MHz boost clock of the Radeon RX Vega 64.
The claimed architecture tweaks in the Radeon VII should amplify any performance gain, and its higher memory bandwidth certainly won't hurt.
The power rating between these two is hardly changed; there's no getting around that 300 watts is high for a modern graphics card.
Even the GeForce RTX 2080 Ti Founders Edition, Nvidia's $1,199 flagship GPU, has a lower 260-watt rating.
Radeon VII Versus GeForce RTX 2080
For months after its launch, the GeForce RTX 2080 was all but impossible to find for its recommended $699 starting price for reference boards.
Given that that's no longer the case, I'm using the specifications for the reference card in the table below to give a fair comparison to the same-priced Radeon VII.
(Nvidia's GeForce RTX 2080 Founders Edition comes with a mild factory overclock for $799, but it otherwise shares the specifications below.)
The architectural differences between these two cards are probably worthy of a doctoral thesis.
For one, AMD's stream processors aren't apples-to-apples comparable with Nvidia CUDA cores, so there's no practical takeaway from that metric.
However, the raw computational power between the Radeon VII and the GeForce RTX 2080 is theoretically similar, given the closeness in their transistor counts.
Lofty power requirements are a weak point for the Radeon VII.
It requires more power connectors than the GeForce RTX 2080, and its board power rating is more than one-third higher.
AMD didn't specify a minimum power-supply rating for the Radeon VII, but based on our experience with the Radeon RX Vega 64 (a 295-watt card), opting for a 750-watt model or higher would be a good idea.
The card actually shut down on us using a 500-watt EVGA power supply in a few demanding tests.
Look to our benchmarks in this article to see how the Radeon VII's massive 1TB (1,024GB) per second memory bandwidth, over double the 448GB per second bandwidth of the GeForce RTX 2080, plays out in gaming.
(Unnecessary spoiler: It doesn't hurt.)
Radeon VII Walk-Around
AMD's reference version of the Radeon VII, under review here, is a true two-slot card.
You still need plenty of clearance to fit in your case, though, as it's 11 inches long...
This top view of the card shows its triple-fan cooling solution, and an aluminum heatsink that spans its entire length.
The silver shroud is also aluminum.
This is one heavy graphics card.
A red-backlit Radeon logo dominates the top edge...
Most of the cooling exhaust is channeled through the heatsink fins around the logo, which means directly into your case.
This isn't a blower-style cooling solution.
The two eight-pin power connectors on the edge show that the Radeon VII means business, especially for your local power authority...
The 300-watt board power rating of this card makes it one of the power-hungriest we've tested in recent memory.
Meanwhile, the aft corner of the card has a cube-like "R" that illuminates in red, a nifty detail that was also on the Radeon RX Vega 64 reference design...
The underside of the card is covered with an aluminum backplate that adds both strength and passive cooling.
A backplate is a common inclusion on mid-level and high-end graphics cards.
The four-prong heatsink mounting bracket for the GPU die is clearly visible on the underside...
Moving to the back, note how the heat shroud extends above the backplate by about half an inch.
You'll need to take this into consideration if you have a cramped case.
The perforations back here are almost purely aesthetic, as the fans send most of their air into the case.
I'll discuss the thermal performance of the Radeon VII after the benchmarks section.
As is the trend with modern gaming graphics cards, the Radeon VII has no legacy ports.
The backplate houses a trio of DisplayPort 1.2 video-out connectors, and a single HDMI 2.0b video-out port.
The lack of a VirtualLink USB Type-C port is surprising.
If you're planning to pick up a next-generation virtual-reality (VR) headset, that could be a point in favor of the green team; the GeForce RTX 2080 has one, as do the Founders Editions of even the lower-tier GeForce RTX 2070 and RTX 2060.
That said, no new VR headsets have been announced so far with such a port.
Testing: It's a Fire-Breather
PC Labs ran through a series of DirectX 11- and 12-based synthetic and real-world benchmarks on the AMD Radeon VII.
Our test rig is equipped with an Intel Core i7-8700K processor, 16GB of G.Skill DDR4 memory, a solid-state boot drive, and an Aorus Z370 Gaming 7 motherboard.
I'll focus mostly on the Radeon VII's performance as it relates to the GeForce RTX 2080, but I have also included the loftily-priced GeForce RTX 2080 Ti Founders Edition in the results, and some lesser-priced cards, such as the GeForce RTX 2060 (represented in the charts by the Zotac GeForce RTX 2060 Amp), and the Radeon RX 590 (represented by the XFX Radeon RX 590 Fatboy).
Feel free to skip to the conclusion if you just want the lay of the land.
Otherwise, the following benchmark-by-benchmark commentary is worth the extra few minutes to read to understand, among other things, the effect of 16GB of video memory in the Radeon VII.
Synthetic Benchmarks
3DMark Fire Strike Ultra
Synthetic benchmarks can be good predictors of real-world gaming performance.
Futuremark's circa-2013 Fire Strike Ultra is still a go-to for 4K-based gaming.
We're looking only at the graphics subscore, not the overall score, here.
The Radeon VII gets a good start by taking the GeForce RTX 2080 Founders Edition by 6 percent in the graphics sub-score.
It's not massively faster, but there's no question the result is outside the margin of error.
The Radeon VII is 17 percent behind the GeForce RTX 2080 Ti Founders Edition, which isn't too shabby, considering it's 42 percent less expensive ($699 versus $1,199).
Next to the Radeon RX Vega 64, the Radeon VII is ahead by an impressive 27 percent.
3DMark Time Spy and Time Spy Extreme
This is Futuremark's DirectX 12-enabled benchmark for predicting the performance of DirectX 12-enabled games.
It uses major features of the API, including asynchronous compute, explicit multi-adapter, and multi-threading.
The promising start didn't last.
The Radeon VII takes a bath here, losing to the GeForce RTX 2080 Founders Edition by 20 percent in the standard test, and 14 percent at the Extreme preset.
This DirectX 12 benchmark is newer than the DirectX 11-based 3DMark Fire Strike, indicating that Nvidia's GeForce RTX-class cards may be better optimized for it, at least for now.
Driver updates over time have been known to narrow gaps like these, although this is quite the gap.
The fact that the Radeon VII didn't lose as badly at the Extreme preset hints that it may prefer higher resolutions.
Unigine Superposition
Our last synthetic benchmark is Unigine's 2017 release, Superposition.
This benchmark does incorporate ray tracing, but it's done in software, not hardware, and thus doesn't utilize the RT cores of the RTX 20 series in these charts.
Here, again, the Radeon VII loses by double-digit percentages to the GeForce RTX 2080 Founders Edition.
It keeps ahead of the old GeForce GTX 1080 Founders Edition$549.00 at Nvidia, though, and doubles the performance of the midrange Radeon RX 590.
The Radeon VII is just over twice as expensive as the latter.
The 16GB of video memory and high memory bandwidth of the Radeon VII likely explain why its performance gets better with resolution.
For example, it's 24 percent faster than the Radeon RX Vega 64 at the 1080p setting, but 33 percent faster in the 8K optimized test.
Watch for this trend to continue.
Real-World Gaming
The following benchmarks are games that you can play.
The charts themselves will list the settings we used (typically the highest in-game presets and, if available, DirectX 12).
Shadow of the Tomb Raider
Square Enix's recent title is our first real-world test.
This game is well-optimized for the PC platform, but very demanding at its higher visual quality settings.
These results average to a dead heat between the Radeon VII and the GeForce RTX 2080 Founders Edition.
Taken separately, however, the Radeon VII has stronger relative performance at 4K against every graphics card here.
Take the GeForce RTX 2080 Ti Founders Edition as an example; its 34 percent advantage over the Radeon VII at 1440p drops to just 22 percent at 4K.
Rise of the Tomb Raider
The 2015...
