AMD Radeon Vega Frontier Edition 16GB Review
-
Page 1:Meet AMD’s Radeon Vega Frontier Edition
-
Page 2:Vega Architecture & HBM2
-
Page 3:Disassembly, Cooler & Interposer
-
Page 4:Board Layout & Components
-
Page 5:2D Workstation Performance
-
Page 6:3D Workstation Performance
-
Page 7:DirectX 11 Gaming Performance
-
Page 8:DirectX 12 Gaming Performance
-
Page 9:Vulkan/OpenGL 4.5 Gaming Performance
-
Page 10:Power Consumption
-
Page 11:Frequency, Temperature & Noise
-
Page 12:Summary & Conclusion
AMD’s goal with its Radeon Vega Frontier Edition seems clear: try to do everything well. Nvidia has to segment its portfolio carefully; it doesn't want to cannibalize sales of the Quadro P6000 with Titan Xp. But AMD doesn't have that problem. Radeon RX Vega 64 isn't here yet, so pros and gamers alike are putting the Frontier Edition board through its paces.
The company even encourages this by allowing the Radeon Vega Frontier Edition card to use AMD's Radeon Pro certified drivers, without which many workstation-class applications won't even start. The same can’t be said for Nvidia’s Titan Xp. No matter if you’re a prosumer who dabbles in high-end hardware for work and play, or a scientist eager to access lots of floating-point compute performance, this card is supposed to be a jack-of-all-trades.
Because we're still waiting for the Vega card meant to address gamers, we're not going to focus on that market exclusively (although we do have gaming benchmarks to show as well). Instead, you'll see us mix it up with workstation-class comparisons and high-end desktop cards. Included in the mix is Nvidia's Quadro P6000, which is around three times more expensive than AMD's Frontier Edition board.
Look, Feel & Connectors
AMD's Radeon Vega FE weighs in at a solid 1050g. Its length, from the slot bracket's edge to the end of the cooler, is 26.8cm. Its height, from the bottom of the PCIe slot to the cooler's top, is 10.5cm. A depth of 3.5cm makes this a true dual-slot card, with the backplate needing another 0.4cm of clearance on the other side.
Both the card’s cover and backplate are made of blue anodized aluminum that has a high-quality feel to it. All of the screws are painted black. A Radeon logo printed on the cover and the logo cube provide the only splashes of dissimilar color. This follows the theme of AMD's other Pro WX cards, though Vega FE's blue is noticeably darker and more saturated.
The card's top sports a pair of eight-pin auxiliary power connectors, as well as the aforementioned Radeon logo cube that lights up yellow. It's made from acrylic and recessed into the card’s cover.
The Radeon Vega FE's end is closed. It sports mounting holes, which are fairly standard for workstation-oriented graphics cards. Around front, the matte black anodized slot cover hosts three DisplayPort connectors and one HDMI 2.0 output. The omission of DVI connectivity makes sense from an airflow perspective; the slot cover also serves as the card's exhaust vent and needs to be as free-flowing as possible.
| Model | Radeon Vega FE | Radeon R9 Fury X | Titan Xp | Quadro P6000 | GeForce GTX 1080 Ti |
|---|---|---|---|---|---|
| GPU | Vega 10 | Fiji XT | GP102 (450-A1) | GP102 (350-K1-A1) | |
| GPU Size | 486mm² | 596mm² | 471mm² | 471mm² | |
| Transistors | 12.5 billion | 8.9 Billion | 12 Billion | 12 Billion | |
| GPU Clock Frequency (Base/Boost) | 1382 MHz 1600 MHz | Unknown 1050 MHz | 1480 MHz 1582 MHz | 1506 MHz 1645 MHz | 1480 MHz 1582 MHz |
| Shaders/SIMD | 4096/64 | 4096/64 | 3840/30 | 3584/28 | |
| Texture Units/ROPS | 256/64 | 256/64 | 240/96 | 224/88 | |
| Pixel Fillrate | 88.4 GPixel/s | 67.2 GPixel/s | 151.9 GPixel/s | 144.6 GPixel/s | 141.7 GPixel/s |
| Texture Fillrate | 353.8 GTexel/s | 268.8 GTexel/s | 379.7 GTexel/s | 361.4 GTexel/s | 354.1 GTexel/s |
| Memory Bus | 2048-bit | 4096-bit | 384-bit | 352-bit | |
| Memory Type | HBM2 | HBM | GDDR5X | GDDR5X | |
| Memory Bandwidth | 483.8 GB/s | 512.0 GB/s | 547.6 GB/s | 432.0 GB/s | 484.0 GB/s |
| Memory Data Rate | 1.89 Gb/s | 1.0 Gb/s | 11.4 Gb/s | 9 Gb/s | 11.0 Gb/s |
| Memory Size | 16GB | 4GB | 12GB | 24GB | 11GB |
| DX12 Feature Level | 12_1 | 12_0 | 12_1 | 12_1 | |
| Aux. Power Connectors | 2× 8-pin | 2× 8-pin | 6 + 8-pin | 8-pin | 6 + 8-pin |
| TDP | 300W | 275W | <250W | <225W | <250W |
Test System & Methodology
We introduced our new test system and methodology in How We Test Graphics Cards. If you'd like more detail about our general approach, check that piece out. Note that, in order to avoid any potential CPU bottlenecks when testing AMD's powerful graphics card, we've upgraded our CPU and its cooling solution in the meantime.
The hardware used in our lab includes:
| Test Equipment & Environment | |
| System | - Intel Core i7-6900K @4.3 GHz - MSI X99S XPower Gaming Titanium - Corsair Vengeance DDR4-3200 - 1x 1TB Toshiba OCZ RD400 (M.2, System) - 2x 960GB Toshiba OCZ TR150 (Storage, Images) - be quiet! Dark Power Pro 11, 850W Power Supply Unit (PSU) |
| Cooling | - Alphacool Eisblock XPX - Alphacool Eiszeit 2000 Chiller - 2x Be Quiet! Silent Wings 3 PWM (Closed Case Simulation) - Thermal Grizzly Kryonaut (Used when Switching Coolers) |
| PC Case | - Lian Li PC-T70 with Extension Kit and Mods - Configurations: Open Benchtable, Closed Case |
| Monitor | - Eizo EV3237-BK |
Power Consumption Measurement | - Contact-free DC Measurement at PCIe Slot (Using a Riser Card) - Contact-free DC Measurement at External Auxiliary Power Supply Cable - Direct Voltage Measurement at Power Supply - 2x Rohde & Schwarz HMO 3054, 500MHz Digital Multi-Channel Oscilloscope with Storage Function - 4x Rohde & Schwarz HZO50 Current Probe (1mA - 30A, 100kHz, DC) - 4x Rohde & Schwarz HZ355 (10:1 Probes, 500MHz) - 1x Rohde & Schwarz HMC 8012 Digital Multimeter with Storage Function |
Thermal Measurement | - 1x Optris PI640 80Hz Infrared Camera + PI Connect - Real-Time Infrared Monitoring and Recording |
Noise Measurement | - NTI Audio M2211 (with Calibration File, Low Cut at 50Hz) - Steinberg UR12 (with Phantom Power for Microphones) - Creative X7, Smaart v.7 - Custom-Made Proprietary Measurement Chamber, 3.5 x 1.8 x 2.2m (L x D x H) - Perpendicular to Center of Noise Source(s), Measurement Distance of 50cm - Noise Level in dB(A) (Slow), Real-time Frequency Analyzer (RTA) - Graphical Frequency Spectrum of Noise |
| Drivers | - Radeon: 17.20.1035 - Quadro: R381 U2 (382.05) |
| Operating System | - Windows 10 Pro (Creators Update, All Updates) |
MORE: Best Graphics Cards
MORE: Desktop GPU Performance Hierarchy Table
MORE: All Graphics Content
- Meet AMD’s Radeon Vega Frontier Edition
- Vega Architecture & HBM2
- Disassembly, Cooler & Interposer
- Board Layout & Components
- 2D Workstation Performance
- 3D Workstation Performance
- DirectX 11 Gaming Performance
- DirectX 12 Gaming Performance
- Vulkan/OpenGL 4.5 Gaming Performance
- Power Consumption
- Frequency, Temperature & Noise
- Summary & Conclusion
I cannot expect to see a huge % boost in games just on a driver update. Maybe 10% tops?
What bothers me the most is the heat/power and throttling concerns.
I dont think those will change with a fx version of the card.
So it would appear that 80-90% of the time Vega 64 will be running closer to 1200mhz, which is a major bummer and why it wont surpass a 1080.
Almost wish there had been a review of the watercooled version as well to see if it was able to maintain higher avg mhz.
It seems like it eats a lot of power for a very similar performance to a 1080 in gaming but at the same time its faster than Titan Xp in some workstations?
Doesn't that suggest driver/optimization issues?
Considering Vega also costs much more, I label it as an april fool
2 games, 1 resolution... and just where does this card fit in the vega lineup? You didn't even list the details of the other vega cards in a handy chart for comparison.
*starts looking for reviews on other sites*
wh-what? Toms stated like four times that this *is not a gaming card* and shouldn't be treated as one. They simply included the gaming benchmarks to be inclusive. Also, there's not really a point in showing benchmarks under 4k for such high end cards, because at that point (like 1080p) it's a CPU bottleneck - the GPU has nothing much to do. Considering listing other Vega cards, those are in a different class and not relevant, however it would be nice to have the details of the other cards, I guess.
I really wonder how could they make a new GPU on a big better process (28nm vs 14nm) and still get more power consumption compared to Fury X. What is more surprising is that they have the same number of CUs, approx same performance with Fury X, a lot more transistors, a lot more frequency and still the improvement is minimal.
They are either hiding some huge block which doesn't do anything in graphics scenarios which consumes a lot of power just idling or they are rubbish at power optimization or maybe the HBM2 consumes a lot of power, don't know.
Everything on this card is strange, starting with performance, specs, power consumption, everything...
Why even bothering rating the gaming benchmarks? Typical Toms... first line of these page should be "Take these with a grain of salt!"
Not everybody uses their computers to play video games. This is a workstation card with workstation drivers for doing work with your computer.
Just let him switch sites.
He obviously doesn't even read the articles as it is.
This is a pro level device and not intended for most users. Keep all this in mind when comparing it to other products.
On the other hand, the gaming performance is frighteningly bad. Yes, this card wasn't intended for gaming, but the numbers here have major implications for AMD's upcoming RX Vega cards. Gaming-oriented drivers are only going to take the Vega GPU so far. Unless there's something AMD disabled on Vega FE that is *severely* holding it down, the future looks pretty grim for RX, considering how those cards are all going to have reduced specs compared to FE.
give it two months. do not forget that this is in fact a new arch. it would not be the first time amd / ati pushed out gains way north of 25% with drivers on a new architecture. certain features are not even active yet in drivers (dsbr - a rasterizer) and that alone will give you some gains (in certain games very substantial ones it looks like) but more than that it shows how "raw" those drivers still are at this point.
the focus seemes to (logically) have been totally on the pro part of vega not the gaming one up to now.
i for one am going to go for the 56 ncu version of vega. since that will likely not be out before october anyway we should have a somewhat better picture of actual performance by then with drivers having matured at least a bit.
16 GB is a lot, that's the amount of my whole system. 8 is fine for gaming.
If you want help choosing a card, make a thread, you'll get more help that way