AMD professional graphics have played second fiddle to NVIDIA's for some years now. Although the Radeon Pro WX 8200 put up a considerable challenge to the Quadro P4000, the RTX 4000 arrived a few months later and redressed the balance back to NVIDIA again. But AMD's early adoption of 7nm process technology has given it a real boost amongst consumer graphics, and now that benefit is being fed into the professional graphics market. The first product to hit the shelves is the Radeon Pro W5700.
The W5700 is, as its model number implies, a professional parallel to the consumer-grade Radeon RX 5700. The core specs are very similar, with a Navi 10 XL / RDNA GPU inside that sports 36 compute unites and 2,304 Stream Processors. That is considerably fewer than the Radeon Pro WX 8200's 3,584, but the W5700 has both a higher base clock and a much higher boost clock, which will go some way in compensation. It also uses GDDR6 memory instead of HBM2, which has a positive benefit for price since HBM is more expensive. You get the same 8GB of frame buffer, too.
This is a future-proofed card in terms of connectivity, with five Mini-DisplayPort connections and USB-C. You can connect up to five 4K screens and even run an 8K screen at up to 60Hz. The TDP is 205W, which isn't too onerous, so there's potential to get a lot of performance per Watt as well.
So while the specs don't imply that the W5700 will be a complete Quadro RTX 4000 killer in terms of performance, it should be reasonably competitive, and the price of $799 makes it potentially better value, particularly now that the Radeon Pro WX 8200 is actually more expensive than the Quadro RTX 4000. Let's find out how it performs.
GPU | AMD Radeon Pro W5700 |
AMD Radeon Pro WX 8200 |
AMD Radeon Pro WX 9100 |
Compute Units |
36 | 56 | 64 |
Stream Processors |
2,304 | 3,584 | 4,096 |
GPU Architecture / Variant | Navi 10 XL | Vega 10 / Vega 10 XT | Vega 10 / Vega 10 XT |
Base Clock | 1,183 MHz | 1,200 MHz | 1,200 MHz |
GPU Boost Clock | 1,930 MHz | 1,530 MHz | 1,500 MHz |
Total Video memory | 8 GB GDDR6 | 8 GB HBM2 | 16 GB HBM2 |
Memory Clock (Effective) |
1,750 (14,000) MHz | 1,000 (2,000) MHz | 945 (1,890) MHz |
Memory Bandwidth | 448 GB/sec | 512 GB/sec | 484 GB/sec |
Bus Width | 256-bit | 2,048-bit | 2,048-bit |
Manufacturing Process | 7nm | 14nm | 14nm |
TDP | 205 W | 230 W | 230 W |
Display Outputs | 5 x Mini-DisplayPort 1.4, USB-C | 4 x Mini-DisplayPort 1.4 | 6 x Mini-DisplayPort 1.4 |
Display Resolution |
5 @ 1920×1080
5 @ 3840×2160
3 @ 5120×2880
1 @ 7680×4320
(all at 60Hz)
|
4 @ 1920×1080
4 @ 3840×2160
3 @ 5120×2880
1 @ 7680×4320
(all at 60Hz)
|
6 @ 1920×1080
6 @ 3840×2160
3 @ 5120×2880
1 @ 7680×4320
(all at 60Hz)
|
Software API Support | DirectX 12, OpenGL 4.6, OpenCL 2.0, Vulkan 1.1 | DirectX 12, OpenGL 4.6, OpenCL 2.0, Vulkan 1.1 | DirectX 12, OpenGL 4.6, OpenCL 2.0, Vulkan 1.1 |
AMD Radeon Pro W5700 Estimated Retail Price: $799
The Radeon Pro W5700 looks almost identical to the WX 8200 and 9100, with a dark blue casing and the same cooling system. It's still a dual-slot card, unlike the NVIDIA Quadro RTX 4000.
The W5700 is so new that GPU-Z doesn't detect all its parameters yet, particularly clock speed and GPU details. Looking at the published specs, although the W5700's GPU has only got 2,304 Stream Processors – 36 per cent fewer than the WX 8200 – they can run at up to 1,930MHz, which is 26 per cent faster. So although the WX 8200 is still likely to win on some benchmarks, it won't be by much.
The GDDR6 memory is also very different in specification. The WX 8200's HBM2 memory runs at an effective 2,000MHz but on a 1,024-bit bus. The W5700's GDDR6 memory runs at 14,000MHz on a 256-bit bus. But the end result is quite similar bandwidth. Where the WX 8200 boasts 512GB/sec, the W5700 still manages 448GB/sec, which is more than the Quadro RTX 4000 and on par with the much more expensive RTX 5000.
Another thing worth mentioning is that the W5700 is “PCI Express 4.0 ready”, so can run in this mode when partnered with a motherboard and processor offering PCI Express 4.0 (currently only AMD's latest CPU generation). This offers 1,969MB/sec per lane, twice as much as PCI Express 3.0. So if you're running a lot of PCI Express peripherals, such as multiple graphics cards and storage devices, you will have twice as much headroom.
This is a great graphics card for multi-monitor configurations, too. There's a very healthy five Mini-DisplayPort connections, plus a future-proofing USB-C port that can even supply 15W power for a screen. All the DisplayPorts are 1.4-capable, so can support up to 4K screens simultaneously.
If you use 5K (5,120 x 2,880) screens, you can drive three of these, or a single 8K (7,680 x 4,320) screen. The latter can run at up to 60Hz, too. Now that 8K video production is beginning to arrive, this could well be a valuable feature, although an 8K screen will set you back at least £3,000.
The W5700 is potentially a capable device for GPGPU as well. Whereas NVIDIA cards support CUDA, AMD's support OpenCL, and in this case version 2.0 of the API. With 2,304 Stream Processors running at up to 1,930MHz, there's a decent amount of grunt for activities like GPU-powered animation rendering or video encoding.
There's more to the RDNA architecture than just a 7nm die shrink compared to the previous 14nm. AMD claims an average 25 per cent more instructions per clock compared to GCN 5.0 used in the WX 8200, and much higher efficiency, with up to 41 per cent higher performance per Watt.
The Radeon Media Engine will decode VP9 or H.265 video up to 8K at 24fps and 4K at 90fps. There's encoding support up to 4K in H.264 and H.265 as well. There's also a feature called Image Boost, which allows the card to render your screen at a higher resolution and then down-sample to the native resolution, providing greater perceived detail.
Although AMD's graphics cards are efficient in terms of instructions per Watt, they do tend to have slightly higher overall TDP ratings than their NVIDIA equivalent. Where the NVIDIA RTX 4000 is rated at 160W, the AMD Radeon Pro W5700 is rated at 205W, although this includes extras like the 15W USB-C power supply ability.
These are just ratings and the more relevant measurement is how much work is done per Watt. However, you do need both 8-pin and 6-pin power connectors with the W5700, whereas the RTX 4000 uses just a single 8-pin.
We will mainly be testing the W5700 against the WX 7100, WX 8200 and NVIDIA Quadro RTX 4000, but we thought it would be good to look at the full Quadro RTX range as a background.
The Quadro RTX 4000 is a single-width card, whilst the RTX 5000 and 6000 take up two expansion slots like the W5700.
You could be deceived into thinking the RTX 4000 has some great similarities to the W5700, since both have 36 compute units and 2,304 cores. But the differing architectures mean that you can't make a direct comparison. In the past, AMD has needed more Stream Processors to provide similar performance to a given number of CUDA cores, and the RTX 4000 also runs its cores at a considerably lower clock.
The RTX 4000 also only offers 416GB/sec memory bandwidth from its GDDR6 memory, due to a slightly lower clock speed compared to the W5700.
Either way, with the RTX 4000 costing close to £1,000 inc VAT, and the R5700 costing $799, if the latter can provide competitive performance with the same level of professional support, the RTX 4000 will have some questions to answer.
GPU | NVIDIA Quadro RTX 4000 | NVIDIA Quadro RTX 5000 |
NVIDIA Quadro RTX 6000 |
Compute Units |
36 | 48 | 72 |
CUDA Cores | 2,304 | 3,072 | 4,608 |
GPU Architecture / Variant | Turing / TU104 | Turing / TU104 | Turing / TU102 |
Base Clock | 1,005 MHz | 1,620MHz | 1,440MHz |
GPU Boost Clock | 1,545MHz | 1,815MHz | 1,770MHz |
Total Video memory | 8GB GDDR6 | 16GB GDDR6 | 24GB GDDR6 |
Memory Clock (Effective) |
1,625 (13,000) MHz | 1,750 (14,000) MHz | 1,750 (14,000) MHz |
Memory Bandwidth | 416 GB/sec | 448 GB/sec | 672 GB/sec |
Bus Width | 256-bit | 256-bit | 384-bit |
Manufacturing Process | 12nm | 12nm | 12nm |
TDP | 160 W | 230 W | 260 W |
Display Outputs | 3 x DisplayPort 1.4, 1 x USB-C | 4 x DisplayPort 1.4, 1 x USB-C | 4 x DisplayPort 1.4, 1 x USB-C |
Display Resolution |
4 @ 4096×2160 @ 120Hz
4 @ 5120×2880 @ 60Hz
1 @ 7680×4320 @ 60Hz
|
4 @ 4096×2160 @ 120Hz
4 @ 5120×2880 @ 60Hz
1 @ 7680×4320 @ 60Hz
|
4 @ 4096×2160 @ 120Hz
4 @ 5120×2880 @ 60Hz
1 @ 7680×4320 @ 60Hz
|
Software API Support | DirectX 12, OpeGL 4.6, OpenCL 1.2, Vulkan 1.1, CUDA 7.5 | DirectX 12, OpeGL 4.6, OpenCL 1.2, Vulkan 1.1, CUDA 7.5 | DirectX 12, OpeGL 4.6, OpenCL 2.0, Vulkan 1.1, CUDA 7.5 |
We tested the AMD Radeon Pro W5700 in an Armari Magnetar workstation. This is a hugely powerful workstation based around an Intel Core i9-10980XE processor, backed with 128GB of 2933MHz DDR4 SDRAM.
For comparison, we pitted the W5700 against the WX 7100, WX 8200 and NVIDIA's Quadro RTX 4000.
Software:
Maxon Cinebench R15 (OpenGL only)
SPECviewperf 13
VRMark Advanced Edition
LuxMark 3.1
Armari Magnetar Specifications:
- Intel Core i9-10980XE @3GHz
- Enermax Liqtech 240 Water Cooling
- 128GB Samsung DDR4-2933 SDRAM @ 2,933MHz
- ASUS WS X299 PRO / SE Motherboard
- 8GB GDDR5 AMD Radeon Pro WX 7100 Graphics
- 1TB Samsung 970 PRO M.2 NVMe PCI Express SSD
- LiteOn Slimline SATA DVD-RW
- 1280W Slimline Platinum Efficiency PSU
- Armari Magnetar AWR820T Chassis
- Windows 10 Professional 64-bit
- 3 Years Warranty (1st Year On-Site, 2nd and 3rd Years RTB Parts and Labour)
CINEBENCH 15 is a cross-platform testing suite that measures hardware performance and is the de facto standard benchmarking tool for leading companies and trade journals for conducting real-world hardware performance tests. With the new Release 15, systems with up to 256 threads can be tested.
CINEBENCH is available for both Windows and OS X and is used by almost all hardware manufacturers and trade journals for comparing CPUs and graphics cards. We only ran the Open GL portion of the test, as graphics power has no effect on rendering performance.
The W5700 beats the elderly WX 7100, and is behind the WX 8200. But this is one test where the NVIDIA Quadro RTX 4000 wins out. So if you're running Maxon Cinebench, the new Radeon Pro gives commendable performance for the money, but the Quadro beats it in raw performance.
SPECviewperf 13
The SPECviewperf 13 benchmark is the worldwide standard for measuring graphics performance based on professional applications. The benchmark measures the 3D graphics performance of systems running under the OpenGL and Direct X application programming interfaces. The benchmark’s workloads, called viewsets, represent graphics content and behavior from actual applications.
The latest version is SPECviewperf 13, released on May 23, 2018. SPECgpc members at the time of V13 release include AMD, Dell, Fujitsu, HP, Intel, Lenovo, and NVIDIA.
SPECviewperf 13 is a comprehensive upgrade of previous versions of the benchmark. Medical and energy viewsets incorporate new models and raycasting for volume visualization; the Maya viewset features new models based on the SPECapc for Maya 2017 benchmark; and the Creo viewset has been updated with fresh application traces. All other viewsets have been recompiled with minor changes. Results from SPECviewperf 13 are not comparable to those from earlier versions.
Other major updates in SPECviewperf 13 include:
- Support for 4K resolution displays.
- New reporting methods, including JSON output that enables more robust and flexible result parsing.
- A new user interface that will be standardized across all SPEC/GWPG benchmarks.
- New workloads and scoring that reflect the range of activities found in real-world applications.
- Various bug fixes and performance improvements.
This is arguably the most important test we ran, and it's very good news for AMD in most cases.
The W5700 outperforms its WX 7100 and 8200 very respectably with the 3dsmax-06 viewset, beating the 8200 by nearly 10 per cent, but NVIDIA's Quadro RTX 4000 is 23 per cent faster still.
However, the catia-05 viewset is a different story. The W5700 is faster than all the other cards, beating the Quadro RTX 4000 by nearly 3 per cent.
Results swing back in NVIDIA's favour with the creo-02 viewset, where the Quadro RTX 4000 is 27 per cent faster, but the W5700 is still the quickest AMD card.
It's virtually neck-and-neck with the energy-02 viewset, whilst medical-02 gives the RTX 4000 another 9 per cent win. However, these tests don't correspond directly to specific branded software applications. The energy-02 viewset is based OpendTect seismic visualisation, whilst medical-02 simulates things like MRI scanner outputs.
With the maya-05 viewset, the W5700 again comes out on top, beating the Quadro RTX 4000 by 5 per cent.
The showcase-02 viewset is another one based on Autodesk software, in this case Showcase. AMD Radeon Pros are clearly well optimised for this, as both the WX 8200 and W5700 gain a jump on the NVIDIA Quadro RTX 4000. The W5700 is a sizeable 32 per cent in the lead.
The final two viewsets – snx-03 and sw-04 – are focused on CAD software used frequently for product design – Siemens NX and Solidworks respectively. It's reassuring to see that the W5700 is the leading performer in both. It's 3 per cent faster than the RTX 4000 in snx-03 and 9 per cent faster in sw-04.
The Radeon Pro W5700 is clearly a great card for CAD, engineering and product design with Siemens NX or Solidworks, and it's ahead of the RTX 4000 with Dassault Systemes CATIA as well. Only PTC Creo puts it behind the NVIDIA alternative.
For 3D animation, if you're using 3ds Max, then NVIDIA's Quadro RTX 4000 is the best choice. But for Maya, the AMD Radeon Pro W5700 takes the lead.
Overall, this is an impressive set of results considering that the W5700 is likely to be quite a bit cheaper than the RTX 4000 when it becomes available to buy. It renders the WX 8200 all but obsolete as well.
VRMark Advanced Edition
The performance requirements for VR games are much higher than for typical PC games. So if you're developing content for the HTC Vive or an Oculus Rift, you will need to do so on a PC that can cope with the added load.
VRMark includes three VR benchmark tests. You can run the tests on your monitor, no headset required, or on a connected HMD. At the end of each test, you'll see whether your PC is VR ready, and if not, how far it falls short.
We performed our tests with the middle Cyan Room run. This is a DirectX 12 benchmark. It features a large, complex scene and many eye-catching effects. Cyan Room shows how using an API with less overhead can help developers deliver impressive VR experiences even on modest PC systems.
The W5700 shows great promise with VR, beating the Quadro RTX 4000, although not by much.
LuxMark 3.1
OpenCL is a platform for harnessing GPU power for activities other than real-time 3D rendering to screen, also known as GPGPU. Unlike NVIDIA’s CUDA platform, OpenCL is open source and can be ported to anything with processing power. So drivers are available for CPUs as well, both from Intel and AMD. NVIDIA's graphics cards also support OpenCL, albeit an earlier version (1.2) than the AMD cards, which support version 2.0.
A popular tool for testing OpenCL performance is LuxMark. We rendered the most gruelling Hotel Lobby scene.
The W5700 may be great for modelling, but the lack of Stream Processors compared to the WX 8200 is clear when it comes to raw GPGPU compute, where it trails by 84 per cent. The NVIDIA Quadro RTX 4000 is even faster still – 220 per cent faster. If your workflow does involve GPU rendering, the W5700 is not the best option.
Power Consumption
Since AMD is claiming that the Radeon Pro W5700 is more power-efficient than previous models, we wanted to test this, so we set the VRMark Cyan Room test running as a loop, then measured total system power consumption. We also tested the system at idle, with nothing running.
When idle, the system draws 79W, and the AMD Radeon WX 8200 is the most thirsty, consuming 417.5W at its peak. The W5700 consumes 14 per cent less peak power, and since it is getting more work done as well (with a 24 per cent higher score), it's much more efficient than the previous model. The most miserly is the WX 7100, drawing just 235W, but it also has a VRMark score of less than half. The NVIDIA Quadro RTX 4000 is the most efficient, posting a 1 per cent lower score than the W5700, but consuming 22 per cent less power to do so. So whilst the W5700 is much more efficient than previous AMD generations, it's not quite competing with NVIDIA in this respect yet.
Temperature
We also checked the temperature of each card before we began the VRMark loops, and then during once it had settled to a constant value.
Here, AMD does beat NVIDIA comfortably, with both the Radeon Pro WX 8200 and W5700 idling around 29-31C, whilst not breaking too much of a sweat at full pelt with 75C. The WX 7100 is the least able to cope with the increased load, reaching 90C and idling at a higher 42C, whilst the NVIDIA Quadro RTX 4000 also hits a hotter 84C under load.
The AMD Radeon Pro W5700 isn't quite an NVIDIA Quadro RTX 4000 killer, but it has some clear strengths, particularly in VR and a number of industry standard CAD and 3D animation applications. When you consider that it's likely to be about £100 cheaper to buy than the RTX 4000, it's clearly a contender.
What's particularly telling is that the Radeon Pro W5700 is based on the lowest-end XL version of Navi 10 XL, not the XT or XTX that have more Stream Processors. There are higher-end Navi GPUs soon to be released too. The W5700 is not a successor to the WX 8200 or WX 9100, but for the lower-end and now quite antiquated WX 7100. There will probably be much more powerful Radeon Pro W5000 series cards hitting the market quite soon.
The W5700 also comes as news arrives that NVIDIA's next generation 7nm Ampere GPU generation has been delayed and is unlikely to hit the scene until the middle of 2020, with professional versions likely to be months after that.
So while AMD's presence in the professional graphics market has become increasingly non-existent, it could have as much as a year to re-establish itself, and the Radeon Pro W5700 is a very promising start. Overall, for quite a few industry-standard applications, the W5700 is the best professional graphics card you can buy for under £1,000, although for others the NVIDIA Quadro RTX 4000 retains its crown.
It's not as clear a win for AMD as the Ryzen 9 / Threadripper / EPYC ranges are becoming in the CPU market, but it looks like AMD is back for professional GPUs, thanks to the company's successful early move to 7nm technology. The AMD Radeon Pro W5700 is a great professional 3D graphics accelerator, and we can't wait to see what AMD releases next in the range.
The W5700 has a MSRP of $799. We don't have a buy link but will update this article when one is available.
Pros:
- The fastest sub-£1,000 professional graphics card with a wide range of industry-standard 3D applications.
- Cheaper than the NVIDIA Quadro RTX 4000 (we expect…)
- Reasonable power consumption.
- Extensive multi-monitor options with five Mini-DisplayPort outputs.
- 8K 60Hz support.
- USB-C graphics connectivity with 15W power delivery.
Cons:
- Mediocre OpenCL GPU compute abilities compared to competitors.
- Not the king in every professional 3D application.
KitGuru says: AMD is back in the professional GPU market, with a very competitive price-performance graphics proposition for many 3D design and animation applications.
Be sure to check out our sponsors store EKWB here