Date: May 20, 2020
Author(s): Rob Williams
AMD released its Navi-based Radeon Pro W5500 a few months ago, and to say we’ve spent a lot of time battering it with tests would be an understatement. We’re taking a look at AMD’s latest $399 professional GPU with the help of many encoding, rendering, viewport, and gaming tasks.
Since late 2016 when we saw AMD transition its workstation graphics series from FirePro to Radeon Pro, the boldly blue cards have spanned three architecture generations. The first out-of-the-gate were the Polaris 4th-gen GCN cards, which were quickly joined by the 5th-gen GCN Vega cards. With the launch of the W5700 in November, RDNA 1.0 became the newest hotness.
It’s an interesting time for AMD workstation graphics, because depending on the workload, either Navi or Vega could be the better choice. For most ProViz use, Navi is a safe bet, especially since it has the latest (and best) version of AMD’s encoder. For those with hungrier data crunching needs, or want the absolutely best performance AMD offers, Vega remains a solid choice. We’ve seen proof with the recent launch of the Radeon Pro VII that AMD is keeping to its word that Vega’s life will continue on.
With its 1,408 cores, the W5500’s specs closely align with the Radeon RX 5500 XT from the gaming side, whereas the W5700 we’ve taken a look at before has closely matched the RX 5700 (non-XT). That means AMD still has a bit of room to later release a W5600 that would likely inherit the 2,304 cores of the RX 5600. The W5700 has fewer cores than the RX 5700 XT, so the company even has room to release another top-end Navi beast later. Or, it could become that our speculations are wrong (as usual).
What we do know for sure is that the W5500 we’re taking a look at here is AMD’s second released Navi card. Thankfully, the fact that AMD just released a new Radeon Pro VII doesn’t impact our performance look here, as the $1,500 delta in the prices of those cards may suggest.
|AMD’s Radeon Pro Workstation GPU Lineup|
|Cores||Base MHz||Peak FP32||Memory||Bandwidth||TDP||Price|
|Pro VII||3840||???||13.1 TFLOPS||16 GB 8||1TB/s||250W||$1899|
|W5700||2304||1183||8.89 TFLOPS||8 GB 1||484 GB/s||205W||$799|
|W5500||1408||1200||5.35 TFLOPS||8 GB 1||224 GB/s||125W||$399|
|WX 9100||4096||1200||12.3 TFLOPS||16 GB 8||484 GB/s||230W||$1399|
|WX 8200||3584||1200||10.8 TFLOPS||8 GB 8||512 GB/s||230W||$999|
|WX 7100||2304||1188||5.73 TFLOPS||8 GB 3||224 GB/s||130W||$549|
|WX 5100||1792||713||3.89 TFLOPS||8 GB 3||160 GB/s||75W||$359|
|WX 4100||1024||1125||2.46 TFLOPS||4 GB 3||96 GB/s||50W||$259|
|WX 3200||640||1295||1.66 TFLOPS||4 GB 3||96 GB/s||50W||$199|
|WX 3100||512||925||1.25 TFLOPS||4 GB 3||96 GB/s||50W||$169|
|WX 2100||512||925||1.25 TFLOPS||2 GB 3||56 GB/s||50W||$129|
|Notes||1 GDDR6; 2 GDDR5X; 3 GDDR5; 4 HBM2|
5 GDDR6 (ECC); 6 GDDR5X (ECC); 7 GDDR5 (ECC); 8 HBM2 (ECC)
Polaris 4th-gen GCN = WX 2100~7100
Vega 5th-gen GCN = WX 8200~9100 & Pro VII
Navi RDNA 1.0 = W5500~5700
In terms of quoted single-precision performance, the W5500 settles a little bit behind the older WX 7100. That technically means that the WX 7100 should win most tests, but it’s of course never that simple. AMD’s newer card will definitely be able to show off its strengths throughout our performance results.
And with that, this card is pretty easy to gauge based on its specs, so we can get right into a quick look at our test setup, and then move into the results.
On the following pages, the results of our workstation GPU test gauntlet will be seen. The tests chosen cover a wide range of scenarios, from rendering to compute, and includes the use of both synthetic benchmarks and tests with real-world applications from the likes of Adobe and Autodesk.
To paint a good overall picture of current performance, we’re using a blend of gaming and workstation GPUs. For some users, the best performance for the buck is the most important factor when scouting out a new GPU, while for others, increased platform stability and optimization through drivers is more integral.
Across these tested GPUs, price ranges of $200~$900 are covered. That means we’re missing out on some higher-end results, but for the purpose of taking a sufficient look at a $400 GPU, those cards weren’t necessary for this review. We do however have additional cards being tested now, and will post a more general (but fuller) performance look in the weeks ahead.
|Techgage Workstation Test System|
|Processor||Intel Core i9-10980XE (18-core; 3.4GHz)|
|Motherboard||ASUS ROG STRIX X299-E GAMING|
|Memory||G.SKILL FlareX (F4-3200C14-8GFX)|
4x8GB; DDR4-3200 14-14-14
|Graphics||AMD Radeon Pro W5700 (8GB, Enterprise 20.Q1.1)|
AMD Radeon Pro W5500 (8GB, Enterprise 20.Q1.1)
AMD Radeon Pro WX 7100 (8GB, Enterprise 20.Q1.1)
AMD Radeon Pro WX 4100 (8GB, Enterprise 20.Q1.1)
AMD Radeon RX 5700 XT (8GB, Adrenaline 20.4.1)
AMD Radeon RX 5600 XT (8GB, Adrenaline 20.4.1)
NVIDIA Quadro RTX 4000 (8GB, Quadro 442.92)
NVIDIA Quadro P2200 (5GB, Quadro 442.92)
NVIDIA GeForce RTX 2060S (8GB, Studio 442.92)
NVIDIA GeForce RTX 2060 (8GB, Studio 442.92)
NVIDIA GeForce GTX 1660 Ti (8GB, Studio 442.92)
|Storage||AMD: Samsung 480GB SATA SSD|
NVIDIA: Samsung 480GB SATA SSD
|Power Supply||Corsair 80 Plus Gold AX1200|
|Chassis||Corsair Carbide 600C Inverted Full-Tower|
|Cooling||NZXT Kraken X62 AIO Liquid Cooler|
|Et cetera||Windows 10 Pro build 18363 (1909)|
Our benchmark results are categorized and spread across the following pages. On page 2, we’re taking a look at renderers with the help of Blender and LuxMark. Page 3 is home to our encoding tests, which are handled by Adobe’s Premiere Pro, MAGIX’s Vegas Pro, BRAW Speed Test, and an application we throw into the same pile: Agisoft’s photogrammetry tool, Metashape. Page 4 handles viewport performance across a range of popular suites, and page 5 tackles some simple gaming. The final page will include some power tests along with our final thoughts.
And with all of that covered, let’s get on with it:
It’s interesting to see how GPU rankings shake up from one project to the next. In the less complex of the two scenes, BMW, NVIDIA has some obvious strengths. But in Classroom, AMD ends up getting the nod overall. In the matchup between the ~$400 Quadro P2200 and Radeon Pro W5500, AMD comes out ahead overall, and with more available VRAM in the end. The W5500 even beats the technically faster WX 7100, which goes to show how architecture plays an important role beyond what the TFLOPS number shows.
Cycles and Eevee are completely different beasts, so testing the latter gives us the opportunity to see if things change once again:
AMD’s RPro W5500 performed well with Cycles, but it falls a fair bit behind with Eevee. The overall results here make the RX 5600 XT and 1660 Ti look pretty attractive for their ~$290 price points.
We don’t like bringing gaming GPUs into the discussion too much in a workstation review, but Blender’s audience generally cares about performance above all, and so the best value is always going to be with the gaming cards. The Radeon RX 5600 XT for ~$290 looks particularly attractive here for its price-point, while the GeForce RTX 2060 SUPER would be an ideal choice at at the $400 mark. It’s important to note that if Cycles will be your primary render engine, NVIDIA’s OptiX/RTX features can add further performance boosts.
To keep things organized better, Blender viewport performance can be found on the dedicated page, which happens to be page four.
LuxMark is revolved around rendering, so the faster the GPU, the quicker the render will complete. We’re once again seeing an example here of how shake-ups can happen between projects. In the Food test, the RTX 4000 soars to the top, but in Hall Bench, AMD’s 5700 XT manages to pounce it. We can even see the same kind of shake-up between AMD’s own W5500 and WX 7100. Between W5500 and P2200, the Radeon Pro takes the cake.
It’s interesting just how often two projects can show starkly different performance results in the same application. Our 1080p project encoded faster on NVIDIA GPUs, to the extent that even the lowest-end NVIDIA card beat out the highest-end AMD. In the 4K test, which uses entirely different assets (all 4K-sourced), most of the GPUs perform the same. That probably means we need a different 4K project to see better scaling, but at least we still have 4K and 8K codec tests to look at:
The above results once again highlight the fact that it’s important to understand your workload, since even the codec choice can sway which graphics card comes out ahead. With AVC source, both AVC and HEVC encodes were faster on NVIDIA’s Quadro P2200, versus W5500, but AMD’s Radeon Pro redeemed itself in both the RED and ProRes transcodes.
It’s clear from the start that AMD’s graphics cards have some strong optimization in Vegas, with the four high-end Radeons soaring straight to the top. In the HEVC test, even the older Polaris-based cards placed ahead of NVIDIA.
Let’s see how the addition of FX affects things:
With either the LUT or Median-infused projects, the AMD stack continues to perform extremely well. We’d draw the line at the older Polaris cards, though. With LUT use, the WX 7100 and 4100 both found themselves at the bottom of the pile, while only the 7100 managed to redeem itself in the Median test.
This really is the final nail in the coffin. No NVIDIA graphics card managed to hit 60 FPS with LUT in our testing, while all of the AMD cards did. As mentioned in our launch look at VP17, Quadro cards still require you to add the application as a profile in NVIDIA’s Control Panel in order to gain full acceleration.
With the Median playback test, it’s proven that if you want the best possible performance with grueling effects use, AMD’s 5700 series is hard to beat. We’ll make sure the RX 590 and Radeon VII are tested for an upcoming look of general ProViz performance.
AMD’s Radeon laid the smack down in Vegas Pro, and NVIDIA has struck back hard with this BRAW test. We’re still planning to introduce real-world DaVinci Resolve tests in the future, but for now, this BRAW speed test tackles a basic, but important performance area.
While depths maps generation is just one part of the photogrammetry recipe, the process is one of the more intensive, and overall, it seems to favor Radeons in Metashape. Both of AMD’s 5700-series cards place at the top, and the RX 5600 XT gaming card places ahead of every other test NVIDIA card – even the Quadro RTX 4000, which is similar in spec to a GeForce RTX 2070.
In the match-up of Quadro P2200 and Radeon Pro W5500, AMD sure does make a strong statement here.
It seems like NVIDIA’s aging Quadro P2200 still has a lot of performance to give when it comes to SolidWorks. It’s tests like these that highlight the differences between gaming and workstation graphics cards quite well. Even the low-end Quadro P2200 manages to outperform AMD’s top-end Navi-based Radeon GPU. Let’s see if SolidWork’s sibling continues this sort of scaling:
In both SolidWorks and CATIA, NVIDIA manages to take the lead in the multiple match-ups we can see, like RTX 4000 vs. W5700 and P2200 vs. W5500. What’s proven by these results is just how far workstation optimizations can go, since technically, the W5500 is the faster of the two in the match-up against P2200. That means that the rendering performance will be a little better on the RPro, so which one is right for you really depends on the weight you put on certain performance angles.
With Creo, AMD’s W5500 fell a bit behind the P2200 and even the RX 5600 XT at 1080p, but it managed to strike back at 4K, inching ahead of the P2200 ever-so-slightly. Overall, the entire stack here performs pretty well for their price-point, although we can really see just how long overdue Creo users equipped with the WX 4100 are for an upgrade.
Siemens NX is a bit of a fun test, since it’s one that seems to have a seething hatred towards gaming-focused graphics cards. You may as well rule those out from the get-go. In this SNX match-up, AMD’s W5500 leaps ahead of the P2200, and at 4K, it also beats out the top-end Polaris-based RPro, WX 7100.
We don’t usually think of workstation GPUs having a performance advantage in either Max or Maya, but that may be changing over time. Versus the RX 5700 XT, the W5700 performs better in all tests. Oddly, though, the W5500 fell to second-last place in Max, but it overtook the P2200 in Maya.
While many of the viewport tests on this page will show clear differences between gaming and workstation graphics cards, Blender is as close to neutral as it gets, offering every vendor an optimized API to use. We saw on our rendering pages that Cycles and Eevee results will show different scaling, but so too is the case with viewport results.
Overall, NVIDIA shows great strengths in our custom Blender viewport tests, to such an extent that it almost feels like the company specifically optimized Turing for it. You’ll note that even the RTX-less 1660 Ti still keeps ahead of the rest of AMD’s entire lineup. We’ll see if anything changes here with the forthcoming 2.83 build of Blender.
Throughout much of our workstation testing, AMD’s Radeon Pro W5500 and NVIDIA’s Quadro P2200 had some surprisingly close fights, considering we know that the former card is the technically more powerful. That’s proven with these 3DMark tests, as the W5500 easily leaps past the P2200 in each one of these tests, encapsulating both DX11 and DX12.
We’re sure one of AMD’s biggest fears at the moment is NVIDIA spinning out a Turing-based 2000-series card, which could level the playing field a bit (and seems overdue at this point).
The strong performance of the W5500 carries through to VRMark, with it again toppling NVIDIA’s Quadro P2200. Meanwhile, the older WX 4100 is really showing its age, which isn’t surprising when the card was considered a lower-end option from the get-go. That said, we do see proof on the previous page where that aging card can still offer surprising viewport results.
As expected, AMD’s W5500 jumps in front of the P2200, and for what it’s worth, a fair bit ahead of the older WX 7100 (which was priced at $799 when released in late 2016).
Different tasks will vary how a graphics card is used, which means that some workloads will draw more power (and generate more heat) than others. It’s safe to say that more often than not, a gaming test is going to push a graphics card the hardest, outside of using a power virus like Furmark.
For the power testing in this article, we rendered the Classroom scene in Blender with a huge number of iterations to generate that result, and then ran a 3DMark Fire Strike loop to see how much further power draw could go. The reported wattages were measured through a Kill-a-Watt, which has only the tower PC plugged into it. Both the rendering and gaming tests run for ten minutes. Temperature monitoring is done with AIDA64, with a room temp as close to 23°C as possible.
As we’d expect, gaming pushes all of these GPUs a lot harder than rendering, with the RX 5600 XT in particular showing a notable delta. It’s a great thing that rendering doesn’t use the same power as gaming, or else many workstations would be a lot louder – and hotter. Speaking of heat:
Almost the entire stack here can hit 80°C without much effort; it just requires a good stress. Thankfully, that’s all within spec; you only really need to be concerned if it reaches temperatures above 90°, especially since that can really start to increase heat in other parts of the system.
AMD’s Radeon Pro W5500 is an attractive ProViz graphics card from multiple angles. For starters, it’s a $399 option that packs an 8GB framebuffer. By comparison, you’ll need to opt for NVIDIA’s $899 Quadro RTX 4000 to get that much memory on the pro-side. The Quadro P2200, which is most closely matched against the W5500, has only 5GB of VRAM. That’s yet another reason why we feel NVIDIA’s 2000-series is due for an update.
The W5500 is an especially great card for CAD users, namely those using suites that directly see performance and general optimizations on workstation cards. We’re largely talking Siemens NX, CATIA, SolidWorks, and Creo, but that’s only of those we tested (through SPECviewperf). If your primary use is a solution like AutoCAD, and you don’t care about gaming performance, then we’d always suggest a pro card over a gaming one, especially since the price-point to get quality performance is more palatable than it is for those who need top-end render or denoising performance.
The fact that the W5500 is a single-slot and single-connector card will also make it attractive to those who for one reason or another can’t or would prefer to not use a dual-slot GPU. To get a dual-slot GPU from Radeon Pro anyway, you’d be moving up to the Vega-based WX 8200 or Pro VII. The only thing that would make us happier about the W5500’s design would be if it could be shortened to fit in a small form-factor PC – such as the Intel Quartz Canyon NUC we recently wrote about.
What’s nice to see is the generational gains delivered from the Navi update. The current-gen W5500 at $399 performs a lot better in most cases than the then-$799 WX 7100 from three-and-a-half years ago. It also does it with a wee bit less power use overall.
Versus the bigger W5700, the W5500 opts for 4x DisplayPort connectors, while the W5700 uses 5x mini-DisplayPort connectors, and also includes a USB-C port. Not even the Radeon Pro VII has that, or any other Radeon for that matter.
It’s also worth pointing out that AMD’s graphics software has improved quite a bit over the years, and the current iteration is really robust. Of note, it offers support for AMD Remote Workstation, which as it sounds, allows you to access your physical rig from anywhere – assuming you have a sufficient enough internet connection. In the past couple of months, interest in remote work has really surged, so it’s good that AMD is already delivering a certified (Citrix and Microsoft) solution.
Copyright © 2005-2020 Techgage Networks Inc. - All Rights Reserved.