Date: April 7, 2021
Author(s): Rob Williams
Equipped with 11 graphics cards, ten game titles, and a bunch of synthetic tests, we’re taking a look at select GPUs from AMD’s and NVIDIA’s previous- and current-gen series to see which models stand out of the crowd, and perhaps offer the best bang-for-the-buck (…at least, as far as SRP goes).
It’s been some time since we’ve taken a good look at 1080p and 1440p gaming performance, so with AMD’s Radeon RX 6700 XT and NVIDIA’s GeForce RTX 3060 having been released, now seems like a great time to get current with the latest drivers and game updates.
As is likely obvious by now, this article is going to focus on 1080p and 1440 gaming, with 11 graphics cards being tested across ten game titles (and synthetics). We’re currently working on a follow-up to this article that will shift the focus to 3440×1440 ultrawide and 4K performance, as well as another that will be laser-focused on ray tracing.
If you’re in the market for a new graphics card, you’re no doubt aware that the industry is suffering from chip shortages, and thus store shelves are largely empty. It’s for that reason that writing this kind of content right now doesn’t feel fulfilling, because the chances of you finding the exact GPU you want is going to be stupidly tough. But, if you do happen to find a model in stock that fits your budget, our benchmark results here should help you decide one way or another whether you should pull the trigger.
Before jumping into our test results, let’s quickly go over AMD’s and NVIDIA’s GPU lineups. To give some extra perspective, these tables also include last-gen models.
|AMD’s Radeon Gaming GPU Lineup|
|Cores||Base MHz||Peak FP32||Memory||Bandwidth||TDP||Price|
|RX 6900 XT||5,120||2,015||23 TFLOPS||16 GB 1||512 GB/s||300W||$999|
|RX 6800 XT||4,608||2,015||20.74 TFLOPS||16 GB 1||512 GB/s||300W||$649|
|RX 6800||3,840||1,815||16.17 TFLOPS||16 GB 1||512 GB/s||250W||$579|
|RX 6700 XT||2,560||2,321||13.21 TFLOPS||12 GB 1||384 GB/s||230W||$479|
|RX 5700 XT||2,560||1,605||9.75 TFLOPS||8 GB 1||448 GB/s||225W||$399|
|RX 5700||2,304||1,465||7.95 TFLOPS||8 GB 1||448 GB/s||180W||$349|
|RX 5600 XT||2,304||1,375||7.19 TFLOPS||6 GB 1||288 GB/s||150W||$289|
|RX 5600||2,048||1,375||6.39 TFLOPS||6 GB 1||288 GB/s||150W||$279|
|RX 5500 XT||1,408||1,717||5.2 TFLOPS||8 GB 1||224 GB/s||130W||$199|
Architecture: RX 6000 = RDNA2; RX 5000 = RDNA
AMD’s GPU lineup is interesting for a few reasons, but a big one is the fact that the last-gen parts slide in cleanly behind the new RDNA2-based parts. In other words, AMD’s current-gen RDNA2 cards have so far only catered to the high-end crowd, with pricing starting at $479 SRP, while the previous-gen RDNA cards take care of the rest, topping out at $399 SRP.
With RDNA2, AMD made great improvements to its architecture, making those cards the first Radeons to officially support accelerated ray tracing. As we’ll see later, the company also did well in improving RDNA’s power efficiency.
As for NVIDIA:
|NVIDIA’s GeForce Gaming GPU Lineup|
|Cores||Base MHz||Peak FP32||Memory||Bandwidth||TDP||SRP|
|RTX 3090||10,496||1,400||35.6 TFLOPS||24GB 1||936 GB/s||350W||$1,499|
|RTX 3080||8,704||1,440||29.7 TFLOPS||10GB 1||760 GB/s||320W||$699|
|RTX 3070||5,888||1,500||20.4 TFLOPS||8GB 2||448 GB/s||220W||$499|
|RTX 3060 Ti||4,864||1,670||16.2 TFLOPS||8GB 2||448 GB/s||200W||$399|
|RTX 3060||3,584||1,780||12.7 TFLOPS||12GB 2||360 GB/s||170W||$329|
|RTX 2080 Ti||4,352||1,350||13.4 TFLOPS||11GB 2||616 GB/s||250W||$1,199|
|RTX 2080 S||3,072||1,650||11.1 TFLOPS||8GB 2||496 GB/s||250W||$699|
|RTX 2070 S||2,560||1,605||9.1 TFLOPS||8GB 2||448 GB/s||215W||$499|
|RTX 2060 S||2,176||1,470||7.2 TFLOPS||8GB 2||448 GB/s||175W||$399|
|RTX 2060||1,920||1,365||6.4 TFLOPS||6GB 2||336 GB/s||160W||$299|
|GTX 1660 Ti||1,536||1,500||5.5 TFLOPS||6GB 2||288 GB/s||120W||$279|
|GTX 1660 S||1,408||1,530||5.0 TFLOPS||6GB 2||336 GB/s||125W||$229|
|GTX 1660||1,408||1,530||5 TFLOPS||6GB 3||192 GB/s||120W||$219|
|GTX 1650 S||1,280||1,530||4.4 TFLOPS||4GB 2||192 GB/s||100W||$159|
|GTX 1650||896||1,485||3 TFLOPS||4GB 3||128 GB/s||75W||$149|
|Notes||1 GDDR6X; 2 GDDR6; 3 GDDR5
RTX 3000 = Ampere; GTX/RTX 1600/2000 = Turing
NVIDIA has so far released five Ampere-based GeForces, ranging from the $329 SRP RTX 3060 up to the $1,499 SRP RTX 3090. Interestingly, that lowbie RTX 3060 has a bigger frame buffer than the middle three SKUs, making it an appealing GPU not just for gaming, but also creation (and yes, we have fresh benchmarks in the works for that angle, as well.)
Whereas AMD has just gained accelerated ray tracing support, the current Ampere-based GeForce generation marks NVIDIA’s second go, and it really shows. Those who care deeply about ray tracing performance need to go green.
As mentioned before, this article focuses on 1080p and 1440p gaming performance. If you’re curious about performance specific to ray tracing, we’ll be dedicating an article to it soon. Also, while we’ve tested only 11 GPUs in our standard game collection, eight more have been added to the synthetics, found on the next page.
|Techgage Gaming GPU Test PC|
|Processor||AMD Ryzen 9 5950X (3.4GHz Base, 4.9GHz Turbo, 16C/32T)|
|Motherboard||ASRock X570 TAICHI (EFI: P4.00 01/19/2021)|
|Memory||G.SKILL TridentZ Royal (F4-3600C16-8GTRG) 8GB x 2
Operates at DDR4-3600 16-16-16 (1.35V)
|AMD Graphics||AMD Radeon RX 6800 XT (16GB; Adrenalin 21.3.1)
AMD Radeon RX 6800 (16GB; Adrenalin 21.3.1)
AMD Radeon RX 6700 XT (12GB; Adrenalin 21.3.1)
AMD Radeon RX 5700 XT (8GB; Adrenalin 21.3.1)
|NVIDIA Graphics||NVIDIA GeForce RTX 3080 (10GB; GeForce 461.72)
NVIDIA GeForce RTX 3070 (8GB; GeForce 461.72)
NVIDIA GeForce RTX 3060 Ti (8GB; GeForce 461.72)
NVIDIA GeForce RTX 3060 (12GB; GeForce 461.72)
NVIDIA GeForce RTX 2070 SUPER (8GB; GeForce 461.72)
NVIDIA GeForce RTX 2060 SUPER (8GB; GeForce 461.72)
NVIDIA GeForce RTX 2060 (6GB; GeForce 461.72)
|Storage||GeForce: WD Blue 3D NAND 1TB (SATA 6Gbps)
Radeon: WD Blue 3D NAND 1TB (SATA 6Gbps)
|Power Supply||Corsair RM850x (850W)|
|Chassis||Fractal Design Define C Mid-tower|
|Cooling||Corsair iCUE H115i RGB PLATINUM (240mm)|
|Et cetera||Windows 10 Pro build 19042.867 (20H2)|
All of our testing is done using the latest versions (at the time of benchmarking) of Windows (10, 20H2) and all of the games tested. Because some games update frequently, our goal is to always clear through our chosen set of GPUs before any of our tested games manage to drop a new update. If an update does roll out, we quickly sanity check a previous result to make sure performance remains the same.
Here are some other general guidelines we follow:
This article includes testing with nine standard games, and one specific to high-performance esports gaming. We wanted to include Rainbow Six Siege as a second esports title, but a recent update causes the game to use a random texture detail setting instead of the one specified, which is obviously a problem. To round our testing out, we have a number of synthetic tests on the next page, thanks to UL and Unigine.
Here’s the full breakdown of our tested games and synthetic tests:
Note: You can download all of the tested setting images at once here (ZIP, 10MB).
This article marks the first time we’ve tested with Assassin’s Creed Valhalla, and we’re fairly surprised with some of the results. In the past, NVIDIA seemed to generally have an edge in AC titles, but AMD’s Radeon performs strongly here. It’s not only with the current generation of cards, either; even the last-gen RX 5700 XT performs great against its competition.
If your aim is 60 FPS, then either the RX 5700 XT or RTX 3060 Ti will deliver it at ultra high detail levels. Valhalla has a decent assortment of graphics settings to tweak, so you will have options if you are struggling to hit 60 FPS.
One option worth noting is Resolution Scale, which lets you render the game at a lower resolution than your native to improve performance at the expense of some sharpness. On some GPUs, this option may default to 70% (something that bit us during testing).
AMD’s Radeons continue to perform great against the NVIDIA competition in Borderlands 3, which isn’t much of a surprise given what we’ve seen in the past. This is one game we’ve previously tested resizeable BAR with, and we can say that’s worth turning on if you have supported hardware. Note, though, that reBAR wasn’t enabled for any of our tested configurations. Since NVIDIA now supports reBAR on its current-gen GPUs, we’re tempted to explore before / after performance in a wider variety of games soon.
Death Stranding is arguably one of the best-looking games in this lineup of ten titles, and thankfully, it doesn’t require a powerhouse GPU to run. Even at 1440p, the smallest GPU in this tested lineup could hit 77 FPS average, and 59 FPS low. That means the higher-end SKUs would have no trouble running this game with good frame rates at 4K.
There are many different battles you can look at here, since the cards tested range from $329 up to $699. Overall, AMD’s Radeons once again perform strongly here, with even the RX 5700 XT outpacing the RTX 3060, at both resolutions.
After falling a bit behind AMD in the previous games, NVIDIA strikes back in Destiny 2. As we’ve seen in the past, AMD’s GPUs suffer a bit with percentile lows, something that seems to be more obvious on the latest-gen RDNA2 chips, for some reason.
It seems NVIDIA has polished performance in this particular title since our last round of testing, as GeForces are pulling ahead slightly more now than they did in the past.
Horizon Zero Dawn is another game that doesn’t require a powerhouse GPU to look great. At 1440p, even the RTX 2060 delivers 60 FPS, with the percentile low still hovering around 50. If you have a high refresh monitor, and would like to take full advantage of it, then the bigger GPUs are going to suit you better.
It’s become quite clear by now that AMD’s current-gen GPUs give NVIDIA the competition we’ve been hoping to see. Monster Hunter: World has historically been a game that’s performed better on NVIDIA’s GPUs, but Radeons strike hard here, in some cases placing ahead of models they didn’t used to. As we’ve seen before, Radeon percentile lows are usually behind NVIDIA’s in this title (mostly with RDNA1).
With The Division 2 continuing to be a pretty popular game, we decided to re-add it to our test suite this go-around, to see where performance currently stands. The results here seem to scale a bit better towards perf-per-dollar than the others have, but at the top-end, AMD’s 6800 XT once again edges out the RTX 3080.
If you’re a competitive gamer equipped with a high-refresh monitor, you’ll probably want to aim towards a model in the middle of the chart here. Note, however, that our tested settings are quite high, so there is a lot of room to tweak and drop some detail to gain FPS.
Like most of the other games tested here, our chosen detail settings for Three Kingdoms is quite high, although there are still a number of settings that could be increased further. Fortunately, this game can look and run great on any one of the GPUs listed here. If you have a high-end GPU, you can push the detail settings a bit higher, unless you’re after 1440/144Hz.
This is the first time we’ve tested Watch Dogs Legion since the game’s launch. It felt like a good time to return and see if anything has changed since launch (and to include new GPUs). Legion is a ray traced title, but for this article, we tested without it (as mentioned earlier, we have a dedicated ray tracing performance article coming.)
When we look back to that performance article from last fall, we can see that scaling really hasn’t changed at all. In that previous testing, the RX 5700 XT placed ahead of the RTX 2070 SUPER at 1080p, but at 1440p, they switched places. We see the same thing happen in this retest.
Overall, for being an NVIDIA-sponsored title, we’re surprised to see AMD performing a fair bit better. Again, this is without ray tracing; as we’ve seen from our myriad tests in the past, anyone who demands the best RT performance for most titles will need to scoot over to the green side of the fence.
Being a high-FPS title, CS: GO benefits from CPUs with faster clock speeds when gaming at lower resolutions (eg: 1080p), with clock speeds becoming less of an issue at higher resolutions (eg: 4K). As you can see in our 1440p graph above, scaling is boring, because the CPU can’t keep up with the amount of work the GPU wants to do. At 4K, we see much more defined scaling.
Almost all of the tested GPUs could support 4K/144 CS: GO gaming, although higher-end GPUs will help avoid the worst lows. AMD’s Radeons do really well here; even the last-gen 5700 XT.
This wraps up our look at performance in real game titles. If you’re interested in synthetic test results with an expanded GPU list, you can head on over to the next page. Otherwise, you can hit up page three for power and our final thoughts.
It’s not too often that we see a 3DMark score directly correlate with real-world gaming performance, and nothing really changes here. Despite matching the RTX 3080 overall in real gaming, the RX 6800 XT manages to surpass the Fire Strike scores of even the RTX 3090. Things change up in the DX12-focused Time Spy, with the RTX 3090 placing well ahead of the RX 6900 XT.
Perhaps these results are not truly relevant, but they at least give you an option to test your current GPU against ours, and get a vague idea of what kind of performance you’ll gain with a move to newer hardware.
Given what we’ve seen in the past, and given what we know of AMD’s RDNA2 ray tracing capabilities, the results from these RT-specific tests don’t come as too much of a surprise. We will say, though, that AMD’s top brass performs quite well, all things considered. That said, a GPU like the RX 6900 XT manages only half of the frame rate of the RTX 3090 in the DXR test.
It’s worth noting that the Turing-based GTX 1660 Ti could run the Port Royal ray tracing test no problem, likely thanks to the fact that NVIDIA at some point added support to the driver. That added support doesn’t enable DXR support, however, as it doesn’t exist in GTX Turing hardware. Even with poor performance, it’s kind of neat to see the 1660 Ti sit among the rest in the Port Royal test.
Here are some other random sub-tests to round things out:
Both AMD and NVIDIA perform great in the VRS and Mesh Shaders test, although in the latter, AMD’s cards seem to suffer a lot more than NVIDIA’s when the shaders are turned off. Turned on, however, and scaling turns into something more expected, for the most part.
The final chart, for DLSS, highlights just what the special NVIDIA feature can do for frame rate. While DLSS is typically tied to ray tracing, it’s not always. We enabled DLSS in Outriders last week, and it doubled our frame rate as hoped. In fact, DLSS has proven far more impressive in that title than the game’s servers have (not to be bitter or anything).
VRMark is a test that’s typically favored NVIDIA’s architecture, in a similar manner as the DX12-based Time Spy test does. Things have changed a little bit this go around, as AMD is keeping up to NVIDIA a lot better, with the top GPUs swapping places depending on whether it’s the Cyan or Blue test being run. If you truly care about VR, you’d not want to opt for a model below the middle of this chart. 6700 XT or RTX 3070 would make for a good starting target.
To wrap up our performance results, Unigine’s Superposition largely repeats what we’ve known already, although NVIDIA clearly has a bit more of an edge here than in some previous tests.
To take a look at GPUs from a power perspective, we’re using UL’s 3DMark Fire Strike 4K stress test, and the PC plugged into a Kill-A-Watt to monitor full system power draw. The PC is left to sit idle for about five minutes, or at least until the PC is idle, at which point that number is recorded. After five minutes of running the Fire Strike test, the load value is recorded (at the same point in the looping test).
Some of the results here are quite interesting. NVIDIA’s new GeForce RTX 3060 uses just a pinch more power than the last-gen RTX 2060, despite being a bit faster and having twice the frame buffer. As for the new Radeon RX 6700 XT from AMD, it’s a fair bit quicker than the last-gen RX 5700 XT, but draws only 25W more.
It became so common back in the day that a new Radeon series would draw more power than NVIDIA’s competition that we simply expected it to continue with each go around. RDNA2 is different, however, as the RTX 3080 draws a bit more power than the RX 6800 XT, and while not listed this time, the 3090 draws more power than the RX 6900 XT. AMD is even winning the power consumption battle on the CPU side, so to say the company managed to turn things around that way would be an understatement.
If time was never an issue, we would like to include as many games as possible in roundups like these, but for this time around, we had to stick to our ten (or what would have been 11 if Siege benchmarked correctly). The games chosen to test with span a wide range of genres, and even partner favoritism (although again, NVIDIA’s sponsored Siege had to be dropped).
From these results, it’s become clear that AMD has really struck back hard in the high-end arena with its RDNA2 graphics cards. When NVIDIA released its first Ampere GeForces last fall, there’s no sugar coating the fact that we were really quite impressed, especially with regards to stack pricing vs. the previous gen. So it was only more impressive when AMD released its latest Radeons, and managed to largely keep up (if not surpass).
Both AMD and NVIDIA have strong feature sets, but ultimately, it can be argued that NVIDIA has the richer one, which means that depending on what you’re after, you may still want NVIDIA even if performance in some place is a bit lower than Radeon. That’s primarily the case if you care greatly about ray tracing, as NVIDIA has a stronghold on performance there.
With RDNA2, AMD technically caught up to the ray tracing performance of NVIDIA’s last-gen Turing, but the current-gen Ampere brings a notable improvement, and that’s not even counting DLSS, which can dramatically improve frame rates in games that can take advantage of it.
Another area NVIDIA has a firm hold on is the creator market, if you’re dealing with software that takes advantage of CUDA. Unfortunately for AMD, CUDA still seems to run the creator world, with there still being a handful of popular renders that exclusively use CUDA/OptiX. If 3D design isn’t your creator focus, then our upcoming article taking a look at general ProViz performance across a similar selection of GPUs should help.
Even if AMD is being held back by weaker ray tracing performance, and lacks some support for creator-type renders, the overall number of people impacted by either will be fairly small. Where AMD does well is delivering cards that gives users the top-end performance they’re after, with pretty favorable power consumption.
To either company’s benefit, what strengths exist barely seems to matter when users have to fight tooth and nail to secure a card from either vendor. While it seems more difficult than ever to find a new GPU to purchase, our perusal of popular hardware forums shows that people are still getting them, and without paying scalper prices (it’s clearly ideal if you don’t let yourself be ripped off by scalpers).
AMD and NVIDIA are going to be fairly happy either way, since all of their product is flying off the shelves before it can gather a speck of dust. For you, all we can do is wish you the best of luck, and hope that you don’t wear your left mouse button out before finally being able to buy one.
As covered earlier, we have other gaming GPU content in the works, including a look at 3440×1440 and 4K performance in a similar set of games that was tested here, ray tracing in a wider range of titles than we’ve tested with before, as well as fresh creator benchmarks. We also want to tackle some reBAR tests once those other content ideas are fulfilled. Stay tuned.
Copyright © 2005-2021 Techgage Networks Inc. - All Rights Reserved.