NVIDIA GeForce RTX 3080 Founders Edition Review
Ampere Arrives With A New $699 Gaming Flagship
When the GeForce GTX 1080 Ti launched back in March 2017, it essentially brought the gaming performance of the $1200 Titan X (Pascal) down to the $699 price level. The pinnacle of the 10 Series introduced in 2016, the 1080 Ti was such a gaming powerhouse that, for many, the Turing-based GeForce RTX launch in October 2018 was underwhelming.
Indeed, Pascal has had such an impact that, three and a half years down the road, the GeForce 10 Series was still considerably more popular among gamers than the RTX 20 Series. It’s pretty safe to say that real-time ray tracing has been adopted more slowly than NVIDIA would have liked, and Turing just wasn’t as compelling as it could have been. And that’s putting it as diplomatically as possible.
Suffice it to say, excitement about the 20 Series family among enthusiasts has been rather tepid for nearly two years now, but that all changed just a couple of weeks ago when NVIDIA revealed the high-level specs and pricing of the new Ampere-based GeForce RTX lineup.
Here’s a look at the specs your $699 could buy from NVIDIA at launch over the past three and a half years (plus the $999+ 2080 Ti for reference):
|RTX 3080||RTX 2080 Ti||RTX 2080 SUPER||RTX 2080||GTX 1080 Ti|
|Tensor Cores||272 (2nd Gen)||544 (1st Gen)||384 (1st Gen)||368 (1st Gen)||–|
|RT Cores||68 (2nd Gen)||68 (1st Gen)||48 (1st Gen)||46 (1st Gen)||–|
|Base Clock||1440 MHz||1350 MHz||1650 MHz||1515 MHz||1480 MHz|
|Boost Clock||1710 MHz||1545 MHz||1815 MHz||1710 MHz||1582 MHz|
|Memory||10GB GDDR6X||11GB GDDR6||8GB GDDR6||8GB GDDR6||11GB GDDR5X|
|Memory Data Rate||19 Gbps||14 Gbps||15.5 Gbps||14 Gbps||11 Gbps|
|Memory Bandwidth||760 GB/s||616 GB/s||496.1 GB/s||448 GB/s||484.4 GB/s|
|Die Size||628 mm2||754 mm2||545 mm2||545 mm2||471 mm2|
|Process Tech||8nm Samsung||12nm TSMC||12nm TSMC||12nm TSMC||16nm TSMC|
A comprehensive breakdown of all specs is available (as always) at TechPowerUp’s GPU Database, and for further reading about the Ampere architecture you can visit NVIDIA’s page which contains a link to the existing technology whitepaper.
This review will focus on performance with the new card, and will not attempt to break down the new Ampere architecture.
The RTX 3080 Founders Edition
Tastes vary, but I have to say, the new industrial design from NVIDIA for the RTX 3080 Founders Edition is as good as any consumer computer product I’ve ever seen. It’s a solid, futuristic design that weighs in at a solid 3.0 lbs. You may not agree, but I think an in-person demo might convince you.
Longer than the previous Founders Edition design, this dual-slot width card measures 11.2″ (285 mm) long, and 4.4″ (112 mm) tall. The card’s design is virtually all heatsink from the front side, with a pass-through fan the only opening on the rear side.
The rear IO bracket offers very large vents, with display connections along the edge consisting of three DisplayPort 1.4a and a single HDMI 2.1. Up to 4 monitors are supported, with a maximum resolution of 7680×4320 (“8K”), and HDCP 2.3 is supported.
Some have looked at both the design and thermal modeling shown by NVIDIA in their design video (YouTube link), and assumed that this unusual dual fan design – in which there are fans on opposite sides of the card – could be problematic to case airflow. That can be measured easily enough, of course.
Of course no overview of the Founders Edition design would be complete without mention of the new compact 12-pin connector, which takes the place of two 8-pin PCIe connectors.
Power supplies featuring dedicated cables with this new 12-pin connector are coming, and optional 12-pin modular cables already on offer from major PSU makers. The $0 solution is simply to make use of the dual 8-pin PCIe to 12-pin Y-adapter included with the Founders Edition (and presumably any other card making use of this new connector).
Before expounding on results we first embark on a journey of component exploration, with the test methodology of some interest here as the hardware is completely new for us, in this context. After talking it over with the staff it seemed to make sense to use an AMD platform for this launch.
Why? Because that’s where the enthusiast DIY market is, of course! Clearly I don’t need to tell our readers that AMD has a killer product lineup with Ryzen, and if that wasn’t enough to get us off of our Intel test platform there’s this little matter of PCI Express 4.0 to consider.
Yes, NVIDIA’s latest GPUs all take advantage of PCIe Gen4, meaning that results with an Intel platform will necessarily reduce throughput to Gen3 levels. Does this matter at all? We will cover the implications with some Gen vs. Gen testing later on, but for now enjoy some all-new benchmarks using a shiny new AMD system based on the Ryzen 9 3900X.
I hope this choice doesn’t end up costing us a week’s worth of benchmarking, but hey. It’s all temporary.
|PC Perspective GPU Test Platform|
|Processor||AMD Ryzen 9 3900X|
|Motherboard||ASUS ROG Crosshair VIII HERO WIFI|
|Memory||HyperX Predator DDR4-3600 CL16 32GB (16GBx2)|
|Storage||Samsung 850 EVO 1TB SSD|
|Power Supply||CORSAIR RM1000x 1000W|
|Operating System||Windows 10 64-bit (1909)|
|Drivers||GeForce Game Ready Driver 452.06 WHQL
Radeon Software Adrenalin 2020 Edition 20.8.3
GeForce Game Ready Driver 456.16 (Press Driver, RTX 3080)
3DMark Time Spy
3DMark doesn’t always reflect real-world results, but it’s still a handy method of estimating relative performance between graphics cards. First we see results from Time Spy, a moderately challenging 2560×1440 resolution test.
With an average graphics score of 17875 over our three test runs, the RTX 3080 FE provided a 66.3% increase over the RTX 2080 FE, and a 21.6% increase over the tested ASUS RTX 2080 Ti Strix OC.
3DMark Time Spy Extreme
Let’s raise the stakes, moving up to the more challenging 3840×2160 Time Spy Extreme benchmark.
Looking at the overall graphics score, the RTX 3080 FE shows a massive 77.7% increase over RTX 2080 FE, with the increase over the 2080 Ti Strix OC growing to 29.3%.
A very tough benchmark even at its “high” preset setting (with both ultra and extreme above that), we ran Metro Exodus using DirectX 12 at both 2560×1440 and 3840×2160.
Results at 1440p are first:
Hitting an average of 115.5 FPS across our three benchmark runs, the RTX 3080 FE shows a 47.3% increase over the RTX 2080 at these settings, with a more modest 15.49% increase over the RTX 2080 Ti Strix OC here.
Next we look at 3840×2160 (“4K”) results:
At 4K the RTX 3080 FE offers a 63.7% increase over the RTX 2080 FE, and a 26.2% increase over the RTX 2080 Ti Strix OC.
The Division 2
We ran Tom Clancy’s The Division 2 at its “high” preset setting using the DirectX 12 API option, at both 2560×1440 and 3840×2160.
We begin with results at 1440p:
At 2560×1440 frame rates with these cards can get quite high as you can see, but I wanted at least one example of something for the 144 Hz monitor users out there. The RTX 3080 FE averages nearly 175 FPS, which is an increase of 62.1% over the RTX 2080 FE, and 27.7% over the RTX 2080 Ti Strix OC.
Now a look at 3840×2160:
At 4K the RTX 3080 FE averages just over 100 FPS, which is an increase of 66.8% over the RTX 2080 FE, and 27.1% over the RTX 2080 Ti Strix OC.
Far Cry 5
Far Cry 5 is where I may have to rethink this Ryzen 3900X testbench – unless there’s some hidden 120 FPS frame limit I missed. Erroneous 1440/Ultra results aside (not published here), the game posed no issue at the GPU-bound 3840×2160 resolution.
At 4K the RTX 3080 FE averages just over under 95 FPS, which is an increase of 57.3% over the RTX 2080 FE, and 20.2% over the RTX 2080 Ti Strix OC.
Before moving on to RTX results, I wanted to add that the above selection of games leaves quite a bit to be desired. While we do get a picture of how the RTX 3080 FE stacks up against a handful of recent GPUs, the original list (including Shadow of the Tomb Raider and F1 2019) had to be pared down due to issues with some titles. These were either related to the new platform, or pre-release drivers.
For example, when I couldn’t get F1 2019 to load with the RX 5700 XT (using the latest Radeon driver) I knew something was wrong, but re-testing multiple games on different driver versions is quite laborious and these will have to be re-visited later on. The entire game benchmark suite is in need of an overhaul.
Performance Part 2: RTX On
The RTX series brought with it dedicated hardware for real-time ray tracing for the first time in any graphics card. And it has been quite popular to disparage Turing’s ray tracing support as being unnecessary, or even some sort of “scam” (preposterous but a great clickbait title – subsequently changed, of course).
It takes an objective look at the technology to appreciate what NVIDIA has been trying to do here, and while I doubt very many readers would agree that RTX made sense in 2018, the performance with the new RTX 3080 is so much more impressive that I think it’s time to take another look here in 2020.
Metro Exodus RTX
We’ll begin our look at RTX performance with a game from the previous benchmarks, Metro Exodus. This offers both NVIDIA RTX and DLSS support, though issues getting DLSS going on the 3080 with the benchmark tool prevented DLSS results today.
While the RTX 2080 Ti was able to average over 60 FPS in this game at 1440/RTX settings previously, the RTX 3080’s nearly 80 FPS was a notably smoother experience, with a 95th percentile average still up in the mid-50s.
Comparing the RTX 3080 to the RTX 2080 here, there was a 67.2% increase in performance with the new GPU at these settings, and a 25.3% improvement over the 2080 Ti Strix OC.
Bright Memory Infinite RTX Benchmark
Bright Memory Infinite is an in-development followup to Bright Memory that currently exists as an RTX-infused benchmark – and it’s stunning. It’s also the most challenging 2560×1440 test I’ve found to date at its highest preset, though a very impressive implementation of DLSS offers a significant boost when toggled.
Some extra attention has clearly been paid to make this demo look as good as possible, and I found DLSS to offer no noticeable quality loss – using the “quality” setting, appropriately enough. The charts display both non-DLSS and DLSS Quality results, each run at the highest “very high” RTX setting at 2560×1440.
All you have to do is look at the average FPS numbers to see just how punishing a fully ray-traced demo like this can be at its highest settings even at 2560×1440, though DLSS provides a significant improvement.
The gains from the 20 Series are massive here, with the RTX 3080 showing a 60.3% increase over the RTX 2080 Ti Strix OC card here, and a 108.9% increase over the RTX 2080. It appears that, depending on the scenario, you can indeed enjoy a 2x advantage over the RTX 2080 with this new GPU.
Here the RTX 3080 FE is a full 30 FPS faster with DLSS – again, at the “quality” level (with “balanced” and “performance” offering even faster results, of course). The RTX 2080 and 2080 Ti also show large gains using DLSS, but the 3080 is the only one to reach the 60 FPS mark.
Boundary Raytracing Benchmark
Another standalone benchmark from a game still in development, the Boundary: Raytracing Benchmark (YouTube link) is nonetheless a good test of DLSS-enabled RTX eye candy.
As you can see from the above chart, this benchmark poses less of a strain on the system compared to the previous test (Bright Memory Infinite).
Wolfenstein: Youngblood Optimizations
For our final look at RTX On performance with the new RTX 3080, we’re testing the card with a new beta build of Wolfenstein: Youngblood that takes advantage of the improved asynchronous compute capabilities of Ampere, and also makes use of the RTX 3080’s 2nd Gen RT cores (and the latest version of DLSS). Ray traced effects are limited to reflections in this title, but it’s still worth looking at as an example of how RTX performance can be improved with optimization.
Here are the results of this (Vulkan API) game running at 4K with the highest detail preset and RTX On, first without DLSS:
Once again we have a 2x performance jump over the RTX 2080 FE (well, an increase of 96.9% anyhow). The 45.7% increase over the RTX 2080 Ti is also quite impressive here.
And now with DLSS enabled:
Pretty consistent results here at the DLSS Quality setting as the RTX 3080 FE puts up another 96% increase over the 2080 FE. The 48.5% increase over the 2080 Ti is also very close to the non-DLSS result. The only real difference is the much higher FPS average, which is kind of the whole point of DLSS when you think about it.
I have to say, using only the “quality” preset – which uses the least aggressive render scaling – I wasn’t seeing any noticeable difference in quality in any of the benchmarks I ran here. I’m sure I could point things out using screenshots, but – subjectively, of course – in real time it looks no different.
PCI Express 4.0 Testing
The short version? The Gen3 vs. Gen4 question becomes much ado about nothing once the benchmarks are run. I didn’t see any difference outside of a normal run-to-run variance in my testing, and yes, I verified that the card was running in Gen3 mode after manually adjusting the settings in the motherboard setup.
GPUZ Screenshots Indicating PCIe Gen4 (left) and Gen3 (right) Interface
Of course there can be a measurable difference (I did see about 2 FPS higher using Gen4 with Metro Exodus at 1440/Ultra), but it’s very small. Until you break out the PCI Express Feature Test in 3DMark, that is!
RTX 3080 FE PCIe Gen4 result on the left, Gen3 result on the right
Results of the PCI Express Feature Test were 26.24 GB/s bandwidth using PCI Express 4.0, dropping to just 13.05 GB/s using PCI Express 3.0. I’m sure there will be applications that can leverage more bandwidth than the games and synthetics benchmarks I ran, but for gamers on current Gen3 Intel platforms there shouldn’t be a noticeable impact.
Power Consumption and Thermals
Leading up to today’s reviews much has been made about power consumption with the RTX 30 Series, and there’s no doubt that NVIDIA chose to pursue higher performance rather than reduced power consumption on this custom Samsung 8nm process.
But one thing is certain after spending the last week benchmarking the RTX 3080 Founders Edition: it’s not a particularly hot or loud card. In fact, it’s pretty cool and quiet. Granted, this was on an open test platform, and enclosure-based testing is planned.
As an example of both power draw and thermals, here are the results from a test involving 10 successive runs of the Metro Exodus benchmark at 4K/High settings using DX12:
Total GPU board power (as reported by GPUZ) topped out at 331.3W, with max temps of 77 C in a 22 C room. That board power measurement can’t be immediately validated via hardware (as I wait for GPU testing equipment), so I will add some trusty wall power measurements to compliment the software reporting here:
Two things stand out here. First, the RTX 3080 Founders Edition is using quite a bit more power than even a factory overclocked RTX 2080 Ti. Second, total system draw at the wall under a full 4K gaming load is still just 462 watts here, and considering the graphics horsepower you’re getting I don’t personally consider this a problem.
AMD fans will be quick to judge NVIDIA on power consumption with the RTX 30 Series, and I daresay the RTX 3080 will not end up being a darling of the small form-factor community. Still, I was expecting more power draw than 462W, which is largely a testament to AMD’s efficient Ryzen 3000 series CPUs. Still, the 750W PSU recommendation seems very safe, even if you have a CPU that consumes 200W on its own.
Ok, here we go. This is the topic that seemed to get more attention with the 20 Series than all of the performance metrics combined. It makes sense. The simple fact is that the cost of NVIDIA graphics cards rose sharply from Pascal (the GTX 10 series) to Turing (RTX 20 series).
Remember, the high-end GTX 1080 Ti launched at $699, which at that time was just $50 more than the outgoing GTX 980 Ti. In late 2018 the RTX 2080 Ti Founders Edition arrived at $1199, a $200 premium over the official (theoretical) $999 launch price. And partner models often commanded significantly more.
The jump from $699 to $1199 for the high-end consumer option was, to put it mildly, not particularly well received on the DIY side of the PC enthusiast community. And now a significant performance improvement over this $1200 card is launching for $699, which is the same launch price as the RTX 2080 and the GTX 1080 Ti.
Needless to say, I think pricing with the RTX 3080 is fantastic. Grumblings about the days of high-end sub-$500 graphics cards aside, I think the 3080 is a bargain at $699, and the upcoming RTX 3070 might just be the sweet spot at $499.
NVIDIA’s GeForce RTX 3080 Founders Edition is the fastest graphics card I’ve ever tested, and it’s an amazing product for the money. Now, actually buying one for $699 might require divine intervention, but we don’t really know until they go on sale.
We know demand will be there because performance is just so damn impressive with this card. No, the leap in performance isn’t 2x over the RTX 2080 outside of certain testing scenarios, but it’s always significant – often 60% or greater. The RTX 2080 was soundly beaten in these benchmarks by an average of 72% (!) between raster and raytraced tests (and the number moves to 76% if you count the beta Wolfenstein: Youngblood results).
I’ll be honest here. The RTX 2080 was a letdown. The Turing launch left a lot of gamers frustrated, and Pascal continued to be the architecture of choice for most GeForce gamers. RTX made for an awesome demo, but outside of a few titles that was about it. DLSS took time to improve, and without it full native rendering with real-time ray tracing was too expensive from a performance standpoint.
I feel like the ray tracing story has changed, if the RTX 3080 is any indication. Suddenly I’m really interested in games that use more RTX features, and excited about the prospect of the RTX 3090’s performance in this department. Frame rates – even without DLSS – are suddenly playable even at very high settings, and the visuals in some of the games and demos are stunning.
With the RTX 3080 we finally have a graphics card that redefines the $699 performance level in a way that eclipses even the GTX 1080 Ti. It’s an exciting time to be an enthusiast, and I can’t wait to get my hands on the RTX 3070 – and (fingers crossed) the RTX 3090 as well.
On the team red side of things AMD does not have anything that comes close to competing with the RTX 3080 – or at least not yet. We know that AMD is announcing new GPUs next month, and we’ve seen some of the leaks. It could be turn into a very interesting autumn, but NVIDIA has set the bar very, very high.
This disclosure statement covers the way the product being reviewed was obtained and the relationship between the product's manufacturer and PC Perspective.
How Product Was Obtained
The GPU is on loan from NVIDIA for the purpose of this review.
What Happens To Product After Review
The GPU remains the property of NVIDIA but will be on extended loan to PC Perspective for the purpose of future testing and product comparisons.
NVIDIA provided the product sample and technical brief to PC Perspective but had no control over the content of the review and was not consulted prior to publication.
PC Perspective Compensation
Neither PC Perspective nor any of its staff were paid or compensated in any way by NVIDIA for this review.
NVIDIA has not purchased advertising at PC Perspective during the past twelve months.
This article contains affiliate links to online retailers. PC Perspective may receive compensation for purchases made through those links.