NVIDIA recently launched the RTX 3080, the flagship gaming GPU in the RTX 3000 series of its Ampere family of graphics cards. The RTX 3080 is the first to launch from the much anticipated lineup, to be followed by the RTX 3090 on September 24th and the RTX 3070 in early October. For more details on NVIDIA’s RTX 3000 series, I recommend you read the blog that Patrick Moorhead and I wrote a couple of weeks ago.
For this review, we will mostly be comparing the RTX 3080 with the last generation’s flagship card, the RTX 2080. Additionally, we will look at the RTX 2080 Ti, the highest-end GeForce card. Once we get our hands on the RTX 3090, we’ll also compare it to the RTX 2080 Ti, but that will have to wait for another column. For now, let’s dive in to the RTX 3080.
For this review, we ran these benchmarks on an AMD Ryzen 9 3900X CPU, running on an ASRock X570 Taichi motherboard with 32GB of DDR4 3600 CL18 memory. The SSD in that system is a Corsair MP600 PCIe Gen4 SSD, which complements the motherboard’s PCIe Gen 4 capabilities. We opted for that chipset because the entire RTX 3000 series uses PCIe Gen 4. Additionally, we attached three monitors to the GPUs—two 144 Hz, 1440P monitors, and one 60hz 4K monitor. This setup allowed us to run multiple monitor resolutions for different games. We ran NVIDIA’s latest 456.16 driver, which came to us directly from NVIDIA for testing the RTX 3080. Additionally, we used NVIDIA’s frame capture tool, FrameView, to capture frame times and frame rates in games that didn’t have benchmarks.
We ran an array of UL Benchmarks’ synthetic graphics benchmarks, namely the 3DMark suite (including TimeSpy and TimeSpy Extreme, the company’s DirectX 12 benchmarks). Additionally, we ran the company’s ray tracing benchmark, Port Royale, which takes advantage of the GPU’s unique ray tracing hardware.
As you can see, the GeForce RTX 3080 is, on average, 50% faster than the RTX 2080 across three tests, both with and without ray tracing. In 3DMark TimeSpy, a DX12 benchmark, the RTX 3080 scores 15,065 points versus the RTX 2080’s 10,893, representing a 38% improvement. With TimeSpy Extreme, a more intensive version of the DX12 benchmark, the RTX 3080 scores 7,950 versus the RX 2080’s 5,254, representing a 51% improvement. Last but certainly not least is the Port Royale benchmark, which takes advantage of the RTX series’ ray-tracing capabilities. In Port Royale, the RTX 3080 scores 10,700 versus the RTX 2080’s 6,543, which translates to a performance improvement of 64%. NVIDIA’s RTX 3000 series shows some clear improvements with ray tracing, at higher graphics quality settings. These synthetic benchmarks generally tend to be a good baseline to compare other benchmarks to when looking at application performance.
Next is VRMark, another synthetic benchmark, with 3 tests designed to simulate the different tiers and resolutions of VR headsets. We only ran VRMark’s Blue Room test, since it simulates a 5Kx3K resolution VR headset—more along the lines of what RTX 3080 will be powering in the real world. That said, we cannot wait to run VR benchmarks with HP’s Reverb G2 when we get our hands on it.
As you can see in VRMark Blue Room, the RTX 3080 scores 5,256 points versus the RTX 2080’s 3532. This represents an increase of 49%, in line with the roughly 50% performance improvement we saw earlier with the high-resolution synthetic benchmarks.
For VFX and 3D rendering benchmarks, we used Blender and Octane. These are some of the most popular rendering tools out in the market today, and they already have quite well-optimized benchmarks for the Ampere series (particularly Octane). As you will see, this category showed the most significant improvement in performance. I suspect it may be only a taste of what’s to come for the RTX 3080’s rendering performance, as drivers and applications are further optimized. This test takes advantage of both the shader cores and the ray-tracing cores on the RTX GPUs and allows you to turn off ray tracing acceleration to see shader core render performance.
Assuming worse than real-world performance, Octane Bench is the company’s most conservative set of benchmarks. Here the RTX 3080 shows an almost 100%, or 2x, performance uplift over the RTX 2080.The RTX 3080 appears to take advantage of its new and improved RT cores, broadening the gap significantly with RTX turned on.
In Blender, we saw the RTX 3080 make huge strides against the RTX 2080 and RTX 2080 Ti across many scenes. In the Blender benchmark, lower is better because it translates to a shorter or quicker render time for each rendered scene. For example, in nearly every scene except for the Victor test scene, the RTX 3080 was at least 2x faster than the RTX 2080—further validating the 2x performance improvement we saw with Octane’s Render benchmark.
We set all games to maximum settings for our gaming benchmarks. None of the games were tested at 1080P because too many games at that resolution are CPU bound. The RTX 3080 is a powerhouse GPU and should be tested at 1440P max settings or 4K for most tests. Since most enthusiasts are still gaming at 1440P, we tested at both resolutions in different games. However, my 4K monitor is not a 120 Hz display, so I will have to wait until my Samsung Odyssey G9 arrives to test Ultrawide/5K @ 240 Hz on the RTX 3080.
One title immediately came to mind for testing when it came to the RTX 3080, and that was Microsoft Flight Simulator 2020, the unexpected hit of the year. The game is notoriously intensive, causing RTX 2080 and 2080 Ti GPUs to crawl on their knees at 4K. So, naturally, this was the perfect performance test for the RTX 3080. We ran all three GPUs at 4K and maximum graphical settings to understand the limitations of single card performance.
In Microsoft Flight Simulator 2020, the RTX 3080 delivered an average frame rate of 49 FPS with a minimum of 22; comparatively, the RTX 2080 had an average frame rate of 31 and a minimum of 17. In my experience, the RTX 3080 was the more playable experience, and with a 58% improvement in the average frame rate, that makes sense. The RTX 3080 easily handled Microsoft Flight Simulator 2020 in 4K and maximum quality settings, answering the question, “Can it Play Flight Simulator 2020?”
Call of Duty: Warzone
Warzone was probably one of the most challenging and most popular games to test the RTX 3080 on, mostly because it is an online game with gigantic maps and considerable variations in performance across them. However, people really want to know what kind of performance it delivers. Minimum and maximum frame rates were all over the place and really did not affect gameplay at all, so I focused mostly on the averages. Considering how buggy Warzone has been for me since it launched, it comes as no surprise that the RTX 3080 is only about 20% faster than the RTX 2080, and about 10% faster than the RTX 2080 Ti. That said, the RTX 3080 is the first time I’ve seen 150+ FPS, and consistent 120 and 130 FPS in-game. Still, until Activision optimizes more for the RTX 3080, the card will only bring marginal performance improvements.
Power & Thermals
NVIDIA put a significant amount of work into redesigning the stock cooler for the RTX 3080 Founder’s Edition (FE). The new RTX 3080 FE has two fans: one mounted on the bottom, to blow exhaust out of the back, and the other mounted at the top to pull air through and out the end of the case. In my experience, this GPU does run considerably quieter than my RTX 2080 Ti and idles at 37C. While gaming for an hour in Call of Duty Modern Warfare: Warzone, I never saw temps over 70C. It did hit 73C while playing Microsoft Flight Simulator 2020, but that is still quite good for a game that is known to bring most GPUs to their knees. In terms of power consumption, the RTX 3080 TDP is 320W, which the GPU should draw from the PSU at stock clocks. Unfortunately, I didn’t get my hands on NVIDIA’s power measurement tool, so GPU-Z was the only testing option for power consumption. GPU-Z reported 322W in Flight Simulator 2020, which is close enough to the 320W TDP. NVIDIA recommends a 750W PSU for RTX 3080 users, which makes sense if you account for the rest of the system, including the CPU, memory, motherboard, hard drives and other components.
One thing to consider with the new RTX 3080 is its increased TDP, from 250 to 320W, which is about 25%. However, the 25% increase in power consumption is somewhat justified by the performance improvements it brings. It will be interesting to see what the RTX 3090 delivers in terms of performance per watt since it has a 350W TD and should be roughly 20% faster than the RTX 3080. In addition to making the GPU more powerful, NVIDIA also made the RTX 3080 quieter. The GPU only draws 22W at idle and doesn’t even spin up the fans until the GPU is put under stress. Even when the fans do spin up, the RTX 3080 is quieter than the RTX 2080 Ti—a welcome improvement.
NVIDIA broadcast is the spiritual successor to RTX Voice. Broadcast adds the ability to do background blur and image insertion and provides a platform-agnostic noise cancellation tool for virtually any mic. In my experience, NVIDIA Broadcast works incredibly well, and I mostly use the noise cancellation tool. Many who have experienced this tool loves it deeply in spite of the fact that it still has some bugs. The most significant bug with RTX Voice was that the noise cancellation would go to static if you ever came out of sleep mode. It appears as though NVIDIA fixed that, but there are still some bugs that need working out. Regardless, I’ll keep using the app because my friends cannot stand the noise of my typing and fans without it.
Overall, the RTX 3000 series delivers impressive performance, especially when it comes to high-resolution gaming, VFX rendering and ray tracing. I believe that the 50% improvement in performance over the RTX 2080 shows that the RTX 3080 is indeed the generational improvement that people were expecting. NVIDIA also demonstrated that with the appropriate hardware, realtime ray tracing can run smoothly and look great on a single GPU. While it appears that there is still room for improvement in some titles, the RTX 3080 is the absolute peak of performance (until the RTX 3090 hits the market). One thing to consider for this review is that while the RTX 2080 Ti is a great GPU, it shouldn’t really be compared to the RTX 3080 for anything other than curiosity’s sake. The RTX 3090 will be a much better comparison to the 2080 Ti considering its pricing and where it sits in the stack.
Speaking of pricing, the pricing for the RTX 3080 at MSRP is fantastic for the performance that you get. The only issue is that obtaining one at that price will be virtually impossible for the foreseeable future. While nobody knows where retail pricing will settle, it’s quite clear that allocations are relatively scarce, and we will probably see shortages for weeks, if not months. NVIDIA has set the bar incredibly high for themselves and AMD, and it will be a matter of weeks until we hear AMD’s response. AMD’s response should be quite impressive when you consider where the 5700XT landed in terms of performance last year and will hopefully be competitive enough to continue to make NVIDIA respond.
NVIDIA hit it out of the park with the RTX 3080, and while some of the hype around the RTX 3000 series may have gone overboard, the overall improvements are still very impressive and welcome. The RTX 3080 will, without a doubt, be my default gaming GPU for the foreseeable future. I’m especially excited to throw my ultrawide Samsung Odyssey G9 at this GPU. While many people cannot wait to see what the RTX 3090 and 3070 can do, now that the 3080’s performance numbers are public, I expect that many gamers will be quite happy with the RTX 3080.