(Image credit: Tom’s Hardware)
Choosing the right graphics card is crucial for optimal performance in gaming, AI workloads like Stable Diffusion, and professional video editing. Your GPU is often the most significant factor determining system performance, overshadowing even the best CPUs for gaming. At compare.edu.vn, we understand the importance of making informed decisions when it comes to hardware. That’s why we provide comprehensive comparisons to help you navigate the complex world of PC components. In this article, we delve into the realm of graphics cards, offering a detailed comparison of various models to assist you in selecting the perfect GPU for your needs.
The GPU market saw a final wave of current-generation refreshes earlier this year. Nvidia introduced the RTX 4070 Super, RTX 4070 Ti Super, and RTX 4080 Super, while AMD launched the RX 7600 XT and the RX 7900 GRE in the US. Looking ahead, major shifts in the GPU hierarchy are anticipated with the arrival of Nvidia Blackwell RTX 50-series, Intel Battlemage, and AMD RDNA 4 GPUs. While these next-gen cards are widely expected in early 2025, some releases might still occur before the end of 2024, promising exciting advancements in graphics technology.
We are also in the process of updating our GPU testing methodology, incorporating new games and transitioning to a new testing platform. Following issues with our Core i9-13900K, we are considering the AMD Ryzen 7 9800X3D for our updated testbench, known for its exceptional gaming performance. This transition will necessitate a complete re-evaluation of all GPUs, and we’re currently determining the breadth of models to include in our future testing. Our latest reviews already utilize the 13900K testbed with an expanded game selection, and these results are incorporated in the performance charts featured below.
Our comprehensive GPU hierarchy is divided into two sections: traditional rendering (rasterization) and ray tracing benchmarks. The ray tracing hierarchy includes GPUs capable of this advanced rendering technique, specifically AMD’s RX 7000/6000 series, Intel’s Arc GPUs, and Nvidia’s RTX series. All benchmark results are obtained at native resolution, without enabling upscaling technologies like DLSS, FSR, or XeSS, or frame generation, ensuring a direct comparison of raw GPU power.
Nvidia’s current RTX 40-series GPUs are built upon the Ada Lovelace architecture, featuring innovations like DLSS 3 Frame Generation and DLSS 3.5 Ray Reconstruction (currently supported in select games). AMD’s RX 7000-series leverages the RDNA 3 architecture, offering a diverse range of seven desktop cards. Intel’s Arc Alchemist architecture marks a significant entry into the dedicated GPU market, positioning itself as a competitor, particularly in the midrange segment, against previous generation offerings.
For historical context, our 2020–2021 benchmark data, utilizing an older test suite on a Core i9-9900K platform, is available on page two of the original article for those interested in comparing older GPUs. Additionally, a legacy GPU hierarchy, sorted by theoretical performance without benchmark data, is provided for reference.
The subsequent tables rank graphics cards based purely on gaming benchmarks, conducted at 1080p “ultra” settings for the primary suite and 1080p “medium” for the DXR (DirectX Raytracing) suite. Factors such as price, graphics card power consumption, overall efficiency, and specific features are not considered in these performance rankings. Our current 2024 benchmarks are performed using an Alder Lake Core i9-12900K testbed. Let’s now examine the benchmark results and performance tables to Compare Gfx Cards in detail.
GPU Benchmarks Ranking 2025
Image 1 of 4
(Image credit: Tom’s Hardware)
For our most recent GPU benchmarks, we’ve rigorously tested nearly every graphics card released in the last seven years, along with select older models. These tests were conducted at 1080p medium and 1080p ultra settings, and the performance table is organized according to 1080p ultra results. Where relevant, we also include benchmark data at 1440p ultra and 4K ultra resolutions to provide a comprehensive performance overview. All scores are normalized relative to the top-performing card at 1080p ultra, which in our test suite is the RTX 4090, particularly evident at 4K and 1440p resolutions.
The summary chart above visually represents the relative performance of GPUs tested across several hardware generations at 1080p ultra. By swiping through the gallery, you can also view charts for 1080p medium, 1440p, and 4K ultra settings, offering a complete visual comparison of GPU performance. While some less common cards like the GT 1030, RX 550, and certain Titan models are excluded for clarity, the hierarchy is largely comprehensive. Detailed data for additional older GPUs is available in the tables below.
Our standard GPU benchmark suite consists of eight demanding games: Borderlands 3 (DX12), Far Cry 6 (DX12), Flight Simulator (DX11 Nvidia, DX12 AMD/Intel), Forza Horizon 5 (DX12), Horizon Zero Dawn (DX12), Red Dead Redemption 2 (Vulkan), Total War Warhammer 3 (DX11), and Watch Dogs Legion (DX12). The FPS score presented is the geometric mean of frame rates across these eight games, providing an equally weighted average performance metric. For detailed specifications of each GPU, refer to the links provided in the specifications column, which direct to our original in-depth reviews.
GPU Rasterization Hierarchy, Key Takeaways
Swipe to scroll horizontally
| Graphics Card | Lowest Price | 1080p Ultra | 1080p Medium | 1440p Ultra | 4K Ultra | Specifications (Links to Review) |
|---|---|---|---|---|---|---|
| GeForce RTX 4090 | $2,529 | 100.0% (154.1fps) | 100.0% (195.7fps) | 100.0% (146.1fps) | 100.0% (114.5fps) | AD102, 16384 shaders, 2520MHz, 24GB GDDR6X@21Gbps, 1008GB/s, 450W |
| Radeon RX 7900 XTX | $869 | 96.7% (149.0fps) | 97.2% (190.3fps) | 92.6% (135.3fps) | 83.1% (95.1fps) | Navi 31, 6144 shaders, 2500MHz, 24GB GDDR6@20Gbps, 960GB/s, 355W |
| GeForce RTX 4080 Super | No Stock | 96.2% (148.3fps) | 98.5% (192.7fps) | 91.0% (133.0fps) | 80.3% (91.9fps) | AD103, 10240 shaders, 2550MHz, 16GB GDDR6X@23Gbps, 736GB/s, 320W |
| GeForce RTX 4080 | $1,699 | 95.4% (147.0fps) | 98.1% (192.0fps) | 89.3% (130.4fps) | 78.0% (89.3fps) | AD103, 9728 shaders, 2505MHz, 16GB [email protected], 717GB/s, 320W |
| … | … | … | … | … | … | … |
The table continues with the rest of the GPUs as in the original article
While the RTX 4090 technically tops the charts at 1080p ultra, its dominance truly shines at 1440p and 4K resolutions. The performance difference between the RTX 4090 and RTX 4080 Super is minimal at 1080p ultra, less than 2%, but expands to 9% at 1440p and a significant 25% at 4K. It’s important to note that our FPS scores are calculated considering both average and minimum frame rates, with a greater emphasis on average FPS to reflect overall gameplay smoothness.
This rasterization hierarchy excludes ray tracing and DLSS benchmarks to ensure a level playing field across all generations of graphics cards. Since DLSS is exclusive to RTX cards (and DLSS 3 to RTX 40-series), including it would skew direct comparisons. For those interested in DLSS 2/3 and FSR 2 upscaling performance, refer to our detailed RTX 4070 review, which includes comprehensive upscaling benchmark results.
The RTX 4090 commands a premium price, yet its cost relative to the previous generation RTX 3090 is arguably justified by the substantial performance gains. The RTX 3090, at its launch, offered only incremental performance improvements over the RTX 3080, albeit with double the VRAM. Nvidia has maximized the RTX 4090’s capabilities, boosting core counts, clock speeds, and power limits to establish a clear performance lead. However, the RTX 4090 faces two main challenges: limited availability at MSRP due to high demand from the AI sector, often exceeding $2000 in price, and concerns regarding the 450W power draw through the 16-pin connector.
Stepping down from the top-tier RTX 4090, the RTX 4080 Super and RX 7900 XTX exhibit a closer performance competition, particularly at higher resolutions, while CPU limitations become more apparent at 1080p. We are preparing to transition to a new testbed in the near future. The current results presented here utilize our 13900K test system and are shown in the charts at the end of this section for your reference.
(Image credit: Intel)
Beyond the latest offerings from AMD and Nvidia, the RX 6000 and RTX 30 series GPUs remain viable options, delivering commendable performance. If you currently own a card from these series, upgrading might not be immediately necessary. Intel’s Arc GPUs also fall into this performance bracket and represent an intriguing wildcard in the GPU landscape.
Our ongoing testing and driver updates for Arc GPUs have resolved previous inconsistencies, allowing them to complete our full benchmark suite without major issues. While Arc GPUs may not excel in power efficiency, the A750, in particular, offers a compelling balance of performance and price.
Looking at older generations, the RTX 20-series and GTX 16-series GPUs, along with the RX 5000-series, are positioned throughout the performance hierarchy. Generally, newer architectures offer a performance uplift equivalent to one or two “model upgrades.” For example, the RTX 2080 Super performs comparably to the RTX 3060 Ti, while the RX 5700 XT closely matches the newer, more affordable RX 6600 XT.
Older graphics cards with limited VRAM, particularly those with 4GB or less, struggle significantly with modern games at ultra settings. We have consistently advised against GPUs with 4GB VRAM for several years, and today, we recommend a minimum of 8GB VRAM for comfortable gaming. For mainstream GPUs, 12GB or more is preferable, and for high-end cards and above, 16GB or more is increasingly becoming the standard. Older cards like the GTX 1060 3GB and GTX 1050 encountered issues running some of our tests, potentially skewing their results, even though they perform better at 1080p medium settings.
Now, let’s shift our focus to the ray tracing performance hierarchy and compare gfx cards in this demanding rendering domain.
(Image credit: Techland)
Ray Tracing GPU Benchmarks Ranking 2025
Enabling ray tracing, especially in graphically intensive games within our DXR test suite, can dramatically reduce frame rates. Our ray tracing benchmarks are conducted using “medium” and “ultra” settings. “Medium” generally utilizes medium graphics presets with ray tracing effects enabled (set to “medium” if available, otherwise “on”). “Ultra” settings maximize all ray tracing options to achieve the highest visual fidelity.
Due to the substantial performance demands of ray tracing, we’ve organized these results based on 1080p medium scores. This is also because entry-level ray tracing capable cards like the RX 6500 XT, RX 6400, and Arc A380 struggle to handle ray tracing even at these settings. Testing beyond 1080p medium would yield largely unplayable frame rates for these lower-end cards. However, we have included 1080p ultra, 1440p Ultra and 4K Ultra results for a more complete picture in the charts at the end of this section.
Our ray tracing benchmark suite comprises five games that heavily utilize the DirectX 12 / DX12 Ultimate API: Bright Memory Infinite, Control Ultimate Edition, Cyberpunk 2077, Metro Exodus Enhanced, and Minecraft. The FPS score is the geometric mean of frame rates across these five games, and the performance percentage is scaled relative to the fastest GPU in this test, which is again the GeForce RTX 4090.
To glimpse the future of ray tracing, explore our Alan Wake 2 benchmarks. This game’s full path tracing implementation pushes even high-end GPUs to their limits, requiring upscaling to achieve playable performance, even on non-Nvidia cards. However, it’s crucial to acknowledge that games where ray tracing fundamentally transforms visual quality remain relatively scarce. For most games, traditional rasterization rendering still offers a more practical balance of performance and visual fidelity.
Image 1 of 4
(Image credit: Tom’s Hardware)
GPU Ray Tracing Hierarchy, Key Takeaways
Swipe to scroll horizontally
| Graphics Card | Lowest Price | 1080p Medium | 1080p Ultra | 1440p Ultra | 4K Ultra | Specifications (Links to Review) |
|---|---|---|---|---|---|---|
| GeForce RTX 4090 | $2,643 | 100.0% (165.9fps) | 100.0% (136.3fps) | 100.0% (103.9fps) | 100.0% (55.9fps) | AD102, 16384 shaders, 2520MHz, 24GB GDDR6X@21Gbps, 1008GB/s, 450W |
| GeForce RTX 4080 Super | No Stock | 86.8% (144.0fps) | 85.3% (116.3fps) | 75.6% (78.6fps) | 70.5% (39.4fps) | AD103, 10240 shaders, 2550MHz, 16GB GDDR6X@23Gbps, 736GB/s, 320W |
| GeForce RTX 4080 | $1,725 | 85.4% (141.6fps) | 83.4% (113.6fps) | 73.1% (76.0fps) | 67.7% (37.8fps) | AD103, 9728 shaders, 2505MHz, 16GB [email protected], 717GB/s, 320W |
| GeForce RTX 4070 Ti Super | $819 | 77.3% (128.2fps) | 73.5% (100.3fps) | 63.5% (66.0fps) | 58.4% (32.6fps) | AD103, 8448 shaders, 2610MHz, 16GB GDDR6X@21Gbps, 672GB/s, 285W |
| … | … | … | … | … | … | … |
The table continues with the rest of the GPUs as in the original article
If you found the RTX 4090’s 4K rasterization performance impressive, its ray tracing capabilities are even more striking. Nvidia has significantly enhanced ray tracing performance in its Ada Lovelace architecture, becoming evident in these benchmarks. Further performance gains in ray tracing are anticipated with advancements like Shader Execution Reordering (SER), Opacity Micromaps (OMM), and Displaced Microgeometry Meshes (DMM), alongside DLSS 3. However, DLSS 3, while boosting frame rates, introduces a mixed experience due to generated frames not incorporating new user input and adding latency.
For a glimpse into extreme ray tracing scenarios, we also tested high-end ray tracing GPUs with Cyberpunk 2077‘s RT Overdrive mode and Alan Wake 2, both implementing full path tracing. Black Myth: Wukong also supports full ray tracing, indicating a trend toward more demanding ray-traced gaming experiences. These games highlight the future direction of gaming visuals and underscore the increasing importance of upscaling and AI-driven techniques like frame generation to maintain playable frame rates.
Even at 1080p medium, a relatively moderate setting for DirectX Raytracing (DXR), the RTX 4090 outperforms all competitors, leading the previous generation RTX 3090 Ti by a substantial 41%. At 1080p ultra, this lead expands to 53%, and at 1440p, it reaches nearly 64%. Nvidia’s pre-launch claims of the RTX 4090 being “2x to 4x faster than the RTX 3090 Ti,” factoring in DLSS 3 Frame Generation, are supported by our findings. Even without DLSS 3, the RTX 4090 demonstrates a 72% performance advantage over the 3090 Ti at 4K resolution.
AMD’s approach to DXR and ray tracing remains secondary, prioritizing rasterization performance improvements and cost reduction through chiplet designs in their RDNA 3 GPUs. Consequently, AMD’s ray tracing performance lags behind Nvidia’s. The top-performing RX 7900 XTX roughly matches Nvidia’s previous-generation RTX 3080 12GB, placing it just ahead of the RTX 4070 in ray tracing benchmarks, and this performance parity is not consistent across all DXR games. While RDNA 3 introduces some ray tracing performance enhancements, exemplified by the RX 7800 XT’s 10% DXR performance lead over the RX 6800 XT (despite similar rasterization performance), AMD still trails Nvidia significantly in ray tracing capabilities.
Intel’s Arc A7-series GPUs present a balanced performance profile, with the A750 outperforming the RTX 3060 overall. With updated drivers, and vsync disabled, Minecraft performance on Arc GPUs now aligns more consistently with other DXR benchmark results.
(Image credit: Tom’s Hardware)
Our RTX 4090 review details the performance gains achieved with DLSS Quality mode in DXR games, summarizing a 78% performance increase at 4K ultra. DLSS 3 frame generation further boosted frame rates by 30% to 100% in our tests, though we advise caution when interpreting FPS figures with frame generation enabled due to potential latency and perceived smoothness issues in actual gameplay.
Overall, with DLSS 2 enabled, the RTX 4090 in our ray tracing test suite achieves nearly four times the performance of AMD’s RX 7900 XTX. AMD’s FSR 2 and FSR 3 technologies offer upscaling and frame generation alternatives, and AMD is actively increasing their adoption rate. However, FSR still lags behind DLSS in both game support and overall image quality. Only two games in our DXR test suite currently support FSR2, while all DXR games tested support DLSS2, with one also supporting DLSS3.
Without FSR2 or DLSS, AMD’s fastest GPUs can only achieve around 60 fps at 1080p ultra ray tracing settings, maintaining decent playability at 1440p with average frame rates of 40–50 fps. However, native 4K DXR gaming remains challenging for most GPUs. Only the RTX 3090 Ti and higher-end cards surpass the 30 fps threshold in our composite score, and even the 3090 Ti falls short in certain individual games at 4K DXR settings.
AMD’s FSR 3 frame generation, similar to DLSS3, introduces latency and necessitates Anti-Lag+ integration in games. Anti-Lag+ is exclusive to AMD GPUs, potentially resulting in higher latency penalties on non-AMD cards. While FSR 3 has shown promise in games like Avatar: Frontiers of Pandora, its quality and latency characteristics remain inconsistent across different titles, performing well in some and less effectively in others like Forspoken and Immortals of Aveum.
Midrange GPUs like the RTX 3070 and RX 6700 XT are generally capable of 1080p ultra ray tracing, but struggle beyond that resolution. Entry-level DXR-capable GPUs barely manage 1080p medium settings, with the RX 6500 XT falling short even at these reduced settings, exhibiting single-digit frame rates in most tests and failing to run one game (Control) at medium ray tracing settings due to VRAM limitations (requiring at least 6GB VRAM for ray tracing in Control).
Interestingly, Intel’s Arc A380 outperforms the RX 6500 XT in ray tracing, despite having fewer Ray Tracing Units (RTUs) – 8 compared to AMD’s 16 Ray Accelerators. Intel’s detailed analysis of its ray tracing hardware highlights the Arc architecture’s potential. However, the limited number of RTUs on even the top-end Arc A770 (32 RTUs) restricts its ray tracing performance ceiling. The Arc A770 narrowly surpasses the RTX 3060 in DXR benchmarks, while Arc A750 and above outperform AMD’s RX 6750 XT in DXR, revealing the relative ray tracing weakness of AMD’s RDNA 2 architecture.
Comparing generational performance within Nvidia’s RTX lineup, the older RTX 2060 slightly outperforms the newer RTX 3050. However, the top-end RTX 2080 Ti falls behind the RTX 3070. While the RTX 2080 Ti doubled the performance of the RTX 2060, the RTX 3090 delivers approximately triple the performance of the RTX 3050, illustrating significant generational performance scaling.
(Image credit: Tom’s Hardware)
Test System and How We Test for GPU Benchmarks
Our GPU benchmarks are conducted using multiple test systems. Our latest 2022–2024 configuration utilizes an Alder Lake CPU and platform, while our previous testbed was based on Coffee Lake and Z390. The most recent benchmark charts (below) are generated using a Core i9-13900K system with an updated game selection. Detailed specifications of our test PCs are provided below.
Tom’s Hardware 2022–2024 GPU Testbed
Intel Core i9-12900K
MSI Pro Z690-A WiFi DDR4
Corsair 2x16GB DDR4-3600 CL16
Crucial P5 Plus 2TB
Cooler Master MWE 1250 V2 Gold
Cooler Master PL360 Flux
Cooler Master HAF500
Windows 11 Pro 64-bit
Tom’s Hardware 2020–2021 GPU Testbed
Intel Core i9-9900K
Corsair H150i Pro RGB
MSI MEG Z390 Ace
Corsair 2x16GB DDR4-3200
XPG SX8200 Pro 2TB
Windows 10 Pro (21H1)
Our testing methodology remains consistent across all graphics cards. We perform an initial benchmark run to warm up the GPU after game launch, followed by at least two benchmark passes at each resolution and setting combination. If the results of two consecutive runs are nearly identical (within 0.5% variation), we record the faster run. Larger discrepancies prompt additional test runs to establish a consistent performance baseline.
We meticulously analyze all benchmark data, identifying and investigating any anomalies. For instance, GPUs within the RTX 3070 Ti, RTX 3070, and RTX 3060 Ti performance tier should exhibit predictable performance deltas (e.g., RTX 3070 Ti ~5% faster than RTX 3070, which is ~5% faster than RTX 3060 Ti). Significant deviations (>10%) trigger retesting to ascertain accurate performance figures.
Given the time-intensive nature of GPU testing, driver updates and game patches inevitably emerge, potentially influencing benchmark results. We routinely re-evaluate a subset of GPUs to ensure the continued validity of our data. Significant discrepancies necessitate retesting affected games and GPUs. We may also incorporate new game benchmarks into our test suite throughout the year, selecting popular and benchmark-friendly titles based on predefined criteria.
GPU Benchmarks: Individual Game Charts
While the preceding tables provide summarized performance metrics, detailed individual game charts for both standard and ray tracing benchmarks are available for users seeking granular performance data. To maintain chart clarity, these individual game charts primarily feature recent GPUs. These charts utilize our new test PC, which may result in slight performance variations compared to the summary tables. The individual game charts reflect our most current benchmark data, although comprehensive testing across older GPUs for these updated charts is ongoing.
These charts are up to date as of November 11, 2024.
GPU Benchmarks — 1080p Medium
Image 1 of 22
(Image credit: Tom’s Hardware)
GPU Benchmarks — 1080p Ultra
Image 1 of 22
(Image credit: Tom’s Hardware)
GPU Benchmarks — 1440p Ultra
Image 1 of 22
(Image credit: Tom’s Hardware)
GPU Benchmarks — 4K Ultra
Image 1 of 22
(Image credit: Tom’s Hardware)
GPU Benchmarks — Power, Clocks, and Temperatures
Beyond raw performance, power consumption, clock speeds, and temperatures are crucial considerations when comparing gfx cards. The following charts provide detailed insights into these aspects for various GPUs.
Image 1 of 4
(Image credit: Tom’s Hardware)
Image 1 of 4
(Image credit: Tom’s Hardware)
Image 1 of 4
(Image credit: Tom’s Hardware)
If you are interested in the legacy GPU hierarchy, please refer to page two of this article. We have relocated it to a separate page to optimize loading times for both our content management system and the main website. For discussions and feedback on our GPU benchmarks hierarchy, we encourage you to visit our forums and join the ongoing conversation.
Choosing a Graphics Card
Which graphics card is right for you? Our comprehensive GPU benchmarks hierarchy, encompassing numerous GPUs from the past four generations, is designed to assist you in making an informed decision. Unsurprisingly, the fastest cards originate from Nvidia’s latest Ada Lovelace and AMD’s RDNA 3 architectures. AMD GPUs excel in rasterization performance but tend to lag behind Nvidia when ray tracing is enabled, particularly with DLSS. However, with GPU prices normalizing, now is an opportune time to upgrade your graphics card.
Gaming is not the sole determinant when choosing a GPU. Many professional applications leverage GPU acceleration, and we provide professional GPU benchmarks in our full GPU reviews. Generally, a strong gaming GPU will also perform well in GPU-intensive computational tasks. Opting for a high-end card ensures smooth high-resolution, high-frame-rate gaming with maximum visual settings and ample performance for content creation workloads. Lower-tier GPUs necessitate dialing down settings to achieve acceptable performance in modern games and benchmarks.
For gaming-centric builds, the CPU remains a critical component. Even the most powerful GPU will be bottlenecked by an underpowered or outdated CPU. Consult our Best CPUs for Gaming guide and CPU Benchmarks Hierarchy to ensure a balanced system configuration tailored to your desired gaming experience.
- 1
- 2
Current page: GPU Benchmarks Hierarchy 2025
Next Page 2020-2021 and Legacy GPU Benchmarks Hierarchy
Stay On the Cutting Edge: Get the Tom’s Hardware Newsletter
Get Tom’s Hardware’s best news and in-depth reviews delivered directly to your inbox.
TOPICS
Jarred Walton
Jarred Walton is a senior editor at Tom’s Hardware, focusing on all aspects of GPUs. With a tech journalism career spanning back to 2004, including contributions to AnandTech, Maximum PC, and PC Gamer, Jarred is a seasoned expert in graphics technology. From early 3D accelerators to modern GPUs, Jarred stays at the forefront of graphics trends and is your go-to source for game performance analysis.
