ASRock Radeon RX 9070 Steel Legend Review: AMD’s Best Mid-Range Card Yet?

AMD has launched the Radeon RX 9070 and RX 9070 XT cards on their RDNA 4 architecture. Unlike the previous two generations, AMD is not looking to compete with Nvidia’s higher-end (RTX 5080 and beyond) GPU segment. Yet, they have marketed “4K performance at 1440P price,” among other things. We’ve received the Steel Legend OC variant of the RX 9070 from AMD, and having tested various Steel Legend products in the near past, we’re excited to take a look at this one!

Like the motherboards, the Steel Legend cards from ASRock are more budget-oriented and feature unique aesthetics under a white theme. They also feature modestly increased boost clock speeds over the reference cards. ASRock has previously released Steel Legend variants for Intel’s Battlemage and AMD’s RX 7000 cards.

---

Key Takeaways

1. The RX 9070 is faster than the RTX 5070 in regular gaming at all resolutions and beats it in most ray-traced games. It has great overclocking potential, boosting performance by up to 10% in Cyberpunk 2077. Plus, thanks to its 16GB of VRAM, it also delivers a smoother gaming experience, especially in the lowest frame rates, making it a better choice than the competition
2. As such, you should buy the ASRock RX 9070 Steel Legend if want excellent rasterization performance at both 1440p and 4K, and you are not very nit-picky about having the best RT performance.
3. The only reason to skip this card is if its costing you over $600, because then you should consider the RX 9070 XT instead.

---

Here are the specifications, compared with some competing GPUs.

Product	ASRock Radeon RX 9070 Steel Legend OC	Nvidia GeForce RTX 5070	Nvidia GeForce RTX 4070 Super	AMD Radeon RX 7900 GRE
GPU Die	Navi 48	GB205	AD104	Navi 31 XL
Manufacturing Process	5nm	5nm	5nm	5nm GPU, 6nm MCD
Base Clock	1330 MHz	2325 MHz	1980 MHz	1287 MHz
Boost Clock	2700 MHz	2512 MHz	2475 MHz	2245 MHz
Memory	16 GB GDDR6	12 GB GDDR7	12 GB GDDR6X	16 GB GDDR6
Bus Width	256-bit	192-bit	192-bit	256-bit
Shaders	3584	6144	7168	5120
TMUs	224	192	224	320
ROPs	128	80	80	160
Tensor Cores | AI Accelerators	112	192	224	160
Ray Tracing Cores | Ray Accelerators	56	48	56	80
L1 Cache	128KB (per Array)	128 KB (per SM)	128 KB (per SM)	256 KB (per Array)
Shared Cache	64 MB (L3)	48 MB (L2)	48 MB (L2)	64 MB (L3)
TDP	220W	250W	220W	260W
MSRP	– (Reference Edition: $550)	$550	$600	$550

With a $550 price tag, the GPU is comparable to Nvidia’s recently released GeForce RTX 5070. Previous-generation GPUs that can expect competition include the RTX 4070 (including Super and Ti models). AMD also marketed up to 42% faster gaming performance (raster and ray tracing combined) than the 7900 GRE at 4K ultra.

Notably, AMD is also using an updated naming convention to represent which Nvidia card their models compete with. So, while we couldn’t expect the RX 7800 to compete with the RTX 4080, we can expect the RX 9070 to compete with the RTX 5070. We’ll compare their performances to really see.

Let’s first look at the new architecture.

Architecture

AMD marketed some big performance gains, and it’s all thanks to the RDNA 4 architecture. The RX 9070, like the XT card, uses the Navi 48 GPU with a diameter of 357 mm², packing 53.9b transistors inside. It uses an optimized 5nm process called N4C, similar to Nvidia’s 4N. The transistor count is much greater thanks to this new node, with only about 2.8b fewer transistors despite a much smaller die.

Following Nvidia, AMD has also added a PCIe 5.0 interface to the cards. The memory used is still GDDR6, probably to keep the prices competitive. The memory clock on the card is set to 1330 MHz for an effective memory clock of 20.1 Gbps. This gives us a 644.6 GB/s bandwidth, which can’t compete with Nvidia’s GDDR7 cards, but the large L3 cache chunk should somewhat make up for it.

Improvements

1. Vector Units: The new CU consists of Dual SIMD32 vector units, one set capable of FMA/INT and the other capable of FMA only.
2. AI Accelerators: AMD’s AI cores are capable of faster matrix operations with 2x 16-bit or 4x 8-bit/4-bit matrix rates. Support for the 8-bit floating point format types is a new introduction. Moreover, the 4:2 Structured Sparsity enables up to twice the speed/performance and lower power consumption. Wave Multiply Matrix Accumulate (WMMA) operations are optimized for better performance.
3. Scalar Unit: The new scalar unit supports the new Float32 ops.
4. Scheduler: The Scheduler responsible for handling the different workloads features Split and Named barriers and shows faster split/fill operations. Prefetching has also been tweaked.
5. Ray Tracing Core: Improvements to the Ray Accelerator include twice as many box and triangle intersection units, better stack management, and improved node-compression to reduce BVH size. The 8-wide BVH lowers memory utilization. Lastly, shader performance has been improved thanks to additional out-of-order queues for memory, leading to out-of-order requests from different shaders to be delivered.
6. Dynamic Registers: AMD has introduced Dynamic Registers, allowing the GPU shaders to allocate them based on load. This reduces the load on the shaders and leads to better handling of memory latency.
7. Media: The GPU’s encoding units show improved performance, with a 25% and 11% improvement in H.264 low latency and HEVC encode quality. AV1 encoding efficiency is improved thanks to B-frames, which use data from preceding and following frames to build the current frame. The context switch overhead and memory write accesses have been reduced for overall improved performance.
8. Display: AMD also introduced hardware flip-metering support, like the RTX 5000 cards, for improved frame pacing. This will also reduce CPU power consumption during video playback. Not only that, idle power consumption for dual-monitor configurations has been reduced. Lastly, the Display Engine tweaks have also enabled Radeon Image Sharpening 2 for better sharpening.

FSR 4 

AMD introduced FSR 4 with RX 9000, and the major change here is an AI-based upscaling technique for improved image quality. This involves custom machine-learning models for different games that use details from these models to upscale low-resolution video game renders. AMD displayed some comparison images in its presentation, with FSR 4 on Performance Mode showing strikingly similar results to native 4K rendering.

Combined with Anti-Lag, FSR 4 can provide higher frame rates and quality with low latency. AMD also mentioned that FSR 4 is ready for Neural Rendering. We briefly discussed how neural rendering works in our recent RTX 5000 reviews. In short, this is a deep learning technique that uses neural networks to apply different rendering techniques. It has the potential to improve image quality drastically.

Fluid Motion Frames, Anti-Lag 2, and AMD Adrenalin

AMD also updated Fluid Motion Frames to its latest iteration: FMF 2.1. This is AMD’s frame-generation technology (similar to Nvidia’s Frame-Generation), which shows improved image quality and temporal tracking with less ghosting.

Anti-Lag 2 was released earlier, but AMD marketed better game support. The latest upcoming one is Apex Legends. In this title, AMD’s figures promise 42% lower latency than applications without Anti-Lag.

The latest edition of AMD’s Adrenalin Software, 25.3.1, is out, adding new features. These include AMD Image Inspector for auto-reporting bugs, artifacts, or other game issues. An AI chatbot called AMD Chat helps you navigate the software. AMD’s AI Apps Manager and Install Manager are some other software components that AMD has advertised.

Packaging

The card is shipped inside a greyish and blacked-themed cardboard box. It looks classy.

The front does not have much more than the card’s name and the VRAM amount. On the back, we get key features and requirements for the card. We also have compliance symbols and certification marks.

Design

The ASRock Steel features the familiar white theme of Steel Legend products. The front features a fairly minimalistic plastic shroud, mixing grey and white and featuring some printed patterns. This is a triple-fan design, and the fans feature ARGB that can be synced with ASRock motherboards through the Polychrome Sync software.

The left and right-end fans feature the Steel Legend logo on the fan hub and the ASRock branding on the middle fan. These are ASRock’s highly-curved Striped Ring fans designed for greater lateral intake and better overall airflow. The fans have temperature-dependent auto-control and won’t spin at low temperatures.

The backside shows a metallic backplate with the Steel Legend logo in the center and some symmetrical shapes printed around it. Again, it’s a very nice and simplistic look. We have the GPU support bracket on the left side (when viewed from the above angle) and three cutouts on the right for heat dissipation. For the same purpose, we see more space left open for heat dissipation on the upper side. Here, we also spot the ARGB Header text and directions for switching on/off the onboard LED. We’ll look at it shortly.

On the top, we first notice an RGB zone on the lower left-hand side. It glows from underneath, while the Steel Legend text and logos themselves do not.

Moving toward the right, we have the LED switch on the upper side. Under it, we see the 3-pin ARGB header to sync the RGB with your supporting ASRock motherboard if you have one. Toward the center, we spot the dual 8-pin PCIe connectors. Each outputs 150W, while the card is rated at just 220W.

We can also see the GPU heatsink here. The GPU’s cooling design uses a nickel-plated copper base, the heat of which is carried by the five heat pipes to the copper heatsink. The three fans then dissipate this heat.

On the right side of the GPU, we can see the five heat pipes and V-shaped air vents contributing to better heat dissipation. With most graphics cards, this region is completely covered by the GPU shroud, but that’s not the case here. On the left side are the four output ports: 3x DisplayPort 2.1a and one HDMI 2.1b. This is a triple-slot design over the dual-slot design of the Reference Edition.

Test Setup

• CPU: AMD Ryzen 7 9700X, PBO Tuned, TDP 170W, Curve Optimizer -5
• RAM: XPG Lancer 16×2 6400MT/s CL28 (Tuned)
• Motherboard: MSI B650 TOMAHAWK WIFI
• Cooler: Deepcool Mystique 360 Liquid Cooler (ARGB fans)
• GPU: ZOTAC RTX 5080 | Colorful iGame Ultra RTX 5070 | ASUS Dual RTX 4070 Super | ASRock RX 9070 Steel Legend OC
• BIOS: 7D75v1L AGESA ComboAM5 1.2.0.2b
• Drivers: GeForce Game Ready 572.65 (RTX 5080, RTX 4070 Super), AMD Adrenalin-Edition-24.30.31.03 (RX 9070)
• Windows Version: Windows 11 24H2 (OS Build 26100.3194)
• Storage: WD SN 770 1TB, Samsung 970 EVO 500GB for OS
• Benchmarks: CapframeX v1.7.4, Nvidia Frame View v1.6.10, HWINFO Pro v8.22

Testing Methodology

Testing Specifications and Setup:

• All games and GPUs were tested using the specifications, drivers, and software listed above.
• All GPUs were tested using the same game version.
• Gaming performance was first evaluated at the native resolution without utilizing upscalers or frame-generation technologies.

Upscaling and Anti-Aliasing:

• For games that lack TAA (Temporal Anti-Aliasing) or any game engine resolution upscalers (such as TAA or TSR), FSR Native was used. This is because FSR is universally available and compatible with all GPUs. An example of such a game is Alan Wake II.

Settings Configuration:

• No control panel settings were used. All configurations were applied directly within the game settings.

Performance and Measurement Tools:

• CapframeX was used to capture performance metrics.
• Nvidia Frame View was utilized for latency measurement in Frame-Generation + DLSS tests for RTX GeForce Graphics Series
• HWINFO Pro was employed to measure effective clocks, average power consumption, and maximum power consumption.
• NvAPP: We have completely disabled the NVIDIA shadow-play for the RTX GeForce Graphics Series.
  

Benchmarking Methods:

• Some games were tested using their built-in benchmarks, while others were tested with custom game settings.

GPU Configuration:

• No GPUs were tuned, overclocked, or had their power limits increased during testing.

Screen Resolutions Tested:

• FHD (Full HD): 1920×1080
• QHD (Quad HD): 2560x1440p
• UHD (Ultra HD): 3840x2160p

Rasterization Performance

Gaming 1080p Benchmarks (Native)

At our raster 1080p native performance numbers, Radeon RX 9070 is leading the newly released NVIDIA’s GeForce RTX 5070 by 14% on Average, and almost the 20% lead over NVIDIA’s GeForce RTX 4070 Super, NVIDIA GPU’s nearly caught up to Radeon RX 9070 in Black Myth Wukong but surprisingly the RDNA 4 GPU is beating the GeForce RTX 5070 by 7% on Average at 1080p. However, if you’re still at the 1080p High-Refresh Rate display, such as 144Hz, RX 9070 can efficiently run without sweat at the maximum hertz in the maxed-out settings. It can deliver smooth gaming performance without lowering the game settings.

Gaming 1440p Benchmarks (Native)

At our raster 1440p performance, RX 9070 is 16% Faster than NVIDIA’s GeForce RTX 5070 across seven titles that we have tested at 1440p, and 23% faster than NVIDIA’s GeForce RTX 4070 Super, RX 9070 averages around 106 FPS in all the titles at the QHD 1440p resolution on the Ultra Settings. In Alan Wake II, RX 9070 posts the strong 31% lead; this title works best on the RDNA architecture; in Monster Hunter world, the lead was around 25%; in Cyberpunk, the AMD GPU was 12% faster; overall, AMD enjoys the great victory at 1440p resolution on ultra-settings.

Gaming 2160p Benchmarks (Native)

At our raster 2160p Performance, RX 9070 is whoopingly 19% Faster here. The gap keeps increasing as we increase the resolution; the extra video memory on the RX 9070 is pumping the numbers up; in Alan Wake II, RX 9070 enjoys a solid lead of 26% over NVIDIA’s GeForce RTX 5070; the closest gap was in Black Myth Wukong where RX 9070 is still leading, but by the tiny margin 7%, which is still a victory for the Radeon Camp, However, Monster Hunter Wilds shows the double-digit performance gains over NVIDIA at whooping 25% performance gain, RX 9070 can comfortably run the game at 2160p Ultra Settings on Raster.

Ray Tracing Performance

Gaming 1080p Benchmarks (RT)

Now, here comes the RTX; AMD wins in our ray-tracing but by a slim margin; in our raster benchmarks at 1080p, AMD was beating NVIDIA by 14%; however, in RTX, things are getting interesting; RX 9070 is only 1.19% Faster than NVIDIA’s GeForce RTX 5070, In Alan Wake II, we did not enable the path-tracing direct lightning, we have only enabled the RT lightning and RT denoiser, and RT transparency, RX 9070 is beating the NVIDIA’s GeForce RTX 5070 by just only 6%.

You might not see the double-digit gain here (except for Monster Hunter Wilds) as we have seen on the raster, but overall, it is the victory for AMD’s camp. However, The Black Myth Wukong is completely broken on the AMD hardware, and NVIDIA’s GeForce RTX 5070 enjoys the massive victory here by the 86% lead. However, this is the only game where AMD is struggling; in Cyberpunk 2077, NVIDIA’s GeForce outperforms the RX 9070 by just 4%.

However, AMD can still provide a smooth gaming experience at 1080p. In Starwars Outlaw, we have enabled the 6 RT settings in the advanced settings; we are not using RTX Direct Lightning in the comparison. However, NVIDIA’s GeForce RTX 5070 was still 1 FPS faster than RX 9070, but the 1% lows on the AMD camp are 3 FPS better; RX 9070 can deliver the smooth ray-tracing gaming experience at 1080p (Except for the Black Myth Wukong).

Gaming 1440p Benchmarks (RT)

At 1440p Ray-Tracing Benchmarks, in overall games, RX 9070 posts the identical results as NVIDIA’s GeForce RTX 5070, averaging 61 FPS; however, RTX 4070 Super faces the defeat here by 9%, Black Myth Wukong still suffers on the RX 9070 at 1440p Ray-Tracing, A small victory in the Starwars Outlaw by just 2 FPS vs. the GeForce RTX 5070, The Extra Video-Memory on the Radeon GPU is doing some magic here. However, there is nothing to discuss as the performance differences are almost identical to the 1080p results. The 4K RT graphs would be fascinating to look at.

Gaming 2160p Benchmarks (RT)

AMD RX 9070 is enjoying a massive victory at 4K Ray-Tracing Benchmarks by 22% over NVIDIA’s GeForce RTX 5070; an extra 4GB of video memory is helping the Radeon Camp here; the 2160p + RT is extremely video memory intensive, which is the reason why RTX 4070 Super and RTX 5070 GPU’s are chocking due to less video memory.

However, 4K Ray-Tracing is still not ideal even on the RX 9070 without the upscaling; we don’t recommend our users buy this card if they intend to use it in ray-tracing on 4K resolution without the FSR. In Alan Wake II with RT, RX 9070 is 300% Faster. In Forza Horizon 5, RX 9070 unlocks 10% more performance.

Of course, you wouldn’t expect much from the mid-range graphics card to perform optimally at the 4K, especially when the ray-tracing is enabled, but it’s fascinating to see AMD has come far, Thanks to improved RT accelerators; while it’s not entirely beating the NVIDIA here but it’s still giving some exciting results for the AMD camp.

Synthetic Test Results

• 3D Mark: Port Royal: 3DMark Port Royal is the world’s first real-time ray tracing benchmark for gamers. It shows you how well your PC handles ray-tracing effects in real time. Use Port Royal to benchmark graphics cards that support Microsoft DirectX Raytracing. (Source 3D Mark)

RX 9070 is 10% Faster than NVIDIA’s GeForce 5070, 20% Faster than RTX 4070 Super in 3D Mark: Port Royal

• 3D Mark: Speed Way: 3DMark Speed Way’s engine is assembled to demonstrate what the latest DirectX API brings to ray traced gaming, using DirectX Raytracing tier 1.1 for real-time global illumination and real-time raytraced reflections, coupled with new performance optimizations like Mesh Shaders. Source (3D Mark)

RX 9070 is 3% slower than NVIDIA’s GeForce 5070, and 10% Faster than RTX 4070 Super in 3D Mark: Speed Way.

• 3D Mark: Steel Nomad: 3DMark Steel Nomad is a cross-platform, non-raytraced benchmark for high-end gaming PCs and Macs. It uses the DirectX 12 API on Windows. (Source 3D Mark)

RX 9070 is 16% Faster than NVIDIA’s GeForce 5070, 26% Faster than RTX 4070 Super in 3D Mark: Steel Nomad

AMD FidelityFX Super Resolution 4 (FSR 4)

AMD’s new iteration of FSR, AMD FidelityFX™ Super Resolution 4, is based on a Machine Learning Algorithm that aims to deliver better image quality, temporal stability and less ghosting, unlike the previous version of FSR that utilized the combination of super-resolution temporal upscaling technology and relied on analytical and algorithmic methods. AMD FSR 4 now leverages AI (Artificial Intelligence) Technology for the upscaling. However, AMD FSR 4 can only be enabled in the game that supports FSR 3.1, which you must allow from the driver settings. Right now, there is no way you can select the FSR 4 natively in the game.

We tested the FSR 4 Performance on Horizon Forbidden West; however, enabling the FSR 4 was tricky. AMD has sent us the press drivers, which were quite buggy while applying the FSR 4 in the drivers setting. However, we then installed the launch day driver and our problem was resolved: to use the FSR 4 you need to ensure you have FSR 4-supported titles. It would have been nice if the FSR 4 had been inside the game settings menu, not the driver settings.

How to Enable FSR 4?

1. To Enable the FSR, Open AMD Software: Adrenalin Edition.
2. Click on the ”Gaming Tab” and move to the ”Graphics” TAB.
3. Enable the 3rd Option ”AMD FidelityFX Super Resolution”
4. Run the game and enable the FSR 3. X in the game (Make sure the game is FSR 4 supported)
5. Press ALT + R during the gameplay and check the FSR4 status indicator.

FSR 4 Benchmarks

As our performance numbers demonstrate, FSR 4 is slightly heavier than FSR 3. Here, we bring you two tests, one with the frame-generation enabled and one without, while using the FSR 4 Upscaller in both. We have gained 22% more performance with FSR 4 quality vs TAA.

However, we see almost a 4 FPS reduction while comparing it against FSR 3.1, which is roughly 4% but at better image stability and less ghosting; likewise, with FSR Balance, we have gained around 27% more performance jump vs TAA but 5% less performance vs FSR 3.1 Balance. FSR 4 algorithm utilizes the AI accelerators on RDNA 4, which could be why we see a little performance drop compared to the previous version of FSR.

Overclocking Results

Overclocking an ASRock 9070 wasn’t difficult, but it was worthwhile; we have gained another 10% more performance in Cyberpunk 2077 by blending the memory and core overclock. If you are familiar with overclocking on AMD cards, you would know it’s slightly
different than NVIDIA; on AMD, you have to set the target frequency offset in the AMD Adrenaline Drivers, and you need to lower the GPU voltage, however increasing the power limit is necessary, which, in our case, on ASRock RX 9070 Steel Legend, you can unlock the 10% more power.

How Did We Overclock An RX 9070?

1. Open the AMD Adrenaline Software
2. At the selection menu on the top, select the Performance tab, and then click on Tuning.
3. At the tuning control, Set it to ‘Custom’
4. Toggle the GPU Tuning Option and Set it to on
5. In Maximum Frequency Offset (MHz), Set your desired frequency; in our case, we did +400MHz
6. Below the Maximum Frequency Offset (MHz) Slider, You will find the Voltage Offset Option.
7. You need to adjust the voltage offset to see the gains. In our case, we set the offset to -100mV
8. After you have tuned your GPU cores, now head to the VRAM tuning option below the Voltage Offset Slider
9. Set your desired memory clock. In our case, we have overclocked our memory from 2518Mhz to 2750MHz
10. Finally, ensure you have unlocked the power limit; you can max out the slider.

• Max Frequency Core Clock Offset (Mhz): +400MHz
• Max Frequency Memory Clock (MHz): 2750MHz
• Voltage Offset (mV): -100mV
• Power Limit: +10% Power

Overclocking Performance Benchmarks

The performance gains are really massive here, almost double digits. We have gained another 10% more free performance in Cyberpunk 2077. AMD offers good overclocking headroom for the extra performance in RX 9070. However, the results may vary on your RX 9070 as the silicons are not binned equally. You may have less overclocking headroom or more.

Power Consumption & Clock Speed

Our sample ASRock Steel Legend RX 9070 has the effective core clock speed of 2,519MHz in 3D Mark: Speed Way, roughly 5% more factory overclock and averaging 244W at the total board power as reported by the HWINFO; we did not have the external tools to test the power consumption. RX 9070 is drawing an extra 20 watts to match the RTX 5070.

In the Gaming Test, we have Cyberpunk 2077 power and clocks testing numbers. The ASRock Steel Legend RX 9070 has an effective core clock speed of 2,463MHz while consuming an average of 243W. In this test, the RX 9070 was matching the RTX 5070 in performance.

Temperatures & Fan Speed

ASRock Steel Legend RX 9070 does exceptionally well in temperature and sound. The GPU core stayed at 57c. However, the HOTSPOT temperatures are 79c here, which is well within the AMD safe range while consuming 243W on average in this specific test. We left the fans at auto and only spun at 1356, which is not really audible if you’re sitting 1 foot next to the system. We’re impressed with the cooling performance of this GPU.

In the gaming test session, we peaked at 55c on the core temp while hitting 90c on the HOTSPOT, the fans are doing more leg work here at 1465RPM than our synthetic test, but it was still quieter, Thanks to ASRock’s highly-curved Striped Ring fans.

Should You Buy It?

Buy It If

✅You want great raster performance: At both 1440p and even 4K, the RX 9070 showed impressive performance, especially considering its price.

✅You are fine with decent RT performance: While not the best, the RX 9070 still handles ray tracing fairly well, so if having the absolute best ray tracing is not your biggest concern, this card is good enough.

✅You want to future-proof your build: In comparison with the RTX 5070, the 16GB VRAM in the RX 9070 means this GPU will last you for more years down the road than its NVIDIA counterpart.

Don’t Buy It If

❌You’re spending over $600: I don’t recommend buying the RX 9070 if it costs more than $600, as at that point, the RX 9070 XT is the way to go.

Conclusion

The launch of the RX 9070/XT was much anticipated, as AMD has managed to achieve what NVIDIA failed to do in the mid-range gaming segment: providing better value for gamers. The RX 9070 is an excellent GPU that outperforms RTX 5070 in raster, and gaming across all resolutions. Historically, AMD struggled with ray tracing performance, but the new RDNA 4 architecture has changed that. The RX 9070 delivers impressive ray tracing capabilities at all three resolutions.

With 16GB of video memory, the RX 9070 excels in games that require more memory at higher resolutions, such as 4K UHD, outperforming the competition when additional video memory is crucial. If you look at the 4K RT charts, AMD has smoked NVIDIA’s GeForce RTX 5070, and AMD has clearly put in the effort to excel in all aspects of this GPU.

Upcoming games will run better on the RX 9070 compared to the RTX 5070 due to its extra video memory. Overclocking worked well on this GPU; I don’t know why AMD left a couple more FPS on the table. ASRock Steel Legend can unlock 10% more power, which could help in the power-limited scenario. The Cooling performance on ASRock Steel Legend 9070 is outstanding, only peaking at 55c, and the noise is hardly audible. 

However, the pricing of the RX 9070 is somewhat confusing. AMD has set an MSRP of $550 for RX 9070, It would have been ideal for it to be $120-$150 cheaper than the XT variant to position it as the optimal choice for gamers. For $50 extra, you can upgrade to the RX 9070 XT, which offers superior computing performance. If you can find the 9070 XT by spending the extra $50 despite its weaker cooler, it’s definitely a worthwhile investment.

In today’s market, securing a card at the Manufacturer’s Suggested Retail Price (MSRP) feels nothing short of miraculous. If you manage to find it for $550, consider it a win. Our tests have shown that the RX 9070 outperforms the RTX 5070 while consuming slightly more power. Unless you’re upgrading from the RX 7900XT or 7800XT, there’s really no reason not to purchase this card.