ASUS ROG Strix RTX 2070 Super A8G Review: Still Worth It in 2026?

Right, let me ask you something. Have you ever bought a GPU, got it home, plugged it in, and immediately felt like you either massively overspent or quietly cheated yourself? I have. More times than I'd like to admit. The GPU market has this nasty habit of sitting you between two stools: the card below your budget leaves you dropping settings in games you actually care about, and the card above it costs so much extra that you'd need to play it for three years straight just to justify the difference. Finding the sweet spot is genuinely hard, and it's made harder when you're looking at a used or older-generation card that was originally positioned as a premium product.

That's exactly the situation with the ASUS ROG Strix GeForce RTX 2070 Super Advanced Overclocked 8G GDDR6 HDMI DisplayPort USB Type-C Gaming Graphics Card (ROG-STRIX-RTX-2070S-A8G-GAMING). This is a card that launched into a very different market, wore a very premium price tag, and now finds itself in the second-hand ecosystem competing against newer budget options and older workhorses alike. I've been running it through its paces for about a month now, testing it across a range of games and workloads, and I want to give you the honest picture of where it sits in 2025.

Before we get into the numbers, I want to set the scene properly. Because context is everything with a card like this. You can't just look at raw FPS figures in isolation. You need to know what else your money gets you right now, what you're giving up, and whether the ROG Strix branding and build quality actually justify the premium over a reference-style 2070 Super. Spoiler: sometimes it does, sometimes it doesn't. Let's get into it.

Memory is 8GB of GDDR6 on a 256-bit bus, hitting 448 GB/s of bandwidth. The display outputs are generous: two DisplayPort 1.4 connections, two HDMI 2.0b ports, and a USB Type-C port that supports VirtualLink (remember when that was going to be the future of VR headset connectivity?). The card is a triple-slot design, measuring around 300mm in length, so you'll want to check your case clearance before buying. It's not a small card. ASUS has fitted it with their triple-fan Axial-tech cooler, which we'll talk about properly in the thermals section.

One thing worth flagging upfront: the PCIe interface is Gen 3 x16. In 2025, with Gen 4 and Gen 5 PCIe slots standard on modern motherboards, this is backwards compatible but you won't be saturating that bandwidth with a 2070 Super anyway. It's a non-issue in practice. What's more relevant is the HDMI 2.0b rather than 2.1, which caps you at 4K60 on that output rather than 4K120. If you're running a high-refresh 4K display via HDMI, that's a genuine limitation. DisplayPort 1.4 handles 4K144 fine though.

The RT cores are first-generation, and I want to be honest about what that means in 2025. They're slower at BVH traversal (the mathematical process of figuring out which rays hit which objects) than the second-gen RT cores in Ampere or the third-gen ones in Ada Lovelace. In practical terms, you'll see a bigger performance hit when enabling ray tracing on Turing compared to a 3070 or 4060. It's not that ray tracing doesn't work, it's that you'll be leaning on DLSS much harder to maintain playable framerates when RT is on. The Tensor cores handle DLSS inference, and being second-generation Tensor cores they support DLSS 2.x properly, which is genuinely good upscaling. DLSS 3 Frame Generation, however, requires Ada Lovelace. So you're capped at DLSS 2.x here.

The 12nm FFN process from TSMC is showing its age compared to TSMC's 4N node used in Ada or AMD's 5nm node in RDNA 3. More transistors per mm squared means newer cards can pack more compute into a smaller, cooler, more power-efficient die. The TU104 has 13.6 billion transistors on a 545mm squared die, which was impressive in 2018 but looks chunky by modern standards. This is partly why the power consumption is what it is, and why thermals require a serious cooler. ASUS clearly knew this when they designed the ROG Strix triple-fan solution for this card.

Compared to the reference RTX 2070 Super Founders Edition, which boosts to around 1770MHz to 1800MHz in practice, the ROG Strix's factory overclock is delivering a genuine real-world advantage. We're talking roughly 7 to 8 percent higher clocks, which translates to a similar percentage uplift in rasterisation performance. It's not transformative, but it's not nothing either. If you're comparing two 2070 Super cards and one is the ROG Strix A8G, you're getting measurably more performance out of the box without touching any overclocking software yourself.

I did try pushing the overclock further using GPU Tweak III. The card has solid power delivery on its PCB, with a 14-phase VRM design that ASUS uses on their ROG Strix line. I managed a stable additional 75MHz on the core and 500MHz on the memory (GDDR6 memory responds well to memory overclocking on Turing). That pushed average gaming clocks to around 2010MHz, which is a nice bonus. But honestly, for most people, the out-of-box performance is already well above reference, and chasing another 5 percent through manual overclocking isn't worth the effort unless you enjoy that sort of thing. I do, but I'm aware I'm in a minority.

At 1080p, 8GB is still comfortable. Most games at 1080p High to Ultra settings sit between 4GB and 6GB of VRAM usage, so you've got headroom. This is actually where the 2070 Super lives most happily in 2025: it's a very good 1080p card, and a capable 1440p card if you're willing to dial back texture quality one notch from maximum. At 4K, I'd be honest and say the VRAM situation is a problem before the shader performance even becomes the limiting factor. You'll hit 8GB before the GPU is fully loaded, which means you're leaving performance on the table and getting stutters on top.

For context, AMD's RX 6700 XT launched with 12GB of GDDR6 on a 192-bit bus, and NVIDIA's own RTX 3070 shipped with 8GB. The 3070's 8GB has aged similarly to the 2070 Super's 8GB, which is to say: it's fine today, questionable in a year or two. If you're buying this card specifically for 4K gaming, I'd steer you elsewhere. But if 1080p to 1440p is your target resolution, the VRAM situation is manageable with a bit of settings tweaking. Just don't expect to run maximum texture quality in every modern title at 1440p without hitting the ceiling.

And that's where DLSS saves the day on this card. DLSS 2.x on the 2070 Super is genuinely good. Running DLSS Quality mode at 1440p output (rendering at roughly 1080p internally) gives you image quality that's very close to native 1440p, and it hands back 40 to 50 percent of your performance. In Cyberpunk with RT Medium and DLSS Quality at 1440p, I was seeing around 58fps average, which is actually playable. Not perfect, but playable. DLSS Balanced mode pushes that closer to 70fps with a slight image quality trade-off that most people won't notice unless they're pixel-peeping screenshots.

What you don't get is DLSS 3 Frame Generation. That's locked to Ada Lovelace (RTX 40 series) and requires the Optical Flow Accelerator hardware that Turing doesn't have. So if you're seeing benchmarks of RTX 4070 cards hitting 120fps in demanding titles with Frame Generation on, that's not a comparison point for this card. You're in DLSS 2.x territory, which is still miles ahead of no upscaling at all, but it's worth knowing the ceiling. FSR 2.x and FSR 3.x are also supported since those are software-based and run on any GPU, and FSR 2 in particular is a decent alternative in games where DLSS isn't available.

What Turing's NVENC doesn't support is AV1 encoding. That arrived with Ada Lovelace (RTX 40 series) and is increasingly relevant for YouTube uploads and Discord streaming. The 2070 Super supports H.264 and HEVC (H.265) encode and decode, which covers the vast majority of use cases in 2025, but if AV1 encoding is important to you (say, you're uploading a lot of content to YouTube where AV1 offers better quality at lower bitrates), you'll need a newer card. NVIDIA's encode/decode support matrix confirms Turing supports HEVC decode up to 4K but lacks AV1 encode entirely.

For content creators who are doing light video editing alongside gaming, the Tensor cores in Turing also accelerate some AI-based tasks in applications like DaVinci Resolve and Adobe Premiere. It's not as fast as Ampere or Ada for these workloads, but it's noticeably quicker than doing the same tasks on a CPU alone. If you're a streamer or occasional video editor who also games, the 2070 Super is a capable all-rounder. Just don't expect it to compete with a 3080 or 4070 on pure encoding throughput.

The dual 8-pin connector setup is standard and completely fine. No 12VHPWR adapter nonsense to worry about here. If you've got a decent modular PSU from the last five years with two 8-pin PCIe cables, you're sorted. I did notice the card pulls close to its full TDP almost immediately under load rather than ramping up gradually, which is typical of Turing. It's not a card that sips power at medium loads; it's fairly binary between idle and full tilt. Idle power draw is around 8W to 10W, which is perfectly reasonable.

Compared to modern equivalents, the 215W TDP looks a bit heavy. An RTX 4060 Ti delivers similar or better 1080p performance at 165W, and AMD's RX 7600 XT does competitive 1080p work at 190W. The 2070 Super's power consumption reflects its older manufacturing process. You're paying a bit more on your electricity bill compared to a modern equivalent, and that's a real cost over time if you game for several hours a day. Not a dealbreaker, but worth factoring in if you're on a prepayment meter or just conscious about running costs.

I deliberately tested in a more restrictive case scenario too, blocking off some of the intake airflow to simulate a poorly ventilated build. GPU die temps climbed to 80 degrees, hotspot to 90 degrees, and I did see the boost clock drop slightly from 1920MHz to around 1875MHz as the card started thermal throttling. So the card is sensitive to case airflow, as you'd expect from a 215W GPU. If you're putting this in a compact case with minimal airflow, you'll want to think carefully about your cooling setup. It's not a card that forgives a stuffy environment.

Idle temperatures are excellent: 38 to 40 degrees with the fans in zero-RPM mode (more on that in the acoustics section). The card doesn't get hot sitting at the Windows desktop or doing light tasks. Memory junction temperatures, which became a talking point with some Turing cards, were fine on this ROG Strix variant. I saw GDDR6 memory temps around 80 to 84 degrees under load, which is within GDDR6's rated operating range. Some cheaper 2070 Super coolers struggled with memory thermals, but the ROG Strix's larger heatsink covers the memory modules properly.

Under gaming load, the fans spin up and the character of the noise matters as much as the volume. At 1440p gaming loads, I measured around 38 to 40 dB(A) at one metre from the case, which is quiet enough that you won't hear it over game audio with headphones on. The Axial-tech fans have a slightly higher-pitched tone than some competitors, but it's not unpleasant. It's more of a gentle whoosh than a whine. I've tested cards that sound like a hairdryer at full load (looking at you, certain Radeon VII coolers), and this is nowhere near that territory.

At maximum fan speed, which you'd only hit if you manually set a more aggressive fan curve or the card is really struggling thermally, it does get noticeably louder, maybe 47 to 48 dB(A). But in normal gaming use with a sensible case, the fans never need to go that high. The default fan curve ASUS ships is well-tuned: it keeps temperatures in check without making the card sound like it's trying to take off. This is one area where the ROG Strix premium is genuinely justified over a reference blower-style cooler, which would be significantly louder and hotter.

At 1080p, the 2070 Super is genuinely excellent. Cyberpunk 2077 at High settings (no RT) averaged 78fps with 1% lows around 58fps. Shadow of the Tomb Raider at Highest settings hit 112fps average. Forza Horizon 5 at Ultra settings managed 95fps average. Apex Legends at High settings was CPU-limited on my 5600X, sitting around 180fps to 200fps. These are strong 1080p numbers. If you're gaming on a 1080p144 monitor, the 2070 Super will keep you above 100fps in most titles at High settings, which is exactly what you want from a card in this tier.

At 1440p, the picture is still good but you need to be a bit more selective with settings. Cyberpunk 2077 at High 1440p averaged 58fps, which is playable but not buttery. Drop to Medium and you're at 74fps with minimal visual difference in most scenes. Shadow of the Tomb Raider at Highest 1440p hit 82fps average. Forza Horizon 5 at Ultra 1440p managed 68fps. With DLSS Quality enabled at 1440p, Cyberpunk jumps to around 75fps at High settings, which is a much more comfortable experience. At 4K, the card struggles: Cyberpunk at Medium 4K averaged 38fps, which is below comfortable gaming territory. The 2070 Super is not a 4K card in 2025, full stop.

The RTX 4060 Ti is the most direct modern comparison. It offers better 1080p and 1440p rasterisation performance (roughly 15 to 20 percent faster), proper DLSS 3 support including Frame Generation, AV1 encoding, and a 165W TDP. It also has 8GB VRAM on the base model (there's a 16GB variant at a higher price), so the VRAM situation isn't dramatically better. The RTX 4060 Ti is the card I'd point most people towards if they're buying new today. The AMD RX 7700 XT is another strong option, offering 12GB of GDDR6 which addresses the VRAM concern, competitive rasterisation performance, and FSR 3 support. It doesn't have DLSS, but FSR 3 is available in a growing number of titles.

Where the ROG Strix 2070 Super can still make sense is in the used market at the right price. If you can find one in good condition for significantly less than a new 4060 Ti, and you're primarily gaming at 1080p to 1440p without heavy RT use, the performance gap doesn't justify the price premium of going new. The build quality of the ROG Strix is excellent, the cooler is proven, and Turing cards have shown good longevity. But you need to be buying it at the right price for that argument to hold up.

Build quality is genuinely premium. The shroud is a mix of plastic and metal with ASUS's ROG aesthetic, which is aggressive and angular in a way that either appeals to you or doesn't. The backplate is metal, which adds rigidity and helps with heat dissipation from the back of the PCB. The ROG logo on the side illuminates with ARGB lighting that's controllable through ASUS's Aura Sync software, and it syncs with other Aura-compatible components if you're into that sort of thing. I'm personally ambivalent about RGB on GPUs, but the implementation here is tasteful rather than garish.

The triple Axial-tech fans use a combination of blade designs: the outer two fans spin in one direction while the middle fan spins in the opposite direction, which ASUS claims reduces turbulence and improves airflow consistency across the heatsink. Whether that specific design choice makes a measurable difference compared to three fans all spinning the same way is debatable, but the end result is a cooler that performs well and sounds decent, so whatever they're doing is working. The fans feel solid and well-mounted, with no rattling or vibration transmitted to the card body during testing.

The most common complaints centre on the size and weight causing GPU sag, which I mentioned above, and the RGB software requiring ASUS's Aura Sync ecosystem to control properly. If you're not running an ASUS motherboard or other Aura-compatible components, the RGB defaults to a rainbow cycle that you can't easily customise without installing the full Armoury Crate software suite, which some people find bloated. A fair criticism. The other recurring theme in user feedback is the VRAM situation, with owners who've held onto the card into 2024 and 2025 noticing it struggling in newer titles at higher settings. That's not a surprise given what we know about 8GB in 2025.

Positive feedback consistently highlights the out-of-box performance versus reference 2070 Super cards, with users noting the factory overclock makes a real difference in day-to-day gaming. The dual BIOS switch (which lets you toggle between Performance and Quiet BIOS modes) gets mentioned positively by users who want to prioritise acoustics over maximum performance. In Quiet mode, the fan curve is more conservative and the card runs a touch warmer but noticeably quieter. It's a small feature but one that shows ASUS thought about different use cases when designing the card.

If you can find one in good condition for a price that's meaningfully below a new RTX 4060 (not 4060 Ti, just the base 4060), it represents reasonable value for a 1080p to 1440p gaming card. The ROG Strix build quality means it's likely to have been well-cooled throughout its life, reducing the risk of buying a used card that's been thermally stressed. But if the asking price is close to a new budget modern card, the maths doesn't work. You'd be paying similar money for older architecture, no DLSS 3, no AV1 encode, and the same 8GB VRAM ceiling.

The tier this card occupies in 2025 is "solid used buy at the right price, questionable at anything approaching new-card money." It's not a card I'd recommend buying new if new stock somehow still exists at original pricing. But as a used purchase from someone who's maintained it well, it's a capable 1080p to 1440p card that will serve you for another two to three years at those resolutions. Just go in with eyes open about the VRAM situation and the RT performance limitations, and make sure you're not overpaying relative to the current used market.

• 8GB VRAM is increasingly tight at 1440p in modern titles and limiting at 4K
• First-gen RT cores mean ray tracing performance is limited compared to Ampere or Ada
• No DLSS 3 or Frame Generation support, capped at DLSS 2.x
• No AV1 encoding for content creators who need it
• 215W TDP is higher than modern equivalents for similar performance
• GPU sag is a real issue given the card's size and weight
• HDMI 2.0b rather than 2.1 limits high-refresh 4K over HDMI

But context matters. The ROG Strix 2070 Super is a well-built card that has aged better than some of its contemporaries precisely because ASUS invested in proper cooling and power delivery. It won't embarrass itself at 1080p in 2025, and with DLSS Quality mode it handles 1440p respectably in most titles. The 8GB VRAM ceiling is the biggest practical concern for longevity, and the first-gen RT cores mean ray tracing is more of a curiosity than a usable feature at higher settings. If you're a streamer or content creator, the lack of AV1 encoding is a genuine gap compared to newer cards.

My honest score for this card in 2025 is 7 out of 10 at the right used price, dropping to 5 out of 10 if you're paying anywhere near what modern alternatives cost new. The build quality and cooling solution are genuinely excellent, the factory overclock is meaningful, and the card has proven longevity. But the architecture is showing its age in ways that matter for modern gaming, and the VRAM situation will only get more pressing as 2026 approaches. Buy it smart, use it at 1080p to 1440p, and don't expect miracles from ray tracing. Do that, and it's still a solid card. Expect it to compete with modern alternatives on features and future-proofing, and you'll be disappointed.

Testing conducted over about a month on a system running a Ryzen 5 5600X, 32GB DDR4-3200, and NVMe SSD. All gaming benchmarks taken at average FPS over a five-minute run with 1% lows recorded. Temperatures measured using HWiNFO64. Power draw measured at the wall using a plug-in energy monitor.

Affiliate disclosure: vividrepairs.co.uk participates in affiliate programmes. If you purchase through links on this page, we may earn a small commission at no extra cost to you. This does not influence our editorial opinions. We test products independently and our reviews reflect genuine findings from our testing process.