AMD first announced the Radeon RX 5500 in early October, back when the RX 5500 Mobile (Apple) and OEM-only "RX 5500" launched, which we reviewed not long ago. Today, the consumer DIY-focused Radeon RX 5500 XT is released. The biggest surprise about this new model is that AMD is using the exact same shader amount on the "XT" SKU as on the "non-XT"—something that never happened in recent history. The Navi 14 silicon does have 24 CUs (1536 shaders) physically, but it looks like either Apple is gobbling up all the production for their new MacBook lineup or wanted exclusivity on the higher shader count. All that is different between the RX 5500 and RX 5500 XT is that the OEM model ticks at 1670 MHz game clock, and the XT cards start at 1717 MHz, a meager +2.8% increase. We've only seen 4 GB variants of the RX 5500, the RX 5500 XT comes in both 4 GB and 8 GB flavors. For today's launch, we also reviewed the MSI RX 5500 XT Gaming X 8 GB to provide a data point for 4 GB vs. 8 GB VRAM.

Sapphire's Radeon RX 5500 XT Pulse is built on the same philosophy as the company's RX 5700 XT Pulse, which had great success on the market: make improvements over the base reference where it matters without increasing pricing too much. For a $10 increase over AMD MSRP, Sapphire managed to include a factory overclock, a much better dual-fan cooler, idle-fan stop, dual BIOS, and even a backplate. Out of the box, the Pulse runs at 1737 MHz rated Game Clock. In reality, we measured an average clock frequency for 1080p gaming of 1833 MHz. Compared to the RX 5500 non-XT, which averaged out at 1795 MHz, that's a 2.1% increase in actual clock. AMD's math promises 4% (1737 MHz Game clock vs. 1670 MHz Game clock). I have no idea how the Game Clock value is measured, calculated, or estimated, but it's safe to say that it's a fantasy that doesn't hold up in real life.

Overall, the Sapphire RX 5500 XT Pulse achieves a 1% performance increase over the RX 5500 non-XT, hardly worth mentioning. Compared to NVIDIA's new GTX 1650 Super, the card ends up just 1% behind, which means they made up a lot of ground since my original RX 5500 OEM review. The secret sauce here is the new 19.12.2 Adrenalin 2020 driver, which significantly improves performance for all Navi cards. That's the reason why I included two RX 5500 non-XT data points in this review, to give you a feel for how big the performance gain from drivers is. AMD certainly did a great job here. This performance increase also enables the RX 5500 XT to beat the aging RX 580, the RX 570 4 GB is 16% behind. AMD's Radeon RX 590, which is Polaris-based, too, is 9% faster than the RX 5500 XT. NVIDIA's GTX 1650 is 25% slower, which is why the green team launched the GTX 1650 Super. The GTX 1660 is 15% faster, but also more expensive. Overall, we can definitely recommend the Radeon RX 5500 Series for all games at 1080p Full HD.

Sapphire's thermal solution is an impressive kit of gear. It looks well-designed without increasing cost too much. The two 100-mm fans run nice and slow at all times when gaming, yet temperatures are excellent, also thanks to the low heat output of the Navi 14 graphics chip. We measured only 64°C during gaming, which is comfortably cool and lower than even the MSI RX 5500 XT Gaming X. At the same time noise levels are unbelievable. The card emits only 28 dBA at full load, which makes it nearly inaudible on an open bench; it's even quieter than the MSI Gaming X. When installed in a case with other active components it'll be impossible to make out the graphics card noise. Remember, we're talking about full gaming load here, other cards are just as loud in idle as the Pulse is in gaming. Idle fan noise of Sapphire's card is perfect because it includes the almost mandatory idle-fan-stop capability, which completely turns off the graphics card fans in idle, Internet browsing, productivity, and light gaming. Sapphire even managed to squeeze in a metal backplate, which greatly improves the overall look and feel of the product—most lower-priced GTX 1650 Super cards come without a backplate.

Sapphire includes a dual-BIOS feature with their card, too, which lets you select between the default "performance" BIOS and a "quiet" mode BIOS. While we can confirm that the quiet mode BIOS runs the fans even slower (and includes fan-stop), the differences are minimal and will be impossible to make out subjectively. While dual BIOS is definitely useful if you want to mess with BIOS flashing, I could imagine Sapphire getting rid of it in the future to save a few bucks on production cost, which could help further improve cost efficiency and is all that matters in this price-sensitive market segment.

Back in July, Navi 10, which is used on the Radeon RX 5700 series, confirmed that AMD has made substantial improvements in power efficiency, and Navi 14 on the RX 5500 is no different. Looking at performance per watt, we see the RX 5500 match the RX 5700 XT almost exactly. Only the RX 5700 non-XT is more power efficient, but it is a special undervolted design. Compared to NVIDIA, this means the RX 5500 is roughly as power efficient as NVIDIA's Pascal architecture, which is a good improvement. NVIDIA's Turing architecture is still more efficient, and NVIDIA is still on 12 nanometer, while Navi uses the more efficient 7 nanometer tech. Nevertheless, looking at what the RX 5500 delivers in terms of power/heat/noise, it seems the differences aren't that major anymore.

Sapphire does have a unique feature for their new Navi graphics cards called "TriXX Boost". It's a new capability of their TriXX overclocking software and lets you precisely set a custom resolution (in small steps) that's active system-wide. That new resolution can now be selected in your games and nets you higher FPS with a small loss in image quality due to the lower rendering resolution. Paired with Radeon Image Sharpening, this can help improve performance without compromising image quality too much.

Overclocking using Wattman worked much better than in my early Navi 10 reviews. It seems AMD is actively working on getting all the issues fixed. Their new Adrenalin 2020 confirms their interest in software improvements. Just like on previous Radeon cards, overclocking is limited to a maximum range AMD decides, no idea why as these cards definitely can take more. Our maximum overclock was decent, reaching around +7% on GPU and +6% on memory, which turned into a 6% real-life performance improvement. NVIDIA's GTX 1650 Super does overclock a bit better though, regularly achieving +10% performance gains or more.

As mentioned before, the Radeon RX 5500 XT comes with memory sizes of 4 GB and 8 GB. For this launch we had the chance to test both variants, and I have to admit I'm surprised by the results. I always thought that there would be no significant difference between 4 GB and 8 GB at 1080p in most titles, and the differences would only show at 1440p or 4K—resolutions the card is simply too slow for. Looking through our benchmark results there are indeed a few cases where 1080p performance is higher on the 8 GB model. For example, Assassin's Creed, Far Cry 5, Gears 5, GreedFall, Tomb Raider, and Wolfenstein see improvements of varying degree from doubling the memory amount. While I'm sure that will be heavily used by marketing, paired with "new consoles will have more memory", I'm not convinced that these gains are worth spending $30 more. NVIDIA's GTX 1650 Super comes in 4 GB variants only, which, with the right cherry-picking, could lure in less educated buyers too, "twice the memory, it must be twice as fast" (like in Wolfenstein 1080p). Personally, if I was shopping in this segment with a limited budget, I'd still opt for the 4 GB variant for the enormous cost savings, and possibly dial down memory intensive settings one notch. What's also important to mention here is that NVIDIA does manage limited VRAM more efficiently than AMD, as indicated by some of our benchmarks.

AMD has set base pricing of RX 5500 XT at $170 (4 GB) and $200 (8 GB), NVIDIA GTX 1650 Super starts at $160. While the $170 price isn't bad, it's not good enough to convince hordes of buyers to opt for AMD's card, especially since NVIDIA's offering is a tiny bit faster and more power efficient. The 8 GB RX 5500 XT is simply way too expensive, charging $30 (or 17%) for additional VRAM that yields 5% in performance makes little sense from a buyer's perspective, I would be willing to maybe pay $15 more. Also, at that price point, the 8 GB RX 5500 goes up against the GTX 1660 at $210, which is 10% faster and has 6 GB VRAM. I have to commend Sapphire for not overpricing their Pulse product, a cost increase of $10 is very reasonable, just the amazing cooler is worth that investment. Even at $170, I could see myself considering the Sapphire Pulse for its low noise levels that beat all the NVIDIA GTX 1650 Super cards we've tested so far. If noise is only secondary to you, the $160 EVGA GTX 1650 Super SC Ultra is extremely strong competition for RX 5500 XT. I talked to various board partners regarding their Polaris (RX 570/580/590) stock levels and they all say they still have significant inventory, so maybe higher RX 5500 pricing will let them sell off those cards quickly, and we'll see price drops for the RX 5500 XT in the new year. What does concern me a little bit here is that the RX 5500 XT board design is very complex, with expensive VRM circuitry, whereas the NVIDIA cards look to be engineered with much better cost optimization in mind.