
The most useful GPU testing of the week did not involve a new graphics card, or even a benchmark run in the usual sense. ComputerBase strapped an LDAT light sensor to seven games and measured the full input to photon chain to find out whether NVIDIA Reflex, AMD Anti-Lag 2, and Intel XeLL actually do the same job, and the answer is messier than any vendor slide deck. Beyond that, Hardware Unboxed settled the RX 9070 versus RX 9070 XT question across 52 games, be quiet! finally shipped an AIO with a screen and got a flurry of reviews, two very different high wattage power supplies landed on the same day, and Lexar quietly put an entire SSD inside one chip package.
NVIDIA Reflex vs AMD Anti-Lag 2 vs Intel XeLL
Written reviews:
- ComputerBase: https://www.computerbase.de/artikel/grafikkarten/reflex-anti-lag-2-xell-benchmark-vergleich.98595/
Consensus summary: ComputerBase picked seven games that support all three technologies and that allow each one to be toggled independently of frame generation: Battlefield 6, Black Myth: Wukong, Death Stranding 2, Dying Light: The Beast, Halo: Campaign Evolved, Hogwarts Legacy, and Marathon. Testing ran at 1440p on a GeForce RTX 5060 Ti 16GB, a Radeon RX 9060 XT 16GB, and an Arc Pro B70, with vendor upscaling set to Quality, and latency measured in hardware with LDAT rather than through each vendor’s own software counter, since those numbers are not comparable to each other. The headline finding is that when the integration is done properly the three are broadly comparable, but that proper integration is not the norm. Reflex and Anti-Lag 2 each failed to do anything at all in one game apiece. XeLL improved latency everywhere, but it mangled frame pacing badly enough to cost 25 percent of the percentile framerate in one title and 14 percent in another. Reflex was better behaved with up to 8 percent off the percent, and Anti-Lag 2 was the gentlest of the three on frame pacing while also delivering the smallest average latency reduction, which ComputerBase suggests may be related. Reflex rides along with DLSS frame generation and is therefore nearly everywhere, while AMD never tied Anti-Lag 2 to FSR frame generation and it goes missing surprisingly often, and XeLL is rarer still, although the newest versions can potentially run on any vendor’s GPU. Their verdict names Reflex the most mature of the three without calling it flawless.
FPS Review take: This is the test somebody should have run two years ago (but probably could not have due to lack of game integration), and the frame pacing finding is the part that matters, because a feature that shaves 10 ms off input lag while chewing 25 percent off your 1 percent lows is not obviously a win. Note the tier ComputerBase chose to test on, which is the same class of card we put through our RTX 5060 Ti 16GB versus 8GB testing and the same competitive bracket as our Radeon RX 9070 GRE versus RTX 5070 comparison. If you own a Radeon, the takeaway is to check whether the game even exposes Anti-Lag 2 before you assume you are getting it.
Radeon RX 9070 vs Radeon RX 9070 XT
Video reviews:
- Hardware Unboxed: https://www.youtube.com/watch?v=eSv8n3seMi0
Consensus summary: Hardware Unboxed ran both Navi 48 cards through 52 games at 1080p, 1440p, and 4K, with the full graph set posted alongside the video. The average gap lands at roughly 12 percent in the XT’s favour, which is about what the specification difference implies given the extra compute units and higher clocks, and it is a single outlet’s dataset rather than a cross site consensus.
FPS Review take: A 12 percent average spread is exactly the range where street pricing decides the argument rather than the silicon, and in the current market that spread moves week to week. If you are shopping this tier, our Radeon RX 9070 GRE review covers the rung below, and our XFX QuickSilver RX 9070 OC testing and XFX Swift RX 9070 XT review are our own numbers on the two cards in question.
6K Gaming Tested: GeForce RTX 5090 and RTX 5070 Ti
Video reviews:
- Hardware Unboxed: https://www.youtube.com/watch?v=Q-DXwZEO21g
Consensus summary: With the Samsung Odyssey G80HS arriving as the first true 6K gaming monitor at 165 Hz, Hardware Unboxed asked whether the fastest consumer GPU can actually drive it, and framed the exercise against NVIDIA’s 2020 claim that the RTX 3090 was an 8K gaming card. 6144 by 3456 is 21.2 million pixels against 8.3 million at 4K, a 2.6 times increase, and older cards are locked out before performance even matters because 40 series display engines cannot drive this panel above 60 Hz at 6K without DisplayPort 2.1. Across seven titles the pattern is consistent, with GPU limited configurations losing roughly half their framerate.
The RTX 5070 Ti is written off outright, needing Low presets and ultra performance upscaling to clear 60 FPS in most titles. Their conclusion is that the RTX 5090 is not useless at 6K but cannot run current games at maximum settings there even with upscaling, and that a 4K 240 Hz OLED is the better buy at a fraction of the combined 4,700 dollar GPU and 1,600 dollar monitor outlay. Frame generation is dismissed as a fix, both because the base framerates are too low and because its overhead and VRAM cost scale with output resolution.
FPS Review take: This is the useful perspective to a monitor category our readers have started asking about, and the finding that matters is the one about display engines rather than shaders, because an RTX 4090 owner is excluded from 165 Hz at this resolution by the output pipeline no matter how much raster performance the card has. The comparison to the 8K marketing of 2020 lands harder than it looks, too: a 5070 Ti is faster than the 3090 that was sold as an 8K card, and it cannot manage 6K. We covered the Odyssey G80HS itself in last week’s Weekender, and the practical advice has not changed since: buy this panel for the desktop and creator work, run games in its 3K dual mode, and treat 6K gaming as a benchmark curiosity for now. For where the requirements ceiling sits at the other end of the stack, our RTX 5060 Ti 16GB versus 8GB testing makes the same point about demands that only ever rise.
be quiet! Light Loop IO LCD 360
Written reviews:
- Guru3D: https://www.guru3d.com/review/bequiet-light-loop-io-lcd-cpu-cooler-review/
- KitGuru: https://www.kitguru.net/components/cooling/james-dawson/be-quiet-light-loop-io-lcd-360-aio-cpu-cooler-review/
- Club386: https://www.club386.com/be-quiet-light-loop-io-lcd-360-review/
- NikkTech: https://www.nikktech.com/main/articles/pc-hardware/cpu-cooling/liquid-cpu-coolers/16626-be-quiet-light-loop-io-lcd-360mm-silent-aio-water-cooler-review
- OC3D: https://overclock3d.net/reviews/cases_cooling/be-quiet-light-loop-io-lcd-360mm-liquid-cooler-review/
- Madshrimps: https://www.madshrimps.be/2026/08/be-quiet-light-loop-io-lcd-360mm-white-aio-review/
Video reviews:
- KitGuru: https://www.youtube.com/watch?v=MQNj-pjrbtk
- WELSHYTECH (240 ARGB White): https://www.youtube.com/watch?v=LK3Vm_q7qs0
- Nicolas11x12: https://www.youtube.com/watch?v=DHT12dSxegc
Consensus summary: be quiet! is late to the screen on a pump housing party, and the Light Loop IO LCD is its answer: a 2.1 inch 480 by 480 IPS panel rated at 500 nits, three daisy chained Light Wings LX fans, and a new IO Controller that carries power, PWM, ARGB, and display signal down one cable. The mechanical platform is shared with the existing Light Loop, and reviewers found the performance shared too. OC3D describes results in line with the non LCD models, strong at 1,000 RPM and stronger at 1,500. Club386 gives the most specific numbers, running a Ryzen 9 9950X3D with its full 170 W unlocked: at low fan speed the peak delta T of 65.0 C works out to a 91 C core temperature, which is in spec but toasty, while at 1,200 RPM the average delta T drops to 50.2 C for a real temperature of 76.8 C. For context Club386 puts the MSI CoreLiquid P22 360 a couple of degrees ahead and the 420 mm Arctic Liquid Freezer III Pro further ahead still. KitGuru’s verdict is that the looks and low noise impress but it is not a slam dunk, and Club386 points out that you still screw on and wire up the fans yourself, which is a step behind rivals shipping pre assembled radiators. At around $225 for the 240mm variant it is priced above the standard Light Loop.
FPS Review take: Five written reviews and three videos in one week for a cooler whose radiator and pump are carried over is a good reminder of what actually sells AIOs in 2026, and it is not delta T. The read is that this is the same competent cooler with a nicer screen and much better cable management, so buy it for the IO Controller and the panel, not for thermals, and step up to a 420 mm if you are running an unlocked 16 core.
InWin Covalent
Written reviews:
Consensus summary: Club386 treats the Covalent as a full tower that genuinely straddles gaming and workstation duty, which is the point of the name. It is enormous at 536 mm tall, 250 mm wide, and 606 mm deep for 81 liters of volume, it takes SSI-EEB motherboards and a pile of 2.5 and 3.5 inch drives, and it has the airflow to match. At just over $200 the value case is strong, with a few small annoyances holding it back from a clean sweep, and the real caveat is regional: UK buyers face roughly 53.60 euros in shipping from InWin’s EU store, which pushes Club386 toward recommending the Arctic Xtender VG locally instead.
FPS Review take: An 81 liter chassis that takes SSI-EEB for a bit over $200 is a specific and underserved pitch, and it is aimed squarely at the reader building a local AI or rendering box rather than a gaming rig. For premium full tower comparison from our own bench, our LIAN LI LANCOOL 217 Black Genuine Walnut Wood case review is the closest reference point, though it plays a very different aesthetic game.
High Wattage Power Supplies: MSI MPG Ai1600TS PCIE5 and FSP DAGGER PM 1200W
Written reviews:
- Igor’s Lab (MSI MPG Ai1600TS PCIE5): https://www.igorslab.de/en/msi-mpg-ai1600ts-pcie5-review-built-in-safety-12v2x6-connector/
- Vortez (FSP DAGGER PM 1200W): https://www.vortez.net/articles_pages/fsp_dagger_pm_1200w_psu_review,1.html
Video reviews:
- Hardware Busters (MSI MPG Ai1600TS Safeguard+ evaluation): https://www.youtube.com/watch?v=dtvrMpW1_Zg
Consensus summary: Two answers to the same question, which is how you feed a modern flagship GPU without setting the connector on fire. Igor’s Lab puts the 1600 W MSI MPG Ai1600TS PCIE5 on a CWT semi digital platform with interleaved PFC and full bridge LLC, measuring 91.07 percent efficiency at 115 V and 92.67 percent at 230 V for Cybenetics Titanium and Platinum respectively, 12 V regulation inside 0.73 percent, noise around 19.5 dBA, and hold up time of 19.6 ms. The distinguishing feature is not the wattage but GPU Safeguard+, which watches power distribution across the two native 12V-2×6 connectors and can trigger an audible alarm, an MSI Center notification, and an Afterburner action that drops the GPU power limit to 75 percent when it sees an imbalance. Igor’s complaints are that it becomes clearly audible past roughly 1,400 W at just over 42 dBA, that the 12 V rail sagged to 11.45 V during the 200 percent transient test which passes spec without much margin, that the convenience features need a USB header and software, and that this power class is simply oversized for a normal single GPU gaming PC.
Vortez comes at it from the opposite direction with the FSP DAGGER PM 1200W, a Platinum rated SFX-L unit in under a litre of volume with a native 12V-2×6 connector, and found excellent 12 V regulation feeding both CPU and GPU, plus semi passive cooling that stayed fanless to surprisingly high loads. At a $259 MSRP they call it an easy recommendation for small form factor builders running an RTX 5080 or RTX 5090, noting the ROG LOKI 1200W SFX-L was sitting near $323 for comparison..
FPS Review take: The interesting development here is not capacity, it is that power supplies have started monitoring the connector rather than trusting it, and MSI shipping an Afterburner hook that throttles the card when it detects an imbalance is a more useful safety feature than another 200 W of headroom. It is also an admission about the state of 12V-2×6. On the other side of the fence, a sub litre 1200 W Platinum unit for $259 is the more relevant product, since SFF builders have had the worst options for the longest. For our own high wattage and small form factor testing, see our Seasonic PRIME PX-1600 1600W Platinum review and our Lian Li SP850 SFX review.
Lexar Play X 1TB
Written reviews:
Consensus summary: This looks like an ordinary PCIe 4.0 M.2 drive and is not one. The Play X is among the first retail products built on the integrated packaged mSSD concept from Lexar parent Longsys, where the controller, NAND, and supporting components all live inside a single package instead of being soldered onto a populated PCB, a change Longsys claims cuts manufacturing cost by around 10 percent and reduces solder joint failures, with the tradeoff that any internal defect means replacing the whole package. The module measures 30 by 20 mm, 2 mm narrower than M.2 2230, and ships with an aluminium adapter frame that converts it to a standard 2280 footprint while adding some cooling. Lexar quotes 7,400 MB/s read and 6,400 MB/s write for the 1TB model on a four channel MaxIO MAP1602 controller with 3D TLC, the same controller as the popular NM790, and warranty is the usual five years and 600 TBW. ComputerBase found the packaging concept works in practice and hit the rated sequential numbers, but performance overall is mid pack at best, with a weak copy test, comparatively high latency at 48 microseconds, and random reads of 67 MB/s that trail the NM790 by about 15 percent, landing the drive in the lower third of the PCIe 4.0 field near the Samsung 990 and Seagate FireCuda 530R. Power draw and thermals are the standout positives at a 4.7 W peak, 64 C bare in the adapter, and a steady 45 C under sustained load with a motherboard heatsink, with no sign of throttling.
FPS Review take: If single package assembly really trims 10 percent off SSD build cost, that matters enormously in a market where DRAM and NAND pricing has been the dominant story all year, and the 2230 to 2280 flexibility is genuinely useful for anyone who moves drives between a handheld and a desktop. We reviewed the conventional sibling in our Lexar PLAY 2280 SE 4TB review, which is the fair comparison for what you give up.
Budget Motherboards: Maxsun Terminator B850M Pro II and MSI Z890 GAMING PLUS WIFI6E
Written reviews:
- Tom’s Hardware (Maxsun Terminator B850M Pro II): https://www.tomshardware.com/pc-components/motherboards/maxsun-terminator-b850m-pro-ii-motherboard-review
- TweakTown (MSI Z890 GAMING PLUS WIFI6E): https://www.tweaktown.com/reviews/11570/msi-z890-gaming-plus-wifi6e-z890-for-under-dollars200/index.html
Consensus summary: Two value boards, two verdicts split by pricing rather than engineering. Tom’s Hardware finds the Maxsun Terminator B850M Pro II a well equipped micro ATX AM5 board on paper, with Wi-Fi 7, 2.5 GbE, three M.2 sockets including two at PCIe 5.0, two full length PCIe slots, and support for flagship Ryzen parts, held back by a 10 Gbps USB ceiling, a mediocre audio codec, and all three M.2 sockets sitting under one plate heatsink beneath the graphics card. Their conclusion is that it is competitive at 199 dollars and a tough sell above that, which is where US pricing currently sits. TweakTown reaches the friendlier verdict on the MSI Z890 GAMING PLUS WIFI6E at 199 dollars, calling out 5 GbE, four expansion slots which is rare at any price, a decent VRM, high speed memory support, and a peak VRM temperature of 57 C that they consider impressive for a sub 200 dollar board, with SSD temperatures the weak spot and the LGA 1851 dead end as the strategic caveat.
FPS Review take: We reviewed this exact MSI board ourselves, so read our MSI Z890 GAMING PLUS WiFi6E review alongside TweakTown’s and note we came to a similar conclusion on where the corners were cut. The broader point both reviews stumble into is that a 199 dollar board is now the interesting part of the market precisely because it is the part the AI buildout has not gotten around to taxing yet. For the step up, our MSI MAG Z890 TOMAHAWK WiFi II review is the closest comparison.
Hell Let Loose: Vietnam PC Performance
Written reviews:
- ComputerBase: https://www.computerbase.de/artikel/gaming/hell-let-loose-vietnam-benchmark-test.98755/
Consensus summary: Expression Games moved the series from Unreal Engine 4 to UE5.5.4, and ComputerBase’s verdict is that it looks good for a pure multiplayer title without being uniformly impressive, with the jungle and water carrying the presentation while characters and LOD transitions let it down. Turning off hardware Lumen does not disable ray tracing, it falls back to software Lumen, and ComputerBase argues hardware Lumen is worth the cost for a competitive reason as much as a visual one: in dense jungle and indoors, software Lumen leaves scenes genuinely dark and enemies harder to spot. That costs 22 percent on an RTX 5070 Ti at 4K, 20 percent on an RTX 4080 Super, 18 percent on an RX 9070 XT, and 24 percent on an RX 7900 XTX. The launch upscaling situation is the real complaint, since DLSS 4 and 4.5 are present while FSR and XeSS are not, leaving Radeon and Arc owners on TSR at roughly a 5 percent deficit to DLSS 4 plus a clear image quality loss, with FSR promised shortly after launch. GeForce cards lead Radeon throughout and the gap widens with resolution, Arc struggles more, and frame pacing and stability were clean with no crashes during testing.
FPS Review take: Shipping a UE5 multiplayer game in 2026 with DLSS but no FSR is a choice, and it is the kind of choice that quietly decides which brand of card feels good in a title for the first month. The Lumen visibility argument is the part worth passing along to anyone who plays this competitively, because the fastest settings are also the settings that hide people in the treeline. For where these tiers land against each other in our own testing, see our Radeon RX 9070 GRE versus RTX 5070 performance review.
G.SKILL Trident Z5 NeoX RGB DDR5-6000 EXPO ULL
Written reviews:
- Tom’s Hardware: https://www.tomshardware.com/pc-components/ram/g-skill-trident-z5-neox-rgb-ddr5-6000-c30-2x16gb-review
- KitGuru: https://www.kitguru.net/components/memory/leo-waldock/g-skill-trident-z5-neox-w-expo-ull/
Video reviews:
- KitGuru (AMD EXPO ULL tested): https://www.youtube.com/watch?v=JwGSpTLo9gM
Consensus summary: EXPO ULL is AMD’s answer to the manually tuned kit, exposing secondary and tertiary sub timings that standard EXPO profiles leave alone for compatibility, and it requires AGESA 1.3.0.1b Path A or newer to load at all. Tom’s Hardware calls the DDR5-6000 CL30 NeoX kit the top echelon of AMD memory and stabilized it at DDR5-6200 on a 0.05 V bump to 1.40 V while holding the same tight timings, but is careful to set expectations: there are no double digit uplifts, the gains vary heavily by hardware combination, and ULL is aimed at vanilla Ryzen because X3D parts are unlikely to see perceptible benefit. KitGuru’s testing puts a number on that, with roughly 3 percent in Assassin’s Creed Mirage at 1440p on a regular Ryzen 7 and effectively nothing on a 3D V-Cache chip, against G.SKILL’s claimed 4 percent. Both land on pricing as the deciding factor in the current memory market.
FPS Review take: Two independent findings that ULL does nothing measurable for X3D owners is the useful result here, and it is worth stating plainly because the sub timing pitch sounds like it should benefit everyone. If you want the same question asked at the other end of the frequency curve, our G.SKILL Trident Z5 Neo RGB DDR5-8000 EXPO review found that raw frequency loses to lower latency in gaming because of the memory controller clock penalty, which is the same lesson from the other side. Our Trident Z5 CK DDR5-8400 CU-DIMM review and our Core Ultra 7 270K Plus RAM scaling testing cover the Intel side.

Discussion
Join Discussion →No replies yet. Be the first to reply on the forum.