This week's release of the Intel Arc 32.0.101.8509 WHQL driver package has benefited many of the company's in-house graphics product lines. Current and past-generations of hardware are enjoying XeSS 3 with multi-frame generation (MFG) support. Earlier today, PC Games Hardware Germany (PCGH.de) published an in-depth article that covers the new feature on Arc "Battlemage" B580 12 GB and an Arc "Alchemist" A770 16 GB desktop graphics card models. The "XeSS Quality" setting was selected across seven game test scenarios, at 2560 x 1440 (QHD) resolutions—involving: Assassin's Creed Shadows, Battlefield 6, Clair Obscur: Expedition 33, Cyberpunk 2077, Dying Light: The Beast, F1 '25, and Hogwarts Legacy.

PCGH.de observed Team Blue's top-end "Alchemist" option doing very well, at least with XeLL (a low latency function) enabled, in particular with the driver overridden MFG three and fours-times modes. Despite launching back in late 2022, their ASRock Arc A770 Phantom OC 16 GB test sample has enjoyed: "something of a second spring thanks to MFG. While the card isn't (yet?) maturing as much as NVIDIA's (GeForce) RTX 2000-series, its usability has been significantly improved overnight. Thanks to 16 GB of memory, there's no need to worry about bottlenecks. The Arc B580 scales even better thanks to the improved hardware, and performs comparably." Despite the pleasing findings, PCGH's reviewer notes that Multi Frame Generation (MFG) is not an all-encompassing property—after all, only a certain selection of titles support XeSS. Due to visual discrepancies being tracked in Battlefield 6 sessions, PCGH preferred the presentation of native frames operation or rival frame generation tech (FSR 4 and DLSS 4.5).

Last week, computer memory industry observers noted the US Government's "very brief" removal of ChangXin Memory Technologies (CXMT) from a trade blacklist. In early 2026, leading North American OEMs were reported as seeking alternative suppliers of consumer-grade DDR5 and DDR4 memory. China's CXMT seems to be an ideal frontrunner for brands that are currently struggling with procurements of more mainstream modules—as produced by Micron, SK hynix, and Samsung. Australia's Hardware Unboxed is in a fortunate geographical position; where CXMT IC-based kits are (kind of) readily available via big box stores. The ongoing "DDR pricing nightmare" has prompted Steve Walton's recent purchase of a cheaper KingBank KFRW DDR5-6000 CL36 32 GB (2 x 16 GB) White RGB kit. The KFRW range seems to be an Aussie market exclusive, derived from KingBank's Soarblade UDIMM product family.

Primarily, Walton compared this less familiar Chinese brand set to a "sweet spot" G.Skill Trident Z5 Neo DDR5-6000 CL30 32 GB (2 x 16 GB) kit. Prior to the present-day crisis, this former price-to-performance champ was available as low as $115 in North America. Fast-forward to February 2026, and the kit's price tag has reached a $490 high. A few weeks ago, Walton managed to pick up one of the cheapest KingBank kits, that happened to feature CXMT's homegrown modules—he did not disclose how much was forked over for review candidate ownership. Today's Hardware Unboxed investigative video notes that KingBank prices have increased around the mid-February period—Australians have to pay a minimum of AU$649 (~$458 USD) for currently in-stock CMXT-based sticks. Curiously, a screenshot of Mwave store listings shows a KingBank Hynix M-die-based kit being sold for a relatively reasonable AU$599 (~$423 USD). Overall, Walton recommends that locals pick up Chinese-made DRAM, if price tags are reasonable.

Since the introduction of Intel's brand-new "Panther Lake-H" mobile processor series, the flagship Arc B390 integrated graphics solution has received the lion's share of media attention. The impressive 12 Xe3-core iGPU is reserved for Core Ultra 300 series X9 and X7 models—likely a sore point for budget-conscious prospective buyers. When considering Team Blue's (roughly) "20% upcharge for a generational leap," customers seek current-gen solace elsewhere (AMD Radeon 890M). Fortunately, the step-down Arc B370 model seems to be relatively performant. The (presumably) much cheaper in-house alternative has been test driven by Notebookcheck. A preview report, published on February 13, provides an initial look at the 10 Xe3-core contender, that was onboard a Core Ultra 5 338H engineering sample. The German publication noted that the 12-core (4P + 8E) processor test subject ran with a "steady power limit of 35 watts." When comparing the Arc B370 to its more capable sibling, Notebookcheck outlined key points of interest: "the maximum GPU clock is also slightly lower at 2.4 vs. 2.5 GHz. Just like models (armed with) the Arc B390 GPU, the Core Ultra 5 338H is only available with soldered RAM (up to 8533 MHz)."

A disappointing Furmark score leaked out late last year, but it is reassuring to see the B370 performing admirably a few months later. Notebookcheck's tester observed this model being: "beaten by the Arc B390 in the synthetic benchmarks, but the power limits also play a role here...The results of the Arc B370 are very promising in general and there is a big advantage over previous iGPUs from Intel (Arc Graphics 140V/140T) as well as AMD (Radeon 890M)." Their 3DMark/Time Spy graphics chart shows the Core Ultra 5 338H engineering sample's Arc B370 iGPU scoring 5933 points. The B390 managed 6679 points, when evaluated on an ASUS ExpertBook Ultra B9406CAA test platform. When analyzing performance at lower TDP values, Notebookcheck noticed that the Arc B370's performance being "virtually identical to the Arc B390 at 20 Watts, while the deficit is just 6 % at 35 Watts." Further comprehensive tests are expected in the coming weeks, as Intel's Core Ultra 5 338H heads to retail in finalized form.

Samsung presented its brand-new Exynos 2600 application processor just before Christmas—this big announcement followed many months of leaks, throughout 2025. Weeks later, a mysterious follow-up model—dubbed "Exynos 2700" by South Korean industry insiders—was tentatively linked to an allegedly more refined 2 nm GAA node process. Officially, Samsung has opted into using its own foundry's SF2 production line for the imminent Exynos 2600 SoC-powered Galaxy S26 smartphone series. Lately, the more advanced "SF2P" node process has been uttered in the same sentence as possible AMD, Qualcomm, and Tesla chip making deals. The South Korean megacorporation is very likely keen to drum up interest from new and returning clients, and their latest Exynos mobile chip is being used as a proof of concept for the regular SF2 node (2 nm GAA), along with proprietary "Heat Path Block" technology.

Unsurprisingly, as the Exynos 2600 AP closes in on its release period, pre-launch samples have posted benchmark data. Last week, TechPowerUp covered this next-gen chip's Geekbench 6 OpenCL score of 24,964 points—achieved by the onboard Xclipse 960 iGPU (aka "JUNO"). Earlier today, a slightly mysterious "Samsung Electronics Co., Ltd. Full Android on S5E9975 ERD" test device posted a not-so-impressive OpenCL score of 15618, in Geekbench 6.2.2 for Android AArch64. A Geekbench browser database entry details a 10-core processor candidate that packs a Samsung Xclipse 970 integrated graphics solution. Presumably, the numerical model climb—from 960 going up to 970—implies the "S5E9975 ERD" validation platform hosting an "Exynos 2700" mobile chipset. At first, Ice Universe—a self-described tech expert of global renown—was excited by the seemingly premature appearance of an "Exynos 2700" engineering sample. A lower than expected benchmark score was blamed on the test setup being an "ERD engineering board."

Intel's "Bartlett Lake-S" (Performance) P-Core-only processor design has made a surprise appearance, after a long period of developmental silence. Earlier, today X86 is dead&back put a spotlight on the intriguing entry of Team Blue's unreleased "Core 7 253PE" SKU within PassMark's CPU benchmark database. The keen watcher of not-yet-released PC hardware believes that this particular unit is possibly the first sample of Team Blue's 10-core "Bartlett Lake-S" variant to appear in the wild. This further refinement of 14th Gen "Raptor Lake Refresh" tech popped up a few times last year, but the TechPowerUp news archive shows a dry period during third and fourth quarters. An August report alluded to an embedded market target, the existence of a flagship 12 P-core model, and a potential launch under the "Intel Core Series 2" banner—minus an Ultra tag, due to lack of onboard NPUs.

All P-Core configurations could appeal to budget-conscious PC gamers. Efficiency-oriented (E and LPE) units are considered superfluous for these purposes. X86 is dead&back was pleased to find PassMark data that signalled improved multicore performance over the related Core i5 14400 "Raptor Lake-R" processor. They exclaimed: "Core 7 253PE (has) more MT performance than Core i5 14400 (6P + 4E)!" Their comparison shows a 20.6% uplift here, but a slight regression in terms of single-core scores (-2.73%). The sole "Core 7 253PE" sample's PassMark multi-thread rating, of 31802, closes in on the 14-core (6P + 8E) Core i5-14500 model's tally (31121). According yesterday's PassMark diagnostic, the newer chip utilizes Intel's UHD Graphics 770 integrated solution (akin to Core i5-14500), whilst the Core i5 14400 makes do with the UHD 730 iGPU. There are apparent differences across the various levels of cache, but the "Core 7 253PE" SKU seems to enjoy a larger overall pool—36 MB—when cross-examined with its older relatives.

Recent Geekbench Browser database entries have outed test scores and basic specifications for two of Intel's upcoming "Arrow Lake Refresh" flagships. On January 13, a not-yet-launched Core Ultra 9 290HX Plus mobile unit was tracked in a new Acer Predator Helios gaming laptop. A few hours earlier, Team Blue's Core Ultra 9 290K Plus desktop chip entered the fray onboard a test rig built around Gigabyte's top-level and world record-holding Z890 AORUS TACHYON ICE mainboard configured with 48 GB DDR5-8000 of system memory. This leak shows the future 24-core "Arrow Lake Refresh" range topper posting overall scores of 3456 single-core and 24610 multicore points in Geekbench 6.5 test scenarios. Naturally, news reports have quickly compared the Core Ultra 9 290K Plus to a nearby non-refreshed "Arrow Lake" part—Core Ultra 9 285K.

The newcomer managed to build upon the current flagship's officially ranked achievements (based upon multiple entries); doing 7% better in single-core stakes, and 9% in multicore tests. When compared to AMD's fanciest "Granite Ridge" desktop offering—Ryzen 9 9950X3D—the Core Ultra 9 290K Plus is 2% faster in terms of single-core results, and 11% above in multicore. Interestingly, the lesser Core Ultra 9 270K Plus model was Geekbenched in December—producing 3235 (single-core) and 21368 (multicore) tallies. The flagship "Arrow Lake-S Refresh" SKU is about 7% and 15% faster (respectively) than its (also) 24-core sibling. According to Geekbench diagnostic data and other specification leaks, the two (8P + 16E) siblings feature minor differences in core clock speeds, so it is possible that Team Blue engineers have tuned the more expensive option to a greater degree. Golden Pig Upgrade has proposed a March/April launch window for the Core Ultra 200 Plus desktop family, consisting of the two aforementioned models and an 18-core (6P + 12E) Core Ultra 5 250K Plus SKU.

Intel's often leaked "Arrow Lake Refresh" processor family was not put under an official spotlight at CES 2026—supposed desktop variants (aka "ARL-S R") took fourth place in TechPowerUp's "Biggest No-shows" of the recently concluded trade event. Team Blue's main focus was the truly new-gen "Panther Lake" mobile series, with a wide swath of manufacturing partners showing off related devices on the showroom floor. Understandably, mid-gen product refreshes are less exciting than the freshest and shiniest products. A very low number of hints were dropped in Las Vegas; reportedly at the Gigabyte and Minisforum booths. Web sleuths are still seeking signs of any pre-launch Core Ultra Plus "Arrow Lake Refresh" information, and a new discovery has brought to light a new "Core Ultra 9 290HX Plus" SKU. Yesterday afternoon, an "Acer Predator PHN18-I71" test system was tracked on the Geekbench Browser database. Initial investigations point to the possible recent evaluation of an upcoming Helios 18 AI gaming laptop.

A Geekbench 6.5 diagnostic session mentions an ARL iX2_ARX board, 64 GB of system memory, and a 24-core "Genuine Intel Family 6 Model 198 Stepping 2" processor. The test platform's socket package is listed as an LGA 1851 type; a big error—since Team Blue's "Arrow Lake-HX" chip line is usually paired up with the BGA "FCBGA2114" format. The not-yet-launched "Core Ultra 9 290HX Plus" model seems to be prepped to take top billing over the current "Arrow Lake-HX" (non-refreshed) Core Ultra 9 285HX mobile chip. The latest leak shows that there is some nomenclature overlap with a flagship desktop equivalent—Core Ultra 290K Plus. Intel's "Plus" designation seems to indicate refreshed status. Yesterday's Geekbench 6.5 results look to be incomplete—with lower than expected scores and clock speeds; VideoCardz proposes that the evaluator ended benchmarking partway through the session.

Maxon is thrilled to announce the highly anticipated release of Cinebench 2026, its latest iteration of the industry-standard benchmarking software, which has been a cornerstone in computer performance evaluation for two decades. Widely used by reviewers, hardware manufacturers, and everyday computer owners, Cinebench utilizes the power of Redshift, Cinema 4D's default rendering engine, to accurately assess and evaluate CPU and GPU performance using real-world 3D rendering workloads.

Cinebench 2026 takes advantage of the latest developments in hardware technology, adding compatibility with NVIDIA's new Blackwell GPUs (5000 series) and AMD 9000 series GPUs on Windows, as well as NVIDIA Hopper and Blackwell data center GPUs. Additionally, Cinebench 2026 now supports Apple M4 and M5 powered systems.

A fairly unique "Gorgon Point" mobile processor has turned up multiple times within the Geekbench Browser database. Instead of discovering the usual Geekbench 6.x-based entries, Olrak29 put a spotlight on scores and diagnostic data produced by Geekbench AI 1.5 benchmark sessions. Six Fujitsu Client Computing test systems posted results on December 9, all featuring an unreleased Ryzen AI 7 445 APU. So far, leaks of slightly refreshed "Strix Point" and "Krackan Point" designs have been compared to direct predecessors, but the leaked Ryzen AI 7 445 SKU seems to be an outlier—in terms of nomenclature and core configuration. The closest forebear would be Team Red's Ryzen AI 5 340 "Krackan Point" mobile chip; sporting a mixture of "Zen 5" and "Zen 5c" cores.

The current-gen 6-core design consists of three standard cores, and three "compact" alternatives. When looking at Geekbench browser entries, the forthcoming Ryzen AI 7 445 model sports two "Zen 5" cores (in cluster 1) and four "Zen 5c" units (in cluster 2). Geekbench AI 1.5 diagnostics do not provide a comprehensive set of specifications, but base frequency was tracked at 2.00 GHz, and the detected integrated graphics solution was AMD's RDNA 3.5-based Radeon 840M. As mentioned earlier, the not-yet-launched CPU's unusual core configuration was not copied from any readily available/current processor products—Team Red could be readying the Ryzen AI 7 445 APU for eventual deployment in efficiency-oriented platforms.

A mysterious "OneXPlayer X1 i" handheld gaming device has leaked out via benchmark result databases. Pre-launch leaks indicate the presence of an Intel "Core Ultra 5 338H" APU within evaluated hardware. PC hardware news outlets have focused on an intriguing FurMark 2 diagnostic—published around mid-November, but brought to light by momomo_us this week. Driver information points to "Panther Lake-H" processor test subjects being engineering samples, and quick interpretations of benchmark data demonstrate performance lagging behind previous-gen equivalents. Tom's Hardware compared the leaked results to archived "Lunar Lake" (Xe-LPG+) and "Arrow Lake" (Xe2-LPG) scores in GPU Magick's database.

"Xe3 iGPU scoring 2383 points in Furmark's benchmark test at a reported max boost clock of up to 2300 MHz, and up to 36 watts of power—possibly for the entire APU. However, the GPU is outperformed by Intel's outgoing Arc 140V and 140T counterparts, with the 140 V scoring 2736 points and the 140T scoring 2398 points, respectively...Both 140-series iGPUs operate at between 10 and 12 watts (below the next-gen design's 35 W TDP), making the B370 worse on energy efficiency as well." The Arc B370 integrated solution is touted as being part of an entry-level package. Curiously, Intel's official slides—from two months ago—only outline a "smaller four-Xe-core GPU tile" and a "larger twelve-Xe-core version." The early "Core Ultra 5 338H" candidates seem to leverage a "somewhere between" 10-core iGPU design. The Arc B370's apparent "B-series" designation adds extra confusion—"B" is normally associated with the "Battlemage" generation, but recent Team Blue declarations have "Panther Lake" Xe3 iGPUs not categorized as using next-gen "Celestial" graphics architecture.

The cadence of AMD Ryzen 7 9850X3D-related leaks has increased over the past week and a half; trickling out from an official outlet (by mistake), and shipping information. This uptick suggests an imminent launch of finalized units, possibly at next month's CES trade show. The highly visible event will provide opportune moments to introduce new and refreshed products alike. Weeks ahead of grand proceedings, Team Red's motherboard partners seem to be taking the (predicted) next 8-core gaming champion for a spin. Geekbench Browser database entries turned up mid-way through November, but were discovered weeks later by Gray/Olrak29. Test rigs utilized Maxsun's MS-eSport B850ITX mainboard model, and Colorful's CVN B850M GAMING FROZEN V14A design.

Both validated sets of Geekbench 6.3 (under Windows 11 64-bit) results indicate performance in the ballpark of AMD's readily available Ryzen 7 9800X3D CPU, with multicore scores tallies lower than anticipated. It should be noted that the Maxsun and Colorful test rigs were not running system memory (both 32 GB in total) at maximum transfer rates. Sub-4800 MT/s levels were recorded during mid-November sessions. VideoCardz notes other possible factors, including the early/in-progress nature of firmware and cautious implementation of power parameters. The Colorful mainboard-based build managed to coax its hosted processor's core clock up to 5602 MHz, whilst the Maxsun's CPU sample trailed behind at 5385 MHz. The two database entries list identical credentials for the much-leaked Ryzen 7 9850X3D SKU: 8 cores, 16 threads, 4.70 GHz base clock, and 96 MB of L3 cache. Both test subjects are identified as "Authentic AMD Family 26 Model 68 Stepping 0" units, belonging to the "Granite Ridge" processor generation.

A list of unannounced Intel Xeon "Granite Rapids-WS" processors turned up online two weeks ago; as a result of momomo_us's detective work. Team Blue's supposed next wave of high-end desktop (HEDT) workstation CPUs are expected to be introduced at CES 2026, but unofficial details have already spread across PC hardware discussion communities. The mid-month leak provided an early look at eleven possible 600-series SKU candidates (supported by the W890 platform), complete with basic specifications—clock speeds and L3 cache capacities. A fresh discovery, again courtesy of momomo_us, grants further insight into the desktop family's sub-flagship chip: "696X."

Lately, a test system entry has turned up on the SiSoftware Official Benchmark Ranker database. The "Xeon 696X" model appears to feature 64 cores, and 128 threads. The lowest measured clock speed was 4.4 GHz, going up to 4.6 GHz (averaging at 4.51 GHz). A 2.4 GHz base clock was listed by momomo_us, a fortnight ago. A 350 W TDP rating was picked up by SiSoftware Sandra's suite. The latest leak reaffirms a presence of 336 MB L3 cache. A total of 128 MB L2 cache is shared across the processor's 64 cores (so 2 MB each). Intel's rumored forthcoming re-entry into HEDT territories could carry Xeon 6 family branding, albeit sporting an added "W"—denoting workstation applications.

Mid-September news reports placed NVIDIA's Chinese market exclusive RTX PRO 6000D model in a dubious position. Insiders believed that this "Blackwell" architecture-based AI accelerator was banned by the nation's Cyberspace Administration (CAC), in favor of boosting the prospects of domestic equivalents. Team Green has gingerly navigated these choppy waters for a number of years; an encouraging stack of sanction-conforming products have been released during trying times. These variants often feature lesser specifications—for example, the aforementioned RTX PRO 6000D is rumored to be a spin-off of the familiar RTX PRO 6000 Blackwell card. Despite facing an uncertain future in China, a curious Geekbench browser entry shines a ray of hope for the latest workstation-grade "D"-affixed product.

Team Green's upcoming NVIDIA RTX 6000D Blackwell GPU—now missing a "PRO" designation—seems to sport 156 compute units (CUs)/SMs; translating to 19,968 CUDA cores. An "AIC CB502-NM" test system was running this 84 GB GDDR7-laden card in conjunction with an AMD EPYC 9654 CPU. In comparison, the unpasteurized RTX PRO 6000 Blackwell model harnesses a largely uncompromised GB202 chip. 24064 CUDA cores are on tap, alongside 96 GB GDDR7 of video memory. The weaker 6000D makes do with a reduced 2430 MHz clock frequency (versus its sibling's 2600+ MHz credentials). Expert analysis posits that the cut-down card makes use of a 448-bit bus interface, rather than the standard unit's 512-bit avenue. VideoCardz has kindly formulated a comparison table of NVIDIA's RTX 6000 "Blackwell" products. The latest addition managed to generate an OpenCL score of 390656 points in Geekbench's 6.5 gauntlet (for Linux AVX2). The evaluated AIC computer was operating on Rocky Linux 9.4 (Blue Onyx).

Just over a week ago, the Radeon RX 9060 XT graphics card became official; introduced as the fastest gaming GPU option for "under $350." It represents the second wave of AMD's RDNA 4 GPU generation, but definitive verdicts are not expected until a lifting of review embargoes—likely happening the day before retail release: June 4. Evaluation samples are very likely in the possession of media outlets and influencers; as evidenced by pre-launch benchmark results appearing within the Geekbench Browser database. A nondescript Radeon RX 9060 XT 16 GB model was put through OpenCL and Vulkan wringers, via "Geekbench 6.2.2 for Windows AVX2." Overall tallies are 109315 and 124251, respectively. The test rig consisted of Team Red's Ryzen 7 9800X3D CPU, a Gigabyte X870E AORUS MASTER motherboard, and 32 GB of DDR5-8000 RAM.

Geekbench results are not the best indicators of gaming performance on modern PC platforms, but semi-useful data can be compared to figures generated by predecessors and current-gen siblings. Quick analysis points to the benchmarked Radeon RX 9060 XT 16 GB sample pulling ahead of its RDNA 3-based elder—the Radeon RX 7600 XT—by 25 to 31%. Stepping up against the Radeon RX 7700 XT 12 GB model, the plucky new candidate trails by 14% in OpenCL stakes and 12% in Vulkan. Naturally, a performance gulf exists between the Radeon RX 9070 16 GB (non-XT) card and its forthcoming smaller sibling—almost a +23% difference in OpenCL, and roughly +32% in Vulkan. Crucially, other Geekbench Browser entries suggest that NVIDIA's competing GeForce RTX 5060 Ti 16/8 GB and RTX 5060 8 GB designs hold slight advantages in terms of OpenCL numbers. AMD's Navi 44 XT GPU-powered card nudges just beyond the RTX 5060's overall Vulkan result. A clearer picture of Radeon RX 9060 XT's standing will be painted next week; stay tuned for TechPowerUp's inevitable in-depth analyses of board partner specimens.

AMD and a select bunch of its board partners are set to launch Radeon RX 9070 GRE 12 GB graphics card models tomorrow; starting as exclusives for China's PC gaming hardware market. Just before an unleashing of retail stock, local media outlets have published reviews—mostly covering brand-new ASUS, Sapphire, and XFX products. The RDNA 4 generation's first "Great Radeon Edition" (GRE) is positioned as a slightly cheaper alternative to Team Red's Radeon RX 9070 (non-XT) 16 GB model; 4199 RMB versus 4499 RMB (respectively, including VAT). In general, Chinese evaluators seem to express lukewarm opinions about the Radeon RX 9070 GRE's value-to-performance ratio. After all, this is a cut-down design—a "reduced" Navi 48 chip makes do with 3072 Stream Processors. The card's 12 GB of GDDR6 VRAM configuration is paired up with a 192-bit memory interface.

Carbon Based Technology's video review presented benchmark results that placed AMD's new contender on par with a previous-gen card: Radeon RX 7900 GRE 16 GB. Considering that this RDNA 3 era Golden Rabbit Edition (GRE) model launched globally with an MSRP of $549, its Navi 48 XL GPU-based descendant's ~$580 (USD) guide price appears to be mildly nonsensical. GamerSky pitched their ASUS ATS RX 9070 GRE MEGALODON OC sample against mid-range and lower level current-gen NVIDIA gaming products: ""through testing, we can find that at 4K resolution, the GeForce RTX 5070 12 GB performs the best, 5% higher than the ASUS RX 9070 GRE Megalodon. As the resolution decreases, its lead also decreases, and at 2K resolution it is only 2% higher. At 1080p resolution, the difference is only 1%. At the same time, compared with RTX 5060 Ti 16G, ASUS RX 9070 GRE Megalodon has a greater advantage. The performance of its competitor's RTX 5060 Ti 16G is only 77% of that of RX 9070 GRE at 4K and 2K resolutions. At 1080p, its performance increased slightly to 79%." AMD and involved AIBs could be testing the waters with an initial Chinese market exclusive release, but Western news outlets reckon that a more aggressive pricing strategy is needed for a (potential) proper global rollout of Radeon RX 9070 GRE cards.

NVIDIA's GeForce RTX 5060 Ti 8 GB graphics card design received little fanfare when review embargoes lifted mid-way through the working week. Reportedly by official instruction, involved board partners sent out 16 GB samples to evaluators. Multiple Western outlets are currently attempting to source GeForce RTX 5060 Ti 8 GB cards—on their own dime—including TechPowerUp. As mentioned in his conclusive rundown of PALIT's GeForce RTX 5060 Ti Infinity 3 16 GB model, W1zzard commented on this situation: "personally, I'm very interested in my results for the RTX 5060 Ti 8 GB, which I'm trying to buy now." The ever reliable harukaze5719 has already stumbled upon one such review. Yesterday, Carbon-based Technology Research Institute (CBTRI) uploaded their findings onto the Chinese bilibili video platform.

Two ASUS options were compared to each other: an 8 GB Hatsune Miku Special Edition card, and a better known property: PRIME RTX 5060 Ti 16 GB. In most situations the two variants perform similarly. A clear difference was demonstrated when CBTRI's lab test moved into a DLSS 4 with Multi-Frame Generation (MFG) phase. Both harukaze5719 and Tom's Hardware noted a significant gulf—the latter's report observed: "in Cyberpunk 2077, for example, the RTX 5060 Ti 8 GB inexplicably performed worse than the RTX 4060 Ti 8 GB at native 1440p resolution. While enabling MFG helped improve performance, pushing it to 4x delivered underwhelming results, with the 16 GB version providing 22% higher performance than the 8 GB card." Rumors have swirled about the late arrival of GeForce RTX 5060 Ti 8 GB cards at retail; potentially a week after the launch of 16 GB siblings. As evidenced by early results, potential buyers should consider paying a little extra ($50) for a larger pool of VRAM. Team Green's introductory material outlined starter price tags of $429 (16 GB) and $379 (8 GB).

Super Micro Computer, Inc. (SMCI), a Total IT Solution Provider for AI/ML, HPC, Cloud, Storage, and 5G/Edge, has announced first-to-market industry leading performance on several MLPerf Inference v5.0 benchmarks, using the 8-GPU. The 4U liquid-cooled and 10U air-cooled systems achieved the best performance in select benchmarks. Supermicro demonstrated more than 3 times the tokens per second (Token/s) generation for Llama2-70B and Llama3.1-405B benchmarks compared to H200 8-GPU systems. "Supermicro remains a leader in the AI industry, as evidenced by the first new benchmarks released by MLCommons in 2025," said Charles Liang, president and CEO of Supermicro. "Our building block architecture enables us to be first-to-market with a diverse range of systems optimized for various workloads. We continue to collaborate closely with NVIDIA to fine-tune our systems and secure a leadership position in AI workloads." Learn more about the new MLPerf v5.0 Inference benchmarks here.

Supermicro is the only system vendor publishing record MLPerf inference performance (on select benchmarks) for both the air-cooled and liquid-cooled NVIDIA HGX B200 8-GPU systems. Both air-cooled and liquid-cooled systems were operational before the MLCommons benchmark start date. Supermicro engineers optimized the systems and software to showcase the impressive performance. Within the operating margin, the Supermicro air-cooled B200 system exhibited the same level of performance as the liquid-cooled B200 system. Supermicro has been delivering these systems to customers while we conducted the benchmarks. MLCommons emphasizes that all results be reproducible, that the products are available and that the results can be audited by other MLCommons members. Supermicro engineers optimized the systems and software, as allowed by the MLCommons rules.

NVIDIA and its laptop manufacturing partners managed to squeeze out higher end models at the start of the week (March 31); qualifying just in time as a Q1 2025 launch. As predicted by PC gaming hardware watchdogs, conditions on day one—for the general public—were far from perfect. Media and influencer outlets received pre-launch evaluation units—Monday's embargo lift did not open up floodgates to a massive number of published/uploaded reviews. Independent benchmarking of Team Green's flagship—GeForce RTX 5090 Mobile—produced somewhat underwhelming results. To summarize, several outlets—including Notebookcheck—observed NVIDIA's topmost laptop-oriented GPU trailing way behind its desktop equivalent in lab tests. Notebookcheck commented on these findings: "laptop gamers will want to keep their expectations in check as the mobile GeForce RTX 5090 can be 50 percent slower than the desktop counterpart as shown by our benchmarks. The enormous gap between the mobile RTX 5090 and desktop RTX 5090 and the somewhat disappointing leap over the outgoing mobile RTX 4080 can be mostly attributed to TGP."

The German online publication was more impressed with NVIDIA's sub-flagship model—two Ryzen 9 9955HX-powered Schenker XMG Neo 16 test units—sporting almost identical specifications—were pitched against each other, a resultant mini-review of benched figures was made available earlier today. Notebookcheck's Allen Ngo provided some context: "3DMark benchmarks...show that the (Schenker Neo's) GeForce RTX 5080 Mobile unit is roughly 10 to 15 percent slower than its pricier sibling. This deficit translates fairly well when running actual games like Baldur's Gate 3, Final Fantasy XV, Alan Wake 2, or Assassin's Creed Shadows. As usual, the deficit is widest when running at 4K resolutions on demanding games and smallest when running at lower resolutions where graphics become less GPU bound. A notable observation is that the performance gap between the mobile RTX 5080 and mobile RTX 5090 would remain the same, whether or not DLSS is enabled. When running Assassin's Creed Shadows with DLSS on, for example, the mobile RTX 5090 would maintain its 15 percent lead over the mobile RTX 5080. The relatively small performance drop between the two enthusiast GPUs means it may be worth configuring laptops with the RTX 5080 instead of the RTX 5090 to save on hundreds of dollars or for better performance-per-dollar." As demonstrated by Bestware.com's system configurator, the XMG NEO 16 (A25) SKU with a GeForce RTX 5090 Mobile GPU demands a €855 (~$928 USD) upcharge over an RTX 5080-based build.

Last Friday, AMD confirmed finalized price points for its upcoming Ryzen 9 9950X3D ($699) and 9900X3D ($599) gaming processors—both launching on March 12. Media outlets are very likely finalizing their evaluations of review silicon; official embargoes are due for lifting tomorrow (March 11). By Team Red decree, a drip feed of pre-launch information was restricted to teasers, a loose March launch window, and an unveiling of basic specifications (at CES 2025). A trickle of mid-January to early March leaks have painted an incomplete picture of performance expectations for the 3D V-Cache-equipped 16 and 12-core parts. A fresh NDA-busting disclosure has arrived online, courtesy of an alleged Ryzen 9 9950X3D sample's set of benchmark scores.

A pre-release candidate posted single and multi-thread ratings of 4739 and 69,701 (respectively), upon completion of PassMark tests. Based on this information, a comparison chart was assembled—pitching the Ryzen 9 9950X3D against its direct predecessor (7950X3D), a Zen 5 relative (9950X), and competition from Intel (Core Ultra 9 285K). AMD's brand-new 16-core flagship managed to outpace the previous-gen Ryzen 9 7950X3D by ~14% in single thread stakes, and roughly 11% in multithreaded scenarios. Test system build details and settings were not mentioned with this leak—we expect to absorb a more complete picture tomorrow, upon publication of widespread reviews. The sampled Ryzen 9 9950X3D CPU surpassed its 9950X sibling by ~5% with its multi-thread result, both processors are just about equal in terms of single-core performance. The Intel Core Ultra 9 285K CPU posted the highest single-core result within the comparison—5078 points—exceeding the 9950X3D's tally by about 7%. The latter pulls ahead by ~3% in terms of recorded multi-thread performance. Keep an eye on TechPowerUp's review section; where W1zzard will be delivering his verdict(s) imminently.

The AMD Ryzen 9 9950X3D processor will head to retail next month—a March 12 launch day is rumored—but a handful of folks seem to have early samples in their possession. Reviewers and online influencers have been tasked with evaluating pre-launch silicon, albeit under strict conditions; i.e. no leaking. Inevitably, NDA-shredding material has seeped out—yesterday, we reported on an alleged sample's ASUS Silicon Prediction rating. Following that, a Bulgarian system integrator/hardware retailer decided to upload Cinebench R23 and PCMark Time Spy results to Facebook. Evidence of this latest leak was scrubbed at the source, but VideoCardz preserved crucial details.

The publication noticed distinguishable QR and serial codes in PCbuild.bg's social media post; so tracing activities could sniff out points of origin. As expected, the leaked benchmark data points were compared to Ryzen 9 9950X and 7950X3D scores. The Ryzen 9 9950X3D sample recorded a score of 17,324 points in 3DMark Time Spy, as well as 2279 points (single-core) and 42,423 points (multi-core) in Cinebench R23. Notebookcheck observed that the pre-launch candidate came: "out ahead of the Ryzen 9 7950X3D in both counts, even if the gaming win is less than significant. Comparing the images of the benchmark results to our in-house testing and benchmark database shows the 9950X3D beating the 7950X3D by nearly 17% in Cinebench multicore." When compared to its non-3D V-Cache equivalent, the Ryzen 9 9950X3D leverages a slight performance advantage. A blurry shot of PCbuild.bg's HWiNFO session shows the leaked processor's core clock speeds; going up to 5.7 GHz (turbo) on a single CCD (non-X3D). The X3D-equipped portion seems capable of going up to 5.54 GHz.

Primate Labs, the developer behind the popular Geekbench benchmarking suite, has launched Geekbench AI—a comprehensive benchmark tool designed to measure the artificial intelligence capabilities of various devices. Geekbench AI, previously known as Geekbench ML during its preview phase, has now reached version 1.0. The benchmark is available on multiple operating systems, including Windows, Linux, macOS, Android, and iOS, making it accessible to many users and developers. One of Geekbench AI's key features is its multifaceted approach to scoring. The benchmark utilizes three distinct precision levels: single-precision, half-precision, and quantized data. This evaluation aims to provide a more accurate representation of AI performance across different hardware designs.

In addition to speed, Geekbench AI places a strong emphasis on accuracy. The benchmark assesses how closely each test's output matches the expected results, offering insights into the trade-offs between performance and precision. The release of Geekbench AI 1.0 brings support for new frameworks, including OpenVINO, ONNX, and Qualcomm QNN, expanding its compatibility across various platforms. Primate Labs has also implemented measures to ensure fair comparisons, such as enforcing minimum runtime durations for each workload. The company noted that Samsung and NVIDIA are already utilizing the software to measure their chip performance in-house, showing that adoption is already strong. While the benchmark provides valuable insights, real-world AI applications are still limited, and reliance on a few benchmarks may paint a partial picture. Nevertheless, Geekbench AI represents a significant step forward in standardizing AI performance measurement, potentially influencing future consumer choices in the AI-driven tech market. Results from the benchmark runs can be seen here.

Game Science, the developer behind the highly anticipated action RPG "Black Myth: Wukong," has released a free benchmark tool on Steam for its upcoming game. This standalone application, separate from the main game, allows PC users to evaluate their hardware performance and system compatibility in preparation for the game's launch. The "Black Myth: Wukong Benchmark Tool" offers a unique glimpse into the game's visuals by rendering a real-time in-game sequence. While not playable, it provides valuable insights into how well a user's system will handle the game's demanding graphics and performance requirements. One of the tool's standout features is its customization options. Users can tweak various graphics settings to preview the game's visuals and performance under different configurations. This flexibility allows gamers to find the optimal balance between visual fidelity and smooth gameplay for their specific hardware setup.

However, Game Science has cautioned that due to the complexity and variability of gaming scenarios, the benchmark results may not fully represent the final gaming experience. This caveat shows the tool's role as a guide rather than a definitive measure of performance. The benchmark tool's system requirements offer a clear picture of the hardware needed to run "Black Myth: Wukong." At a minimum, users will need a Windows 10 system with an Intel Core i5-8400 or AMD Ryzen 5 1600 processor, 16 GB of RAM, and either an NVIDIA GeForce GTX 1060 6 GB or AMD Radeon RX 580 8 GB graphics card. For an optimal experience, the recommended specifications include an Intel Core i7-9700 or AMD Ryzen 5 5500 processor and an NVIDIA GeForce RTX 2060, AMD Radeon RX 5700 XT, or Intel Arc A750 graphics card. Interestingly, the benchmark tool supports DLSS, FSR, and XeSS technologies, indicating that the final game will likely include these performance-enhancing features. The developers also strongly recommend using an SSD for storage.

G.SKILL International Enterprise Co., Ltd., the world's leading brand of performance overclock memory and PC components, is pleased to announce that CENS from Germany won the title of OC Champion in the 8th Annual OC World Cup 2024 overclocking competition. After a grueling day of extreme overclocking with liquid nitrogen cooling, using the latest Intel Core i9-14900KF processor and ASUS ROG Maximus Z790 Apex Encore motherboard, CENS is walking away with the grand cash prize of $10,000 USD!

After the first 3 days of exhausting extreme overclocking in the Live Qualifier Stage of the OC World Cup 2024, biso biso, CENS, and Seby qualified for the Grand Final Stage on June 7. The Grand Final round is comprised of a maximum memory frequency challenge and 3 benchmarks with CPU speed fixed at a maximum of 7 GHz, demonstrating the incredible capability of the latest 14th Gen Intel Core processors and the extraordinary overclock skills of each OC World Cup 2024 Grand Final contestant.

Zhaoxin finally released its oft-delayed KX-7000 CPU series last December—the Chinese manufacturer claimed that its latest "Century Avenue Core" uArch consumer/desktop-oriented range was designed to "deliver double the performance of previous generations." Freshly discovered Geekbench 6.2.2 results indicate that Zhaoxin has succeeded on that front—Wccftech has pored over these figures, generated by an: "entry-level Zhaoxin KX-7000 CPU which has 8 cores, 8 threads, 4 MB of L2, and 32 MB of L3 cache. This chip was running at a base clock of 3.0 GHz and a boost clock of 3.3 GHz which is below its standard 3.6 GHz boost profile."

The new candidate was compared to Zhaoxin's previous-gen KX-U6780A and KX-6000G models. Intel's Core i3-10100F processor was thrown in as a familiar Western point of reference. The KX-7000 scored: "823 points in single-core, and 3813 points in multi-core tests. For comparisons, the Intel's Comet Lake CPU with 4 cores and 8 threads plus a boost of up to 4.3 GHz offers a much higher score. It's around 75% faster in single and 17% faster in multi-core tests within the same benchmark." The higher clock speeds, doubled core counts and TDPs do deliver "twice the performance" when compared to direct forebears—mission accomplished there. It is clear that Zhaoxin's latest CPU architecture cannot keep up with a generations old Team Blue design. Loongson's 3A6000 processor is a very promising prospect—reports suggest that this chip is somewhat comparable to mainstream AMD Zen 4 and Intel Raptor Lake products.

The Moore Threads MTT S80 first attracted wider media attention last summer due to it being introduced as the world's first PCIe Gen 5 gaming graphics card. Unfortunately, its performance prowess in gaming benchmarks did not match early expectations, especially for a 200 W TDP-rated unit with 4096 "MUSA" cores. Evaluators discovered that driver issues have limited the full potential of MTT GPUs—it is speculated that Moore Threads has simply repurposed existing PowerVR architecture under their in-house design: "Chunxaio." The Chinese firm has concentrated on driver improvements in the interim—mid-February experimentations indicated 100% performance boosts for MTT S80 and S70 discrete GPUs courtesy of driver version 240.90. Germany's ComputerBase managed to import Moore Threads MTT S80 and S30 models for testing purposes—in an effort to corroborate recently published performance figures, as disclosed by Asian review outlets.

The Moore Thread MTT S80—discounted down to $164 last October—was likely designed with MMO gamers in mind. VideoCardz (based on ComputerBase findings) discussed the card's struggles when weighed against Team Red's modern day integrated solutions: "S80 falls short when compared to the Ryzen 5 8600G, featuring the Radeon 760M iGPU with RDNA 3 graphics. A geometric mean across various titles reveals the S80's lag, but there are exceptions, like DOTA 2, where it takes the lead in framerate. It's clear that MTT GPUs (have a) less emphasized focus on supporting AAA titles." ComputerBase confirmed that DirectX 12 API support is still lacking, meaning that many popular Western games titles remain untested on the Moore Threads MTT S80 graphics card. The freshly launched entry-level MTT S30 card produced "1/4 of the performance" when compared to its flagship sibling.