Tag

Benchmarks

All articles tagged with #benchmarks

Tensor G6 Delivers Some CPU Gains but Trails Snapdragon and iPhone in Pixel 11 Pro XL
technology4 days ago

Tensor G6 Delivers Some CPU Gains but Trails Snapdragon and iPhone in Pixel 11 Pro XL

The Pixel 11 Pro XL’s Tensor G6 brings roughly 20% higher single- and multi-core CPU performance vs the G5, but Google’s chip still lags behind Qualcomm’s Snapdragon 8 Gen 5 and Apple’s A-series in raw speed, especially for gaming where the Galaxy S26 Ultra and iPhone 17 Pro Max pull ahead by about 2.5x at launch and maintain a sizable lead after throttling. Benchmarks show only modest gains in AI/ML and GPU performance, with Tensor’s 3nm process not delivering the kind of power efficiency or speed parity hoped. Storage tests note the Pixel is slightly faster than its predecessor but slower than the Galaxy, and battery life results are still pending. Overall, Tensor G6 improves efficiency but does not close the performance gap against its main rivals, though Google may gain if the efficiency benefits translate to real-world longevity.

Rethinking AI Intelligence: From Benchmarks to Real Understanding
technology5 days ago

Rethinking AI Intelligence: From Benchmarks to Real Understanding

Quanta’s interview with Melanie Mitchell argues AI cognition isn’t human-like thinking but alien intelligence that requires new testing methods borrowed from developmental and comparative psychology; she outlines six principles to assess machine cognition, warns against overreliance on benchmarks and anthropomorphism, discusses transfer and 'task tyranny', uses Clever Hans as a cautionary tale, and champions mechanistic interpretability to reveal how AIs work and where they fail, shaping safe deployment and the future of science.

technology7 days ago

Benchmarking a $46k Xeon HP Workstation: CachyOS in the Linux Performance Spotlight

Phoronix compares Windows 11 Pro and several Linux distributions—Ubuntu 26.04 LTS, AlmaLinux 10.2, CachyOS, and Fedora Workstation 44—on an HP Z4 G6i with a 48-core/96-thread Xeon 678X, 4x32GB DDR5, and RTX Pro 6000. AlmaLinux uses the Intel P-State performance governor by default, with a separate CachyOS “Performance” run to mirror that setup. The goal is to assess which OS delivers the best overall throughput for demanding workloads on a high-end workstation, with tests spanning multiple creator and compute workloads across a roughly $46k system configuration.

Pixel 11 Pro XL benchmarks hint at Tensor G6 boost, but rivals still ahead
technology8 days ago

Pixel 11 Pro XL benchmarks hint at Tensor G6 boost, but rivals still ahead

Early Geekbench results for the Pixel 11 Pro XL show a roughly 14% single‑core and 20% multi‑core uplift over the Pixel 10 Pro XL, thanks to the new Tensor G6 and Titan M3, but the Pixel still trails the Samsung Galaxy S26 Ultra (about 3,082/10,381 in scores). The Pixel 11 line, including the non‑XL, shares core specs, and Google is targeting an August 20 launch with pre‑order promos like a $200 gift card for Pro models; as with all early benchmarks, take the numbers with caution until official reviews arrive.

Google rolls out Gemini 3.7 Flash with coding gains and a selective rollout
technology12 days ago

Google rolls out Gemini 3.7 Flash with coding gains and a selective rollout

Google releases Gemini 3.7 Flash three weeks after 3.6, touting notable coding and document-processing improvements (FrontierCode 1.1 Main: 34.4%→43.6%; DeepSWE 49%→65.3%; GDP.pdf 22%→34%; AutomationBench 17%→30.4%). It also cuts the introductory price roughly in half, to $0.75/1M input tokens and $3.75/1M output tokens, but access is limited (Gemini API, AI Studio, Gemini Enterprise; Spark/Pro/Ultra plan required for individual use) and analysts question whether the cadence signals a real breakthrough versus marketing.

Gears of War E-Day Beta benchmarks map PC hardware demands across GPUs and CPUs
technology16 days ago

Gears of War E-Day Beta benchmarks map PC hardware demands across GPUs and CPUs

Open Beta benchmarks for Gears of War E-Day Multiplayer test PC hardware across Nvidia/AMD GPUs (from RTX 3060 to 5090 and RX 6750 XT to 9070 XT) and CPUs from AMD/Intel, outlining minimum and recommended 64‑bit Windows 10/11, a 130 GB SSD and DirectX 12 with ray‑tracing support. Using Unreal Engine 5 with Nanite and Lumen, the tests span 1080p, 1440p and 4K with max settings and DLSS/FSR, showing that comfortable 60 FPS at 1080p generally requires top-tier GPUs (e.g., RTX 5090 or RX 9070 XT) while 4K remains demanding and VRAM/RAM usage scales with resolution. The beta also benchmarks network code and balance ahead of the full release, with multiple maps and modes highlighted as part of the testing process.

Surface Laptop Ultra leaks in early benchmarks
technology26 days ago

Surface Laptop Ultra leaks in early benchmarks

A prototype Microsoft Surface Laptop Ultra with Nvidia RTX Spark N1X hardware surfaced near Redmond and underwent hands-on benchmarks, showing a 20-core Arm CPU, 6,144 CUDA cores, 24GB unified RAM, and a 512GB SSD. Early tests revealed unfinished status (screen issues, HDR, G‑Sync) and driver-related variability, with idle power rising significantly on newer drivers. In benchmarks, RTX Spark trails Apple's M5 Pro but delivers identical performance whether plugged in or on battery; build quality is praised, and price predictions place top configurations around $4,500. The device appears not final, with drivers and firmware still evolving ahead of a launch later this year.

Geekbench 7 Overhauls CPU/GPU Benchmarks for Real-World Performance
technology1 month ago

Geekbench 7 Overhauls CPU/GPU Benchmarks for Real-World Performance

Geekbench 7 debuts with a redesigned multi-core benchmark that better mirrors real-world multi-threaded workloads and a revamped GPU benchmark focused on machine-learning and content-creation tasks, including real-time video effects, image upscaling, LUT-based color grading, path tracing, and fluid simulations. It adds new tests for RAW image processing and supports broader image formats, while expanding media workloads (audio/video encoding/decoding, Whisper, AV1, Opus) and keeping CUDA, OpenCL, Vulkan, and Metal as APIs. Available on iOS, macOS, and other platforms, with a free personal-use tier and a Pro version discounted 20% through August 6.

Tiny gains, big trade-offs: Windows 11 on Steam Machine
technology1 month ago

Tiny gains, big trade-offs: Windows 11 on Steam Machine

ETA Prime’s benchmarks show Windows 11 on a Steam Machine yields small wins in some tests (e.g., Geekbench multi-core) and is competitive in 4K gaming, but the overall gains across 1080p–4K are minor and fall within margin of error. Valve’s Windows drivers offer no dual-boot support, so installing Windows requires wiping SteamOS and sacrifices HDMI-CEC, Steam Controller pairing, official support, and easy return. A similarly specced DIY Windows PC can beat a Steam Machine on price/performance, while SteamOS on a PC remains a solid option. Xbox Mode doesn’t deliver a clear performance boost; Windows is only worth it for specific Windows-only games or protections.

Steam Machine Reboot: Windows 11 Proves Feasible, Parity With SteamOS in Key Tests
technology1 month ago

Steam Machine Reboot: Windows 11 Proves Feasible, Parity With SteamOS in Key Tests

Valve’s Steam Machine now has Windows drivers, and a Windows 11 install tested by ETA Prime shows the box can run as a capable PC. In CPU benchmarks, Windows edges SteamOS by about 22% in Geekbench 6 multi‑core; gaming results are mixed across titles and resolutions, with Windows sometimes leading at 4K/1080p and SteamOS catching up or edging ahead at other settings (e.g., Cyberpunk 2077 and Horizon Zero Dawn Remastered). Overall, Windows on the Steam Machine is feasible and broadly competitive with SteamOS, though SteamOS remains attractive for its living‑room optimizations and ongoing driver/community support.)

technology1 month ago

Razer Blade 18 RTX 5090: Ubuntu 26.04 LTS, CachyOS, and Windows 11 Benchmark Face-Off

Phoronix benchmarks a $5,399 Razer Blade 18 laptop across Ubuntu 26.04 LTS, Windows 11, and the rolling-release CachyOS to compare performance on an RTX 5090 setup. Each OS uses its NVIDIA driver from the distro’s repo, with CachyOS on a 7.1 kernel, GCC 16.1.1, and NVIDIA R610, while Ubuntu 26.04 runs Linux 7.0 and Windows 11 is configured out-of-the-box by the vendor. A separate NVIDIA vs Nouveau comparison is planned for a future piece, and results are presented across multiple benchmark pages.

Fanless but Fierce: Why the MacBook Air Stays Cool Without a Fan
technology1 month ago

Fanless but Fierce: Why the MacBook Air Stays Cool Without a Fan

Apple’s fanless MacBook Air stays cool using a thick aluminum heat spreader and the metal chassis to passively dissipate heat from the M-series chip, delivering near‑Pro performance for most tasks; sustained heavy workloads will throttle as cooling becomes less efficient. Benchmarks show the M5 Air approaching the base M5 Pro in single-core, multi-core, and GPU tests, though the Pro can still pull ahead in long, demanding sessions. The design traces back to the M1 era, with iFixit noting the removal of a fan, and reviewers acknowledging cooling tweaks exist but don’t remove throttling under peak load.

Steam Machine energy boosts with a second memory stick, delivering up to 20% more in some tasks
technology1 month ago

Steam Machine energy boosts with a second memory stick, delivering up to 20% more in some tasks

Valve ships Steam Machines with a single 16GB DDR5-5600 module, keeping the system in single-channel mode; GamersNexus benchmarks show adding a second identical stick to enable dual-channel can yield up to ~20% gains, especially in CPU-heavy tasks and certain games (e.g., Baldur’s Gate 3, Resident Evil 4, Outer Worlds 2), while some titles see little difference. Valve previously suggested negligible impact, but all units ship with one module and only gain the bandwidth benefits once a second stick is added.

OpenAI Releases GPT-5.6 Trio—Soul, Terra, and Luna—With Speed Gains, But Access Is Limited
technology1 month ago

OpenAI Releases GPT-5.6 Trio—Soul, Terra, and Luna—With Speed Gains, But Access Is Limited

OpenAI unveils GPT-5.6 featuring three variants—Soul for complex reasoning, Terra for general use, and Luna for high-speed tasks—achieving up to 750 tokens/sec and about 40% lower costs via Cerebras chips. While the models outperform rivals on benchmarks like Terminal and Exploit Bench, OpenAI is restricting access to manage safety, ethical, and regulatory risks, including potential cybersecurity threats and bio-research misuse, and is collaborating with government bodies to ensure responsible deployment.