Skip to content

Author / Systems & Technical Editor / 247 articles

Marcus Thorne

Marcus writes Forge PC's step-by-step build, troubleshooting and system optimisation guides. His background spans Windows kernel and power-state diagnostics, UEFI/BIOS configuration, NVMe storage arrays, ATX 3.1 power delivery and running local AI and LLM workloads on consumer graphics cards.

  • How to Fix Audio Crackling and DPC Latency on Windows With LatencyMon

    How-To

    How to Fix Audio Crackling and DPC Latency on Windows With LatencyMon

    Audio crackling and dropouts on Windows are often caused by a driver running too long at high priority. The free LatencyMon Home Edition measures ISR and DPC execution times and hard pagefaults, and in most cases names the driver responsible. Updating, rolling back or temporarily disabling that driver, plus installing chipset drivers and the latest BIOS, fixes most cases.

    Marcus Thorne5 min

  • How to Stress Test Your PC: CPU, GPU and RAM Stability Checks With Free Tools

    How-To

    How to Stress Test Your PC: CPU, GPU and RAM Stability Checks With Free Tools

    To stress test a PC, use Prime95 or OCCT for the CPU, 3DMark's stress test for the graphics card, MemTest86 or OCCT for memory and OCCT's Power test for the whole system, while watching temperatures in HWiNFO. UL says a 3DMark stress test passes if all loops complete with at least 97% frame-rate stability.

    Marcus Thorne5 min

  • How to Install AMD's AI Bundle: ComfyUI, Ollama, LM Studio and PyTorch in One Go

    How-To

    How to Install AMD's AI Bundle: ComfyUI, Ollama, LM Studio and PyTorch in One Go

    AMD's AI Bundle is an optional part of the Adrenalin driver installer that sets up PyTorch on Windows, ComfyUI, Ollama, LM Studio and Amuse in one go. AMD says it supports Radeon RX 9000 cards, RX 7700 and above from the RX 7000 series, and Ryzen AI 300, 400 and Max processors, and needs about 35GB for everything plus an internet connection.

    Marcus Thorne4 min

  • How to Run a Local LLM on an 8GB Graphics Card: Models, Settings and Limits

    How-To

    How to Run a Local LLM on an 8GB Graphics Card: Models, Settings and Limits

    An 8GB graphics card can run useful local AI models if you choose carefully. Ollama lists qwen3:8b at 5.2GB and gemma3:4b at 3.4GB, both of which fit, while 12B to 14B models spill into system RAM. Use ollama ps to confirm the model shows 100% GPU, keep the context window modest and set the K/V cache to q8_0 to roughly halve its memory use.

    Marcus Thorne5 min

  • How to Reduce Input Lag in PC Games: Reflex, Anti-Lag, VRR and Frame Caps

    How-To

    How to Reduce Input Lag in PC Games: Reflex, Anti-Lag, VRR and Frame Caps

    To reduce input lag on a gaming PC, run your monitor at its full refresh rate, enable NVIDIA Reflex or AMD Anti-Lag in supported games, and pair G-SYNC or FreeSync with a frame rate cap at least 3fps below your refresh rate, as Blur Busters recommends. Windows 11's optimisations for windowed games help borderless DirectX 10 and 11 titles.

    Marcus Thorne4 min

  • How to Update Your SSD's Firmware Safely (Samsung, Kingston and Crucial)

    How-To

    How to Update Your SSD's Firmware Safely (Samsung, Kingston and Crucial)

    To update SSD firmware, back up your data and use the drive maker's free tool: Samsung Magician, Kingston SSD Manager or Crucial Storage Executive. Each can check for and install firmware updates on supported drives. Keep the drive connected directly to the motherboard rather than via a USB adapter, and do not interrupt the update.

    Marcus Thorne4 min

  • How to Generate AI Video on Your Own PC With ComfyUI and Wan 2.2

    How-To

    How to Generate AI Video on Your Own PC With ComfyUI and Wan 2.2

    You can generate short AI videos at home with the free ComfyUI app and the open Wan 2.2 5B model, which ComfyUI says fits on an 8GB graphics card with its built-in offloading. This guide covers the three model files, where they go, text-to-video and image-to-video, and settings to keep render times down.

    Marcus Thorne4 min

  • How to Make Local AI Models Run Faster in Ollama: 7 Settings That Matter

    How-To

    How to Make Local AI Models Run Faster in Ollama: 7 Settings That Matter

    The biggest speed-up for local AI in Ollama is keeping the whole model on your GPU: run ollama ps and aim for 100% GPU. Ollama's documented settings can help, including a smaller context window, flash attention, a quantised KV cache (q8_0 roughly halves its memory use), fewer parallel requests and keep_alive.

    Marcus Thorne4 min

  • How to Connect a 12V-2x6 GPU Power Cable Safely (and Avoid Melting)

    How-To

    How to Connect a 12V-2x6 GPU Power Cable Safely (and Avoid Melting)

    To connect a 12V-2x6 or 12VHPWR graphics card cable safely, push it straight in until the latch engages with no visible gap, keep about 30mm of straight cable before any bend as Corsair advises, and avoid tight 90-degree turns. Re-check it after a few days and whenever the PC is moved, and replace the cable at the first sign of heat damage.

    Marcus Thorne4 min

  • How to Check VRAM Usage on Windows 11 (Including for Local AI)

    How-To

    How to Check VRAM Usage on Windows 11 (Including for Local AI)

    Windows Task Manager shows how much dedicated and shared GPU memory your PC is using, and per-app columns show which program is using it. For local AI, the ollama ps command shows whether a model fits fully on the GPU, and the free GPU Taskman tool gives a per-process breakdown.

    Marcus Thorne4 min

  • How to Connect Front Panel Connectors to Your Motherboard

    How-To

    How to Connect Front Panel Connectors to Your Motherboard

    Front panel cables connect to a small motherboard header. Power and reset switches work either way round, but the power LED must go the right way, with the white ground wire on the ground pin. Use a Q-Connector if your board has one, and line up USB and audio plugs before pushing gently. Steps follow ASUS's official installation guide; check your own manual.

    Marcus Thorne4 min

  • How to Use Vision Models in Ollama: Ask a Local AI About Your Images

    How-To

    How to Use Vision Models in Ollama: Ask a Local AI About Your Images

    Ollama can run vision models such as Gemma 4 that answer questions about images on your own PC. Pick a size your graphics card can hold (the default Gemma 4 is listed at 6.6GB to 9.5GB), then add an image in the Ollama app, at the command line or in the API's images field. Steps follow Ollama's documentation.

    Marcus Thorne4 min

  • How to Set Up HDR on Windows 11 (and Make It Look Right)

    How-To

    How to Set Up HDR on Windows 11 (and Make It Look Right)

    To set up HDR in Windows 11, enable HDR on your display, switch on Use HDR in Settings > System > Display and adjust the SDR content brightness slider. Then run Microsoft's free Windows HDR Calibration app to tune HDR for your screen. If HDR will not turn on, dxdiag shows whether your display and driver support HDR10. Steps follow Microsoft Support.

    Marcus Thorne5 min

  • How to Move Steam Games to Another Drive Without Redownloading

    How-To

    How to Move Steam Games to Another Drive Without Redownloading

    Steam can move installed games to a new drive without redownloading them. Add a Steam Library folder in Settings > Storage, select the games and click Move. Moving the whole Steam installation is possible but riskier, so back up steamapps first. If a game misbehaves afterwards, Verify integrity of game files usually fixes it. Steps follow Valve's Steam Support pages.

    Marcus Thorne4 min

  • How to Run a Local AI Model Across Two GPUs with Ollama, LM Studio or llama.cpp

    How-To

    How to Run a Local AI Model Across Two GPUs with Ollama, LM Studio or llama.cpp

    Two graphics cards can hold a larger local AI model than one. Ollama spreads a model across GPUs automatically only when it will not fit on one card; LM Studio 0.3.14 and later let you choose which cards to use and how to fill them; llama.cpp gives full control with --split-mode and --tensor-split. This guide follows each app's official documentation.

    Marcus Thorne4 min

  • How to Install Chipset Drivers on Windows 11 (AMD Ryzen and Intel)

    How-To

    How to Install Chipset Drivers on Windows 11 (AMD Ryzen and Intel)

    On an AMD Ryzen PC, install AMD's Ryzen Chipset Driver (8.09.28.1124 at the time of writing) from AMD or your board maker and restart. On Intel, Intel says its chipset software is only INF files, needed when installing an operating system or if Device Manager shows unknown devices. This guide covers both, plus AMD's known issue when downgrading drivers.

    Marcus Thorne3 min

  • How to Test Your SSD Speed with CrystalDiskMark (and Read the Results)

    How-To

    How to Test Your SSD Speed with CrystalDiskMark (and Read the Results)

    CrystalDiskMark 9.0.3 is the free, official way to check that an SSD is running as fast as it should. Download it from Crystal Dew World, choose the NVMe SSD profile for an M.2 drive, run All with the default five passes and 1GiB test, and compare the SEQ1M result with your drive's official rating. This guide explains each row and what to check if results are low.

    Marcus Thorne4 min

  • How to Generate AI Images with Fooocus

    How-To

    How to Generate AI Images with Fooocus

    How to generate images locally with official Fooocus: download the Windows package from the lllyasviel/Fooocus GitHub Releases page, extract it to a short path with plenty of free disk, run the official launch script and wait for the first model download. Check the official README for current GPU and RAM requirements. It sits between Amuse and ComfyUI as a prompt-first local app.

    Marcus Thorne3 min