Editor's note: This post is part of the AI Decoded series, which demystifies AI by making the technology more accessible, and showcases new hardware, software, tools and accelerations for GeForce RTX PC and NVIDIA RTX workstation users.
Large language models (LLMs) are reshaping productivity. They're capable of drafting documents, summarizing web pages and, having been trained on vast quantities of data, accurately answering questions about nearly any topic.
LLMs are at the core of many emerging use cases in generative AI, including digital assistants, conversational avatars and customer service agents.
Many of the latest LLMs can run locally on PCs or workstations. This is useful for a variety of reasons: users can keep conversations and content private on-device, use AI without the internet, or simply take advantage of the powerful NVIDIA GeForce RTX GPUs in their system. Other models, because of their size and complexity, do no't fit into the local GPU's video memory (VRAM) and require hardware in large data centers.
However, Iit i's possible to accelerate part of a prompt on a data-center-class model locally on RTX-powered PCs using a technique called GPU offloading. This allows users to benefit from GPU acceleration without being as limited by GPU memory constraints.
Size and Quality vs. Performance There's a tradeoff between the model size and the quality of responses and the performance. In general, larger models deliver higher-quality responses, but run more slowly. With smaller models, performance goes up while quality goes down.
This tradeoff isn't always straightforward. There are cases where performance might be more important than quality. Some users may prioritize accuracy for use cases like content generation, since it can run in the background. A conversational assistant, meanwhile, needs to be fast while also providing accurate responses.
The most accurate LLMs, designed to run in the data center, are tens of gigabytes in size, and may not fit in a GPU's memory. This would traditionally prevent the application from taking advantage of GPU acceleration.
However, GPU offloading uses part of the LLM on the GPU and part on the CPU. This allows users to take maximum advantage of GPU acceleration regardless of model size.
Optimize AI Acceleration With GPU Offloading and LM Studio LM Studio is an application that lets users download and host LLMs on their desktop or laptop computer, with an easy-to-use interface that allows for extensive customization in how those models operate. LM Studio is built on top of llama.cpp, so it's fully optimized for use with GeForce RTX and NVIDIA RTX GPUs.
LM Studio and GPU offloading takes advantage of GPU acceleration to boost the performance of a locally hosted LLM, even if the model can't be fully loaded into VRAM.
With GPU offloading, LM Studio divides the model into smaller chunks, or subgraphs, which represent layers of the model architecture. Subgraphs aren't permanently fixed on the GPU, but loaded and unloaded as needed. With LM Studio's GPU offloading slider, users can decide how many of these layers are processed by the GPU.
LM Studio's interface makes it easy to decide how much of an LLM should be loaded to the GPU. For example, imagine using this GPU offloading technique with a large model like Gemma 2 27B. 27B refers to the number of parameters in the model, informing an estimate as to how much memory is required to run the model.
According to 4-bit quantization, a technique for reducing the size of an LLM without significantly reducing accuracy, each parameter takes up a half byte of memory. This means that the model should require about 13.5 billion bytes, or 13.5GB - plus some overhead, which generally ranges from 1-5GB.
Accelerating this model entirely on the GPU requires 19GB of VRAM, available on the GeForce RTX 4090 desktop GPU. With GPU offloading, the model can run on a system with a lower-end GPU and still benefit from acceleration.
The table above shows how to run several popular models of increasing size across a range of GeForce RTX and NVIDIA RTX GPUs. The maximum level of GPU offload is indicated for each combination. Note that even with GPU offloading, users still need enough system RAM to fit the whole model. In LM Studio, it's possible to assess the performance impact of different levels of GPU offloading, compared with CPU only. The below table shows the results of running the same query across different offloading levels on a GeForce RTX 4090 desktop GPU.
Depending on the percent of the model offloaded to GPU, users see increasing throughput performance compared with running on CPUs alone. For the Gemma 2 27B model, performance goes from an anemic 2.1 tokens per second to increasingly usable speeds the more the GPU is used. This enables users to benefit from the performance of larger models that they otherwise would've been unable to run. On this particular model, even users with an 8GB GPU can enjoy a meaningful speedup versus running only on CPUs. Of course, an 8GB GPU can always run a smaller model that fits entirely in GPU memory and get full GPU acceleration.
Achieving Optimal Balance LM Studio's GPU offloading feature is a powerful tool for unlocking the full potential of LLMs designed for the data center, like Gemma 2 27B, locally on RTX AI PCs. It makes larger, more complex models accessible across the entire lineup of PCs powered by GeForce RTX and NVIDIA RTX GPUs.
Download LM Studio to try GPU offloading on larger models, or experiment with a variety of RTX-accelerated LLMs running locally on RTX AI PCs and workstations.
Generative AI is transforming gaming, videoconferencing and interactive experiences of all kinds. Make sense of what's new and what's next by subscribing to the AI Decoded newsletter.
Most recent headlines
17/02/2025
TVBEurope talks to the team who brought a piece of broadcasting history back to life on BBC Wales last night
By Jenny Priestley
Published: February 14, 2025 ...
17/02/2025
Eligible studios will receive the relief directly from their local authorities, with no need to submit a claim
By Matthew Corrigan
Published: February 17, 20...
17/02/2025
As DEI initiatives come under pressure in the USA, supporters of Rise, which promotes diversity in the media and entertainment sector, discuss the continuing im...
17/02/2025
EIB Global loan will partially fund SES's continued O3b mPOWER medium earth orbit (MEO) fleet expansion enabling additional capacity worldwide. O3b mPOWER s...
15/02/2025
At Nielsen, Business Resource Groups (BRGs) play a critical role in nurturing a ...
15/02/2025
Innovation to empower creativity: Explore the latest in Lightroom
Rob Christensen February 15, 2025
0 Comments
Whether you want to remove distracting ...
15/02/2025
Wisycom, a worldwide leader in RF technology and provider of wireless solutions, announces a new distribution agreement with EQUAPHON. An Argentinian profession...
15/02/2025
DPA Microphones, the leading manufacturer of high-quality microphone solutions, is pleased to welcome Nicholas Nick Mariano and Vincent Vince Divine to the ...
15/02/2025
Tuxera, a leading provider of quality-assured file systems and networking solutions, will showcase how Tuxera Fusion SMB enables unprecedented performance for m...
15/02/2025
Bitmovin's latest Video Developer Report indicates an increasing role of AI, the slow adoption of next-generation codecs, and continuing challenges when it ...
15/02/2025
When Michelin Enterprises set out to promote its prestigious Hotel Guide in the new Sleepless in the Kitchen , campaign, veteran German/Slovenian gaffer Robert...
15/02/2025
Formula E, the world's first all-electric racing series, revolutionizing motorsport with cutting-edge technology and a commitment to sustainability has sele...
15/02/2025
Agile Content, a global leader in TV and video technology solutions, announces its rebranding to AgileTV.
The transition to AgileTV stems from a commitment to ...
15/02/2025
Hiltron Communications will promote the latest additions to its range of satellite communication products, systems and supporting services at GovSatCom 2025. Th...
15/02/2025
Following the success of the original Airglow Booklight, DoPchoice debuts a new addition to their inflatable line the Float. This lightweight, inflatable, round...
15/02/2025
LAS VEGAS Marshall Electronics has announced that it will debut its newest camera controller, the RCP Plus, at the 2025 NAB Show at Booth N2649 between April 5 ...
15/02/2025
ATLANTA Gray Media has named Blake Sebo as the next general manager of KTUU and KYES, Gray's NBC and CBS affiliates in Anchorage, Alaska....
15/02/2025
KOKKEDAL, Denmark DPA Microphones will feature its new CORE+ technology for distortion-free microphone sound at the 2025 NAB Show, April 5-9, in Las Vegas....
15/02/2025
SWANSEA, U.K. QuickLink will launch StudioPro-NDI and StudioPro-3, two new models of its QuickLine StudioPro video production platform, at the 2025 NAB Show, Ap...
15/02/2025
Live From 2025 NBA All-Star Game Weekend: After 25 Years, Turner Sports All-In F...
15/02/2025
NASCAR Driver Cam on Max Embraces Cloud-Based Workflow for 2025 Season Chris Brown of WBD Sports says Mobii will play key part in delivering signals from the tr...
14/02/2025
Last month, we officially launched the Spotify Partner Program, a powerful new m...
14/02/2025
From an island off the coast of Lutruwita to the global stage in France SBS & ...
14/02/2025
The future of electro-optical/infrared (EO/IR) land systems is evolving so that ...
14/02/2025
Here are 5 reasons why you should attend ISE 2025
Discover more about Calrec: Learn more about Calrec's legacy of innovation in broadcast audio solutions ...
14/02/2025
World Radio Day: On the same wavelength
By Graham Murray, International Sales Manager at Calrec
To mark World Radio Day, Calrec's International Sales Ma...
14/02/2025
DALLAS and TYSONS, Va. Tegna has concluded a broadcast rights agreement that makes Tegna's KFAA in Dallas the exclusive local television partner of the Dall...
14/02/2025
Amazon's Prime Video has released a new app for Apple TV that the streamer said delivers an improved streaming experiences with faster scrolling, improved s...
14/02/2025
CLEVELAND Telos Alliance announced a software update for its Axia Quasar XR and SR line of audio-over-IP (AoIP) mixing consoles and will show the offering at th...
14/02/2025
Puget Systems Returns to HPA Tech Retreat 2025 with New Analysis Tools for Puge...
14/02/2025
Glass Animals: Tour of Earth Uses Blackmagic Live Production Workflow
Brie Clayton February 14, 2025
0 Comments
80six deploys Blackmagic URSA Broadcas...
14/02/2025
A dispute between Paramount and Youtube TV that would have pulled CBS stations and Paramount's cable channels from the streaming service has been temporaril...
14/02/2025
Colourist Deidre (Dee) McClelland details how she helped shape the distinct look and feel of whimsical Oscar-nominated stop-motion film, Memoir of a Snail
By C...
14/02/2025
BitFire said it intends to introduce forward-looking services that simplify the launch of live production workflows, and enhance real-time collaboration
By Jen...
14/02/2025
The claim by the liquidators is the latest twist in the ongoing saga of the companys collapse at the end of 2021
By Matthew Corrigan
Published: February 14, ...
14/02/2025
A carriage dispute between YouTube TV and Paramount could come to a head today (Feb. 13), with the virtual pay TV provider threatening to pull all of the progra...
14/02/2025
ENGLEWOOD CLIFFS, N.J. LG Electronics USA has opened a new in-house, state-of-the-art virtual production studio at its 360,000-square-foot North American headqu...
14/02/2025
NEW YORK In a bid to help local broadcasters find new ways of monetizing their content, the Coalition for Innovative Media Measurement (CIMM) and TVB have relea...
14/02/2025
WASHINGTON Three Democratic U.S. Senators have sent a stern letter blasting FCC Chair Brendan Carr for the weaponization of the agency for attacking broadcast...
14/02/2025
Years after launching the streaming service Apple TV+ in November of 2019, Apple has finally released a version of its steaming app for Android devices....
14/02/2025
FOOTHILL RANCH, Calif. Red Digital Cinema will make its V-RAPTOR [X] and KOMODO-X camera systems available with Nikon's premier Z Mount, providing filmmaker...
14/02/2025
DAYTONA, Fla. The Fox Sports production of this year's Daytona 500 will see the addition of a few new tech elements and the continued unfolding of the broad...
14/02/2025
Watch Bryson Battles Blind Audition on The Voice Get a Four-Chair Turn Michael Bubl said the Boston Conservatory at Berklee musical theater senior might have...
14/02/2025
2025 Daytona 500: FOX Sports Looks To Ride Super Bowl LIX Momentum Into NASCAR S...
14/02/2025
14 Feb 2025
VEON's Kyivstar and the Come Back Alive Foundation Raise USD 2....
14/02/2025
Max Renews THE PITT, Starring Noah Wyle, For A Second Season
Max
Warner Bros. Television Group
The Pitt
New Episodes Of The Medical Drama From John...
14/02/2025
This year, Technicolor proudly celebrates 110 years of shaping the magic of cinema pioneering color, visual effects, and innovation that bring unforgettable l...
14/02/2025
NBA All-Star: More Games, More Teams, More Music, More Sound From the Court Turner Sports' team will deploy familiar tech to capture all the audio By Dan D...
14/02/2025
LIV Golf Continues To Innovate With Drones, Data-Driven Graphics, and All-Access...
Editor's note: This post is part of the AI Decoded series, which demystifies AI by making the technology more accessible, and showcases new hardware, software, tools and accelerations for GeForce RTX PC and NVIDIA RTX workstation users.
