How to Accelerate Larger LLMs Locally on RTX With LM Studio

How to Accelerate Larger LLMs Locally on RTX With LM Studio

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.

Read More

How to Accelerate Larger LLMs Locally on RTX With LM Studio

How to Accelerate Larger LLMs Locally on RTX With LM Studio

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.

Read More

How to Accelerate Larger LLMs Locally on RTX With LM Studio

How to Accelerate Larger LLMs Locally on RTX With LM Studio

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.

Read More

How to Accelerate Larger LLMs Locally on RTX With LM Studio

How to Accelerate Larger LLMs Locally on RTX With LM Studio

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.

Read More

How to Accelerate Larger LLMs Locally on RTX With LM Studio

How to Accelerate Larger LLMs Locally on RTX With LM Studio

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.

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, 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.

Read More

How to Accelerate Larger LLMs Locally on RTX With LM Studio

How to Accelerate Larger LLMs Locally on RTX With LM Studio

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.

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, 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.

Read More

How to Accelerate Larger LLMs Locally on RTX With LM Studio

How to Accelerate Larger LLMs Locally on RTX With LM Studio

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.

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, 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.

Read More

How to Accelerate Larger LLMs Locally on RTX With LM Studio

How to Accelerate Larger LLMs Locally on RTX With LM Studio

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.

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, 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.

Read More

Denmark Launches Leading Sovereign AI Supercomputer to Solve Scientific Challenges With Social Impact

Denmark Launches Leading Sovereign AI Supercomputer to Solve Scientific Challenges With Social Impact

NVIDIA founder and CEO Jensen Huang joined the king of Denmark to launch the country’s largest sovereign AI supercomputer, aimed at breakthroughs in quantum computing, clean energy, biotechnology and other areas serving Danish society and the world.

Denmark’s first AI supercomputer, named Gefion after a goddess in Danish mythology, is an NVIDIA DGX SuperPOD driven by 1,528 NVIDIA H100 Tensor Core GPUs and interconnected using NVIDIA Quantum-2 InfiniBand networking.

Gefion is operated by the Danish Center for AI Innovation (DCAI), a company established with funding from the Novo Nordisk Foundation, the world’s wealthiest charitable foundation, and the Export and Investment Fund of Denmark. The new AI supercomputer was symbolically turned on by King Frederik X of Denmark, Huang and Nadia Carlsten, CEO of DCAI, at an event in Copenhagen.

Huang sat down with Carlsten, a quantum computing industry leader, to discuss the public-private initiative to build one of the world’s fastest AI supercomputers in collaboration with NVIDIA.

The Gefion AI supercomputer comes to Copenhagen to serve industry, startups and academia.

“Gefion is going to be a factory of intelligence. This is a new industry that never existed before. It sits on top of the IT industry. We’re inventing something fundamentally new,” Huang said.

The launch of Gefion is an important milestone for Denmark in establishing its own sovereign AI. Sovereign AI can be achieved when a nation has the capacity to produce artificial intelligence with its own data, workforce, infrastructure and business networks. Having a supercomputer on national soil provides a foundation for countries to use their own infrastructure as they build AI models and applications that reflect their unique culture and language.

“What country can afford not to have this infrastructure, just as every country realizes you have communications, transportation, healthcare, fundamental infrastructures — the fundamental infrastructure of any country surely must be the manufacturer of intelligence,” said Huang. “For Denmark to be one of the handful of countries in the world that has now initiated on this vision is really incredible.”

The new supercomputer is expected to address global challenges with insights into infectious disease, climate change and food security. Gefion is now being prepared for users, and a pilot phase will begin to bring in projects that seek to use AI to accelerate progress, including in such areas as quantum computing, drug discovery and energy efficiency.

“The era of computer-aided drug discovery must be within this decade. I’m hoping that what the computer did to the technology industry, it will do for digital biology,” Huang said.

Supporting Next Generation of Breakthroughs With Gefion

The Danish Meteorological Institute (DMI) is in the pilot and aims to deliver faster and more accurate weather forecasts. It promises to reduce forecast times from hours to minutes while greatly reducing the energy footprint required for these forecasts when compared with traditional methods.

Researchers from the University of Copenhagen are tapping into Gefion to implement and carry out a large-scale distributed simulation of quantum computer circuits. Gefion enables the simulated system to increase from 36 to 40 entangled qubits, which brings it close to what’s known as “quantum supremacy,” or essentially outperforming a traditional computer while using less resources.

The University of Copenhagen and the Technical University of Denmark are working together on a multi-modal genomic foundation model for discoveries in disease mutation analysis and vaccine design. Their model will be used to improve signal detection and the functional understanding of genomes, made possible by the capability to train LLMs on Gefion.

Startup Go Autonomous seeks training time on Gefion to develop an AI model that understands and uses multi-modal input from both text, layout and images. Another startup, Teton, is building an AI Care Companion with large video pretraining, using Gefion.

Addressing Global Challenges With Leading Supercomputer

The Gefion supercomputer and ongoing collaborations with NVIDIA will position Denmark, with its renowned research community, to pursue the world’s leading scientific challenges with enormous social impact as well as large-scale projects across industries.

With Gefion, researchers will be able to work with industry experts at NVIDIA to co-develop solutions to complex problems, including research in pharmaceuticals and biotechnology and protein design using the NVIDIA BioNeMo platform.

Scientists will also be collaborating with NVIDIA on fault-tolerant quantum computing using NVIDIA CUDA-Q, the open-source hybrid quantum computing platform.

Read More

Denmark Launches Leading Sovereign AI Supercomputer to Solve Scientific Challenges With Social Impact

Denmark Launches Leading Sovereign AI Supercomputer to Solve Scientific Challenges With Social Impact

NVIDIA founder and CEO Jensen Huang joined the king of Denmark to launch the country’s largest sovereign AI supercomputer, aimed at breakthroughs in quantum computing, clean energy, biotechnology and other areas serving Danish society and the world.

Denmark’s first AI supercomputer, named Gefion after a goddess in Danish mythology, is an NVIDIA DGX SuperPOD driven by 1,528 NVIDIA H100 Tensor Core GPUs and interconnected using NVIDIA Quantum-2 InfiniBand networking.

Gefion is operated by the Danish Center for AI Innovation (DCAI), a company established with funding from the Novo Nordisk Foundation, the world’s wealthiest charitable foundation, and the Export and Investment Fund of Denmark. The new AI supercomputer was symbolically turned on by King Frederik X of Denmark, Huang and Nadia Carlsten, CEO of DCAI, at an event in Copenhagen.

Huang sat down with Carlsten, a quantum computing industry leader, to discuss the public-private initiative to build one of the world’s fastest AI supercomputers in collaboration with NVIDIA.

The Gefion AI supercomputer comes to Copenhagen to serve industry, startups and academia.

“Gefion is going to be a factory of intelligence. This is a new industry that never existed before. It sits on top of the IT industry. We’re inventing something fundamentally new,” Huang said.

The launch of Gefion is an important milestone for Denmark in establishing its own sovereign AI. Sovereign AI can be achieved when a nation has the capacity to produce artificial intelligence with its own data, workforce, infrastructure and business networks. Having a supercomputer on national soil provides a foundation for countries to use their own infrastructure as they build AI models and applications that reflect their unique culture and language.

“What country can afford not to have this infrastructure, just as every country realizes you have communications, transportation, healthcare, fundamental infrastructures — the fundamental infrastructure of any country surely must be the manufacturer of intelligence,” said Huang. “For Denmark to be one of the handful of countries in the world that has now initiated on this vision is really incredible.”

The new supercomputer is expected to address global challenges with insights into infectious disease, climate change and food security. Gefion is now being prepared for users, and a pilot phase will begin to bring in projects that seek to use AI to accelerate progress, including in such areas as quantum computing, drug discovery and energy efficiency.

“The era of computer-aided drug discovery must be within this decade. I’m hoping that what the computer did to the technology industry, it will do for digital biology,” Huang said.

Supporting Next Generation of Breakthroughs With Gefion

The Danish Meteorological Institute (DMI) is in the pilot and aims to deliver faster and more accurate weather forecasts. It promises to reduce forecast times from hours to minutes while greatly reducing the energy footprint required for these forecasts when compared with traditional methods.

Researchers from the University of Copenhagen are tapping into Gefion to implement and carry out a large-scale distributed simulation of quantum computer circuits. Gefion enables the simulated system to increase from 36 to 40 entangled qubits, which brings it close to what’s known as “quantum supremacy,” or essentially outperforming a traditional computer while using less resources.

The University of Copenhagen and the Technical University of Denmark are working together on a multi-modal genomic foundation model for discoveries in disease mutation analysis and vaccine design. Their model will be used to improve signal detection and the functional understanding of genomes, made possible by the capability to train LLMs on Gefion.

Startup Go Autonomous seeks training time on Gefion to develop an AI model that understands and uses multi-modal input from both text, layout and images. Another startup, Teton, is building an AI Care Companion with large video pretraining, using Gefion.

Addressing Global Challenges With Leading Supercomputer

The Gefion supercomputer and ongoing collaborations with NVIDIA will position Denmark, with its renowned research community, to pursue the world’s leading scientific challenges with enormous social impact as well as large-scale projects across industries.

With Gefion, researchers will be able to work with industry experts at NVIDIA to co-develop solutions to complex problems, including research in pharmaceuticals and biotechnology and protein design using the NVIDIA BioNeMo platform.

Scientists will also be collaborating with NVIDIA on fault-tolerant quantum computing using NVIDIA CUDA-Q, the open-source hybrid quantum computing platform.

Read More