Live from GTC: Game Development with Paul Logan

Summary

Podcast Summary: Reshaping Workflows with Dell Pro Precision and NVIDIA RTX PRO GPUs

Episode: Live from GTC: Game Development with Paul Logan
Date: March 17, 2026
Host: Logan Lawler
Guest: Paul Logan, Senior Product Marketing Manager for Enterprise Game Development, NVIDIA

---

Overview

This episode, recorded live from NVIDIA’s GTC 2026, dives into the latest advancements in game development workflows powered by Dell Pro Precision workstations and NVIDIA RTX PRO GPUs. Host Logan Lawler interviews Paul Logan from NVIDIA, exploring groundbreaking innovations, new releases, and the tools transforming the landscape for developers, artists, and studios. The discussion focuses on NVIDIA's latest technologies, including neural rendering and a revolutionary blueprint for AI-driven code development, as well as virtualization workflows for studios.

---

Key Discussion Points & Insights

1. GTC and the State of Innovation in Game Development (00:20–01:15)

• Setting the Scene: Logan Lawler notes GTC as "where all the hotness happens, all the new things" and asks what’s new for game development at NVIDIA.
• Paul Logan's Role: Senior Product Marketing Manager for Enterprise Game Development, NVIDIA.
  • “I help companies like Activision, like Sony, like Blizzard, accelerate their game development and reduce costs so that they can get to market quicker.” (Paul Logan, 00:36)

2. Major Announcements from NVIDIA at GTC (01:15–02:14)

• DLSS 5 Neural Rendering:
  • Announced as a major leap, applying advanced AI-driven upscaling for game engines.
  • “We just announced DLSS 5, that's neural rendering for game engines. … That’s probably the biggest thing holistically in terms of how we're going to affect the gaming industry.” (Paul Logan, 01:15)
• Blueprint for MCP Agentic Code Development:
  • A new framework, developed in partnership with industry leaders like EA, to transform code generation for studios, targeting both efficiency and cost reduction.
  • “An upcoming blueprint for MCP Agentic code development … is going to be really impactful in terms of reducing token costs for entire studios…" (Paul Logan, 01:32)
• RTX Pro Server & NVIDIA vGPU Solutions:
  • Virtualizing the entire development workflow, enabling remote and distributed teams to work seamlessly.
  • “All of that can be done on virtualized GPU infrastructure.” (Paul Logan, 01:58)

3. Deep Dive: MCP Agentic Blueprint & Automated Context-Aware Code Prompts (02:14–04:01)

• Purpose & Workflow:
  • Designed for studios with massive legacy codebases (e.g., “500 gigabytes of legacy code in C”).
  • Addresses cost and complexity in programming and codebase management.
• Technical Approach:
  • Integration with Claude code and large language models:
    • Acknowledges that “Anthropic's models produce some of the best results,” but cost is a major concern.
    • “You've looked at the prices of Claude code. … It's expensive.” (Logan Lawler & Paul Logan, 02:46–02:49)
  • Novel Solution – MCP Server:
    • Uses “quencoder next on Quadco” combined with a custom MCP server.
    • MCP indexes the entire code repository and makes "informed prompting decisions" rather than "raw prompts."
    • Employs vector search to fetch highly relevant, context-rich code segments—e.g., whole functions rather than arbitrary code blocks.
    • “We do automated syntax chunking. So instead of just grabbing random blobs... we grab context-aware chunks.” (Paul Logan, 03:19)
    • Dramatic lift in code accuracy: “That increases code accuracy from anywhere from 25% on a base model to up to 90% on a fully fine-tuned model.” (Paul Logan, 03:48)

---

Notable Quotes & Memorable Moments

• Logan Lawler’s humor:

  • “So senior, is that you just adding that because you're trying to feel special? Did you actually get promoted?” (00:53)
  • Sets a casual, engaging tone.

• On the significance of neural rendering:

  • “That’s probably the biggest thing holistically in terms of how we're going to affect the gaming industry.” (Paul Logan, 01:20)

• On advancing code accuracy:

  • “...grab context aware chunks, so you grab the entire function for collision on bullets, or ... for day and night cycles … that ... increases code accuracy from ... 25% ... up to 90%.” (Paul Logan, 03:41–03:48)

• Podcast wrap, capturing excitement:

  • “I'm blown away. Paul Logan, this is why I flagged you down.” (Logan Lawler, 04:01)

---

Timestamps for Key Segments

• 00:20 – Introduction; Logan and Paul’s banter about roles and promotions
• 01:15 – Major NVIDIA announcements: DLSS 5, MCP Agentic blueprint, virtualized workflows
• 02:14 – Deep dive into MCP Agentic agent: technical challenges, workflow, and benefits
• 03:19 – Explanation of automated syntax chunking and code accuracy improvements
• 04:01 – Host reaction and episode close

---

Style & Tone

• The conversation is informal, energetic, with friendly banter.
• Technical details are made accessible, demystifying advanced concepts for a broad audience.
• Focuses on both real-world impacts and behind-the-scenes innovation.

---

Conclusion

This episode spotlights the synergy between Dell Pro Precision workstations and NVIDIA RTX PRO GPUs in reshaping game development. It highlights NVIDIA’s pioneering moves in neural rendering, AI-powered code workflows, and complete virtualization for studios. Paul Logan provides deep yet accessible insights, making this a must-listen for anyone interested in the future of high-performance computing in game development.

Transcript

[A]

Welcome to reshaping workflows with Dell Pro Precision and Nvidia, where innovation meets real world impact in high performance computing.

[B]

Live from gtc. This is reshaping workflows with Dell Pro Precision and Nvidia RTX GPUs. I'm here with my new favorite friend and I'm losing my voice because I've done 100 of these. Paul Logan. So Paul Logan, tell everyone what you do at Nvidia.

[A]

Well, Logan, thank you so much for asking. I am the senior Product Marketing Manager for Enterprise Game Development at Nvidia. What that means is I help companies like Activision, like Sony, like Blizzard, accelerate their game development and reduce costs so that they can get to market quicker.

[B]

So senior, is that you just adding that because you're trying to feel special? Did you actually get promoted?

[A]

I do feel special and I've always had that role, actually, Logan. So maybe you just didn't recognize.

[B]

Honestly, I feel kind of bad. I didn't know you were a senior. So GTC is where all the hotness happens, all the new things. What are the big things from Nvidia with game dev happening at gtc, what are the big announcements?

[A]

Okay, in terms of big announcements, obviously we just announced DLSS5, that's neural rendering for aim engines. You can see all the information about it in the GTC keynote. And there's also going to be a lot of information coming out with for press. That's probably the biggest thing holistically in terms of how we're going to affect the gaming industry. For enterprise gaming specifically, we are also announcing an upcoming blueprint for MCP Agentic code development. That's something that we actually have a talk with EA here at the conference. It's going to be really impactful in terms of reducing token costs for entire studios that are trying to generate code across the entire development workflow. And finally, we're also, we have a big pitch, a big push here on RTX Pro Server for NV and Nvidia VGPU for virtualizing your entire game. Game studio's development workflow, from artists to game developers to AI researchers to quality assurance, all of that can be done on virtualized GPU infrastructure.

[B]

I love that. So tell us a little bit more about the MCP Agentic agent. Is this, you know, making NPC type characters? Tell us a little bit more about the workflow and the blueprint.

[A]

Yeah, yeah, definitely. So the workflow itself is basically, let's say you're a large game dev studio, you have maybe 500 gigabytes of legacy code in C. You've been working in Unreal or a custom branch for a long time and right now all of your programmers are programming on Claude code. I don't know if you've probably programmed on claude.

[B]

I have.

[A]

You've looked at the prices of Claude code.

[B]

It's expensive.

[A]

It's expensive. And we absolutely love our partners here. And obviously anthropics models produce some of the best results. But what we have done is we've wired up kind of quencoder next on Quadco and then directed it towards an MCP that we have custom developed. So an MCP server is it basically indexes all of the code in these large repos and then makes informed prompting decisions instead of actually just doing the raw prompts. So we enrich the prompts and then we go and we do a vector search on those ingress prompts to get full context. And one of the things that's really important here is we do automated syntax chunking. So instead of just grabbing random blobs of code from the entire database or from the entire code base, we go in and we grab context aware chunks. So you grab the entire function for collision on bullets or we grab the entire function for day and night cycles. And so that way when you actually are going and querying, instead of getting like a blob or a trunk that might not have everything you absolutely need to know, you are getting exactly what you need from an enriched prompt. And that increases code accuracy from anywhere from 25% on a base model to up to 90% on a fully fine tuned model.

[B]

I'm blown away. Paul Logan, this is why I flagged you down. So Nvidia GTC Game Dev Paul Logan, always a pleasure. Check it out and we'll see you on the next one. Do what you want. Do what you want. Do it.

[A]

This podcast was produced in partnership with Amaze Media Labs.