Probably neither case will have a measurable impact on your SEO.
If I had to pick one, based on this information, I would pick the new option. If multiple smaller files are requested in parallel, that can make the website load faster.
The problem with page speed in general
You mention something called "page load speed." This is an abstract concept. I understand you have a tool that measures this, but it is just a metric the tool made up. And other tools might come to other results based on their made-up metrics. Also, users do not care what some tool defines as "page load speed."
It does not mean the tools are wrong. But they try to express something complex in a single metric. And that rarely leads to good results.
Many years ago, you could actually measure "page load speed." It was obviously the time a website took to load fully. But with JavaScript, AJAX, etc. this is no longer true. As a user, once I can start using a website, I do not care if it takes two more minutes to load pixels and cookies in the background. Or image website that is using lazy loading. When is it loaded? Or an interactive website - how do you even define when it is loaded? When we do not know when a page is loaded, we cannot possibly measure it and say it takes 2.9 seconds vs 3.0 seconds.
Google's solution: Core Web Vitals
Because of the issues mentioned above, Google has introduced the Core Web Vitals (CWV). This are three page speed metrics that are well-aligned with how humans perceive the page load experience. Whenever you talk about page speed for SEO, only the three CWV metrics matter. They are:
- Largest Contentful Paint (LCP) to approximate the time to visually load
- First Input Delay (FID) to approximate the delay to interact (click buttons, etc.)
- Cumulative Layout Shift (CLS) to approximate the time until a website has reached visual stability (no more images or text blocks jumping around)
What does it mean for you?
Measure the Core Web Vitals with both versions. Then make your decision.