News AMD introduces free and open-source Frame Latency Meter utility — works with all GPUs, no high-speed camera or sensor required

[Admin]

AMD released its Frame Latency Meter software utility for measuring mouse click to display latency that works with all GPUs, without the specialized hardware or manual frame counting that's normally required.

AMD introduces free and open-source Frame Latency Meter utility — works with all GPUs, no high-speed camera or sensor required : Read more

 

[Makaveli]

Tool looks good when will THG incorporate it into your reviews? 😎

 

[DougMcC]

Physical reality constrains this to be a measure of mouse to ouput frame, right? You can't know how long from there to the result is reflected on the screen?

 

[JarredWaltonGPU]

Physical reality constrains this to be a measure of mouse to ouput frame, right? You can't know how long from there to the result is reflected on the screen?

Yeah, without a sensor on the display, you don't know for certain when it shows up. So this is really more of a tool to determine if drivers or whatever are adding latency AFAICT. Because some displays (especially older monitors) would add one or even two frames of processing latency. I remember testing this back in the day.

Tool looks good when will THG incorporate it into your reviews? 😎

Honestly, with FrameView already capturing similar data on some games, I think we're fine. I average the latency over all frames of the benchmark runs that I run, and it doesn't require a separate workload that takes additional time and effort to generate a number that may or may not be fully accurate. And since it's in software, we're basically only looking at latency measurements not counting the display (which is actually good unless you're doing a monitor review).

 

[Makaveli]

Yeah, without a sensor on the display, you don't know for certain when it shows up. So this is really more of a tool to determine if drivers or whatever are adding latency AFAICT. Because some displays (especially older monitors) would add one or even two frames of processing latency. I remember testing this back in the day.

Honestly, with FrameView already capturing similar data on some games, I think we're fine. I average the latency over all frames of the benchmark runs that I run, and it doesn't require a separate workload that takes additional time and effort to generate a number that may or may not be fully accurate. And since it's in software, we're basically only looking at latency measurements not counting the display (which is actually good unless you're doing a monitor review).

Thank you for the quick response.

Will you guys post an article about

AMD Announces Full Support For Llama 3.1 AI Models Across EPYC CPUs, Instinct Accelerators, Ryzen AI NPUs & Radeon GPUs

 

[JarredWaltonGPU]

AMD Announces Full Support For Llama 3.1 AI Models Across EPYC CPUs, Instinct Accelerators, Ryzen AI NPUs & Radeon GPUs

I pitched it as a news story, it got turned down. Not really surprising, as stuff about AI and LLMs and such really doesn't do much traffic for us. Feels like unless it's a Really Big Deal (or something really funny or surprising), people are getting tired of hearing about AI. That Stack Overflow ban of OpenAI? That did a lot of traffic. Most other AI stories are about 2~5 percent of what that one article did. :-\

 

[Makaveli]

I pitched it as a news story, it got turned down. Not really surprising, as stuff about AI and LLMs and such really doesn't do much traffic for us. Feels like unless it's a Really Big Deal (or something really funny or surprising), people are getting tired of hearing about AI. That Stack Overflow ban of OpenAI? That did a lot of traffic. Most other AI stories are about 2~5 percent of what that one article did. :-\

lol I had no idea thanks for the insight.

On a side note I just got it up and running on my Radeon 7900XTX.

 

[hotaru251]

any chance to do a side by side comparison and see if its accurate vs the hardware alternative?