Added OV-2025_2_AI-PC_10_TPS

[mfhansen]

Adding results for OV-2025.2 for AI-PC with 10+ tokens/sec performance on 2nd token latency

Details:

• CSV files for LNL-7-258V, ARL-7-255H and MTL-7-155H
• Updated release date
• Updated system description PDF file

Tickets:

• n/a

[rkazants]

not sure that we need this binary file. previously, we seemed not to have similar one

[mfhansen]

The binary file is a PDF version of an Excel sheet that describes the system configurations of the platforms used for generating the data we display on that web page. It is an Intel Legal requirement that this information be supplied on the web page. The PDF is a requirement and must be present.

[rkazants]

Hi @mfhansen, btw do you have table with comparison of previous and current results?
Do we have improvements and regressions there? How many of each of them?

[mfhansen]

Hi @mfhansen, btw do you have table with comparison of previous and current results? Do we have improvements and regressions there? How many of each of them?

Hi @rkazants , we usually do not make any comparisons as there are no defined criteria related to this for releasing these numbers. They are a snapshot or "current status" and not a promotion of any AI-PC on any workload. However, I took up your suggestion and for our own benefit we will create this comparison spreadsheet from now on.
In summary:
We have 92 measurements of a model_precision_input_token_length and we see 31 improvements and 11 regressions on 1st token latency. 2nd token latency show 1 improvement and one regression.
Improvements on 1st token latency, flat on 2nd token latency. See uploaded spreadsheet for details.
OV-2025.2_to_OV-2025.1_comparison_of_10_TPS.xlsx