Background: There are few contemporary evaluations of directly measured LDL (dLDL) assays. We evaluated the performance of the Roche Gen.3 dLDL assay and compared it to the Friedewald LDL (cLDL) in a large cohort, tested on the Cobas c702 analyser. Methods: We evaluated assay precision, linearity, and limit of detection (LOD). To compare cLDL/dLDL, lipid panels (TC/TG/HDL/cLDL) from 2017- 2019 (n=117,090) were tested for dLDL. Samples with TG >400mg/dL (4.5mmol/L) (n=605) and negative cLDL (n=32) were excluded. We examined the difference between cLDL/dLDL (n=116,453), the influence of increasing levels of TG/LDL on their measurements, and how cLDL/dLDL classified cardiovascular risk by LDL levels. Results: The Roche dLDL assay has a CV of 1.0%/0.9% at 58.4/106.4mg/dL, is linear from 19.4-374mg/dL, and has a verified LOD of 4.2mg/dL. Despite close agreement between dLDL/cLDL [Pearson r=0.98 (95%CI 0.9795-0.9800)], cLDL underestimates dLDL in 98.5% (n=114,750) of subjects across all levels of LDL/TG. The underestimation increases with LDL/TG levels. cLDL classified more subjects (63.5%) as having a desirable LDL (<100mg/dL) than dLDL (46.9%). Conclusions: The Cobas c702 dLDL assay performs well, and contemporary cLDL results underestimate dLDL across all levels of TG/LDL. cLDL classifies more patients into lower cardiovascular risk categories than dLDL