The performance is so bad the test fails on SPR tesbeds, as it cannot sustain even 9k pps rate. The observed reason is the abysmal telemetry [0], half a million cycles per packet.
Those tests were added in [1], three of them use SHA_256_128 integrity algorithm, and only those three have reduced performance, depending on testbed. On 3n-snr the tests were even crashing VPP, fixed by so I guess similar fix is also needed for Xeon processors.
Summary of trending performance on various testbeds: [3],

[0] https://s3-logs.fd.io/vex-yul-rot-jenkins-1/csit-vpp-perf-report-iterative-2406-3na-spr/57/log.html.gz#s1-s1-s1-s1-s3-t1-k3-k7-k1-k1-k1-k8-k10-k1-k1-k1-k12
[1] 40820: feat(ipsec): Add suites for more algs | https://gerrit.fd.io/r/c/csit/+/40820
[2] 41141: crypto-native: fix CPU detection for SHA2 code | https://gerrit.fd.io/r/c/vpp/+/41141
[3] https://csit.fd.io/trending/#eNrllk1ugzAQhU9DN9VI2MSFTRdNuUdlxtNilR_XdlLo6esQJBMprLpBhQUs3sjz6T08tvO9pTdHzXMijkl-THiuVXgl2ctj-JyNgawD2Xg4pB_EDRuanKWgjPoEtKPxPTDBigoYAvlam4M2jpCl4fFd477hpAxIC0GppCPQnQdJjvECK4S6lcjFk6slGGUvffnrpa86-RuIqJh6jMoaWiyXlmSsvxJH1ZNb9PkTf1z03cqWnP6huPJkUazA4PlCxFsgP5qFOtuSl1PFWkAaBwjMkw1UsBS_NhTRCtweQ1LAxewDz3BbGd1h211ErrMLF_imttF9tt1F5F0dpsl14Au-paPoHtne4pHgTPhN52mPQ34m6aFtBrGJjNbx9hZUNTmx1TvDKt2_jEmUD11v2-kKLspfA5lShg