diff --git a/cm-mlops/script/get-sys-utils-cm/do_pip_installs.sh b/cm-mlops/script/get-sys-utils-cm/do_pip_installs.sh
new file mode 100644
index 0000000000..55a1492492
--- /dev/null
+++ b/cm-mlops/script/get-sys-utils-cm/do_pip_installs.sh
@@ -0,0 +1,6 @@
+#!/bin/bash
+
+PIP_EXTRA=`python3 -c "import pkg_resources; print(' --break-system-packages ' if int(pkg_resources.get_distribution('pip').version.split('.')[0]) >= 23 else '')"`
+cmd="python3 -m pip install -r ${CM_TMP_CURRENT_SCRIPT_PATH}/requirements.txt ${CM_PYTHON_PIP_USER} ${CM_PYTHON_PIP_COMMON_EXTRA} ${PIP_EXTRA}"
+echo $cmd
+eval $cmd
diff --git a/cm-mlops/script/get-sys-utils-cm/run-arch.sh b/cm-mlops/script/get-sys-utils-cm/run-arch.sh
index 1f786fd8cd..1a5338c58a 100644
--- a/cm-mlops/script/get-sys-utils-cm/run-arch.sh
+++ b/cm-mlops/script/get-sys-utils-cm/run-arch.sh
@@ -31,4 +31,5 @@ ${CM_SUDO} ${CM_PACKAGE_TOOL} -Syu && \
            xz \
            zip 
 
-python3 -m pip install -r ${CM_TMP_CURRENT_SCRIPT_PATH}/requirements.txt ${CM_PYTHON_PIP_USER}
+. ${CM_TMP_CURRENT_SCRIPT_PATH}/do_pip_installs.sh
+test $? -eq 0 || exit $?
diff --git a/cm-mlops/script/get-sys-utils-cm/run-debian.sh b/cm-mlops/script/get-sys-utils-cm/run-debian.sh
index 4c35bd52ba..c7de244641 100644
--- a/cm-mlops/script/get-sys-utils-cm/run-debian.sh
+++ b/cm-mlops/script/get-sys-utils-cm/run-debian.sh
@@ -52,4 +52,5 @@ ${CM_SUDO} ${CM_APT_TOOL} update && \
            libgl1 \
            libncurses5
 
-python3 -m pip install -r ${CM_TMP_CURRENT_SCRIPT_PATH}/requirements.txt ${CM_PYTHON_PIP_USER}
+. ${CM_TMP_CURRENT_SCRIPT_PATH}/do_pip_installs.sh
+test $? -eq 0 || exit $?
diff --git a/cm-mlops/script/get-sys-utils-cm/run-macos.sh b/cm-mlops/script/get-sys-utils-cm/run-macos.sh
index 39e256cecf..1e5eab4ad5 100644
--- a/cm-mlops/script/get-sys-utils-cm/run-macos.sh
+++ b/cm-mlops/script/get-sys-utils-cm/run-macos.sh
@@ -35,4 +35,5 @@ brew update && \
            zlib \
            python3
 
-python3 -m pip install -r ${CM_TMP_CURRENT_SCRIPT_PATH}/requirements.txt ${CM_PYTHON_PIP_USER}
+. ${CM_TMP_CURRENT_SCRIPT_PATH}/do_pip_installs.sh
+test $? -eq 0 || exit $?
diff --git a/cm-mlops/script/get-sys-utils-cm/run-rhel.sh b/cm-mlops/script/get-sys-utils-cm/run-rhel.sh
index 0fcfe252e0..9e3959d36d 100644
--- a/cm-mlops/script/get-sys-utils-cm/run-rhel.sh
+++ b/cm-mlops/script/get-sys-utils-cm/run-rhel.sh
@@ -34,4 +34,5 @@ ${CM_SUDO} ${CM_PACKAGE_TOOL} update && \
            xz \
            zip 
 
-python3 -m pip install -r ${CM_TMP_CURRENT_SCRIPT_PATH}/requirements.txt ${CM_PYTHON_PIP_USER}
+. ${CM_TMP_CURRENT_SCRIPT_PATH}/do_pip_installs.sh
+test $? -eq 0 || exit $?
diff --git a/cm-mlops/script/get-sys-utils-cm/run-sles.sh b/cm-mlops/script/get-sys-utils-cm/run-sles.sh
index 1ea118f0e0..32cfdbabc9 100644
--- a/cm-mlops/script/get-sys-utils-cm/run-sles.sh
+++ b/cm-mlops/script/get-sys-utils-cm/run-sles.sh
@@ -34,4 +34,5 @@ ${CM_SUDO} ${CM_PACKAGE_TOOL} update && \
            xz \
            zip 
 
-python3 -m pip install -r ${CM_TMP_CURRENT_SCRIPT_PATH}/requirements.txt ${CM_PYTHON_PIP_USER}
+. ${CM_TMP_CURRENT_SCRIPT_PATH}/do_pip_installs.sh
+test $? -eq 0 || exit $?
diff --git a/cm-mlops/script/get-sys-utils-cm/run-ubuntu.sh b/cm-mlops/script/get-sys-utils-cm/run-ubuntu.sh
index a5f7648de6..c87dbeb728 100644
--- a/cm-mlops/script/get-sys-utils-cm/run-ubuntu.sh
+++ b/cm-mlops/script/get-sys-utils-cm/run-ubuntu.sh
@@ -56,4 +56,5 @@ ${CM_SUDO} ${CM_APT_TOOL} update && \
            libgl1-mesa-glx \
            zlib1g-dev
 
-python3 -m pip install -r ${CM_TMP_CURRENT_SCRIPT_PATH}/requirements.txt ${CM_PYTHON_PIP_USER} ${CM_PYTHON_PIP_COMMON_EXTRA}
+. ${CM_TMP_CURRENT_SCRIPT_PATH}/do_pip_installs.sh
+test $? -eq 0 || exit $?
diff --git a/docs/mlperf/inference/README_a100.md b/docs/mlperf/inference/README_a100.md
index c60e0fdb9d..92035ce01a 100644
--- a/docs/mlperf/inference/README_a100.md
+++ b/docs/mlperf/inference/README_a100.md
@@ -8,7 +8,7 @@ cmr "generate-run-cmds inference _performance-only"  \
 --model=bert-99 --implementation=nvidia-original \
 --device=cuda --backend=tensorrt --category=edge \
 --division=open --quiet --scenario=Offline --execution_mode=fast \
---target_qps=1000 --rerun --gpu_name=a100 \
+--target_qps=3560 --rerun --gpu_name=a100 \
 --adr.nvidia-harness.tags=_sxm
 ```
 
@@ -20,7 +20,7 @@ cmr "generate-run-cmds inference _submission _all-scenarios"  \
 --model=bert-99 --implementation=nvidia-original \
 --device=cuda --backend=tensorrt --category=edge \
 --division=open --quiet --execution_mode=valid \
---target_qps=1000 --rerun --gpu_name=a100 \
+--target_qps=3560 --rerun --gpu_name=a100 \
 --adr.nvidia-harness.tags=_sxm
 ```
 
@@ -38,7 +38,7 @@ cmr "generate-run-cmds inference _performance-only"  \
 --model=resnet50 --implementation=nvidia-original \
 --device=cuda --backend=tensorrt --category=edge \
 --division=open --quiet --scenario=Offline --execution_mode=fast \
---target_qps=1000 --rerun --gpu_name=a100 \
+--target_qps=43000 --rerun --gpu_name=a100 \
 --adr.nvidia-harness.tags=_sxm
 ```
 
@@ -50,7 +50,7 @@ cmr "generate-run-cmds inference _submission _all-scenarios"  \
 --model=resnet50 --implementation=nvidia-original \
 --device=cuda --backend=tensorrt --category=edge \
 --division=open --quiet --execution_mode=valid \
---target_qps=1000 --rerun --gpu_name=a100 \
+--target_qps=43000 --rerun --gpu_name=a100 \
 --adr.nvidia-harness.tags=_sxm
 ```
 
@@ -62,7 +62,7 @@ cmr "generate-run-cmds inference _performance-only"  \
 --model=retinanet --implementation=nvidia-original \
 --device=cuda --backend=tensorrt --category=edge \
 --division=open --quiet --scenario=Offline --execution_mode=fast \
---target_qps=100 --rerun --gpu_name=a100 \
+--target_qps=715 --rerun --gpu_name=a100 \
 --adr.nvidia-harness.tags=_sxm
 ```
 
@@ -74,7 +74,7 @@ cmr "generate-run-cmds inference _submission _all-scenarios"  \
 --model=retinanet --implementation=nvidia-original \
 --device=cuda --backend=tensorrt --category=edge \
 --division=open --quiet --execution_mode=valid \
---target_qps=100 --rerun --gpu_name=a100 \
+--target_qps=715 --rerun --gpu_name=a100 \
 --adr.nvidia-harness.tags=_sxm
 ```
 
@@ -87,7 +87,7 @@ cmr "generate-run-cmds inference _performance-only"  \
 --model=rnnt --implementation=nvidia-original \
 --device=cuda --backend=tensorrt --category=edge \
 --division=open --quiet --scenario=Offline --execution_mode=fast \
---target_qps=1000 --rerun --gpu_name=a100 \
+--target_qps=14000 --rerun --gpu_name=a100 \
 --adr.nvidia-harness.tags=_sxm
 ```
 
@@ -99,7 +99,7 @@ cmr "generate-run-cmds inference _submission _all-scenarios"  \
 --model=rnnt --implementation=nvidia-original \
 --device=cuda --backend=tensorrt --category=edge \
 --division=open --quiet --execution_mode=valid \
---target_qps=1000 --rerun --gpu_name=a100 \
+--target_qps=14000 --rerun --gpu_name=a100 \
 --adr.nvidia-harness.tags=_sxm
 ```
 
@@ -112,7 +112,7 @@ cmr "generate-run-cmds inference _performance-only"  \
 --model=3d-unet-99 --implementation=nvidia-original \
 --device=cuda --backend=tensorrt --category=edge \
 --division=open --quiet --scenario=Offline --execution_mode=fast \
---target_qps=1000 --rerun --gpu_name=a100 \
+--target_qps=3.7 --rerun --gpu_name=a100 \
 --adr.nvidia-harness.tags=_sxm
 ```
 
@@ -124,7 +124,7 @@ cmr "generate-run-cmds inference _submission _all-scenarios"  \
 --model=3d-unet-99 --implementation=nvidia-original \
 --device=cuda --backend=tensorrt --category=edge \
 --division=open --quiet --execution_mode=valid \
---target_qps=1000 --rerun --gpu_name=a100 \
+--target_qps=3.7 --rerun --gpu_name=a100 \
 --adr.nvidia-harness.tags=_sxm
 ```