diff --git a/tests/tt_metal/tt_metal/perf_microbenchmark/dispatch/filt_pgm_dispatch.pl b/tests/tt_metal/tt_metal/perf_microbenchmark/dispatch/filt_pgm_dispatch.pl
index 29863d6a850..a1083042cde 100755
--- a/tests/tt_metal/tt_metal/perf_microbenchmark/dispatch/filt_pgm_dispatch.pl
+++ b/tests/tt_metal/tt_metal/perf_microbenchmark/dispatch/filt_pgm_dispatch.pl
@@ -22,7 +22,8 @@
         $j = 0;
     }
 
-    if ($line =~ /per iteration/) {
+    # output line from test_pgm_dispatch
+    if ($line =~ /us per iteration/) {
         my @parts = split(' ', $line);
         my $us = $parts[8];
         my $index = index($parts[8], ".");
@@ -32,6 +33,22 @@
         $data->[$j][$i] = $us;
         $j++;
     }
+
+    # output line from test_bw_and_latency
+    if ($line =~ /BW:/) {
+        my @parts = split(' ', $line);
+        my $bw = $parts[7];
+        $data->[$j][$i] = $bw;
+        $j++;
+    }
+
+    # output latency from test_bw_and_latency
+    if ($line =~ /Latency:/) {
+        my @parts = split(' ', $line);
+        my $bw = $parts[7];
+        $data->[$j][$i] = $bw;
+        $j++;
+    }
 }
 
 for (my $y = 0; $y < $maxj; $y++) {
diff --git a/tests/tt_metal/tt_metal/perf_microbenchmark/dispatch/kernels/bw_and_latency.cpp b/tests/tt_metal/tt_metal/perf_microbenchmark/dispatch/kernels/bw_and_latency.cpp
index 731ac407e0f..fec10557331 100644
--- a/tests/tt_metal/tt_metal/perf_microbenchmark/dispatch/kernels/bw_and_latency.cpp
+++ b/tests/tt_metal/tt_metal/perf_microbenchmark/dispatch/kernels/bw_and_latency.cpp
@@ -15,7 +15,6 @@ void kernel_main() {
         uint32_t read_ptr = cb_addr;
         uint32_t write_ptr = cb_addr;
         for (int j = 0; j < PAGE_COUNT; j++) {
-
 #if DRAM_BANKED
             uint64_t noc_addr = get_dram_noc_addr(j, page_size, 0);
 #else
@@ -24,8 +23,11 @@ void kernel_main() {
 
 #if ISSUE_MCAST
             uint64_t dst_noc_multicast_addr =
-                get_noc_multicast_addr(NOC_ADDR_X, NOC_ADDR_Y, MCAST_NOC_END_ADDR_X, MCAST_NOC_END_ADDR_Y, NOC_MEM_ADDR);
-            noc_async_write_multicast(write_ptr, dst_noc_multicast_addr, page_size, NUM_MCAST_DESTS);
+                get_noc_multicast_addr(NOC_ADDR_X, NOC_ADDR_Y, MCAST_NOC_END_ADDR_X, MCAST_NOC_END_ADDR_Y, write_ptr);
+            noc_async_write_multicast(read_ptr, dst_noc_multicast_addr, page_size, NUM_MCAST_DESTS, LINKED);
+#elif WRITE
+            uint64_t noc_write_addr = NOC_XY_ADDR(NOC_X(NOC_ADDR_X), NOC_Y(NOC_ADDR_Y), write_ptr);
+            noc_async_write(NOC_MEM_ADDR, noc_write_addr, page_size);
 #elif READ_ONE_PACKET
             noc_async_read_one_packet(noc_addr, read_ptr, page_size);
 #else
@@ -33,16 +35,29 @@ void kernel_main() {
 #endif
 
 #if LATENCY
-            noc_async_read_barrier();
+#if WRITE
+#if LINKED
+            noc_async_write_multicast(cb_addr, dst_noc_multicast_addr, page_size, NUM_MCAST_DESTS, false);
+#endif
             noc_async_write_barrier();
+#else
+            noc_async_read_barrier();
+#endif
 #endif
-
             read_ptr += page_size;
             write_ptr += page_size;
         }
     }
 #if !LATENCY
-    noc_async_read_barrier();
+#if WRITE
+#if LINKED
+    uint64_t dst_noc_multicast_addr =
+        get_noc_multicast_addr(NOC_ADDR_X, NOC_ADDR_Y, MCAST_NOC_END_ADDR_X, MCAST_NOC_END_ADDR_Y, cb_addr);
+    noc_async_write_multicast(cb_addr, dst_noc_multicast_addr, page_size, NUM_MCAST_DESTS, false);
+#endif
     noc_async_write_barrier();
+#else
+    noc_async_read_barrier();
+#endif
 #endif
 }
diff --git a/tests/tt_metal/tt_metal/perf_microbenchmark/dispatch/run_bw_and_latency.sh b/tests/tt_metal/tt_metal/perf_microbenchmark/dispatch/run_bw_and_latency.sh
deleted file mode 100755
index 90a4972019b..00000000000
--- a/tests/tt_metal/tt_metal/perf_microbenchmark/dispatch/run_bw_and_latency.sh
+++ /dev/null
@@ -1,80 +0,0 @@
-#!/bin/bash
-
-if [ "$ARCH_NAME" = "grayskull" ]; then
-    echo "Configured core range for grayskull"
-    max_x=11
-    max_y=8
-elif [ "$ARCH_NAME" = "wormhole_b0" ]; then
-    echo "Configured core range for wormhole_b0"
-    max_x=7
-    max_y=6
-elif [ "$ARCH_NAME" = "blackhole" ]; then
-    echo "Configured core range for blackhole"
-    max_x=12
-    max_y=9
-else
-    echo "Unknown arch: $ARCH_NAME"
-    exit 1
-fi
-
-function get_half_way_away_core_x() {
-    half_way_away_core_x=$(( ($1 + (($max_x + 1) / 2)) % ($max_x + 1) ))
-    echo $half_way_away_core_x
-}
-
-function get_half_way_away_core_y() {
-    half_way_away_core_y=$(( ($1 + (($max_y + 1) / 2)) % ($max_y + 1) ))
-    echo $half_way_away_core_y
-}
-
-function read_from_half_way_away_core() {
-    half_way_away_core_x=$(get_half_way_away_core_x $1)
-    half_way_away_core_y=$(get_half_way_away_core_y $2)
-    echo "./build/test/tt_metal/perf_microbenchmark/dispatch/test_bw_and_latency -m 2 -rx $1 -ry $2 -sx $half_way_away_core_x -sy $half_way_away_core_y"
-    ./build/test/tt_metal/perf_microbenchmark/dispatch/test_bw_and_latency -m 2 -rx $1 -ry $2 -sx $half_way_away_core_x -sy $half_way_away_core_y
-}
-
-function mcast_write_to_half_way_away_core() {
-    half_way_away_core_x=$(get_half_way_away_core_x $1)
-    half_way_away_core_y=$(get_half_way_away_core_y $2)
-    echo "./build/test/tt_metal/perf_microbenchmark/dispatch/test_bw_and_latency -m 6 -rx $1 -ry $2 -sx $half_way_away_core_x -sy $half_way_away_core_y -tx $half_way_away_core_x -ty $half_way_away_core_y"
-    ./build/test/tt_metal/perf_microbenchmark/dispatch/test_bw_and_latency -m 6 -rx $1 -ry $2 -sx $half_way_away_core_x -sy $half_way_away_core_y -tx $half_way_away_core_x -ty $half_way_away_core_y
-}
-
-function mcast_write_to_adjacent_core() {
-    adj_core_y=$(($2 + 1))
-    if [ $adj_core_y -gt $max_y ]; then
-        adj_core_y=$(($2 - 1))
-    fi
-    echo "./build/test/tt_metal/perf_microbenchmark/dispatch/test_bw_and_latency -m 6 -rx $1 -ry $2 -sx $1 -sy $adj_core_y -tx $1 -ty $adj_core_y"
-    ./build/test/tt_metal/perf_microbenchmark/dispatch/test_bw_and_latency -m 6 -rx $1 -ry $2 -sx $1 -sy $adj_core_y -tx $1 -ty $adj_core_y
-}
-
-function mcast_write_from_core_after_curr_core_to_half_way_away_core() {
-    half_way_away_core_x=$(get_half_way_away_core_x $1)
-    half_way_away_core_y=$(get_half_way_away_core_y $2)
-    mcast_start_y=$(($2 + 1))
-    echo "./build/test/tt_metal/perf_microbenchmark/dispatch/test_bw_and_latency -m 6 -rx $1 -ry $2 -sx $1 -sy $mcast_start_y -tx $half_way_away_core_x -ty $half_way_away_core_y"
-    ./build/test/tt_metal/perf_microbenchmark/dispatch/test_bw_and_latency -m 6 -rx $1 -ry $2 -sx $1 -sy $mcast_start_y -tx $half_way_away_core_x -ty $half_way_away_core_y
-}
-
-for ((x=0; x<=max_x; x++)); do
-    for ((y=0; y<=max_y; y++)); do
-        read_from_half_way_away_core $x $y
-        mcast_write_to_half_way_away_core $x $y
-        mcast_write_to_adjacent_core $x $y
-        mcast_write_from_core_after_curr_core_to_half_way_away_core $x $y
-
-        if [ $y -eq 0 ]; then
-            mcast_start_y=$(($y + 1))
-            echo "./build/test/tt_metal/perf_microbenchmark/dispatch/test_bw_and_latency -m 6 -rx $x -ry $y -sx 0 -sy $mcast_start_y -tx $max_x -ty $max_y"
-            ./build/test/tt_metal/perf_microbenchmark/dispatch/test_bw_and_latency -m 6 -rx $x -ry $y -sx 0 -sy $mcast_start_y -tx $max_x -ty $max_y
-        fi
-
-        if [ $y -eq $max_y ]; then
-            mcast_end_y=$(($y - 1))
-            echo "./build/test/tt_metal/perf_microbenchmark/dispatch/test_bw_and_latency -m 6 -rx $x -ry $y -sx 0 -sy 0 -tx $max_x -ty $mcast_end_y"
-            ./build/test/tt_metal/perf_microbenchmark/dispatch/test_bw_and_latency -m 6 -rx $x -ry $y -sx 0 -sy 0 -tx $max_x -ty $mcast_end_y
-        fi
-    done
-done
diff --git a/tests/tt_metal/tt_metal/perf_microbenchmark/dispatch/sweep_bw_and_latency.sh b/tests/tt_metal/tt_metal/perf_microbenchmark/dispatch/sweep_bw_and_latency.sh
new file mode 100755
index 00000000000..ccc3ec1e8d2
--- /dev/null
+++ b/tests/tt_metal/tt_metal/perf_microbenchmark/dispatch/sweep_bw_and_latency.sh
@@ -0,0 +1,108 @@
+#/bin/bash
+
+# Run this test w/
+#   sweep_bw_and_latency.sh 2>&1 | tee log
+# the run
+# filt_pgm_dispatch.pl
+# then paste the results into the BW spreadsheet
+
+if [ "$ARCH_NAME" = "grayskull" ]; then
+    echo "Configured core range for grayskull"
+    max_x=11
+    max_y=8
+elif [ "$ARCH_NAME" = "wormhole_b0" ]; then
+    echo "Configured core range for wormhole_b0"
+    max_x=7
+    max_y=6
+elif [ "$ARCH_NAME" = "blackhole" ]; then
+    echo "Configured core range for blackhole"
+    max_x=12
+    max_y=9
+else
+    echo "Unknown arch: $ARCH_NAME"
+    exit 1
+fi
+
+function get_half_way_away_core_x() {
+    half_way_away_core_x=$(( ($1 + (($max_x + 1) / 2)) % ($max_x + 1) ))
+    echo $half_way_away_core_x
+}
+
+function get_half_way_away_core_y() {
+    half_way_away_core_y=$(( ($1 + (($max_y + 1) / 2)) % ($max_y + 1) ))
+    echo $half_way_away_core_y
+}
+
+hx=$(get_half_way_away_core_x 0);
+hy=$(get_half_way_away_core_y 0);
+
+function run_one() {
+    echo "Running $@"
+    build/test/tt_metal/perf_microbenchmark/dispatch/test_bw_and_latency $@
+}
+
+function bw_test() {
+    run_one -bs 8 -p 16 $@
+    run_one -bs 8 -p 32 $@
+    run_one -bs 16 -p 64 $@
+    run_one -bs 32 -p 128 $@
+    run_one -bs 64 -p 256 $@
+    run_one -bs 128 -p 512 $@
+    run_one -bs 256 -p 1024 $@
+    run_one -bs 256 -p 2048 $@
+    run_one -bs 256 -p 4096 $@
+    run_one -bs 256 -p 8192 $@
+    run_one -bs 256 -p 16384 $@
+    run_one -bs 256 -p 32768 $@
+    run_one -bs 256 -p 65536 $@
+}
+
+function latency_test() {
+    run_one -bs 8 -p 16 -l $@
+}
+
+echo "###" read pcie
+bw_test "-m 0"
+latency_test "-m 0"
+
+echo "###" read dram
+bw_test "-m 1"
+latency_test "-m 1"
+
+echo "###" read drams
+bw_test "-m 3"
+latency_test "-m 3"
+
+echo "###" read l1 adjacent
+bw_test "-m 2"
+latency_test "-m 2"
+
+echo "###" read l1 far halfway away
+bw_test "-m 2  -rx 0 -ry 0 -sx $hx -sy $hy"
+latency_test "-m 2 -rx 0 -ry 0 -sx $hx -sy $hy"
+
+echo "###" read local
+bw_test "-m 2 -rx 0"
+latency_test "-m 2 -rx 0"
+
+echo "###" write l1 far halfway away
+bw_test "-m 2 -rx 0 -ry 0 -sx $hx -sy $hy -wr"
+latency_test "-m 2 -rx 0 -ry 0 -sx $hx -sy $hy -wr"
+
+echo "###" mcast write to adjacent
+bw_test "-m 6 -rx 0 -ry 0 -sx 1 -sy 0 -tx 1 -ty 0"
+latency_test "-m 6 -rx 0 -ry 0 -sx 1 -sy 0 -tx 1 -ty 0"
+
+echo "###" mcast write to halfway away
+bw_test "-m 6 -rx 0 -ry 0 -sx $hx -sy $hy -tx $hx -ty $hy"
+latency_test "-m 6 -rx 0 -ry 0 -sx $hx -sy $hy -tx $hx -ty $hy"
+
+echo "###" mcast write to all
+bw_test "-m 6 -rx 0 -ry 0 -sx 0 -sy 1 -tx $max_x -ty $max_y"
+latency_test "-m 6 -rx 0 -ry 0 -sx 0 -sy 1 -tx $max_x -ty $max_y"
+
+echo "###" mcast write to all, linked
+bw_test "-m 6 -rx 0 -ry 0 -sx 0 -sy 1 -tx $max_x -ty $max_y -link"
+latency_test "-m 6 -rx 0 -ry 0 -sx 0 -sy 1 -tx $max_x -ty $max_y -link"
+
+echo "###" done
diff --git a/tests/tt_metal/tt_metal/perf_microbenchmark/dispatch/test_bw_and_latency.cpp b/tests/tt_metal/tt_metal/perf_microbenchmark/dispatch/test_bw_and_latency.cpp
index d13994ded6e..34283f67937 100644
--- a/tests/tt_metal/tt_metal/perf_microbenchmark/dispatch/test_bw_and_latency.cpp
+++ b/tests/tt_metal/tt_metal/perf_microbenchmark/dispatch/test_bw_and_latency.cpp
@@ -48,6 +48,8 @@ bool page_size_as_runtime_arg_g; // useful particularly on GS multi-dram tests (
 bool hammer_write_reg_g = false;
 bool hammer_pcie_g = false;
 bool hammer_pcie_type_g = false;
+bool test_write = false;
+bool linked = false;
 
 void init(int argc, char **argv) {
     std::vector<std::string> input_args(argv, argv + argc);
@@ -67,10 +69,12 @@ void init(int argc, char **argv) {
         log_info(LogTest, "  -sy: when reading from L1, Y of core to read from. when issuing a multicast write, Y of start core to write to. (default {})", 0);
         log_info(LogTest, "  -tx: when issuing a multicast write, X of end core to write to (default {})", 0);
         log_info(LogTest, "  -ty: when issuing a multicast write, Y of end core to write to (default {})", 0);
+        log_info(LogTest, "  -wr: issue unicast write instead of read (default false)");
         log_info(LogTest, "  -c: when reading from dram, DRAM channel (default 0)");
         log_info(LogTest, "  -f: time just the finish call (use w/ lazy mode) (default disabled)");
         log_info(LogTest, "  -o: use read_one_packet API.  restricts page size to 8K max (default {})", 0);
         log_info(LogTest, "  -z: enable dispatch lazy mode (default disabled)");
+        log_info(LogTest, "-link: link mcast transactions");
         log_info(LogTest, " -hr: hammer write_reg while executing (for PCIe test)");
         log_info(LogTest, " -hp: hammer hugepage PCIe memory while executing (for PCIe test)");
         log_info(LogTest, " -hpt:hammer hugepage PCIe hammer type: 0:32bit writes 1:128bit non-temporal writes");
@@ -104,6 +108,14 @@ void init(int argc, char **argv) {
     }
     page_count_g = size_bytes / page_size_g;
 
+    test_write = test_args::has_command_option(input_args, "-wr");
+    if (test_write && (source_mem_g != 2 && source_mem_g != 6)) {
+        log_info(LogTest, "Writing only tested w/ L1 destination\n");
+        exit(-1);
+    }
+
+    linked = test_args::has_command_option(input_args, "-link");
+
     worker_g = CoreRange({core_x, core_y}, {core_x, core_y});
     src_worker_g = {src_core_x, src_core_y};
 
@@ -200,7 +212,7 @@ int main(int argc, char **argv) {
             break;
         case 2:
             {
-                src_mem = "FROM_L1";
+                src_mem = test_write ? "TO_L1" : "FROM_L1";
                 CoreCoord w = device->physical_core_from_logical_core(src_worker_g, CoreType::WORKER);
                 noc_addr_x = w.x;
                 noc_addr_y = w.y;
@@ -245,6 +257,7 @@ int main(int argc, char **argv) {
                 noc_addr_y = start.y;
                 mcast_noc_addr_end_x = end.x;
                 mcast_noc_addr_end_y = end.y;
+                test_write = true;
             }
             break;
         }
@@ -259,6 +272,8 @@ int main(int argc, char **argv) {
             {"READ_ONE_PACKET", std::to_string(read_one_packet_g)},
             {"DRAM_BANKED", std::to_string(dram_banked)},
             {"ISSUE_MCAST", std::to_string(issue_mcast)},
+            {"WRITE", std::to_string(test_write)},
+            {"LINKED", std::to_string(linked)},
             {"NUM_MCAST_DESTS", std::to_string(num_mcast_dests)},
             {"MCAST_NOC_END_ADDR_X", std::to_string(mcast_noc_addr_end_x)},
             {"MCAST_NOC_END_ADDR_Y", std::to_string(mcast_noc_addr_end_y)}
@@ -284,27 +299,29 @@ int main(int argc, char **argv) {
 
         CoreCoord w = device->physical_core_from_logical_core(worker_g.start_coord, CoreType::WORKER);
         log_info(LogTest, "Master core: {}", w.str());
+        string direction = test_write ? "Writing" : "Reading";
         if (source_mem_g == 3) {
-            log_info(LogTest, "Reading: {}", src_mem);
+            log_info(LogTest, "{}: {}", direction, src_mem);
         } else if (source_mem_g == 4) {
-            log_info(LogTest, "Reading: {} - core ({}, {})", src_mem, w.x, w.y);
+            log_info(LogTest, "{}: {} - core ({}, {})", direction, src_mem, w.x, w.y);
         } else if (source_mem_g == 5) {
-            log_info(LogTest, "Writing: {} - core ({}, {})", src_mem, w.x, w.y);
+            log_info(LogTest, "{}: {} - core ({}, {})", test_write, src_mem, w.x, w.y);
         } else if (source_mem_g == 6) {
-            log_info(LogTest, "Writing: {} - core grid [({}, {}) - ({}, {})]", src_mem, noc_addr_x, noc_addr_y, mcast_noc_addr_end_x, mcast_noc_addr_end_y);
+            log_info(LogTest, "direction: {} - core grid [({}, {}) - ({}, {})]", direction, src_mem, noc_addr_x, noc_addr_y, mcast_noc_addr_end_x, mcast_noc_addr_end_y);
         } else {
-            log_info(LogTest, "Reading: {} - core ({}, {})", src_mem, noc_addr_x, noc_addr_y);
+            log_info(LogTest, "{}: {} - core ({}, {})", direction, src_mem, noc_addr_x, noc_addr_y);
         }
         if (source_mem_g < 4 || source_mem_g == 6) {
             std::string api;
+            string read_write = test_write ? "write" : "read";
             if (issue_mcast) {
-                api = "noc_async_write_multicast";
+                api = "noc_async_" + read_write + "_multicast";
             }
             else if (read_one_packet_g) {
-                api = "noc_async_read_one_packet";
+                api = "noc_async_" + read_write + "_one_packet";
             }
             else {
-                api = "noc_async_read";
+                api = "noc_async_" + read_write;
             }
             log_info(LogTest, "Using API: {}", api);
             log_info(LogTest, "Lazy: {}", lazy_g);