diff --git a/src/pcm-pcie.cpp b/src/pcm-pcie.cpp
index 335c24d1..35914dcd 100644
--- a/src/pcm-pcie.cpp
+++ b/src/pcm-pcie.cpp
@@ -222,18 +222,22 @@ int mainThrows(int argc, char * argv[])
         // for non-CSV mode delay < 1.0 does not make a lot of practical sense:
         // hard to read from the screen, or
         // in case delay is not provided in command line => set default
-        if ( ((delay<1.0) && (delay>0.0)) || (delay<=0.0) ) delay = PCM_DELAY_DEFAULT;
+        if ( ((delay < 1.0) && (delay > 0.0)) || (delay <= 0.0) ) {
+            cerr << "For non-CSV mode delay < 1.0s does not make a lot of practical sense. Default delay 1s is used. Consider to use CSV mode for lower delay values\n";
+            delay = PCM_DELAY_DEFAULT;
+        }
     }
 
     cerr << "Update every " << delay << " seconds\n";
 
+    // Delay in miliseconds
     unique_ptr<IPlatform> platform(IPlatform::getPlatform(m, csv, print_bandwidth,
-                                    print_additional_info, (uint)delay)); // FIXME: do we support only integer delay? ; lgtm [cpp/fixme-comment]
+                                    print_additional_info, (uint)(delay * 1000)));
 
     if (!platform)
     {
         print_cpu_details();
-        cerr << "Jaketown, Ivytown, Haswell, Broadwell-DE Server CPU is required for this tool! Program aborted\n";
+        cerr << "Jaketown, Ivytown, Haswell, Broadwell-DE, Skylake, Icelake, Snowridge and Sapphirerapids Server CPU is required for this tool! Program aborted\n";
         exit(EXIT_FAILURE);
     }
 
diff --git a/src/pcm-pcie.h b/src/pcm-pcie.h
index 1fb061e7..533ffa66 100644
--- a/src/pcm-pcie.h
+++ b/src/pcm-pcie.h
@@ -131,8 +131,8 @@ class LegacyPlatform: public IPlatform
         int eventsCount = 0;
         for (auto &group : eventGroups) eventsCount += (int)group.size();
 
-        m_delay = uint32(delay * 1000 / (eventGroups.size()) / NUM_SAMPLES);
-        if (m_delay * eventsCount * NUM_SAMPLES < delay * 1000) ++m_delay;
+        // Delay for each multiplexing group. Counters will be scaled.
+        m_delay = uint32(delay / eventGroups.size() / NUM_SAMPLES);
 
         eventSample.resize(m_socketCount);
         for (auto &e: eventSample)
@@ -182,11 +182,12 @@ void LegacyPlatform::getEventGroup(eventGroup_t &eventGroup)
     m_pcm->programPCIeEventGroup(eventGroup);
     uint offset = eventGroupOffset(eventGroup);
 
-    for (auto &run : eventCount) {
-        for(uint skt =0; skt < m_socketCount; ++skt)
+    for (int run = before; run < total; run++) {
+        for (uint skt = 0; skt < m_socketCount; ++skt)
             for (uint ctr = 0; ctr < eventGroup.size(); ++ctr)
-                run[skt][ctr + offset] = m_pcm->getPCIeCounterData(skt, ctr);
-        MySleepMs(m_delay);
+                eventCount[run][skt][ctr + offset] = m_pcm->getPCIeCounterData(skt, ctr);
+        if (run == before)
+            MySleepMs(m_delay);
     }
 
     for(uint skt = 0; skt < m_socketCount; ++skt)