diff --git a/README.md b/README.md
index bfb3ed4..d9fc6a7 100644
--- a/README.md
+++ b/README.md
@@ -76,6 +76,6 @@ The Transport has these options:
 - **DeferSend**: This defers sends to the Transport Thread which increases chances for multiple sends to coalesce. This options
 defaults to true.
 
-- **ThreadCount**: Specifies the number of Transport Threads. This defaults to the number of logical processors in the system divided by 4.
+- **ThreadCount**: Specifies the number of Transport Threads. This defaults to the number of logical processors in the system, maxed to 16.
 
 - **AioSend/AioReceive**: Uses Linux AIO system calls to batch send and receive calls. AioSend implies DeferSend. These options default to true.
diff --git a/src/RedHatX.AspNetCore.Server.Kestrel.Transport.Linux/TransportOptions.cs b/src/RedHatX.AspNetCore.Server.Kestrel.Transport.Linux/TransportOptions.cs
index dc9eaa7..fed354c 100644
--- a/src/RedHatX.AspNetCore.Server.Kestrel.Transport.Linux/TransportOptions.cs
+++ b/src/RedHatX.AspNetCore.Server.Kestrel.Transport.Linux/TransportOptions.cs
@@ -20,11 +20,10 @@ public LinuxTransportOptions()
             AioSend = true;
             AioReceive = true;
 
-            // Benchmarking Techempower Json on 24-core machine with hyper threading (ProcessorCount = 48)
-            // shows best performance at ThreadCount 12.
-            // TODO: what happens if hyperthreading is disabled? Perhaps this should be half the cores?
-            // TODO: benchmark more scenarios to validate this is a good default.
-            ThreadCount = Math.Max((Environment.ProcessorCount + 2) / 4, 1);
+            // Use a default ThreadCount that wont cause the number of threads
+            // to be a bottleneck.
+            // Users that want to optimize, should do their own benchmarks.
+            ThreadCount = Math.Min(Environment.ProcessorCount, 16);
         }
 
         internal bool ReceiveOnIncomingCpu