EUC Score Test Results

[List of Test Results Obtained in EUC Score Experiments]

EUC Score Test Results - Thin Clients

Testing different thin client devices with IGEL OS and an some cases comparing the results with the reference Windows client NUC2. The test results presented here are a subset of the complete datasets collected during multiple thin client test sequences.

If you want to learn how the EUC Score results were produced then check out EUC Score Test Methodology and EUC Score Toolset.

Thin Client and AVD versus Windows Client and Local Host

Benny Tritsch Markus Zehnle Florian Lauer Date of test: 26 January 2023

Testable question: Is there a difference when comparing a remote connection from a thin client to an AVD session with a remote connection from the Windows reference client NUC2 to the local reference host PC Lancelot?

Thin Client and AVD

Windows Client and Local Host

System Under Test: Azure West Europe, AVD NV6 VM, Windows 10 Enterprise for Virtual Desktops, Intel Xeon E5-2690 v3 6vCPUs @ 2.60GHz, 56GB RAM, Virtual HD ATA Device 340GB, NVIDIA M60 GPU (1/2 Card), 8GB VRAM.

System Under Test: Lancelot, Physical Lab Machine, Windows 11, Intel i7-11700K 16Threads @ 3.6GHz, 64GB RAM, Crucial MX500 SSD 1TB, NVIDIA Quadro M5000, 8GB VRAM.
 

Connection: AVD client on HP t640 with IGEL OS 11.8, RDP-UDP, 15ms round trip time.

Connection: RDC client on Windows 11, RDP-UDP, 20ms round trip time.

 

Sync Player Clip View and Type Observation
Sync Player SL1-ChromeWaterWebGL SxS View - GPU Almost similar user experience due to almost identical host hardware specs, except for the number of CPU cores and clock speed.
Sync Player SL1-ShadingDX11 SxS View - GPU Almost similar rendering frame rate due to comparable GPU types, even if number of CPU cores and clock rate is different.

Result: The HP t640 thin client with IGEL OS 11.8 connected to an Azure Virtual Desktop NV6 VM provides a user experience that is very similar to a remote desktop session from the Windows reference client NUC2 to the local reference host PC Lancelot.

Thin Client Hardware Comparisons

Benny Tritsch Markus Zehnle Florian Lauer Date of test: 26 January 2023

Testable question: What is the performance of different thin client types under IGEL OS 11.8 when connected to an Azure Virtual Desktop session?

HP t640

LG CL-600i-6N or IGEL UD3

System Under Test: Azure West Europe, AVD NV6 or NV4as v4 VM type.

System Under Test: Azure West Europe, AVD NV6 or NV4as v4 VM type.

Connection: AVD client on HP t640 with IGEL OS 11.8, RDP-UDP, 15ms round trip time.

Connection: AVD client on either LG CL-600i-6N or IGEL UD3 with IGEL OS 11.8, RDP-UDP, 15ms round trip time.

 

Sync Player Clip View and Type Observation
Sync Player SL1-TessMarkOpenGL SxS View - GPU HP t640 versus LG CL-600i-6N: Without client GPU, reduced user experience and disconnects on the LG CL-600i-6N device (right media tile).
Sync Player SL1-RollercoasterDX9 SxS View - GPU HP t640 versus IGEL UD3: Almost identical user experience due to almost identical client hardware specs.
 
Sync Player SL1-DominoOpenGL SxS View - GPU HP t640 versus IGEL UD3: When things go wrong on both side due to insufficient capacity of the NV4as v4 VM, including disconnect/reconnect situations.

Result: Thin clients without a GPU cannot keep up with thin clients with an embedded GPU under demanding graphics workloads. But there are also cases where bad performance is caused by insufficient AVD VM capacity.

Win365 CPC - Windows Client versus Thin Client

Benny Tritsch Markus Zehnle Florian Lauer Date of test: 26 January 2023

Testable question: Is there a difference in performance when connecting to a Windows 365 Cloud PC session from a Windows client or from a thin client?

AVD Client on NUC2 with Windows 11

AVD Client on HP t640 with IGEL OS

System Under Test: Azure West Europe, Windows 365 Business, Windows 11, Intel Xeon Platinum 8272CL 2vCPUs @ 2.6GHz, 8GB RAM, Microsoft Virtual Disk 128GB, no GPU, Remote Desktop Services SxS Network Stack.

System Under Test: Azure West Europe, Windows 365 Business, Windows 11, Intel Xeon Platinum 8272CL 2vCPUs @ 2.6GHz, 8GB RAM, Microsoft Virtual Disk 128GB, no GPU, Remote Desktop Services SxS Network Stack.

Connection: AVD client on NUC2 with Windows 11, SxS RDP UDP, 30ms round trip time.

Connection: AVD client on HP t640 with IGEL OS 11.8, SxS RDP UDP, 30ms round trip time.

 

Sync Player Clip View and Type Observation
Sync Player SL1-MSOPptSimple SxS View - GPU Windows 365 Cloud PC shows good enough performance for Task Worker activities, independently of the endpoint device.
Sync Player SL1-ChromeWaterWebGL SxS View - GPU Windows 365 Cloud PC is not suited for graphics-intensive applications, independently of the endpoint device.

Result: The client device doesn't make a difference as long as the thin client and the Windows client have similar hardware specs.

Thin Client over FRP versus Windows Client over RDP

Benny Tritsch Markus Zehnle Florian Lauer Date of test: 26 January 2023

Testable question: How good is the remote connection over Frame Remoting Protocol (FRP) from the Chromium browser of a thin client under IGEL OS 11.8?

Thin Client over FRP

NUC2 over RDP

System Under Test: Azure West Europe, NC4asT4 v3 VM, Windows 10 21H2, AMD EPYC 7V12 (Rome) 4vCPUs @ 2.4GHz, 28GB RAM, 256GB + 176GB P-SSD, NVIDIA T4, 16GB VRAM, Frame Display Driver.

System Under Test: Lancelot, Physical Lab Machine, Windows 11, Intel i7-11700K 16Threads @ 3.6GHz, 64GB RAM, Crucial MX500 SSD 1TB, NVIDIA Quadro M5000, 8GB VRAM.
 

Connection: HP t640, IGEL OS 11.8, Chromium browser, FRP8 UDP, 7ms latency.

Connection: RDC client on NUC2 with Windows 11, RDP-UDP, 20ms round trip time.

 

Sync Player Clip View and Type Observation
Sync Player SL1-FurMarkOpenGL SxS View - GPU The performance of a "Cloud Workstation" with NVIDIA T4 GPU on a thin client over FRP is better than a local remote desktop with NVIDIA M5000 GPU over RDP.
Sync Player SL1-MSEdgeAquariumWebGL SxS View - GPU The "Cloud Workstation" with NVIDIA T4 GPU on a thin client over FRP shows higher frame rate and better rendering performance when compared with a local remote desktop with NVIDIA M5000 GPU over RDP.

Result: The remote connection over FRP from the Chromium browser of the thin client outperforms the local reference setup with NUC2 client and Lancelot host. The major reason is the stronger GPU in the NC4asT4 v3 VM type used for the Frame session.

Frame on IGEL OS (Chromium)

Benny Tritsch Markus Zehnle Florian Lauer Date of test: 26 January 2023

Testable question: What is the the performance impact on an Azure VM connection over Frame FRP from a Chromium browser in IGEL OS 11.8 when the assignment of the client GPU didn't work?

Chromium with GPU Acceleration

Chromium without GPU Acceleration

System Under Test: Azure West Europe, NC4asT4 v3 VM, Windows 10 21H2, AMD EPYC 7V12 (Rome) 4vCPUs @ 2.4GHz, 28GB RAM, 256GB + 176GB P-SSD, NVIDIA T4, 16GB VRAM, Frame Display Driver.

System Under Test: Azure West Europe, NC4asT4 v3 VM, Windows 10 21H2, AMD EPYC 7V12 (Rome) 4vCPUs @ 2.4GHz, 28GB RAM, 256GB + 176GB P-SSD, NVIDIA T4, 16GB VRAM, Frame Display Driver.

Connection: HP t640, IGEL OS 11.8, Chromium browser with Hardware Video Acceleration, FRP8 UDP, 7ms latency.

Connection: HP t640, IGEL OS 11.8, Chromium browser without Hardware Video Acceleration, FRP8 UDP, 7ms latency.

 

Sync Player Clip View and Type Observation
Sync Player SL1-TessMarkOpenGL SxS View - GPU Without GPU acceleration on the thin client, performance sucks. This means that the client GPU acceleration really matters.
Sync Player SL1-ShadingDX11 SxS View - GPU Without GPU acceleration on the thin client, performance sucks and the session disconnects. This means that the client GPU acceleration really matters.

 

Use IGEL Setup for configuration disabled

Use IGEL Setup for configuration enabled

 

Result: The default settings in IGEL OS 11.8 overwrite the assignment of the client GPU to the Chromium browser. Without GPU, the client cannot decode the incoming remote desktop datastream without regular freeze frames. Disabled default settings (left) result in a significantly better user experience while the overall load on the backend is the same on both sides. Optimizing the thin client settings can mage a huge difference!