[List of Test Results Collected with EUC Score Experiments]
Dizzion Frame
Analyzing Dizzion Frame Desktop as a Service performance on Azure, Amazon Web Services (AWS) and Google Cloud Platform (GCP). Remoting protocols are Frame Remoting Protocol (FRP) and Microsoft Remote Desktop Protocol (RDP). The test results presented here are a subset of the complete data sets collected during multiple test sequences.
- Dizzion Frame DaaS on Azure VMs with GPU
- Dizzion Frame DaaS on AWS VMs with GPU
- Dizzion Frame DaaS on GCP VMs with GPU
It is important to note that with default settings the Frame Remoting Protocol (FRP) delivers 60 frames per second while Microsoft RDP delivers 30 frames per second to the endpoint device. This may result in higher network consumption on the FRP side.
If you want to learn how the EUC Score results were produced then check out EUC Score Test Methodology and EUC Score Toolset.
Dizzion Frame DaaS on Azure VMs with GPU
Benny Tritsch
Ruben SpruijtTest Goal
Compare an RDP session on a physical Windows workstation to Frame Remoting Protocol (FRP) connections to Windows VMs with GPU on Azure.
Selected Sync Player Clips
| Sync Player Clip | View and Type | Observation | |
|---|---|---|---|
 |
SL1-MSEdgeAquariumWebGL | SxS View - GPU | Azure NC4asT4-v3 VM with NVIDIA T4 GPU: 650 FPS vs 800 FPS and smoother animation on the Azure side (40% CPU load). |
|
SL1-TessMarkOpenGL | SxS View - GPU | Azure NC4asT8-v3 VM with NVIDIA T4 GPU: Smoother animation and better rendering quality on the Azure side. Less overall CPU load (20%) when compared to the NC4asT4 v3 VM with approx. 40% load. |
|
SL1-FurMarkOpenGL | SxS View - GPU | Azure NV12adsA10-v5 VM with NVIDIA A10-8Q vGPU: Slightly better user experience on the Azure side, despite the partitioned GPU. |
|
SL1-WMPlayer1080pMP4 | SxS View - GPU | Azure NV4as-v4 VM with 1/8 AMD Radeon MI25 GPU: Without a video decoder in the GPU, video playback falls apart. |
|
SL3-FractalsPythagoras | SxS View - GPU | Azure NV4as-v4 VM with 1/8 AMD Radeon MI25 GPU: The CPU/GPU combo is not powerful enough for the fractals score. |
Summary of Test Observations
On Azure NC4asT4-v3, NC4asT8-v3 and NV12adsA10-v5 VM types, the overall performance is better than on the physical lab machine Lancelot. But when the combination of CPU and GPU is less powerful, the physical lab machine Lancelot performs better.
Test Setup Details
RDP Connection to Lancelot
FRP Connection to Azure VM
System Under Test: Physical Lab Machine Lancelot, Windows 11, Intel i7-11700K 16Threads @ 3.6GHz, 64GB RAM, Crucial MX500 SSD 1TB, NVIDIA Quadro M5000, 8GB VRAM.
System Under Test: Azure West Europe, different VM types with NVIDIA and AMD GPUs, Frame Display Driver.
Connection: RDP-UDP from Remote Desktop Connection Client, 20ms round trip time.
Connection: FRP8 UDP from Chrome browser, 4ms latency (= 8ms round trip time).
Endpoint: NUC2, Intel NUC 8i7HNK with Windows 11, i7-8705G CPU @ 3.10GHz, 16GB RAM, 500GB Samsung SSD 850 EVO M.2, AMD Radeon RX Vega M GL, 4GB VRAM.
Endpoint: NUC2, Intel NUC 8i7HNK with Windows 11, i7-8705G CPU @ 3.10GHz, 16GB RAM, 500GB Samsung SSD 850 EVO M.2, AMD Radeon RX Vega M GL, 4GB VRAM.
Dizzion Frame DaaS on AWS VMs with GPU
Test Goal
Compare an RDP session on a physical Windows workstation to Frame Remoting Protocol (FRP) connections to Windows VMs with GPU on Amazon Workspaces.
Selected Sync Player Clips
| Sync Player Clip | View and Type | Observation | |
|---|---|---|---|
|
SL1-ChromeWaterWebGL | SxS View - GPU | AWS g4ad.xlarge VM with AMD Radeon Pro V520 MxGPU: The quality of the animation is much better on the AWS side. |
|
SL1-ChromeCarVisualizer | SxS View - GPU | AWS g4dn.2xlarge VM with NVIDIA T4 GPU: Car rotates significantly faster on the AWS side. |
|
SL1-FurMarkOpenGL | SxS View - GPU | AWS g4dn.2xlarge VM with NVIDIA T4 GPU: Very smooth rendering and animation of the fur donat on the AWS side. |
Summary of Test Observations
The overall performance of the AWS DaaS VM types is better than on the physical lab machine Lancelot. The CPU on Lancelot shows less load due to higher CPU thread (= core) number and higher clock rate.
Test Setup Details
RDP Connection to Lancelot
FRP Connection to AWS VM
System Under Test: Physical Lab Machine Lancelot, Windows 11, Intel i7-11700K 16Threads @ 3.6GHz, 64GB RAM, Crucial MX500 SSD 1TB, NVIDIA Quadro M5000, 8GB VRAM.
System Under Test: AWS eu-central-1, EC2 g4ad.xlarge with AMD Radeon Pro V520 MxGPU or EC2 g4dn.2xlarge VM with NVIDIA T4 GPU, Frame Display Driver.
Connection: RDP-UDP from Remote Desktop Connection Client, 20ms round trip time.
Connection: FRP8 UDP from Chrome browser, 4ms latency (= 8ms round trip time).
Endpoint: NUC2, Intel NUC 8i7HNK with Windows 11, i7-8705G CPU @ 3.10GHz, 16GB RAM, 500GB Samsung SSD 850 EVO M.2, AMD Radeon RX Vega M GL, 4GB VRAM.
Endpoint: NUC2, Intel NUC 8i7HNK with Windows 11, i7-8705G CPU @ 3.10GHz, 16GB RAM, 500GB Samsung SSD 850 EVO M.2, AMD Radeon RX Vega M GL, 4GB VRAM.
Dizzion Frame DaaS on GCP with GPU
Test Goal
Compare an RDP session on a physical Windows workstation to Frame Remoting Protocol (FRP) connections to Windows VMs with GPU on Google Cloud Platform.
Selected Sync Player Clips
| Sync Player Clip | View and Type | Observation | |
|---|---|---|---|
|
SL1-MSEdgeAquariumWebGL | SxS View - GPU | GCP N1-Standard-4-GPU-T4-Windows VM with NVIDIA T4 GPU: Higher frame rate and smoother animation on the GCP side, despite the much less powerful CPU. |
Summary of Test Observations
The performance of the GCP DaaS VM types is better than on the physical lab machine Lancelot. The CPU on Lancelot shows less load due to higher CPU thread (= core) number and higher clock rate.
Test Setup Details
RDP Connection to Lancelot
FRP Connection to GCP VM
System Under Test: Physical Lab Machine Lancelot, Windows 11, Intel i7-11700K 16Threads @ 3.6GHz, 64GB RAM, Crucial MX500 SSD 1TB, NVIDIA Quadro M5000, 8GB VRAM.
System Under Test: GCP europe-west4, N1-Standard-4-GPU-T4-Windows VM with NVIDIA T4 GPU, Frame Display Driver.
Connection: RDP-UDP from Remote Desktop Connection Client, 20ms round trip time.
Connection: FRP8 UDP from Chrome browser, 4ms latency (= 8ms round trip time).
Endpoint: NUC2, Intel NUC 8i7HNK with Windows 11, i7-8705G CPU @ 3.10GHz, 16GB RAM, 500GB Samsung SSD 850 EVO M.2, AMD Radeon RX Vega M GL, 4GB VRAM.
Endpoint: NUC2, Intel NUC 8i7HNK with Windows 11, i7-8705G CPU @ 3.10GHz, 16GB RAM, 500GB Samsung SSD 850 EVO M.2, AMD Radeon RX Vega M GL, 4GB VRAM.