NVIDIA
GRID K1/K2 Comparison
GRID K1 and K2 are targeted at quite different users/use cases. While GRID K2 focuses on ultimate graphics performance for the most demanding users runnging 2D&3D professional applications, GRID K1 provides the best user scalability at the entry level. GRID K2 is equipped with two high-end Kepler GPUs with each 1526 CUDA cores and 4GB FB per GPU. The performance is comparable to two Quadro K5000s. In contrast, GRID K1 has 4 entry-level Kepler GPUs with each 192 CUDA cores, but a combined 16GB of FB per board. In combination with our GRID vGPU technology this allows to provide a virtual GPU to up to 32 users per board. Typical servers contain 2 GRID K1 boards which equates to up to 64 users per server. In summary: GRID K1 - ultimate user scalability GRID K2 - ultimate performance
GRID K1 and K2 are targeted at quite different users/use cases. While GRID K2 focuses on ultimate graphics performance for the most demanding users runnging 2D&3D professional applications, GRID K1 provides the best user scalability at the entry level.

GRID K2 is equipped with two high-end Kepler GPUs with each 1526 CUDA cores and 4GB FB per GPU. The performance is comparable to two Quadro K5000s.

In contrast, GRID K1 has 4 entry-level Kepler GPUs with each 192 CUDA cores, but a combined 16GB of FB per board. In combination with our GRID vGPU technology this allows to provide a virtual GPU to up to 32 users per board. Typical servers contain 2 GRID K1 boards which equates to up to 64 users per server.

In summary:

GRID K1 - ultimate user scalability
GRID K2 - ultimate performance

Milan Diebel, Product Manager NVIDIA GRID

#1
Posted 03/21/2014 08:23 PM   
Hey Milan, Thanks for the above explanation but I need clarification on how this applies to my specific scenario. I have Session Host RDS server with multiple users; will I have to use K1 or K2? I am guessing based on you explanation above K1 is ideal for multiple VMs and not multiple users as stated. Multiple VMs would mean that I can virtualize the GPU but my scenario is that of multiple users connecting to different RDP sessions (not VMs) on the same server. Do you think K2 is the ideal product for my scenario since it would mean that all workload would be on the Host Server and not virtualized across VMs.
Hey Milan,

Thanks for the above explanation but I need clarification on how this applies to my specific scenario.
I have Session Host RDS server with multiple users; will I have to use K1 or K2?
I am guessing based on you explanation above K1 is ideal for multiple VMs and not multiple users as stated.
Multiple VMs would mean that I can virtualize the GPU but my scenario is that of multiple users connecting to different RDP sessions (not VMs) on the same server.
Do you think K2 is the ideal product for my scenario since it would mean that all workload would be on the Host Server and not virtualized across VMs.

#2
Posted 08/23/2018 09:16 PM   
Both boards are EOL for almost 2 years now. Really wondering why you bring up this topic. For your use case you should look at Tesla M10. Regards Simon
Both boards are EOL for almost 2 years now. Really wondering why you bring up this topic. For your use case you should look at Tesla M10.

Regards

Simon

#3
Posted 08/24/2018 07:56 AM   
Scroll To Top

Add Reply