Artificial Intelligence Computing Leadership from NVIDIA
Xendesktop Dual Screen bad performence
Hi! Have a strange issue when running with two screens inside a session. The experience is not good, but when i power off the second screen, it works much better. Is there some setting inside the VDA that i missed or what can i check. Xendesktop 7.6, Xenserver 6.5 with K1 card(k120q) Windows 7 Sp1 VDA. Xenserver vgpu version 346.68 Tried with both driver 348.07 and latest on VDA.
Hi!

Have a strange issue when running with two screens inside a session.
The experience is not good, but when i power off the second screen, it works much better.
Is there some setting inside the VDA that i missed or what can i check.

Xendesktop 7.6, Xenserver 6.5 with K1 card(k120q) Windows 7 Sp1 VDA.
Xenserver vgpu version 346.68
Tried with both driver 348.07 and latest on VDA.

#1
Posted 09/03/2015 02:44 PM   
The vGPU profile will drive up to 2 displays at 2560x1600 in the VM, however you have to then encode and transmit that down to the client device, where is then decoded and displayed. Does the VM have sufficient resource to encode dual screens in HDX? What's the load on ctxgfx.exe in the VM, how many CPU's, what's the CPU clock speed? And on the client side, the same thing Does it have enough power to decode the HDX stream for 2 displays?
The vGPU profile will drive up to 2 displays at 2560x1600 in the VM, however you have to then encode and transmit that down to the client device, where is then decoded and displayed.


Does the VM have sufficient resource to encode dual screens in HDX?

What's the load on ctxgfx.exe in the VM, how many CPU's, what's the CPU clock speed?

And on the client side, the same thing

Does it have enough power to decode the HDX stream for 2 displays?

Jason Southern, Regional Lead for ProVis Sales - EMEA: NVIDIA Ltd.

#2
Posted 09/04/2015 10:53 AM   
[quote="Jason_Southern"] The vGPU profile will drive up to 2 displays at 2560x1600 in the VM, however you have to then encode and transmit that down to the client device, where is then decoded and displayed. Does the VM have sufficient resource to encode dual screens in HDX? What's the load on ctxgfx.exe in the VM, how many CPU's, what's the CPU clock speed? And on the client side, the same thing Does it have enough power to decode the HDX stream for 2 displays?[/quote] Hi! Thanks for answering! The VM runns on a dual 2.30Ghz Xeon E5 processor with 4GB ram. Thin client is a HP t520, AMD 1.2Ghz, 4GB ram. When looking at the Ctxgfx.exe, at CPU-Total in the resource manager its betweeen 50-60% utilzitation. Is the client spec enough? / Johan
Jason_Southern said:
The vGPU profile will drive up to 2 displays at 2560x1600 in the VM, however you have to then encode and transmit that down to the client device, where is then decoded and displayed.


Does the VM have sufficient resource to encode dual screens in HDX?

What's the load on ctxgfx.exe in the VM, how many CPU's, what's the CPU clock speed?

And on the client side, the same thing

Does it have enough power to decode the HDX stream for 2 displays?


Hi!

Thanks for answering!

The VM runns on a dual 2.30Ghz Xeon E5 processor with 4GB ram.
Thin client is a HP t520, AMD 1.2Ghz, 4GB ram.
When looking at the Ctxgfx.exe, at CPU-Total in the resource manager its betweeen 50-60% utilzitation.
Is the client spec enough?

/ Johan

#3
Posted 09/28/2015 12:58 PM   
CPU's are a little slow. Recommendation for 3D applications is at least 2.8Ghz and 4 vCPU's in the VM. 2 vCPU's will be a bottlenech as the encoders (Citrix and VMware) wil each consume a full CPU core to deliver good frame rates.
CPU's are a little slow.

Recommendation for 3D applications is at least 2.8Ghz and 4 vCPU's in the VM. 2 vCPU's will be a bottlenech as the encoders (Citrix and VMware) wil each consume a full CPU core to deliver good frame rates.

Jason Southern, Regional Lead for ProVis Sales - EMEA: NVIDIA Ltd.

#4
Posted 09/29/2015 08:15 PM   
Hi Johan, The T520 is not HDX 3D Pro certified (only HDX). You need the T620 or T820 from the HP series. https://citrixready.citrix.com/category-results.html?category=c1-thin-clients/g-hdx-level/f-hdx-3d-pro This is easily verified first however by connecting from a standard desktop device or your laptop - still the same issues? As for the 2.3Ghz I assume you have the v3 gen modell (2680v3 I think it was) that has turbo up to 3+ Ghz, be advised that turbo is not very reliant under heavy VDI load and with a hypervisor in-between to add, so you might get good POC results but run into problems in production later on - depending on the load (concidering both applications and number of VDIs/host) you are planning ofc. Unfortunately it is still marketed by lots of OEM sales people as a good fit. The 2687W is often a better option with higher base clock frequency, just make sure that you use a configuration that is certified to handle the power and thermal from housing both that and dual GRID cards. Good luck!
Hi Johan,

The T520 is not HDX 3D Pro certified (only HDX). You need the T620 or T820 from the HP series.

https://citrixready.citrix.com/category-results.html?category=c1-thin-clients/g-hdx-level/f-hdx-3d-pro

This is easily verified first however by connecting from a standard desktop device or your laptop - still the same issues?
As for the 2.3Ghz I assume you have the v3 gen modell (2680v3 I think it was) that has turbo up to 3+ Ghz, be advised that turbo is not very reliant under heavy VDI load and with a hypervisor in-between to add, so you might get good POC results but run into problems in production later on - depending on the load (concidering both applications and number of VDIs/host) you are planning ofc. Unfortunately it is still marketed by lots of OEM sales people as a good fit. The 2687W is often a better option with higher base clock frequency, just make sure that you use a configuration that is certified to handle the power and thermal from housing both that and dual GRID cards.

Good luck!

#5
Posted 10/01/2015 06:29 AM   
Scroll To Top

Add Reply