vGPU Rating for XenServer and XenDesktop on NVIDIA GRID - using Unigine Tropics Benchmark

10:42 AM
vGPU Rating for XenServer and XenDesktop on NVIDIA GRID - using Unigine Tropics Benchmark -

Benchmarking vGPU .... Seeing is believing, a cautionary tale frame rates!

  • vGPU is the newest addition to the Citrix portfolio of technologies designed to deliver rich applications 3-D on a variety of devices and remotely. This blog is about the vGPU benchmarking for CAD applications and a broader overview of the technology that I recommend you read my colleague Tim's blog announcing the technology. vGPU is a technology developed jointly with the leading NVIDIA GPUs, NVIDIA other information can be found here, available for Citrix XenServer and XenDesktop.

For severe CAD or GPU intensive 3-D graphics application, I came to the conclusion that a set of reference as Unigine all useful for evaluating GPU sharing technologies such and that vGPU vSGA, need not understood and used in conjunction with other landmarks. Unigine Tropics is not a particularly intensive application GPU and as such probably best suited for benchmarking for workers / VDI knowledge.

I blog about some of the factors to consider when benchmarking applications and rich graphics and CAD tools to help you do it, but with each landmark or evaluation, I seem to Learn more about the subtlety of technologies and factors customers need to understand to avoid being fooled by sellers.

Spot the Difference

the Unigine benchmarks are really busy! There's a lot going on and quite fast: flying pirate ships, oceans, clouds, palm trees and dragons. If you look at a single pass and the frame rate you might well miss subtleties. If you evaluate vSGA against vGPU or similar, it can be very insightful to put in place running side by side and play spot the difference, screenshots make it easier. It was only after chatting with the guys laboratory of our solutions and some OpenGL experts I enjoyed the tours these benchmarks can play.

Consider this image of roughly the same part of the Tropics benchmark Unigine on the left is vGPU on XenServer and on the right technology from a competitor for the GPU sharing relying on their drivers.

Watch the shadows. vGPU on the left are shadows that contain subtle colors, this is known as a soft shadow where the shadows reflect the color of an object. On the right, however, the shadows are flat and without any dark shadow original color of the object. namely the reference code is following a code path and APIs for some completely different rendering

Apparently, there are a few likely reasons for what's happening :.

  1. The right hand solution is aware that it is becoming a limited resource and is passed to a simple rendering to compensate
  2. right solution is in default of simple paths code on more subtle features to raise the frame rate score given by the benchmark, making it less and you can get a higher frame-rates.
  3. soft shadow is a relatively new feature, it is possible the solution on the right and the seller simply have not implemented the latest OpenGL / DirectX API

is not an effect you will find prevented you do a lot of modeling CAD however most organizations want to use the best they can made for marketing brochures and capture screen above shows, it would be really disappointing and frustrating if you were the guy made trying to explain to your boss why your shades on your brochure just does not seem to avant-garde.

There were other effects that seemed to follow different paths of code I'm not sure I would have noticed especially on one demo screen in the busy city and conference-user certainly not without comparing side by side. Below, the quality of the ocean in vGPU (again on the left) just looks much better and more realistic

There are a lot of points landmark that does not have a user interface or users to run automated marks on the registration of the scroll speed. This particular survey was a cautionary tale not just count the frame rate, but looking at what is in the frames themselves. I saw some pretty synthetic benchmarks running the imprint of OpenGL / DirectX libraries for specific applications, but now I seriously doubt if I can trust the frames per second (fps) metrics they rely on that without display the content of these frameworks. I was rather annoyed to find suppliers drivers can fold completely different code paths this way. Seeing really believe in this case!

Further reflections on Unigine Tropics

This is a reference that seems to be heavy on CPU usage and more suited to users that may seriously PLM CAD VDI or users as such, I think to make a fair assessment of the scalability of a solution that you really need to reference both the CPU and GPU, I blogged about the tools you can use to do this.

If I was to evaluate the purchase of a particular solution but I would also play spot the difference and ask a provider to activate the subtle features to see if the frame rate is artificially inflated, but also whether a solution effectively supports the latest technologies used by these highly visual applications correctly.

Unigine Heaven is a little more intensive the use of the GPU, but also a game application as the imprint. Gunnar Berger Gartner detailed and YouTube'd vSGA some surveys over vGPU using Unigine Heaven, you can see the links on this page.

For CAD users, it is probably a good idea to investigate modeling oriented benchmarks (eg Redway3d or otherwise) that reflect the use of a geometric model and core. game marks generally do not reflect the workload HLR (Hidden Line rendering), boolean intersections or B common geometry parameterization in most CAD software such as Dassault Solidworks, Ansys Workbench or Siemens PLM SolidEdge.

Thanks

My last glimpse GPU calibration was provided by a large number of fruitful discussions with and technical education of our laboratory solutions and technical marketing team (people very tolerant - eager to leave a CAD blogger drag annoy them) who blogged about the different GPU sharing solutions available for XenDesktop on both XenServer and vSphere, and also on how you can automate these tests (make sure just watching the frame content too!). The images in this blog are the results of some of their tests. I expect to see some articles very interesting and reference data and much more important published by them soon, I just touched on a number of things such as CAD engineer I thought were interesting. Do watch out for shadows anyway!

Previous
Next Post »
0 Komentar