Lucid Virtu Z68- My Benchamrks

[nemipc]

Hi Everyone,

I recently upgraded to a Z68 system and after reading the reviews (of Lucid/Virtu) I thought I would share my own benchmarks as there seems to be a lot of conflicting info out there and the situation is changing fast. I hope this helps.

system
CPU: 2500k i5, stepping 7, revision D2 ($180)
MB: Gigabyte Z68X-UD3H-B3 ($105) (microcenter bundle deal + rebate)
RAM: 2x 4GB XMP-1600 9-9-9-24-33
GPU: Intel HD 2000 + Nvidia GTX 460, 768MB
HD: Agility (v1) 120GB
OS: Windows 7 Pro 64bit
Montor: 2x 1080p

Notes:
For all tests the Nvidia graphics is set as primary GPU.
If Lucid is activated it is in Discrete mode.
My 2x1080p displays are connected to the Nvidia card at all times.

There was a clean OS install and virtu was left uninstalled.

GPU with lucid missing/installed "off" / installed "on":

the flowing benchmarks are with crysis warhead "FBWH Bench Tool" set to DirectsX10, "Ranked benchmark", Display 1080p, AA 0x, Loops 3, Map: Ambush, time of day 0

the benchmark runs cover me increasing GPU and CPU overclocks and installing new BIOS versions, finally followed by installing Lucid drivers for the very first time on the clean OS install and configuring the on/off state. This was over a period of a few days and some extra background apps and services were probably added.

GPU Core GPU Shader GPU Ram CPU RAM timing FPS Min max av notes notes 2
720 1440 1800 100.4x34 1600 9-9-9-24 14.8 54.3 45.5 stock (BIOS F3) before vvitu 1st install
810 1620 1900 101 x 42 1600 9-9-9-24 36.9 59.9 49.7 stock (BIOS F3) before vvitu 1st install
810 1620 1900 101 x 45 1600 9-9-9-24 36.7 61.2 49.8 stock (BIOS F3) before vvitu 1st install
810 1620 1900 101 x 45 1600 9-9-9-24 36.46 60.58 49.28 bios F4 before vvitu 1st install
810 1620 1900 101 x 45 1600 9-9-9-24 36.47 60.19 49.16 bios f5 before vvitu 1st install
810 1620 1900 101 x 45 1600 9-9-9-24 36.48 59.81 49.29 bios f5 vurtu installed but "OFF"
810 1620 1900 101 x 45 1600 9-9-9-24 36.8 59.93 49.28 bios f5 vurtu (ON)
810 1620 1900 101 x 45 1600 9-9-9-24 36.73 60.94 49.28 bios f5 vurtu (ON)

..as you can see Lucid has no effect on FPS in this configuration.



then i tried media encoding with Cyberlink MediaEspresso.

the interestign thing is that it can see the Nvidia and HD graphics at the same time and allow you to video decode on the nvidia and video encode on the intel together, which gave quite a fantastic speed boost. the following numbers are convertign a 1:50 min 1080p MT2S type file to a 720p Mt2S (H264) file (single pass) at 3MBit/s:

Time File Size (GB) CPU (%/100 x clock) Nvidia GPU % notes notes
0:39:15 2.8 0.352 33 hardware encoding ON, hardware Decoding ON, Lucid OFF = CUDA to MT2S
0:39 estiamte 4.077 12 hardware encoding OFF, hardware Decoding ON, Lucid OFF = CPU x86/CUDA to MT2S
0:47 estimate 3.624 1 hardware encoding OFF, hardware Decoding OFF, Lucid OFF = CPU x86 to MT2S
0:16 estimate 2.718 63 hardware encoding ON, hardware Decoding ON, Lucid ON = QuickSync + CUDA? to MT2S

Since the CPU energy efficieny was enabled the normalized clock speed is given by 4 core average utilization (%) x CPU frequency.

To put the above into perspective Handbrake (0.9.5) takes ~40 minutes with same file and settings with 1 pass and ~85 minutes on 2pass.

HOWEVER on fast moving scenes the handbrake 2 pass at 3mibit (and even at 2.3Mbit) was visually far superior to the above files.

Hope that helps!

 

[ViRGE]

The primary reason for the conflicting information is that it depends on which mode is used. If you use i-Mode (the iGPU is the primary GPU), the Virtu has to copy over the frame buffer from the dGPU, which has a performance hit. If you use d-mode (the dGPU is the primary GPU) then there's no performance hit because the only thing the iGPU is being used for is QuickSync encoding which doesn't have any messy frame buffer copying. What you're seeing is consistent with d-Mode results.

Anyhow, welcome and thanks for posting the results.