[FFmpeg-user] Nvidia professional cards for li e transcoding

[Pavel Koshevoy]

On Thu, Aug 23, 2018, 11:06 Pedro Daniel Costa <portalnet2 at outlook.com.br>
wrote:

> This is the datasheet list
>
> For comparison on both models i am thinking
>
>
> 1st
>
> Quadro P6000
>
> SPECIFICATIONS
> GPU Memory 24 GB GDDR5X
> Memory Interface 384-bit
> Memory Bandwidth Up to 432 GB/s
> #################################
> #####   NVIDIA CUDA® CORES 3840      ####
> #################################
> System Interface PCI Express 3.0 x16
> Max Power Consumption 250 W
> Thermal Solution Active
> Form Factor 4.4”H x 10.5” L, Dual Slot,
> Full Height
> Display Connectors 4x DP 1.4 + DVI-D DL
> Max Simultaneous Displays 4 direct, 4 DP1.4 MultiStream
> Max DP 1.4 Resolution 7680 x 4320 @ 30 Hz
> Max DVI-D DL Resolution 2560 x 1600 @ 60 Hz
> Graphics APIs Shader Model 5.1,
> OpenGL 4.54
> ,
> DirectX 12.05
> ,
> Vulkan 1.04
> Compute APIs CUDA, DirectComput
>
>
>
>
>
>
> 2nd choise  Tesla K80
>
>
> TECHNICAL SPECIFICATIONS
> Tesla K801
> Peak double-precision floating point performance (board)  1.87 Tflops
> Peak single-precision floating point performance (board) 5.6 Tflops
> GPU 1 x GK110B 2 x GK210
> #############################################################
> #######################    CUDA cores 4,992    ######################
> #############################################################
>
> Memory size per board (GDDR5) 24 GB
> Memory bandwidth for board (ECC off)2 480 Gbytes/sec
> Architecture features SMX, Dynamic Parallelism, Hyper-Q
> System Servers and workstations Servers
>
>
>
>
>
>
> 3rd choice Tesla V100
>
> GPU Architecture NVIDIA Volta
> NVIDIA Tensor
> Cores 640
> #################################
> #######   NVIDIA CUDA® Cores 5,120    ###
> #################################
> Double-Precision
> Performance 7 TFLOPS 7.8 TFLOPS
> Single-Precision
> Performance 14 TFLOPS 15.7 TFLOPS
> Tensor
> Performance 112 TFLOPS 125 TFLOPS
> GPU Memory 16 GB HBM2
> Memory
> Bandwidth 900 GB/sec
> ECC Yes
> Interconnect
> Bandwidth 32 GB/sec 300 GB/sec
> System Interface PCIe Gen3 NVIDIA NVLink
> Form Factor PCIe Full
> Height/Length SXM2
> Max Power
> Comsumption 250 W 300 W
> Thermal Solution Passive
> Compute APIs CUDA, DirectCompute,
> OpenCL™, OpenACC
>
>
>
>
> -----Mensagem original-----
> De: ffmpeg-user [mailto:ffmpeg-user-bounces at ffmpeg.org] Em nome de Dennis
> Mungai
> Enviada em: quinta-feira, 23 de agosto de 2018 13:27
> Para: FFmpeg user questions
> Assunto: Re: [FFmpeg-user] Best Nvidia professional cards for li e
> transcoding
>
> Hello there,
>
> Does your budget allow for a newer line of NVIDIA GPUs, such as the
> Quadros based on Pascal?
>
> If so, get the Quadro P6000.
>
> Plenty of VRAM, + all the NVENC encoder features you may need, such as
> HEVC high-depth encoding and vastly better encoder performance overall.
>
> Refer to this:
> https://developer.nvidia.com/video-encode-decode-gpu-support-matrix
>
> If budget is not an issue, go for the jugular with the Tesla P100 (if you
> don't need HEVC 8k encoder support) OR the Tesla P40 (if you want/need all
> the features).
>
> I cannot speak for the Volta (GV100) line of GPUs as I'm yet to evaluate
> them in production.
>
> On 23 August 2018 at 19:02, Pedro Daniel Costa <portalnet2 at outlook.com.br>
> wrote:
>
> > Hi guys i am looking for best Nvidia cuda professional card for live
> > transcoding 100Channels, 50Channel HD mpeg4 aac 1980x1080p, and
> > 50Channel sd mpeg2 576x480.
> >
> > I am thinking TESLA K80, 4992GPU cuda cores, is there a more powerfull
> > card?
>
>

I would confirm that K80 is indeed newer.  I think K stands for Kepler,  M
for Maxwell, P for Pascal, and V for Volta.  Volta is the newest on that
list, and Kepler is oldest.  Kepler's NVENC doesn't support HEVC encoding,
I'd avoid it just for that.

Pavel.