GPU Systems

Image and signal processing by ELTEC for maximum performance

CPUs are Faster

GPUs that do more than just process images: they outperform the latest CPUs by factors 3 through 30 when it comes to speed. ELTEC delivers completely configured systems with digital interface or frame grabber as well as compatible GPU–with industry-adequate cooling solutions. ELTEC uses Fermi-based GeForce cards sourced from Nvidia. The system is fitted with a GTS 450 with a GF106 chip, 192 processor elements and a memory bandwidth of 57 GB/s.

Programming in High-Level Language (HLL)

The GPUs are CUDA-programmed. In conjunction with this solution, ELTEC offers its system clients a program framework that includes the delivery of the source code. Within this scope, every user has the option to set up his/her own algorithms. This ensures that the core know-how remains confidential and proprietary and the entry level of GPU programming is facilitated. ELTEC also has the capability of handling the highly critical optimization. In image processing applications, this encompasses the programming of the entire process from image recording to the delivery of the GPU data to the CPU.

Protect your Core Know-how

For customers who do not plan to implement their own CUDA programming, ELTEC offers to handle the complete implementation of existing algorithms. The support for OpenCL programming, which promises to provide an even better universal solution than CUDA is in the works. As a result, it will then also be possible to use AMD-GPUs and the Intel 2000-series.

We will Handle your Programming Work

ELTEC provides hardware, operating system and basic software support to its system clients. As far as all other system requirements are concerned, we keep you completely flexible: from special form factors affiliated with the housings to special cooling measures required because of unique heat development conditions to system programming for Linux and Windows–we are ready to handle it all.

ELTEC offers

  • Complete systems with GPU card
  • GPU: Nvidia GTS 450 with GF106, 192 processor elements and 57 GB/s memory bandwidth
  • Programming in CUDA, OpenCL in development
  • Integrated camera interface or frame grabber

Blob Labelling
Blob Labelling