Ukratko o tome kako gpu pomaze i rasterecuje cpu na modernim masinama ... moze pomoci i u radu kod starijih kompova ako se barem primeni nova grafa koja ima ove mogucnosti .. generalno od Rideon cini mi se serije HD7000 pa na dalje ... za Nvidiju ili Intel ne zna tacno od kojih modela..
5 reasons to offload your CPU with a GPU
Posted on
January 15, 2014 by
Marc Burkels
In my role as a Product Manager at Leaseweb, I constantly interact with our customers to find out how we can improve our products to fit their needs better. Quite often, I receive requests from customers for faster processors or even GPUs to accelerate their hosting platform.
In an effort to expand our product portfolio and meet our customers’ demands better, Leaseweb now offers selected
high-performance GPU servers. Before I explain the specifications, here’s a look at the top 5 ways you can benefit from a GPU.
What gives GPUs the extra edge over CPUs?
3D processing: GPUs were designed for 3D rendering. When assigned to a CPU-only application, the processing is slow due to linear request handling. Inserting a GPU into your server boosts the performance multiple times as they simultaneously process and compute large blocks of data. With the repetitive compute tasks offloaded from the CPU, it is free to process sequential tasks.
Accelerating Speed: Yes, by now we all know GPUs accelerate speed, but what makes them work faster? Composed of multiple cores, a GPU is built to simultaneously handle hundreds of threads–accelerating the application speed by tenfold over a CPU-only application. A CPU uses cache to reduce memory access latency, costing a lot of die-space. A GPU amplifies its bandwidth with cache memory. Where a CPU would wait for RAM to become available to process a thread, a GPU will switch to another thread that is ready for processing, thus reducing the latency and delivering faster results.
Number cruncher: When it comes to number crunching and graphics processing (involving millions of calculations per second), a GPU can make a high-end CPU look like a Commodore 64! This is because of the high amount of cores that GPUs have–high-end graphics cards have up to 2880 cores. A CPU supports 1-2 threads per core. In contrast, the multiprocessor of a NVIDIA CUDA core can execute an astonishing 1024 threads. This is also one of the reasons that mining cryptocurrencies (Bitcoin, Litecoin, etc.) deliver faster results when a GPU is used instead of a CPU. Although ASICS chips now even outsmart a GPU when it comes to mining coins.
Big data analytics: To make better real-time business decisions, GPUs are increasingly being used for big data analytics. Shazam, with a database of over 27 million tracks, uses GPU’s to identify a song from a snippet of track captured by its mobile users. The use of GPUs at Salesforce.com help companies such as Dell, Cisco, and Gatorade analyze and monitor over 500 million tweets daily. Real-time insights are delivered 10 minutes faster compared to a CPU-based system.
VDI environment: GPU hardware acceleration can be shared between virtual desktops–up to 32 users can share a graphics board. NVIDIA GRID is a powerful tool for providing superior graphics performance when sharing a GPU among multiple users. The optimized multi-GPU design with sufficient memory and low-latency remote display maximizes user density for applications that are graphics intensive.
GPU – A better fit for various industries
GPUs are traditionally used to process complex algorithms and massive data set for engineering and computer science applications. More and more companies are exploring various other uses of GPUs–audio search, image recognition, and big data analytics are good examples.
We use NVIDIA Quadro 4000 and 6000 in our
Leaseweb GPU servers for customers in the space of data mining and numerical analysis; heavy content producers such as advertising agencies; and web design agencies developing interactive applications, games, and 3D content.
https://blog.leaseweb.com/2014/01/15/5-reasons-offload-cpu-gpu/
