TurboCache

TurboCache was an NVIDIA graphics-memory technology designed to reduce the cost of entry-level discrete graphics cards by combining a relatively small amount of memory installed on the graphics card with additional memory allocated from the computer's main system RAM. The system memory was accessed through the high-bandwidth PCI Express interface, allowing a compatible graphics processor to use more memory for graphics workloads than was physically present on the board itself. The technology was introduced with the GeForce 6200 family and was positioned as a way to improve the cost-to-performance balance of low-cost PCI Express graphics products. By reducing the amount of dedicated memory required on the card, board manufacturers could produce less expensive designs while still advertising a larger aggregate graphics-memory capacity. NVIDIA presented TurboCache-equipped GeForce 6200 products as substantially faster than Intel's GMA 900 integrated graphics in the relevant comparison, although actual performance depended on the application, the amount of local video memory, system memory capacity, and platform configuration. TurboCache's use of shared memory created an important distinction between advertised total memory and dedicated local memory. Early product marketing and retail packaging could emphasize the combined capacity of onboard graphics memory and system RAM, while the physical amount of memory installed on a particular card could vary substantially. Early cards associated with the technology could contain between 16 and 128 MiB of local memory, with the remainder of the advertised capacity supplied by system memory. This caused confusion among buyers who interpreted the headline capacity as the amount of memory physically installed on the graphics card. NVIDIA subsequently changed its marketing and packaging practice to identify the local video-memory amount rather than presenting only the combined figure. TurboCache could be useful in applications that needed additional memory for textures and other graphics data. It also had significant limitations. Accessing system RAM through the graphics bus was slower than using dedicated memory on the card, and cards using the technology generally performed below otherwise comparable cards with a full complement of discrete video memory in three-dimensional benchmarks. Games that already consumed a large share of system RAM could experience particularly severe performance degradation, especially on computers with 1 GB of RAM or less. Some users sought to disable TurboCache, but NVIDIA's official drivers did not provide a standard control for doing so. Workarounds involving Windows XP registry settings or X.Org configuration files were reported for some systems. The technology became less relevant as graphics software and operating systems developed more general mechanisms for virtualizing and managing GPU memory. With Direct3D 10 and the associated Windows graphics-driver architecture, the system could move graphics data between video memory and system RAM when local memory became constrained. That reduced the need for a separately branded mechanism whose principal purpose was to extend a card's usable memory pool. TurboCache is therefore best understood as a transitional graphics-memory design associated with early PCI Express-era, entry-level NVIDIA cards rather than as a current NVIDIA product line.

TurboCache emerged during the transition from older graphics interfaces to PCI Express as NVIDIA sought to improve the economics of entry-level discrete graphics hardware. The basic design used a small amount of memory located on the graphics card and supplemented it with allocations from the computer's main memory. PCI Express provided the data path through which the graphics processor could access that additional memory. This made it possible to build a graphics card with less dedicated memory while retaining access to a larger aggregate memory pool for workloads such as texture storage. The first widely associated implementation was the GeForce 6200 with TurboCache. NVIDIA promoted the product as a way to offer graphics performance above contemporary integrated graphics while keeping hardware costs down. The approach was not equivalent to installing the advertised total capacity as dedicated video memory: a portion was local to the board and the rest remained ordinary system RAM. The distinction was significant because local graphics memory offered lower latency and more predictable bandwidth than memory reached through the system bus. Early marketing of TurboCache products led to buyer uncertainty. Product descriptions could state a combined memory total without making the amount of onboard memory sufficiently prominent. Depending on the board design, local memory could range from 16 to 128 MiB. Consumers therefore could purchase cards carrying similar TurboCache branding but with materially different amounts of physical video memory. NVIDIA eventually adjusted its packaging and marketing to list the onboard video-memory figure rather than relying solely on the combined total. Performance varied by workload. TurboCache could help applications that needed more graphics memory than the card's local allocation, particularly when additional space for textures improved rendering quality. However, the extra memory was slower to access than dedicated VRAM. Cards using the system-memory approach were generally slower in three-dimensional benchmarks than similar cards equipped with more discrete memory. The impact was even more pronounced in games that consumed substantial amounts of the host computer's RAM. On systems with 1 GB of RAM or less, those workloads could compete directly with TurboCache for resources and produce major slowdowns. The technology also generated interest among technically advanced users who wanted to turn it off. NVIDIA's official drivers did not offer a normal user-facing switch, but configuration workarounds were reported for Windows XP and X.Org-based Linux environments. These workarounds reflected the trade-off at the center of TurboCache: its ability to enlarge the available graphics-memory pool depended on taking capacity and bandwidth from the rest of the system. TurboCache's distinct role diminished as later graphics APIs and operating-system driver models adopted broader virtual-memory management for GPUs. With Direct3D 10-era graphics support, the Windows graphics subsystem could move data between local video memory and system RAM when necessary. That made the underlying concept more integrated into the platform and reduced the need for a separate, prominent technology label. TurboCache consequently became associated with an earlier generation of entry-level PCI Express graphics products and is no longer a current NVIDIA offering.

A cost-reduction and capacity-expansion technology for entry-level discrete graphics cards, using system RAM alongside onboard video memory.

• Discrete graphics card

  GeForce 6200 with TurboCache

  The GeForce 6200 with TurboCache was the principal product associated with the technology. It used a combination of onboard graphics memory and system RAM accessed over PCI Express, allowing NVIDIA and its board partners to offer an entry-level discrete graphics solution with a larger aggregate memory capacity than the physical memory installed on the card. Its positioning emphasized affordability and performance above contemporary integrated graphics, while its results depended heavily on local-memory capacity, host RAM, and the application workload.

• Shift from combined-memory marketing to local-memory disclosureStrategy

  Background

  Early TurboCache product descriptions could lead buyers to interpret the combined local-and-system memory figure as the amount of dedicated memory physically installed on the graphics card.

  Action taken

  NVIDIA subsequently revised its marketing and packaging practice to identify the card's onboard video-memory amount rather than presenting only the aggregate memory figure.

  Aftermath

  The change addressed product-capacity ambiguity, although the underlying performance distinction between dedicated video memory and system RAM remained.

  Source

1. 1TurboCacheen.wikipedia.org