With graphics cards free to use system memory with less of a performance penalty, board manufacturers have been able to build competitive cards with only between 16MB and 64MB of memory on-board. Having less on-board memory allows these cards to be sold at extremely low prices, bringing DirectX 9-class PCI express offerings down to around $50.
ATI and NVIDIA's latest budget graphics cards have embraced using system memory over PCI Express, with ATI calling its approach HyperMemory, and NVIDIA branding its TurboCache. Join us as we explore the intricacies and performance of each company's approach, and crown a new budget PCI Express champion.
ATI's HyperMemory cards are based on the RV370 graphics chip, and members of the Radeon X300 family. With 325MHz core clock speeds and RV370's four full pixel pipelines, the HyperMemory cards actually pack a decent amount of pixel filling power for budget solutions. The RV370's R300 roots also give the cards DirectX 9-class capabilities, including support for Shader Model 2.0.
Radeon X300s are normally available with either a 64- or 128-bit memory bus, but the local memory interface for HyperMemory is limited to 64 bits. With local memory clocked at 300MHz (an effective 600MHz thanks to DDR's double data rate), HyperMemory cards should have 4.8GB/sec of local bandwidth. Combine that with 6.4GB/sec of system memory bandwidth, and HyperMemory cards boast over 11GB/sec of bandwidth overall, although there will invariably be a latency penalty associated with accessing system memory.
To minimize the impact of higher latency system memory access, HyperMemory cards intelligently organize data to keep high priority items in local memory. Data that doesn't make the cut for local memory is stored GART memory, which is non-paged system memory allocated to the graphics card. When GART memory becomes full, pageable system memory is dynamically allocated to store additional data. HyperMemory can shuffle data between local, GART, and pageable system memory as needed, and is capable of accessing both local and system memory in parallel. Data can also be assigned directly to system memory, and may never actually reside in the graphics card's local memory.
|Core clock (MHz)||Memory clock (MHz)||Memory bus width (bits)||On-board memory (MB)|
|Radeon X300 SE 128MB HyperMemory||325||600||64||32|
|Radeon X300 SE 256MB HyperMemory||325||600||64||128|
Unlike NVIDIA, which offers a handful of TurboCache configurations, ATI has only two HyperMemory cards. The first, and likely most popular, is the Radeon X300 SE 128MB HyperMemory, which features 32MB of memory on-board. ATI is also offering a 256MB HyperMemory card with 128MB of memory on-board, although it's unlikely that the card will have enough horsepower to make effective use of the extra memory. Currently, only the 128MB HyperMemory card is available on the market, with prices ranging between $49 and $55.
|The TR Podcast is live, so come ask us stuff!||0|
|AMD shows off DirectX 12 performance with new 3DMark benchmark||33|
|Intel and Micron sampling 3D NAND based on floating gates||16|
|Report: Microsoft to build an Intel-powered, non-Pro Surface tablet||47|
|Toshiba's 3D flash spreads 16GB over 48 layers||6|
|Cougar's 300M gaming mouse looks awfully familiar||14|
|Wednesday Night Shortbread||69|
|EVGA Hybrid liquid-cools the GeForce GTX 980||20|
|Rowhammer attack exploits shrinking process size in DRAM||40|