IBM researchers say they have set a speed record for a type of computer memory that promises a fundamental performance rise in a coming generation of microprocessors with multiple computing engines.
The announcement, to be made at a conference on Wednesday, sets up a potential confrontation between IBM and Intel over the design of microprocessors that will begin to be available commercially next year.
While IBM now appears to be planning to integrate ultrafast memory directly into its processors, Intel has been hinting that it will instead stack memory chips on top of its processors to achieve similar performance.
Both companies are struggling with the challenge of quickly moving vast amounts of data inside processors that increasingly have multiple processing engines, or cores, to achieve faster performance.
The IBM achievement will be described in a paper presented on the third day of the International Solid State Circuits Conference, a gathering used by chip and computer companies to highlight technology advances.
On Monday, Intel demonstrated a Teraflop chip that integrated 80 separate processing engines.
The IBM researchers said they had been able to reduce memory cycle times to less than 2 nanoseconds, roughly 10 times the performance of off-the-shelf dynamic random access memory, or DRAM, used in PCs.
The IBM memory, called embedded dynamic random access memory, or eDRAM, can be used as a sort of simple scratchpad or temporary storage unit that speeds operations by keeping data inside a microprocessor for reuse while it is being processed.
Cache memories like these vastly hasten computation by reducing processor waiting time to perform new calculations.
This is huge, said David Lammers, director of Wesrch.com, a portal for technical professionals, who was briefed by IBM in advance of the papers presentation.
My question is, Does Intel have the answer to respond to IBMs eDRAM? He said he had spoken to Intel executives and that they had been skeptical about the technology, which in the past has been more expensive than other embedded memory.
SOURCE : THE TIMES OF INDIA
The announcement, to be made at a conference on Wednesday, sets up a potential confrontation between IBM and Intel over the design of microprocessors that will begin to be available commercially next year.
While IBM now appears to be planning to integrate ultrafast memory directly into its processors, Intel has been hinting that it will instead stack memory chips on top of its processors to achieve similar performance.
Both companies are struggling with the challenge of quickly moving vast amounts of data inside processors that increasingly have multiple processing engines, or cores, to achieve faster performance.
The IBM achievement will be described in a paper presented on the third day of the International Solid State Circuits Conference, a gathering used by chip and computer companies to highlight technology advances.
On Monday, Intel demonstrated a Teraflop chip that integrated 80 separate processing engines.
The IBM researchers said they had been able to reduce memory cycle times to less than 2 nanoseconds, roughly 10 times the performance of off-the-shelf dynamic random access memory, or DRAM, used in PCs.
The IBM memory, called embedded dynamic random access memory, or eDRAM, can be used as a sort of simple scratchpad or temporary storage unit that speeds operations by keeping data inside a microprocessor for reuse while it is being processed.
Cache memories like these vastly hasten computation by reducing processor waiting time to perform new calculations.
This is huge, said David Lammers, director of Wesrch.com, a portal for technical professionals, who was briefed by IBM in advance of the papers presentation.
My question is, Does Intel have the answer to respond to IBMs eDRAM? He said he had spoken to Intel executives and that they had been skeptical about the technology, which in the past has been more expensive than other embedded memory.
SOURCE : THE TIMES OF INDIA
0 comments :
Post a Comment