- Joined
- Jul 20, 2002
I was reading through an IBM whitepaper (from 2009) on tuning system X servers and came across a reference to MetaSDRAM DIMM's:
This technology’s primary benefit is to allow high capacity DIMM sizes without the exponentially higher prices always commanded
by the largest DIMM sizes.
MetaSDRAM accomplishes this by using a custom-designed chipset on each
DIMM module which makes multiple, inexpensive SDRAMs appear as a single,
large capacity SDRAM. Because this chipset acts as an onboard buffer and
appears as a single electrical load on the memory bus, higher frequencies can
often be achieved using MetaSDRAM-based DIMMs than would be possible
using standard DIMMs, which often have to switch to slower speeds as multiple
DIMMs are added per channel.
MetaSDRAM DIMMs offset the additional power gain imposed by the additional
SDRAMs by implementing intelligent power management techniques, allowing
two to four times the memory capacity to fit within a typical server’s power and
cooling capability.
Due to the inclusion of the additional MetaSDRAM chipset between the memory
bus and the DDR2 or DDR3 SDRAMs, there is some latency added to the
memory transactions. The impact of this latency, however, has been measured
across a number of commercial application environments to impact overall
application performance by just 1% to 2%. Because the technology has the
potential to double a server’s memory size, this small impact is well under the
potential gains achievable from the added memory capacity. However, if
MetaSDRAM is being used only to reduce cost in a server’s memory subsystem,
it is important to realize that performance can be slightly lower than standard
DIMMs, depending on configuration.
Has this tech already been made obsolete?
This technology’s primary benefit is to allow high capacity DIMM sizes without the exponentially higher prices always commanded
by the largest DIMM sizes.
MetaSDRAM accomplishes this by using a custom-designed chipset on each
DIMM module which makes multiple, inexpensive SDRAMs appear as a single,
large capacity SDRAM. Because this chipset acts as an onboard buffer and
appears as a single electrical load on the memory bus, higher frequencies can
often be achieved using MetaSDRAM-based DIMMs than would be possible
using standard DIMMs, which often have to switch to slower speeds as multiple
DIMMs are added per channel.
MetaSDRAM DIMMs offset the additional power gain imposed by the additional
SDRAMs by implementing intelligent power management techniques, allowing
two to four times the memory capacity to fit within a typical server’s power and
cooling capability.
Due to the inclusion of the additional MetaSDRAM chipset between the memory
bus and the DDR2 or DDR3 SDRAMs, there is some latency added to the
memory transactions. The impact of this latency, however, has been measured
across a number of commercial application environments to impact overall
application performance by just 1% to 2%. Because the technology has the
potential to double a server’s memory size, this small impact is well under the
potential gains achievable from the added memory capacity. However, if
MetaSDRAM is being used only to reduce cost in a server’s memory subsystem,
it is important to realize that performance can be slightly lower than standard
DIMMs, depending on configuration.
Has this tech already been made obsolete?
