we need would make it hugely expensive –
there’s a reason why most systems make
do with around 16GB of RAM. Secondly, and
more importantly, system memory isn’t
able to retain its data when power is lost: it’s
volatile. As such, non-volatile storage such as
hard drives and SSDs is required to retain data
when you switch off your PC.
This is one area where Intel’s recently
developed Optane technology can help.
Optane is a storage medium that’s both
non-volatile and yet fast enough to not
need a faster intermediary. Data can be
loaded straight out of Optane storage and
written back again. However, Optane still
isn’t as fast as conventional memory, and it’s
expensive as a long-term storage medium,
so for most PC applications, the RAM/SSD
dichotomy is here for the foreseeable future.
There are also further layers to this
collecting of data into faster mediums. Once
data goes from memory into the CPU (or
other processing units), there are generally
several further layers of ever faster, ever
smaller memory known as caches. For mostCPUs, you get three levels of cache. Level
one (L1) cache is the fastest and smallest,
and it sits right inside individual processing
cores. Then we have L2 cache, which is also
generally dedicated to a single CPU core
but offers a larger capacity. Finally, there’s
L3 cache, which is much larger again and is
generally shared across several cores.Meanwhile, for GPUs, even main
system memory isn’t fast enough for the
amount of data it needs to process, so it
has its own dedicated video RAM (VRAM).
This memory intercepts any data relevant
to graphics processing and takes over the
job of providing that data for the GPU.
It’s this hierarchy of data access that allows
our computers to both store huge volumes
of data indefinitely and still provide lightning-quick data processing, such as the processing
required for rendering 3D graphics in games.
Crucially, you can also see from this
setup how the roll of memory is intrinsically
linked with the overall speed of a PC. If
your memory, or the internal caches of a
processing unit, aren’t large or fast enough,
they’ll hold back performance. However,equally we can see that buying ever faster
memory doesn’t make a CPU or GPU itself
faster – it simply eliminates a bottleneck.WHAT IS DRAM?
By and large, each of the types of memory,
from on-chip caches through system
memory, to long-term storage, all use
fundamentally different technologies for
storing data. For instance, caches tend toSRAM IS EXTREMELY QUICK, SO IT’S
THE OBVIOUS CHOICE FOR A MEMORY
TECHNOLOGY THAT SITS CLOSEST TO
THE PROCESSING UNITS OF A CPUWhether your memory has a heatspreader, RGB lighting or a bare PCB,
most memory modules look near-identical underneath their coversweneedwouldmakeit hugelyexpensive–
there’sa reasonwhymostsystemsmake
dowitharound16GBofRAM.Secondly,and
moreimportantly,systemmemoryisn’t
abletoretainitsdatawhenpoweris lost:it’s
volatile.Assuch,non-volatilestoragesuchas
harddrivesandSSDsis requiredtoretaindata
whenyouswitchoffyourPC.
Thisis oneareawhereIntel’srecently
developedOptanetechnologycanhelp.
Optaneis a storagemediumthat’sboth
non-volatileandyetfastenoughtonot
needa fasterintermediary.Datacanbe
loadedstraightoutofOptanestorageand
writtenbackagain.However,Optanestill
isn’tasfastasconventionalmemory,andit’s
expensiveasa long-termstoragemedium,
soformostPC applications, the RAM/SSD
dichotomy is here for the foreseeable future.
There are also further layers to this
collecting of data into faster mediums. Once
data goes from memory into the CPU (or
other processing units), there are generally
several further layers of ever faster, ever
smaller memory known as caches. For most
CPUs,yougetthreelevels of cache. Level
one (L1) cache is the fastest and smallest,
and it sits right inside individual processing
cores. Then we have L2 cache, which is also
generally dedicated to a single CPU core
but offers a larger capacity. Finally, there’s
L3 cache, which is much larger again and is
generallysharedacrossseveralcores.Meanwhile,forGPUs,evenmain
system memory isn’t fast enough for the
amount of data it needs to process, so it
has its own dedicated video RAM (VRAM).
This memory intercepts any data relevant
to graphics processing and takes over the
job of providing that data for the GPU.
It’s this hierarchy of data access that allows
our computers to both store huge volumes
of data indefinitely and still provide lightning-quick data processing, such as the processing
required for rendering 3D graphics in games.
Crucially, you can also see from this
setup how the roll of memory is intrinsically
linked with the overall speed of a PC. If
your memory, or the internal caches of a
processing unit, aren’t large or fast enough,
they’llholdbackperformance.However,equallywecanseethatbuyingever faster
memory doesn’t make a CPU or GPU itself
faster – it simply eliminates a bottleneck.WHAT IS DRAM?
By and large, each of the types of memory,
from on-chip caches through system
memory, to long-term storage, all use
fundamentally different technologies for
storing data. For instance, caches tend toSRAM IS EXTREMELY QUICK, SO IT’S
THE OBVIOUS CHOICE FOR A MEMORY
TECHNOLOGY THAT SITS CLOSEST TO
THE PROCESSING UNITS OF A CPUWhether your memory has a heatspreader, RGB lighting or a bare PCB,
most memory modules look near-identical underneath their covers