Maximum PC - UK (2019-09)

(Antfer) #1

Quality Control


Latency & Timings


THE JOINT ELECTRON DEVICE ENGINEERING COUNCIL (JEDEC) is
the body responsible for maintaining industry standards when
it comes to numerous solid-state components, including RAM.
With over 300 members, including Microsoft, Intel, and Apple,
JEDEC pools resources to push hardware development forward.
Its most important contribution to RAM standards is the
adoption of a unified connection format for motherboards
(the DIMM slot) to ensure compatibility, but that’s not all.
Standards for data transfer rates and latencies are determined
by JEDEC, so consumers choose between speeds such as DDR4-
3000 and DDR4-3200, instead of 3060, 3111, 3187, and so on. It’s
simple enough in principle, but these constraints ensure memory
is easily put into straightforward brackets for our convenience.
JEDEC’s members are also responsible for producing new
hardware within the RAM universe. The up and coming memory
standard is DDR5, which was ratified in March 2017 (not to be
confused with GDDR5, a memory standard for GPUs). Some


manufacturers (such as SK Hynix) have confirmed that their
DDR5 memory is complete, and desktop DRAM modules should
be available later this year. This next-gen memory will be able to
reach higher speeds than any units from DDR4, and draw less
power per module, thanks to integrated voltage regulators.
While it might sound like a minor improvement, this shift in
standards has implications beyond RAM design. With voltage
controlled on the memory sticks, motherboards will no
longer require voltage regulators for RAM, which will enable
manufacturers to focus on other features, but could cause
compatibility issues in the not-too-distant future. DDR5 will
double the bandwidth available, so 128GB may become more
mainstream, and 256GB builds will become a possibility for
high-end tweakers. Memory speeds will hit a staggering new
maximum at 6,400MT/s, the fastest available within DDR5.

QMemory speed isn’t the only
thing that affects your RAM.
Latency and memory timings
encompass a huge variety
of key statistics that affect
memory performance, and
many of those numbers will
be meaningless unless you’re
a die-hard hardware-head.
The one to look out for is CAS
(column address strobe)
latency, which is an indicator
of how quickly the RAM stick
responds to a memory call.
Bear with us, because it’s
going to get mathematical.
CAS latency by itself isn’t
necessarily a helpful figure—
in order to get a genuinely
useful number, there’s a
formula you need to use.
Real-world latency = (2,000/
RAM speed) x CAS latency,
with speed measured in MT/s
and latency measured in
nanoseconds. So, if you’ve
got a memory kit that runs
at 3,000MT/s and has a CAS

latency of 16ns, the kit’s
real latency is 10.6ns.
However, faster memory
tends to come with greater
CAS latencies, so the actual
performance gains can be
minimal, even when you seek
out higher frequencies. The
ideal memory kit has a high
frequency and a low CAS
latency, and overclocking
is an option, but there’s a
lot of potential messiness
there. If you’re running an
Intel system, variation in
speed isn’t going to make
much difference; real-
world latency in DDR4 and
beyond is starting to become
increasingly homogenized,
regardless of frequency, so
buying high-speed RAM can
become a waste of money.
The impact on some tasks is
more notable; gaming and 3D
rendering is less affected by
memory speed than large-
scale data analysis.

JEDEC doesn’t actually require heatsinks to be integrated on RAM.

Problems? Chances
are it’s your RAM.
Try manually setting
the frequency and
timings in the BIOS.

maximumpc.com SEP 2019 MAXIMUMPC 27

Free download pdf