Push your SSD to the Limits

Introduction

This article is about how you can make an SSD bench more
quickly!

For those of you that are puzzled by why your SSD does not
bench as fast as it does for someone else – this article is for you. 

In truth all SSDs of the same Make, Model and NAND type are
created equal and they are all as fast as each other; if you see a bench result
which is faster or slower than yours then the difference is being caused by the
environment the SSD is running in.  By environment I mean the
Motherboard/chipset, CPU, CPU OC, driver, power plan, BIOS settings...

In this article I am going to show you how you may make your
bench results go faster and faster.

Please note that a 'How To Guide' for applying the ‘boost
factors’ mentioned in this article can be found on the Myce SSD forum, please
click here.

Let's move to the next page to look at the Testing
Approach...

Testing Approach

Hardware

The hardware used for testing was –

The Vectors were plugged into the two native Intel SATA III
(6GBs) ports.

Software

The software used for testing was -

 

Approach

The article presents a succession of incremental ‘boost factors’
that will make your SSD bench faster.

It starts with a fresh install of Windows 7 with the SATA mode
set to RAID rather than AHCI even though the OCZ Vectors are going to be run as
single drives and not in a RAID configuration.  I feel there are some
advantages to this for a number of reasons – firstly, it makes it easier to
implement a RAID array later on (without editing the registry to switch into
RAID mode or without reinstalling Windows) and, secondly, I believe SATA drivers
run more quickly in RAID mode even for single drives (particularly, I feel,
when using an Intel Rapid Storage Technology (‘IRST’) driver) and, thirdly, one
can then boot in AHCI mode if/when desired.

As the article progresses through the stages of incremental
‘boost factors’ you will see that gains diminish.  After each ‘boost factor’ is
applied, AS SSD is run three times in succession and the three overall scores
are recorded, along with an AS SSD screenshot for the highest result.

I wish to make it clear that I do not for one moment suggest
that all of the ‘boost factors’ should be applied to anyone’s day to day
‘production’ system.  Indeed, I suggest that anything beyond Page 5 – CPU Over
Clock – should only be applied when one is trying to bench as fast as one can
(and even then I would not suggest that anyone runs with a 4900MHz OC on a day
to day basis).

However, if you feel the need for speed and/or want to post
the highest benches – you’ll now know how to do it.

Please note that installing additional applications, particularly
those that run continually in the background, such as Anti-Virus software, may
well affect the results of SSD performance benchmarks.  For this reason I
suggest you build a specific ‘SSD Benching System Image’ that you can then
restore and use when you choose to.  Whatever you do, please make sure you
backup your day to day production system before you start.  In particular,
please do not disable Windows services on your day to day production system.

Please note this article is in the main relevant to Intel
based systems only.  Although Windows 7 has been used for testing, the boost
factors are also applicable to Windows 8. 

Let's move to the next page to look at the effect of
Windows Power Plans...

High Performance Power Plan

Following the fresh installation of Windows 7 onto an OCZ
Vector, the default RAID driver had been implemented for use by the Vector and
the default ‘Balanced’ Windows Performance Plan was in force.

The three AS SSD overall scores at this stage were – 1111,
1105 and 1117 and here is a screenshot of the highest one –

The Windows High Performance Power Plan was then engaged.

The three AS SSD overall scores at this stage were – 1140,
1136 and 1140; here is a screenshot of the highest one –

A significant improvement over the default ‘Balanced’
Windows Power Plan can clearly be observed.

Let's move to the next page to look at the effect of
using an Intel RST driver...

Intel Rapid Storage Technology (IRST)

There are many versions of the IRST driver that can be
implemented.  The safest option, if one’s overriding concern is stability, is
to install the latest IRST driver listed on the Support Download page, for your
specific motherboard, by your motherboard provider.  However, I’ve never
experienced a stability problem when trying out different versions.

For this article I have chosen to use version 11.6.0.1030,
which based on my experience is currently the fastest for a single SSD (by the
way I find version 11.2 is currently fastest for a RAID 0 array).

So, the IRST 11.6 driver was then installed.

The three AS SSD overall scores at this stage were – 1180,
1188 and 1189; here is a screenshot of the highest one –

A significant improvement upon the default driver can be
observed.

Let's move to the next page to look at the effect of
applying a CPU Over Clock...

CPU Over Clock

At this stage the Core I7-3770K was over clocked to 4900MHz.

Please note that in my experience there is little, or no,
improvement in AS SSD performance to be gained beyond 4800MHz.

Please also note that you can try a smaller Over Clock (OC)
if you are not happy with going this far.

I am also compelled to make the usual disclaimer by saying
that applying any over clock is strictly at your own risk.  If you are not
familiar with over clocking your rig I’m sure you can use Google to find helpful
guides - click here
for an ‘Easy Over Clocking Guide’ for the implementation of an OC on Sandy and
Ivy Bridge rigs that you can use on a day to day basis (this is what I use on a
day to day basis).

The three AS SSD overall scores at this stage were – 1172,
1191 and 1197; here is a screenshot of the highest one –

Not such a dramatic difference in this round, but our peak
score is still on the up...

Let's move to the next page to look at the effect of
disabling C1E...

Disabling C1E

Please click here for an introduction
to C1E and other CPU power states.

At this stage C1E was disabled in the CPU Power Management
section of the UEFI/BIOS.

The three AS SSD overall scores at this stage were – 1170,
1207 and 1209; here is a screenshot of the highest –

Another improvement and through the 1200 barrier, things are
getting truly fast.

Let's move to the next page to look at the effect of
disabling all other C states and EIST...

Disabling all C States and EIST

At this stage all the other C States (C3, C6 and Package C
State Support) and Enhanced Intel Speedstep Technology (EIST) were disabled in
the CPU Power Management section of the UEFI/BIOS.

The three AS SSD overall scores at this stage were – 1208,
1204 and 1207; here is a screenshot of the highest one –

No significant change in scores can be observed.  At this
stage we are into diminishing returns and it appears that disabling C1E is the
most significant factor when it comes to CPU Power Management features.

You can probably sense by now that the game has become one
of removing possible distractions that may impact upon a run of AS SSD.

Let's move to the next page to look at the effect of
disabling all non-essential onboard devices...

Disabling non-essential onboard devices

At this stage all non-essential onboard devices (i.e. that
are not needed whilst running AS SSD) were disabled in the Onboard Devices
section of the UEFI/BIOS.

For me this included: HD Audio Controller; VIA 1394 Controller;
Marvell Storage Controller; ASM1061 Storage Controller; Intel LAN1 and LAN2
controller; ASMedia USB3 Controller.  I’m far from certain that disabling all
these devices makes a difference but my attitude when ‘going for gold’ is to be
obsessive rather than falling short. 

The three AS SSD overall scores at this stage were – 1206,
1204 and 1210; here is a screenshot of the highest one –

Again, not a significant difference, but a new peak level of
1210 was achieved.

Let's move to the next page to look at the effect of
disabling all non-essential Windows Services...

Disabling non-essential Windows Services

Now we are reaching a stage that only the true SSD Speed
Freak will consider.  It involves disabling all non-essential Windows Services,
so that the ‘last ounce’ of distraction to AS SSD is removed.

For me, the authority on this subject is known on the net as
‘Black Viper’ – please click here
for his guide to which services can be disabled.  I have followed his
suggestions in the ‘bare bones’ column.

Please note, disabling services is not something
one should do to your day to day system – what I do is build a benching
system (based on a fresh install) and then take an image back up so that I can
restore it when I choose to in the future.  If you do want to build a benching
system image make sure you backup up your day to day system first!

The three AS SSD overall scores at this stage were – 1209,
1208 and 1213; here is a screenshot of the highest one –

Again, not a significant increase, but a new peak level of
1213 was achieved – but this is now getting into record setting territory when
benching a drive as a Windows boot drive.

Let's move to the next page to look at using all we have
achieved so far when benching a spare drive (i.e. not the boot drive)...

Benching a Spare

So far, all of the AS SSD bench runs have been run against a
boot drive.

At this stage I switched to running the bench against my
second OCZ Vector, which is not the boot drive (commonly called a ‘spare’
drive).

(I realise not many people will conveniently have a second
SSD to bench as a spare, so it will be necessary to move the benching image
that you have built to a spare drive so that one can boot from it and test the
target SSD as a spare)

The three AS SSD overall scores at this stage were – 1233,
1238 and 1238; here is a screenshot of the highest –

Clearly benching a spare produces significantly better
results.

So the final benchmark in my incremental application of ‘boost
factors’ is 1238.  The truth is that at this stage hitting a new high score
requires a bit of luck and patience; you will need to run AS SSD time and time
again (perhaps with a brief pause between runs to let your SSD recover peak
performance).  You will notice that the key deciding factor in what overall
score is hit will most often be the 4K writes result.

 

Let's move to the next page to look at the Conclusions...

Conclusions and World Record

The fastest AS SSD bench that I have seen posted on the net
for a single Vector 256GB is 1243 and being the speed freak that I am, I
couldn’t stop when I was so close. So, I pumped up my CPU OC to 5000MHz and
after a dozen or so runs of AS SSD I got lucky and hit the following score.

As I have not seen higher I’m going to claim this as a World
Record AS SSD score for a single Vector 256GB running in Windows 7 and because
I don’t know of any consumer drive that benches faster in AS SSD I’m going to
go a step further and claim this as a Windows 7 World Record for any single
consumer SSD.  I expect someone will soon prove me wrong :o)

At the end of this quest for speed, I feel it is necessary
to admit that as user I do not feel any difference in speed from where I
started and where I ended up – the difference is, for me, only observable in
benchmarks.

Note that to be in with a shout of hitting really high
benchmark scores, one needs to use the native Intel SATA III (6GBs) ports on a
P67, or later, motherboard.

For those of you that give it a go – as a fellow
gladiator, I wish you every success.

I feel there may soon come a time where SSD manufacturers
allow users some freedom to configure the firmware on their SSD.  For example,
if one is only using an SSD as a boot and applications launch platform it is, I
feel, unlikely that one will ever threaten the endurance/life expectancy limit
of the drive, so could there be some room to allow endurance parameters to be
relaxed in favour of greater speed and for users to be allowed some level of
self selection?  I hope so.

If you know of any other ‘boost factors’ please let us know.

Lastly, I thank Alex Intelligent Software for the
availability of the excellent AS SSD benchmark.  AS SSD can be downloaded from here.

You may comment on this article below...