Overclockers.com-Overclocking FAQ

“Overclocking FAQ”

by Frank Monroe

10/3/98

Contents:

Introduction and
Disclaimer
So you
want to overclock a Celeron?
Why is the
Celeron so overclockable?
What does stepping
mean?
What is an S-code?
What
is ‘multiplier locking’ and ‘bus locking’?
Which
type of Celeron should I buy?
Which
motherboard should I use?
What kind
and how much memory?
What about cooling?
How do I overclock?
What if it doesn’t
work?
Will
it damage my CPU or other components?
Where can I learn
more?

Introduction
and Disclaimer

I created this FAQ because, after reading
literally thousands of posts, I still see the same requests for basic instructions
over and over in each Newsgroup and forum. There are many web sites
with similar information, but many people either can’t find these
sites or they don’t have web access. Since I have never seen a FAQ
like this posted in any of the Newsgroups I read, I took it upon myself
to offer this small contribution to novice overclockers everywhere.
I don’t claim to have all the answers and I can’t guarantee that everyone
will be able to overclock their Celeron, but after reading this FAQ you
should be well on your way to a successful experience.

Overclocking is not recommended by any
manufacture (especially Intel) and will void your warranty. I do
not advise anyone to follow these instructions unless they are willing
to assume all associated risks. I have consolidated in this document
information that I’ve learned while overclocking my own system or that
I have read about the experiences of others. Overclocking can damage
your system. Working inside your power supply or wiring 110 volt
fans can cause serious personal injury if done by the inexperienced or
without the proper precautions. If you’re unsure or in doubt about
any of these procedures, seek professional advice. I am providing
this document for informational purposes only.

If any one out there in Net-land has suggestions,
comments or contributions for this FAQ, feel free to contact me.
Frank Monroe email: [email protected]

So
you want to overclock a Celeron?

You’ve read a few post, maybe visited a
few web sites. Everyone is reporting their success and claiming fantastic
speeds from a lowly 266 or 300 mHz CPU. You’re excited at the prospect
of a high performance CPU for, essentially, small change and you want to
get in on the action. The speed of a P2-400 or -450 for $90 or $150
sounds too good to be true. But wait, they’re talking about S-codes,
multiplier locking, Pin B21, CAS-2, and other esoteric terms. Names
like Deschutes, Klamath and Mendocino are bandied about while you wonder
what these words have to do with computers. Now you’re confused.
How hard is this going to be? Is it worth it? Do you
need to be an Electrical Engineer to overclock a Celeron? In a word,
no. With the right hardware and a little luck, it should be a snap.

Why
is the Celeron so overclockable?

As you may know, a given chip design is
used for CPU’s of many different speeds. The P2 and Celeron designs
are named after Western US counties: Deschutes, Klamath and Mendocino.
More on this later.

In theory, a CPU is tested first at it’s
maximum speed. The ones that pass the testing process at this speed
are marked as such and sold as top-of-the-line CPU’s. Those that
fail at the fastest speed are tested at successively lower and lower speeds
until they run reliably. These slower cores are then marked with
the speed at which they passed the testing process and sold as slower processors.
At least, that’s the theory. No one really knows how Intel decides
which cores get marked for a given speed. Several other factors,
such as customer demand and production quality, affect how many processors
of each speed are produced.

A CPU of any given speed can usually be
made to run somewhat faster if one is willing to play around with the motherboard
settings. This is the overclocker’s bread and butter. Now, through
a convenient turn of events, Intel has produced a CPU with an unusually
high capacity for overclocking.

Intel has long controlled the high-end
CPU market while it’s competitors, Cyrix and AMD were gaining market share
in the low- and mid-price range because of the popularity of lower priced
PC’s. Intel finally realized what was happening and wanted to recover
the low ground while also keeping the high end market (can you say “total
market domination”?). When Intel designed the CPU core for their
newest line of processors, the P2, they changed the way the CPU was mounted.
All P2’s are mounted on a circuit board, called an SECC (Single Edge Contact
Cartridge), that plugs into a special, patented CPU slot (Slot 1) similar
to a PCI slot.   [Intel calls the Celeron packaging a SEPP (Single
Edge Processor Package) but it’s still compatible with the Slot 1 connector,
go figure.]   AMD and Cyrix do not have a Slot 1 CPU, so if you
want high-end speed, you need to buy an Intel processor.   Thus
the high-end market is preserved for Intel. Now, Intel needed a cheap
Slot 1 CPU to corner the low-cost PC market.

page 2

Enter the Celeron line. To reduce production
costs, Intel left out the expensive Level 2 cache. Also, to eliminate
design costs, the original Celerons (C266 and C300) used the same CPU core
as the new 350-450 mHz P2’s (code name Deschutes). [Remember, design
costs account for a huge percentage of the total cost of a CPU. Once
in production, it costs exactly the same to manufacture a core destined
for use as a 266 mHz processor as it does to use that same core in a 450
mHz processor.] Many media pundits immediately dubbed the Celeron
a backward-stepping piece of crap because of the lack of the L2 cache.
Later, perhaps due to the poor reviews from hardware critics, Intel released
the Celeron 300A and 333 with 128 Kb of built-in cache. Again, they used
basically the same core design with some modifications to incorporate the
on-die cache. The C300A and the C333 modified Deschutes core carries
the code name Mendocino. Since Celerons use a Slot 1
motherboard, you can’t upgrade to one of Cyrix’s or AMD’s fast new CPU’s
later, when prices come down. They don’t have Slot 1 CPU’s and Intel
has the patent. Now Intel has again regained a foothold in the below-$1000
PC market and insured that the upgrade dollars also come home to Papa Intel
too.

Here’s where it gets interesting.
The fastest P2 CPU’s (350 to 450) require a relatively new type of Slot
1 motherboard with the BX chipset. The BX motherboard runs at a bus
speed of 100 mHz. They can also run at 66 mHz bus which allows them
to accept slower P2 CPU’s (233, 266, 300 and 333) and Celerons. The Celerons
are supposed to be used on the earlier EX and LX generation of Slot 1 motherboards
which run at 66 mHz only. Since the Celerons have the exact same
core as the new architecture P2 CPU’s, there’s nothing to stop you
from setting the bus to 100 mHz and running a Celeron at 400 or 450 mHz.

People started buying BX motherboards and
Celerons and overclocking the hell out of them by setting the bus speed
to 100 mHz. A chip meant to run at 266 running at 400 mHz and more was
unheard of previously. It’s all because Intel is trying to capture
the low-cost CPU market without the R & D costs of a new chip.
It’s really a marketing stroke of genius when you think about it.
Produce one type of CPU. Take the best ones, add 512 kb of fast,
expensive cache and sell it as the top-of-the-line CPU for $700+.
Take the rejects, leave off the expensive L2 cache and sell them as cheap
Celerons. Except they’re too smart for their own britches.
The production yield of 450 mHz cores is too good and the “rejects” are
too few and far between. Because they want to flood the market with
$100 CPU’s, they have to mark them as 266 to 333 mHz Celerons and sell
them cheap anyway. It doesn’t cost them any more since both chips
came off the same production line. Because the P2-450 market is relatively
small compared to the low- and mid priced market, the demand is greater
for Celerons.

What
does stepping mean?

Celerons come in four flavors. The
C266 and C300 without L2 cache and the C300A and C333 with 128 Kb L2 cache.
Each type of Celeron has several slightly different variations, called
a “stepping”. Stepping 0 (zero) cores are the original production
run. When minor imperfections (bugs) are found in the instruction
programming (micro-code) of the core or in other parameters of the chip,
they are fixed and the next batch of cores will incorporate the changes.
This batch will be identified as stepping 1. If another change is
required later, the stepping number will be incremented again.
As each successive refinement to the chip is made, the next higher stepping
number will be assigned. For many reasons, one stepping may be easier
to overclock than another, but usually the higher stepping cores make the
best, most stable CPU’s.

What is an
S-code?

An S-code (Intel actually calls it an S-Spec.)
is a 5 character designation beginning with ‘S’ used to identify the various
different types, stepping, voltage and packaging of Celerons and other
Intel processors. There are currently 14 (as of 27 Sep 98)
different S-codes for the Celeron family of CPU’s. OEM packaging
is just the SEPP in a plastic container. There is no heatsink/fan
attached, so you need to buy your own. The warranty, if any, is usually
only for 30 days and from the vendor, not Intel. The retail Celeron (sometimes
also called a “boxed” Celeron) comes in a cardboard box with a pretty
good heatsink and fan already attached. You also get a Certificate
of Authenticity and an Installation Notes booklet in 11 languages, a cute
sticker for the front of your computer and, most important, a three year
warranty from Intel. The S-code can be found on one end of the retail
box or on the back, left side of any Celeron SEPP printed circuit board.

C266 (Deschutes core without cache)

———-

SL2SY Stepping 0
OEM

SL2QG Stepping 1
Retail

SL2TR Stepping 1
OEM

SL2Y3 Stepping 2
Retail

SL2YN Stepping 0
Retail

C300 (Deschutes core without cache)

———-

SL2YP Stepping 0 OEM

SL2Y2 Stepping 1
Retail

SL2X8 Stepping 1 OEM

SL2Y4 Stepping 2
Retail

SL2Z7 Stepping 0
Retail

C300A (Mendocino core with
128 Kb L2 cache)

———-

SL2WM Stepping 0 OEM

SL32A Stepping 0
Retail

page 3

C333 (Mendocino core
with 128 Kb L2 cache)

———-

SL2WN Stepping 0 OEM

SL32B Stepping
0 Retail

What
is ‘multiplier locking’ and ‘bus locking’?

No processors since the early 80486 CPU’s
have taken the motherboard bus clock and used it internally at the same
speed. Remember the 486DX2? It took the 33 mHz bus clock from
the motherboard, multiplied it by 2 and ran at an internal speed of 66
mHz. Modern, BX-chipset motherboards now provide a 66 or 100 mHz
bus clock to the Slot 1 connector. Today’s Celeron and Pentium II
processors multiply this to achieve their designated speed. Without
multiplier locking, circuitry inside the processor reads the multiplier
jumpers on the motherboard via the Slot 1 connector. Depending on
the setting of these jumpers (or BIOS setting for the Abit boards) the
CPU then multiplies the clock by 3.5, 4, 4.5 or 5. Multiplier locking
forces the CPU to use a multiplier that is pre-determined by Intel, ignoring
the settings on the motherboard. All Celerons are multiplier locked.
The C266 is multiplier locked at 4; the C300 and C300A is locked at 4.5;
and the C333 is locked at 5.

Multiplier limiting (only affects P2-350/400
processors made before mid-August ’98) uses a signal from the motherboard
to detect the bus speed and then places an upper limit on the multiplier
based on the bus clock speed. For example, with the bus set to 66
mHz, the processor can be set to a higher multiplier than it can when the
bus clock is set to 100 mHz. In effect, this limits the CPU to a
maximum internal speed while allowing lower speeds. With a 66 mHz
bus, a “multiplier limited” P2 would accept higher multipliers than at
100 mHz. [The BH6 BIOS has a setting under the SoftCPU menu called
100/66#SEL. With the LOW setting you can defeat the clock limitation
on certain P2 processors. It will not work on the newer 400’s and
450’s and it will not unlock the Celerons.] Intel says it uses multiplier
locking and multiplier limiting to prevent unscrupulous retailers from
re-marking processors to higher speeds.

Bus locking is a myth, at least at the
present time. If it was implemented, it would prevent a processor
from being used at a higher bus speed than it was designed for. For
example, since all Celerons are meant to use a 66 mHz bus clock, bus locking
would prevent the CPU from running at any other bus speed. Since
bus speed is set on the motherboard, Intel would need to design and
incorporate special circuitry in the CPU to detect the bus speed and compare
it to the “proper” clock rate. Below is a newsgroup post about bus
locking from a member of the Intel support team. As far as I was
able to determine, this is a genuine post. [I have edited format
but not content.]

From: “Randy S.” <[email protected]>

Subject: Re: Bus locking and clock locking

Date: Mon, 14 Sep 1998 11:52:16 -0700

Newsgroups: intel.microprocessors.celeron

Organization: Intel Corp. – 100003

OK.

First of all, “bus locking?” how feasible
is this in a processor when the bus timing and clock generation is sourced
on the motherboard ? Think about that.

Yes, of course, you could specify
something in the design guidelines for the processor. Given the current
state of art in board design, I do not think this would happen on the processor
side. In fact, what are you currently doing now? Would not
a better description be “Overbussing” (i.e. using a bus speed higher than
specified) .

You can do what you want with your processor
like anything else, but as long as there are those who wish to misuse and
misrepresent Intel products to others for their own personal gain, we will
take whatever means needed to prevent or foil these attempts.

To be honest, the needs of ‘overclockers’
pale to insignificance when compared to the specter of remarking and fraud.
Intel is acutely aware of activity in this regard, and thus, the accommodation
of variable multipliers and the like must be sacrificed to preserve the
basic integrity of our product.

That said, perhaps you would like to convince
the average consumer who was duped into buying an overclocked system of
the
need for this ability.

It seems to me your discontent is misplaced.
It is the unscrupulous people that should demand your attention in this
regard, not Intel.

To paraphrase a quote I read somewhere:

“Chaos offers multiple solutions” – but
who will support it? Perhaps you?

—

Randy S.

Intel Internet Technical Support

*All other brands and names are property
of their respective owners.

Which
type of Celeron should I buy?

The key, of course, is getting the right
stepping of the right type of Celeron. First of all, the “wrong”
Celerons are the C300 (without the ‘A’ designation) and the C333.
If you plan to overclock your Celeron, stay away from these two altogether.
Because of multiplier locking [see “What is ‘multiplier locking’ and ‘bus
locking’?”], the C333 severely limits your options. Reports of successful
overclocking are very rare with these CPU’s. While it’s true that
some Celerons have successfully run at speeds up to 500 mHz, these instances
are rare [estimated at 5-10%]. If your C333 won’t operate at 500
mHz (100 X 5), your only option (since the multiplier is locked), is lowering
the bus speed to 83 mHz (415 mHz) or 75 mHz (375 mHz). Seventy-five
and 83 mHz require the PCI and AGP devices to operate at higher than normal
speeds since motherboards only provide the correct clock step-down dividers
for the 66 mHz and 100 mHz clock speeds…

page 4

[Some motherboards may correctly
divide the 133 mHz bus speed, but then SDRAM timing becomes a major problem.]
Many devices do not react well to this condition. Obviously, you are more
likely to have problems at 83 mHz than at 75 mHz. The C300 and the
C300A should theoretically have the same likelihood of reaching 450 mHz
[estimated at 60-75%]. However, there is very little information
on the success rate of the C300. [The reason may be that improvements
were incorporated into the C300A that were not made to the C300.
Remember the C300A uses the Mendocino core while the C300 uses the cache-less
Deschutes core.] If you’re are going to for 450 mHz with a 300 mHz
CPU, why get a C300 when the C300A has the L2 cache for a few dollars more.

Between the C266 and the C300A, it becomes
a matter of your needs and your wallet. Both Celerons are quite likely
to run well on a 100 mHz bus. The success rate for the stepping 1
C266 at 400 mHz (estimated at 85-90% for the SL2QG and the SL2TR) is somewhat
higher than for the C300A at 450 mHz (estimated at 60-75%).
The C266 is cheaper, roughly 1/2 the price of the C300A, but it doesn’t
have any L2 cache. Though the lack of the L2 cache may not make as
much difference as you might expect, it does affect overall performance
some. In most 3D games (except Unreal) the lack of secondary cache
makes almost no difference. Most other applications do use the L2,
but remember, SDRAM at 100 mHz is pretty damn fast and you do still have
the L1 cache. After you factor in the wasted CPU cycles looking in
cache for data that’s sometimes not there (cache miss), the speed disadvantage
is less than you might imagine.

That said, the C300A seems to be the current
favorite. Many people have decided that the extra cost and added
risk is more than outweighed by the [IMHO, slight] performance gain offered
by the 128 Kb of cache. A word of caution, the C300A does seem to
be a little trickier to overclock. It occasionally needs more than
the default 2.0 volts to be stable and often needs more than a stock CPU
heatsink and fan. I’ve seen many more cries for help from people
with an unstable C300A at 450 mHz than first-time success stories.
Most people usually achieve their goal but some just never make it, no
matter what they do. There are no guarantees with either CPU.

OK. We’ve narrowed the field somewhat,
but there are still the different stepping variations to consider.
For the C266 the best stepping is stepping 1 carrying S-codes SL2QG and
SL2TR. Since the stepping 2 C266 is so new, there is very little
known about the success rate of the SL2Y3. It may very well be as
good or better than the stepping 1 CPU’s. Before the release of the
C300A, the S-code with the most successful reports was by far the SL2QG.
The fan and heatsink provided by Intel with the retail boxed SL2QG has
proved to provide more than enough cooling and there is very little price
difference between it and the OEM SL2TR.

Both the of the C300A CPU’s, the OEM SL2WM
and the retail boxed SL32A, are stepping 0. Many people have reported
that additional cooling helps improve the success rate of reaching 450
mHz. With this in mind, you may want to consider the cheaper OEM
version and buy your own quality heatsink and fan setup rather than find
out later that you need to remove the Intel-provided heatsink on the retail
version. There is one additional point in choosing your C300A that
bares mentioning. There seems to be a consensus that certain manufacturing
plants have a higher success rate. The Malaysia plant has gotten
the most attention as producing the best CPU’s, but few vendors even let
you request specific S-codes, let alone trying to specify country of origin.
[This last thing about the country of fabrication is unscientific at best
and most likely just wild speculation.]

Which
motherboard should I use?

There are basically two motherboards of
choice for overclocking a Celeron, the Asus P2B and the Abit BH6.
The BH6 is by far the most popular with owners of the C300A for a couple
of reasons. First, it’s the only motherboard that doesn’t need a
BIOS upgrade to recognize the C300A. Secondly, it allows you to increase
the CPU voltage from the BIOS SoftMenu. This feature is particularly
attractive since many C300A CPU’s need a voltage higher than the default
setting of 2.0 volts to be completely stable. Additional features
that make the BH6 popular are it’s lower cost, BIOS SoftMenu setup for
all settings and an additional PCI slot.

In all fairness, the Asus P2B is also a
very good board. Though it costs about $40 (US) more, some staunch
Asus supporters maintain that the P2B is more stable and has a higher success
rate when overclocking. The P2B does not have any built-in provision
for changing the CPU voltage if it’s necessary to do so, but it does have
3 ISA slots for those legacy ISA cards while the BH6 has only 2 ISA slots.
The biggest drawback to the P2B is that you need to flash the BIOS to the
newest version (1005) in order for it recognize the C300A. This can
be problematic, to say the least, since you need a CPU to flash the
BIOS and it won’t recognize your CPU until you flash the BIOS. Catch
22.

There are other motherboards that can be
used, however, I recommend at least considering one of these two if at
all possible.

What
kind and how much memory?

If you are going to use a 100 mHz bus speed,
you should plan on getting PC-100 SDRAM memory. Many people have
reported successfully using their old PC-66 memory, however, if you do
try it and have problems overclocking, memory would be a likely suspect.
You should also plan on starting with a minimum of 64 Mb. It’s best
to get 64 or 128 Mb DIMM modules since the number of memory sockets is
usually only 3, or 4 at the most. Two 32 Mb DIMM modules will
limit your ability to upgrade memory in the future.

page 5

Not all PC-100 is created equal and you
usually get what you pay for. Fortunately, if you only plan to try
the 100 or 112 mHz bus speeds, most good quality PC-100 memory will work.
You only need the high-priced, premium “CAS-222” memory if you want to
try 124 or 133 mHz bus speeds. CAS stands for Column Address Strobe
-OR- Column-Address Select. CAS latency refers to the number of processor
cycles from when a Read command is registered to when the data from that
Read command becomes available. CAS-222 certainly won’t hurt if price
is no object, but the added cost will only buy you a few tenths of a percentage
point of overall performance.

There is an excellent web site, called
PC100, that will tell you more than you’ll ever want to know about memory.
The URL is listed in the learning resources section at the end of this
FAQ. Here are some excerpts from that page (used with permission).

From the PC100 web page–

The fastest cas 2 parts that operate at
cas 2 at either 66 mHz or 100 mHz will have a PC100 label on them that
says PC100-222-xxx. Cas 2 parts that operate at cas 2 at 66 mHz and cas
3 at 100 mHz will have a PC100 label on them that says PC100-322-xxx. The
222 or 322 refers to the actual data programed into the spd eeprom chip
located on the memory module that informs the motherboard exactly what
the SDRAM module is capable of.

brand	nano-sec	rated max	real world	cas @ 66mHz	cas @100mHz	PC100-label
many	-12	83 MHz	66 MHz	3	n/a	n/a
many	-10	100 MHz	83 MHz	3	n/a	n/a
most	-8	125 MHz	100 MHz	2	3	322-620
Samsung	-G8	125 MHz	100 MHz	2	2	222-620
Gold Star	-7J	100 MHz	100? MHz	2	3	322-620
Gold Star	-7K	100 MHz	100? MHz	2	2	222-620

What about
cooling?

There are two aspects of system cooling
that need to be considered, case cooling and CPU cooling. The power
supply fan alone normally does not provide sufficient air flow to eliminate
heat build-up inside your case. Hot air trapped in the case forces
all components to operate at higher temperatures and reduces the effectiveness
of convection cooling throughout your system. Many overclockers find
that heat is their main enemy, especially if you find that you need to
raise the CPU voltage. There are several things you can do to ensure
that your case stays cool.

First, check the direction of air flow
from the power supply fan. The best cooling is obtained by having the power
supply fan draw air out of the case. If it draws air into the case,
you may want to try reversing it. It’s a simple procedure than can make
a significant difference in case temperature. [Caution: Capacitors
in the power supply can store a charge even after the power has been off
for several hours. Make sure that the unit has been unplugged for
24 hours or more.] Remove the supply from the case and remove the
cover. Most power supply fans are held in place by four screws.
Remove these four screws and flip the fan over. Generally, both sides
of the fan will have a set of holes so you should be able to re-attach
the fan with the same screws. Reassemble and install the power supply.
You should see a drop of several degrees inside the case just from this
simple, free procedure. Opening the power supply will probably void
the warranty on it so, if you’re worried about that sort of thing, you’ll
be relieved to know that there are other things you can do to lower your
case temperature.

Adding a second fan is a good idea even
if you aren’t overclocking. Many cases provide a location at the
lower front that is designed for a second fan. Even if your case
doesn’t have a ready-made mounting point, you should be able to find a
spot to install a second fan. Depending on the type of connector
your fan has, you can plug it into the motherboard fan connector or use
one of the extra drive power cables for it’s 12 volt supply.

Leaving the case cover off is also a possible
solution to overheating. Though not ascetically pleasing, it is a
free solution that many overclockers employ.

Now that your case is maintaining a near
ambient temperature, you need to think about the CPU. If your system
crashes or seems to become unstable after a few minutes of operation, you
may find that heat build-up is the problem. The fan and heatsink
that is attached to retail Celerons is usually adequate to achieve the
400 or 450 mHz speed with the C266 or C300A. If you bought an OEM
Celeron or if you’re having suspected heat problems with your CPU, you’ll
need to buy a good heatsink and fan combination and install it on the CPU.
Many vendors offer cooling packages with heatsinks and one, two or even
three fans. One vendor (STEP-ThermoDynamics) even offers an electronic
peltier system ($85 US) and another (Kryo-Tech, mentioned at Tom’s hardware
site) offers a $500 refrigeration system. While these expensive cooling
systems work very well, most people find that a simple heatsink setup with
one or two ball-bearing fans will provide all the heat dissipation that
your CPU needs. A number of vendor and cooling information web sites are
listed in the learning resources section at the end of this FAQ.

page 6

Be sure to remove all of the original thermal
tape from the CPU if you’re replacing an old heatsink. It’s important
to apply a very thin layer (.003 to .005 inch) of thermal interface material
between the heatsink and the CPU. A little goes a long way since
you’re only trying to eliminate air gaps, not frost a cake. Too much
is worse than not enough. Some interface materials are conductive
so you’ll want to be careful not to get any on the CPU pins or circuit
board. There are several different types of thermal compound and some work
better than others. Thermal greases are made by dispersing thermally
conductive ceramic fillers in silicone or hydrocarbon oils to form a paste.
Thermally conductive compounds are an improvement on thermal grease as
these compounds are converted to a cured rubber film after application
at the thermal interface. Thermally conductive adhesive tapes are double-sided
pressure sensitive adhesive films filled with sufficient ceramic powder
to balance their thermal and adhesive properties. As you can see
from the table below, thermal tape is only slightly better than nothing
at all. Radio Shack and most electronic supply stores sell small
tubes of the thermal grease or thermal compound for $2-$5 (US).

Thermal resistance of CPU to heatsink (Source:
Electronics Cooling Magazine, Sept. ’96)

———————————————-

Thermal compound = 0.8

Thermal grease = 0.9

Thermal Tape = 2.7

Dry joint = 2.9

(Lower numbers are better)

How do I overclock?

First, set up your system and get it running
at it’s normal speed. Set the SoftMenu or the jumpers as directed
in your motherboard manual. Install all your peripheral cards and
software and test out the system. Run a few benchmarks at the standard
speed so you can compare the before and after results. Only when
you’re satisfied that the system is behaving as it should and that it’s
stable at the rated speed, should you begin to push the performance envelope
of your system.

Now you’re ready. Neither the Asus
or the Abit motherboards require you to cover pin B21 on the Slot 1 edge
connector. It is ignored by the motherboard. Most other motherboards
do require you to cover this pin to fool the bus speed setting circuitry
into selecting the 100 mHz speed.

On the Abit boards you should reset your
system and enter BIOS setup. Change the following options in the
CPU SoftMenu:

CPU Operating Speed: User Define

– External Clock: 100 Mhz

– Multiplier Factor: x4
(or 4.5 for the C300A)

– AGP/CLK: 2/3

Speed Hold Error: Disabled

[Note: I don’t have an Abit motherboard
so the SoftMenu setup is unfamiliar to me. These are the settings
recommended by Andy Drake’s web site. If anyone has something to
add or correct, please contact me.]

On the Asus P2B you have to change a jumper
on the motherboard. If you are set for the 66 mHz bus speed (as you
should be if you followed the advice at the beginning of this section),
you should only need to change one jumper. Power down and unplug
the power cord. The jumper block that you need to set is located
just above the primary IDE connector and it should be labeled “BUS FREQ”.
Your current setting for 66 mHz should be: FS0 1-2 (pins closest to CPU),
FS1 1-2 (pins closest to CPU), FS2 2-3 (pins away from CPU). To set
the bus to 100 mHz you need to change FS2 to 1-2 also. Now all three
jumpers should be on pins 1-2, the pins closest to the CPU.

That’s it. If it works when you power
up, you’ll be at your new, overclocked frequency. A C266 will be
400 mHz; If you’re lucky, a C300 or C300A will be at 450 mHz; and, if you’re
really, really lucky, the C333 will be at 500 mHz. If your
computer completes POST (Power On Self-Test), boots into Windows
and seems stable, try running some applications. Run a benchmark
or two. Let it stay on for several hours, cycling a game demo or
benchmark. If it acts normally, except FASTER, of course, congratulations!

If it doesn’t work at first, don’t worry
(yet), there are several things you can try before you give up and admit
that you’ve got an “unlucky” CPU. Read on.

What
if it doesn’t work?

There are many things that can be done
to coax a stubborn CPU into working. I’ll try to mention as many
as I can here. Above all, don’t give up until you have exhausted
all of your options. Some of the things you can try are free or low
cost, while others may require replacing some expensive components.
Whenever possible, try to eliminate the cheaper options first. Then,
if you suspect you may need to buy a new DIMM or video card, try to borrow
one from a friend first or try your CPU in another, successfully overclocked
system. Remember, it might not be the CPU at all, but something else
in your system that’s giving you problems.

Heat

As mentioned in the section on cooling,
heat build-up is one of the most common problems. It manifests itself
usually after several minutes to an hour after start up, especially when
running CPU intensive applications. If you system won’t POST (Power
On Self-Test), heat is probably not one of your problems. 1.
Try leaving the case open with a table fan blowing into the case.
If the system stays up longer or seems more stable with the table fan and
open case, try some of the cooling methods mentioned above.

page 7

2. Check the temperature of the
CPU (front and back sides) and your video card by gently placing your finger
on the heatsink. Be gentle and only touch the heatsink while your
system is running. If it’s too hot to leave your finger in place,
you definitely have a heat problem.

3. If the heatsink is applied with
thermal tape, try removing it (be sure to get all of the tape off) and
using some thermal grease and/or a better HS and fan. [If you have
a retail Celeron that came with an Intel fan, I’d save this for last since
many, many people have used this fan without problems. Lately though,
I have seen cases reported where changing the HS/fan did help a retail
C300A. It may be that the thermal tape does not provide enough heat
transfer to the heatsink. You could try just replacing the thermal
tape with good old thermal grease and reinstalling the same Intel HS/fan
setup.]

Burn-in

There have been many reports of what is
being called “burn-in effect”. After running the CPU at an elevated
voltage or even at the normal voltage while “exercising” the CPU (cycling
CPU intensive applications), the processor somehow becomes more likely
to run at the desired speed. The time required varies and it doesn’t
always work, but it’s worth a try. If it doesn’t run at 100 mHz bus,
try 75 mHz or 83 mHz for a few days. Leave the computer on for several
days straight. Give it a workout, then try it again at 100 mHz.
[I have found some documentation (see web links below) on this effect in
respect to audio components. The author suggested that it may have
something to do with dopant stabilization and the dielectric properties
fully forming in the tiny, in-circuit semi-conductor junctions, capacitors
and other components.]

Voltage

Sometimes a little extra voltage is all
that’s required to encourage your recalcitrant CPU to straighten up and
fly right. You can adjust the voltage quite easily with the BH6 SoftMenu.
It’s a little harder with the P2B but you can still do it. With the
BH6 you can increase the voltage in small increments. Put your system
through it’s paces after each step. If it still crashes, bump the
voltage a little more. You can fry your CPU by increasing the voltage too
much. Use some caution and common sense here.   If voltage
is your stumbling block, 2.2 volts usually does the trick, though some
have required as much as 2.3 or more. The BH6 BIOS will not let you set
the voltage higher than 2.3 volts without a special procedure (found on
Andy Drake’s site).   All Slot 1 motherboards read the required
CPU voltage through contacts on the SEPP (CPU board). By selectively
blocking certain contacts, you can “tell” the P2B (or any motherboard)
to raise the CPU voltage. Teflon tape is one of the best materials
to use, however, some people have used nail polish (dry 24 hours before
inserting the CPU) or other non-conducting varnish. You can even
cut the trace, but this technique, while effective, is difficult to reverse
(you need to re-solder over the cut).   This procedure is fraught
with pitfalls and, if done incorrectly, can jolt your CPU with 2.6 volts
or more. Since I can’t include pictures with this text document,
I recommend examining a web site that illustrates and details this procedure,
if possible (see the list of web resources at the end of this FAQ).
Your choices are quite limited with this method, but if you have a P2B
or other motherboard, it may be your only option. Notwithstanding
those words of caution, here is the list of pins to cover to get the specified
voltages.

Cover up these pins to attain:

2.2v– A121,A119,B119 (if A119 breaks through,
you get 2.6!)

2.4v– A121,A120,B119 (if A120 breaks
through, you get 2.6!)

2.6v– A121,B119 (not recommended)

2.8v– A121,A119,A120 (don’t even try
it)

Heat production increases when you increase
voltage, so don’t forget about needing more than ordinary cooling
if you need to raise the voltage. One note of encouragement,
there have been many reports of users being able to revert back to the
normal 2.0 volts after a few days and still maintaining stability (see
the section on “burn-in” above).

Drivers and Peripherals

Try different versions of drivers.
Try new video drivers, try old video drivers, try Direct X 5.0 and 6.0.
Remove all your cards except the video card. Disconnect the harddrive
and boot from a floppy. If you have two DIMMS, try with each one
individually. Borrow better, or at least different RAM. Borrow
or use an old video card. Overclocking pushes your whole system to
the edge. There is no predicting what device may be extra sensitive
to slight timing errors, data errors or excess heat. Many of the
new video cards, especially the AGP cards run very, very hot. Be
sure that video chip overheating is not what is keeping you from your desired
speed. [My G200 AGP card has hit 145 F when I forgot to turn on my
extra fans. You can mount an old 486 fan on it’s heatsink for added
insurance.] If you find a card, harddrive or device that’s keeping
you from running at 100 mHz, you’ll need to replace it.

BIOS

Try getting it to run after disabling
the L2 cache in the C300A (L2 should be disabled anyway for the C266 and
C300 which have no L2 cache). Set memory delay settings to higher
values. Be sure the AGP setting (Abit motherboards only) is at 2/3.
Set harddrive mode to PIO 4 or 3. [This is an area I hope to
receive comments on from other experienced Celeron users. How about
it?]

Will
it damage my CPU or other components?

There are a number of things that could
happen to your system through overclocking but, with a little common sense,
you can remove most of the risk. Heat is the primary concern.
The CPU is the system component that is most likely to suffer from excess
heat. Higher clock speeds and/or higher voltage create more heat.
Extreme heat can literally fry the CPU. Keeping the voltage as close
as possible to the default of 2.0 volts and using a quality heatsink and
fan will keep the CPU temperature within reasonable limits.

page 8

Long term effects of higher current produced
by faster clock speeds can have more subtle effects. A process
called electro-migration can slowly erode the microscopic circuits inside
the CPU–causing the traces to spread and the semi-conductor junctions
to break down–until the CPU eventually fails. This is a very slow
process and it takes years. A modern CPU has a design life of 10
to 15 years. While the life of your overclocked CPU may be somewhat
shortened, do you really expect to be using your current 300 mHz processor,
even overclocked to 450 mHz, in even as little as 5 years? By then
we’ll all be overclocking 1000 mHz (giga-Hz) CPU’s.

Since BX motherboards like the BH6 and
P2B are designed for a 100 mHz bus speed, you are not likely to hurt the
mainboard with speeds up to 100 mHz. The other components, however,
can be negatively affected by bus speeds other than the standard speeds
of 66 or 100 mHz. At all other speeds the PCI and AGP clocks are
higher than normal (see the table below). Besides the obvious effects
of increased heat generation, some peripheral devices are especially sensitive
to timing problems when the PCI bus is over clocked. Some harddrives
will trash your data if the PCI bus is clocked too high .

FSB	PCI	FSB Ratio	AGP	FSB Ratio
66 mHz	33 mHz	2 to 1	66 mHz	1 to 1
75 mHz	37.5 mHz	2 to 1	75 mHz	1 to 1
83 mHz	41.5 mHz	2 to 1	83 mHz	1 to 1
100 mHz	33.3 mHz	3 to 1	66 mHz	3 to 2
112 mHz	37.3 mHz	3 to 1	74.6 mHz	3 to 2
124 mHz	41.3 mHz	3 to 1	82.6 mHz	3 to 2
P2B 133 mHz	44.3 mHz	3 to 1	88.6 mHz	3 to 2
BH6 133 mHz	33 mHz	4 to 1	88.6 mHz	3 to 2

Where
can I learn more?

There is an almost inexhaustible supply
of information available on the Internet. Web pages are an invaluable
source of additional FAQ’s, reviews and hardware specifics.
Any of several Newsgroups and Forums can provide feedback and answers to
specific questions. Below I have categorized some of the more popular
and informative resources that deal with overclocking. This FAQ is
only meant to be an introduction to your overclocking experience.
If you still have problems overclocking after following this FAQ, you may
find additional suggestions and ideas at many of the following sites.

Newsgroups

——————

alt.comp.hardware.overclocking

alt.comp.periphs.mainboard.asus

alt.comp.periphs.mainboard.abit

comp.sys.intel

intel.microprocessors.celeron

comp.sys.ibm.pc.hardware.chips

Forums (you need to join Delphi, but it’s
free)

——————

http://www.delphi.com/overclocker

http://forums.delphi.com/m/main.asp?sigdir=overclocking