| -
Home
- Tutorials
- Cd Covers
- Hardware
- Protections
- Email
me
- Links
- Stats Icq:
20042334 |
This page is quite large, please be patient.
|
|
The term 'overclocking' describes the
process of running your CPU at a clock and/or bus speed that the CPU hasn't been specified
for - logically, that speed is usually higher. |
|
|
The tempting idea behind overclocking is to increase system
performance at very little cost. In many cases you only need to change a few settings on
your motherboard to make your system run faster. In other cases you only have to add a few
components (usually for cooling) to achieve the performance increase.
In the past, overclocking was usually nothing more than increasing a CPU's clock speed to
that of the next higher model, e.g. a Pentium 120 to a Pentium 133. Now, with new bus
speeds available on several motherboards, you can change the clock and bus speed of a CPU
to values that don't officially exist. This new way of overclocking is yielding an even
higher performance increase than the classic one. It even gives you the ability to
increase the performance of the fastest model of a particular CPU production line (e.g.
P200 to 250 MHz, PPro 200 to 233 Mhz). |
|
|
Although there are millions of tales of damaged CPUs and other
system components, in most cases overclocking is completely harmless. There are, however,
a few things to take into consideration.
- Your CPU could be damaged by so-called 'electromigration'.
Electromigration takes place on the actual silicon chip of your CPU in areas which operate
at a very high temperature, and can cause permanent damage to the chip. Before you start
to panic, you should first realize a few things. CPUs are designed to run at temperatures
between -25 and 80 degrees Celsius. To give you an idea, 80 degrees Celsius is a
temperature that nobody is able to touch for longer than 1/10 second. I have never come
across a CPU at this temperature. There are plenty of ways to keep the CPU case at less
than 50 degrees Celsius which increases the probability of keeping the chip inside at less
than 80 degrees. Also, electromigration does not immediately damage your chip. It is a
slow process, which more or less shortens the life span of a CPU running at a very high
temperature. A normal CPU is meant to live for about 10 years. However, in ten years
nobody is going to be using a CPU with today's technology. I won't even use my CPU anymore
in 2 months. If you want to be kept free from this electromigration scare, you have to do
as much as possible to cool the CPU. Cooling is the Numero Uno Oncho in overclocking!!!
Never ever forget that!
These terms don't necessarily apply for Cyrix, IBM, and AMD CPUs. Because of the already
high rate of heat production at their original clock rate, you must work extra hard to
keep them cool in overclocked conditions. I've come across several dead Cyrix 6x86 CPUs so
far, so be careful or just let it be!
- Nobody likes system crashes or hangs, but in a professional business
environment, avoiding a system crash or hang can be most crucial. It certainly is a fact
that you are increasing the probability of system faults by overclocking your CPU. But
this is only the probability !! If you have just overclocked your system and the first
thing you do is use it to start writing your dissertation, don't be surprised if a system
crash occurs which causes you to lose all your data. After finishing the overclocking
process you have to put your system through a tough and thorough testing procedure. If the
system passes all the testing, only then can you talk of successful overclocking and feel
confident everything is working well. I'm using Winstone and the BAPCo Suite for
reliability testing. You may not have the BAPCo, but it certainly is worth getting the new
Winstone 97.
- The third debate against overclocking is that your father, brother,
best friend, neighbour, or boss thinks it's immoral. Well, I always enjoyed living in sin,
but if you've got problems with that, read the next chapter.
|
|
|
My dear visitors, I can tell you that I have come across a lot of
animosities from the really straight-up type of law book fellows who have told me how
irresponsible and immoral my website is. There are quite a few different ways to respond
to this topic:
- If you are unwilling or afraid to overclock your CPU, there is no
reason to annoy other people with your opinion. Just because the CPU manufacturers have an
opinion on overclocking does not make it the right one. People who are overclocking their
CPUs appreciate and respect this point of view, but choose to see it differently and are
benefiting from their opinion. As long as there is no sentence against overclocking, we
are not doing anything against the law and hence there is no pleasure in denunciation.
Period!
- It is also fairly questionable if there actually is a difference
between Pentium chips with different official clock speeds. The best example of this is
the P150 and the P166. Isn't it strange, that all P150s are standard voltage chips and
almost all P166s are VRE voltage chips? Doesn't it sound like Intel is using the same chip
in both of them, but it only runs stable enough at 166 MHz with VRE voltage. Intel is
selling the P150 only to satisfy the market and probably gets a good chuckle at the
stupidity of the general public who don't realize this.
- In Europe there are a lot remarked Pentium chips around, as
recently discovered when all over Europe there were several concurrent razzias against
criminal organizations that re-marked thousands of P133s to P166. The proud owners of
these CPUs are convinced they have a real P166, just because it's written on the chip. Do
you think such things do not occur in the US? Hahaha!! I'm wondering how many people own
faked P166s, who would NEVER overclock their CPU! Isn't that a funny thought? Did you know
that Intel isn't interested in marking their chips reliably via a software readout at all?
As long as they sell enough chips, they don't mind the re-marking of Pentiums. They even
tried to avoid the publication of the recent events in Europe and I bet hardly anybody in
the US ever heard about that.
- The main idea behind sensible overclocking is simply to use your
brain, which brings me back to my 'car driving' introduction. If you want to successfully
overclock your system without any loss in reliability, you will have to take care of
proper cooling, do decent testing, and stay within the bounds of common sense. Don't try
to overclock a P100 to 200 Mhz or anything crazy like that. Just use your brain!
- Is smoking immoral?
|
|
|
Well, so far I've only talked about Intel CPUs. These CPUs are in
the majority, but there are also CPUs from Cyrix, IBM and AMD. What about overclocking
these?
Because Cyrix, IBM, and AMD have always lagged behind Intel and its Pentium CPUs in terms
of performance, they have had to release their CPUs at the highest possible
specifications. When Intel produces a Pentium silicon chip, they are not as concerned if
it can run at 166 or 200 Mhz. Because of its design and the high production rate, there
will be enough chips which run at both 166 and 200 MHz. Also Intel was the first
mainstream CPU manufacturer for PCs that released a CPU with the performance of the P200.
There was no rush for Intel to release the P200 any earlier, since there weren't any
competitors.
Cyrix, IBM and AMD have always been one step behind Intel's latest CPU release and
therefore they all have reasons to produce chips which run at the highest possible
performance level. They also have to be competitive, resulting in a lower price, which is
yet another reason to produce high performance CPUs rather than even cheaper slow ones.
This is, in my eyes, the reason why the official clock speeds of Cyrix, IBM and AMD CPUs
are fairly close to their maximum clock speed. All these CPUs run very hot, much hotter
than Pentiums. It is tough enough to cool the CPU properly at its original clock speed,
much less than at overclocked speed.
Although there are now new and 'cooler' 6x86 and K5 chips on the
horizon, I claim the following:
Cyrix, IBM and AMD CPUs are already 'overclocked' at their original
clock speeds. The CPU reliability demands of Intel for specifying the original clock speed
are much much higher than those of Cyrix, IBM or AMD. Hence, we should even thank Intel
for its tough CPU testing procedure because it makes Intel CPUs the easiest chips to
overclock available. You already paid a lower price for your AMD, Cyrix, or IBM CPU than
you would have for a comparable Pentium, so you shouldn't expect even more performance out
of a less expensive chip. This is also the reason why Intel can live quite well with my
website - they would still prefer you buy an Intel CPU and overclock it, instead of buying
a Cyrix, IBM, or AMD CPU.
It's just like tuning engines, there is still a lot of performance
to get out of a big Mercedes or BMW engine, but you can't gain performance by tuning the
already maxed out engine of a 750er Ninja (Kawasaki ZX-7R)? |
|
|
Three things are necessary for overclocking:
- The CPU
- So far, Intel manufactures the CPUs with the highest quality, hence
the probability of a successful overclocking is highest with Intel CPUs.
- In case you want to overclock a Pentium 133, try to avoid the
'SY022'. 33% of the Pentium 133 with this S-spec have been disabled for multiplier
settings of more than x2. In my database there are 22 of 66 entries with the 'SY022',
where the CPU wouldn't support multipliers of x2.5 and x3. A small number of SU073 also
seem to suffer from the same problem.
- Check to make sure your Pentium isn't faked. If you can peel off a
black sticker underneath the CPU, it's definitely a re-marked one. In this case your CPU
is most likely already overclocked.
- The Motherboard
- The quality of the motherboard is crucial for successful
overclocking! Due to the fact that the CPU produces fewer 'clean' signals in overclocked
mode, reflections and other flaws on the bus can cause the system to crash or hang. The
reverse situation is also true - in overclocked mode the CPU is more sensitive to unstable
signals from the bus and will crash if the motherboard can't deliver clean signals. Always
go for a branded motherboard!
- You will have to decide if you want to go for a higher bus speed or
if you will stick to a maximum of 66 MHz. Motherboards with 75 MHz bus speed support are
hard to find and motherboards with 83 MHz bus speed support are even more rare. Refer to
my The 75-83 MHz Bus Speed
Project to find out which boards support these higher bus speeds. Read
in The Bus Speed Guide
why busspeed is so important!
- The board should obviously support a wide range of CPU supply
voltages. Minimum are 3.3 and 3.45 V, for STD and VRE voltage. If you want to use P55C, M2
(the new M1/6x86), or the new K5/K6 CPUs, you will need support for 'split voltage'. This
means that the core of the CPU requires a lower supply voltage than the I/O ports of the
CPU. The latest boards all support 2.5 up to 2.9 V in 0.1 Volt steps. If the board offers
you an even higher voltage than 3.45 as well, you should be happy, because this might be
the last trick to get your CPU successfully overclocked.
- The RAM
- This topic is new on this page, but it is very important indeed. You
will have to consider decent RAM if you want to run your system at bus speeds of more than
66 MHz. If you want to run an HX board, such as the Asus P/I-P55T2P4 at 83 MHz bus speed,
you will require high-end EDO. I've experienced myself, that the marking of the RAM is
less important than it's brand. My official 45ns EDO is not as stable as Fredi's 60ns EDO
of a good brand. I will try to find out the best brands of RAM, but so far I know Siemens
(well, after all one German company is in the computer industry, hehe) and naturally
Micron to be of outstanding quality. Be careful, however, that you don't get second-rate
chips from the manufacturers being sold in some stores. These chips still say Siemens,
Micron, or whatever on them, but their quality won't live up to your expectations.
In the case of high bus speeds always go for SDRAM if you can. SDRAM relieves a lot of the
worries of running at 75 or especially 83 MHz, and runs flawlessly in any case.
- The Cooling
- I can't proclaim it often enough, the cooling of the CPU is extremely
important ! If you have been able to boot your system with an overclocked CPU but it
crashes within the first minutes, it's most likely due to insufficient cooling of your
processor. Don't think the average small heat sinks with their small fans designed for a
Pentium are able to do this job properly! Their job is only to keep a normally clocked CPU
cooler in case you have very hot surroundings (e.g. SCSI or Video cards, which can get
very hot as well). They are not designed to save your overclocked system from crashes due
to overheating. This doesn't mean you always have to have better cooling. If you've got a
new SSS CPU, using the 0.35µm die, it just won't get that hot. If your CPU is of the old
0.6µm die size type, however, you will require decent cooling. To accomplish this, you
can use heat sinks, fans, or both, peltiers, or peltiers with fans. I personally don't
believe in peltiers. Peltiers are elements which transport heat using an electrochemical
method from one side of the element to the other, consuming energy. You will still need a
heat sink to dissipate the heat from the non-CPU side of the peltier and most likely will
also require a fan. My opinion is that you should go for a heat sink, and most importantly
THINK BIG !! If a big heat sink still can't do the job, add a fan on top of it. My
overclocked P133 -> 180 has a temperature of about 30°C, which is hand warm or less.
If you achieve this cooling effect, you can be sure that any crashes which do occur are
not a result of overheating.
So how to get a decent heat sink ? Don't even think of finding anything in a normal
computer shop. You'll find professional heat sinks only in professional shops which sell
electronic equipment such as transistors, resistors, chips, etc. (e.g. Hobby Electronic
Stores). You can tell how good a heat sink is by looking at the K/W value. K/W means
degree Kelvin per Watt of power dissipation . K/W tells how hot the heat sink gets per
each Watt of heating power of the device it's meant to cool. If you were able to follow
that, you will understand that the smaller the value, the better the heat sink. If you can
get a heat sink which has a value below 1K/W, you've found a good one. You'll need to make
the surface of the heat sink that will attach to the top of the CPU match the size of your
CPU (maybe the electronic shop will cut it for you, otherwise you'll have to do some
sawing and grinding). Be careful that this surface stays completely flat,
so that there are no gaps between the heat sink and the CPU surface. Finally, you only
need to affix the heat sink to the CPU which is best done with some thermal
compound (also available in every electronic shop). You can also use super glue,
but it should be applied very sparingly with just enough to attach the heat sink. Do
realize that you might not be able to remove the CPU from the heat sink if the super glue
is good stuff. If required, attach a good (powerful + quiet) fan to the top of the heat
sink (how, I will leave up to your imagination).
I will try to find a heat sink manufacturer that produces large, cool heat sinks for our
overclocking community. I'll keep you posted about that.
|
|
|
First and foremost, we want to improve overall system performance.
Also, we want the system to be just as stable as it was before it was overclocked.
Finally, we want to keep our CPU alive.
As you may have learned from my bus speed guide, the best way to
improve system performance is to increase the bus speed. If you can't do that, either
because your motherboard doesn't support higher bus speeds or your RAM or your PCI devices
aren't up to it, you can change the multiplier instead. Don't expect much gain in
performance, however, if you increase the multiplier but you decrease the bus speed!!! For
example, changing from 166 @ 2.5x66 MHz to 180 @ 3x60 MHz will actually decrease your
overall performance. The same rule applies to changing from 133 @ 2x66 to 150 @ 3x50.
These types of changes will not make your system any faster!!
This is some touchy news for 6x86 users, who should really only
overclock their CPUs to a slightly higher speed than the original. The 6x86 only has
multiplier options for x2 and x3. Don't let yourself be told otherwise!! Of course you can
try jumpering the board to all of the different Intel Pentium settings, but it won't make
a difference for the 6x86 CPU. |
|
- Changing the bus Speed
To understand how you can overclock a Pentium, Pentium Pro, 6x86, or K5 CPU, it helps
to realize that the internal clock in these CPUs runs at a different speed than the
external clock or bus speed. The external clock is the speed at which the cache and the
main memory run and when divided by two yields the speed of the PCI bus. There are only
three different official bus speeds used by the Intel Pentium, Pentium Pro, and the AMD K5
CPUs - 50, 60 and 66 Mhz. The 6x86 uses five bus speeds: 50 MHz, 55 MHz, 60 MHz, 66 MHz,
and 75 MHz. There are also new boards available which support the unofficial bus speed of
83 MHz.
To change the bus speed, look in your motherboard manual for something like 'CPU External
(BUS) Frequency Selection' - these are the jumpers you will have to change. If you are
lucky and happen to have a motherboard with the new SoftMenuTM technology, you can
change these settings in the BIOS setup menu right from the comfort of your chair. You
don't even have to open the case.
Always go slowly and increase the bus speed one step at a time (e.g. go from 60 Mhz to 66,
not 60 Mhz to 75). This is the most successful way to overclock. Using this method, almost
every P150 CPU runs at 166 MHz and most all 6x86 P150+ CPUs run at a P166+ level or 133
Mhz.
- Changing the Multiplier
The internal clock is controlled by an internal clock multiplier in each CPU which is
programmed via CPU pins. Intel Pentium CPUs support the following multipliers: x1.5, x2,
x2.5 and x3. Intel Pentium Pro CPUs support x2.5, x3, x3.5, x4. 6x86 CPUs so far only
support x2 and x3, but the upcoming M2 will support x2, x2.5, x3, x3.5. The K5 is kind of
a difficult fellow here, because it doesn't seem to be affected by the external settings
of its multiplier. So far it only uses the x1.5 multiplier for each of the PR75, PR90,
PR100, PR120, PR133 CPUs. The new PR150 and PR166 K5 CPUs will use the x2 multiplier, but
it seems fairly likely that you won't be able to change it - although I have not been able
to verify this.
To change this setting, find something like 'CPU to BUS Frequency Ratio Selection' in your
motherboard manual. There are usually two jumpers used to change these settings. Again,
you can do all of this in the BIOS setup menu if you have a SoftMenuTM motherboard such as the
new Abit motherboards.
- Changing the CPU Supply Voltage
You might not like it, but this is something which is often required to make the CPU run
more reliably.
First of all, I'd like to stress that the Intel Pentium and Pentium Pro CPUs can
run at a supply voltage of up to 4.6 V. This, of course, requires serious cooling because
the chip is producing a lot more heat than usual. I have tried this with my own P166 and
the chip is still doing fine. It didn't really help my problem either, however, since it
was the memory that would not run reliably at 208 MHz, not the CPU.
Often the change from STD to VRE voltage is the whole trick to successful overclocking.
This is due simply to a bigger voltage difference between the digital HIGH and LOW
conditions, which results in 'cleaner' signals for the CPU and other motherboard devices.
If you can't run your CPU reliably at one particular clock speed, it's always worth
considering changing to a higher supply voltage. The silicon of STD and VRE CPUs is
identical, so you won't damage your STD voltage CPU with VRE voltage, the CPU will only
run a little hotter. The Abit IT5 motherboards offer a voltage which is even higher than
VRE, of 3.6 V in their SoftMenuTM
BIOS CPU setup. I'm running my CPU at this voltage and it runs completely stable at 205
MHz @ 2.5 x 68, even as I type this document.
|
|
|
Have a look at the The 75-83 MHz Bus Speed Project
To be honest, the easiest and most sensible answer to that is: TRY
IT OUT !!! Switch the multiplier to the lowest setting and then put the bus speed jumpers
in all the different configurations. For non mathematical people, there are 2 to the power
of the number of jumpers configurations. 4 configurations for two jumpers, 8
configurations for 3 jumpers, and so on. Just boot up to the DOS prompt and run ctcm, or a
similar program which will tell you the CPU speed. You then only have to divide it by the
multiplier setting to find your bus speed. Again, if your motherboard uses SoftMenuTM BIOS, you only have to
look in your BIOS setup menu to find all of the different bus speed settings from which
you can choose.
There is also, however, a different approach:
- Motherboards with the PLL chip PLL52C59-14 can run
at up to 75 MHz and they also support the 'turbo frequency' feature, which
increases the bus speed by 2.5% (officially approved by Intel's CPU specifications).
- For 75 MHz
Pin 8 via 2.2 k Ohm to 0 V ('0 V' means 'ground' and NOT 'disconnected'!)
Pin 12 via 10 k Ohm to 5 V
Pin 13 via 10 k Ohm to 5 V
- For 68 MHz, the 'turbo frequency' for 66 MHz
Pin 8 via 2.2 k Ohm to 0 V
Pin 12 via 10 k Ohm to 0 V
Pin 13 via 10 k Ohm to 5 V
- For 61.5 MHz, the 'turbo frequency' for 60 MHz
Pin 8 via 2.2 k Ohm to 0 V
Pin 12 via 10 k Ohm to 5 V
Pin 13 via 10 k Ohm to 0 V
Measured on the Abit boards IT5H, IT5V, PR5, which all use the
PLL52C59-14. The PR5 also comes in a version with the PLL52C61-01, however, the setting
below seems to set the board at 61.5 MHz instead of 83 MHz.
- Motherboards with the PLL chip PLL52C61-01 can run at 83 MHz bus
speed as well and also theoretically support the 'turbo frequency'. There obviously is,
however, a way of configuring this chip (circuitry), so that it would not run at 83 MHz.
I'm working on that.
These are the conditions:
- For 83 MHz or 61.5 MHz 'turbo frequency' for 60 MHz.
Unfortunately this depends on the circuitry on the motherboards.
Pin 5 via 10 k Ohm to 0 V ('0 V' means 'ground' and NOT 'disconnected'!)
Pin 12 via 10 k Ohm to 5 V
Pin 13 via 10 k Ohm to 5 V
- For 75 MHz
Pin 5 via 10 k Ohm to 5 V
Pin 12 via 10 k Ohm to 0 V
Pin 13 via 10 k Ohm to 5 V
- For 68 MHz, the 'turbo frequency' for 66 MHz
Pin 5 via 10 k Ohm to 5 V
Pin 12 via 10 k Ohm to 5 V
Pin 13 via 10 k Ohm to 5 V
I have measured this on the Asus P/I-P55T2P4 rev. 3 board and was
able to verify the exact same settings on the FIC PA-2006 board. The FKI SL586VT II or
Magic Pro MP-586VIP board also use this chip, but you can't get to 83 MHz. Instead you get
the pathetic speed of 61.5 MHz.
In terms of jumpers, this means that you will have to find out which of the three jumpers
is connected to the particular pins. It's circuited via the pull up/down resistor of 10 k
Ohm. In case you only have jumpers with ON/OFF positions instead of 1-2/2-3 positions, the
ON condition is the condition for 0 V, and the OFF or open condition is for 5 V.
This information is only for real freaks who
want to get their boards to 75 or maybe even 83 MHz by all means. This way, you might be
able to use the higher bus speeds on boards that only have 4 bus speeds, as long as this
board uses one of these PLL chips. You do the necessary soldering at your own risk and
should only attempt this if you know exactly what you are doing!!!
There are several motherboards that use one of the two PLL chips,
depending on what was available when the motherboard was assembled. The Abit PR5 is such a
fellow (of course I was unlucky enough to receive a board with the PLL52C59-14) and it
seems the shuttle HOT 557 is another. There are a few reports of 83 MHz bus speeds with
these boards, but most simply can't run at this faster bus speed. This is a real pity,
because both of these two boards can use SDRAM - the best RAM for the 83 MHz bus speed. |
|
|
Using this higher bus speeds includes some important restrictions
which you should be aware of.
- The PCI bus runs at 37.5 or even 41.6 MHz. This can lead to several
problems with PCI devices. Typical trouble makers are SCSI controllers, some video cards,
and network cards. SCSI controllers and network cards often refuse to work at the faster
speed, but some video boards just get much hotter than usual. If you find a way to cool
these video cards, you shouldn't have any trouble. My Diamond Stealth 64 Video VRAM isn't
affected at all by those higher bus speeds. I hope the 75/83 MHz bus speed survey will
help us find out which PCI devices run at higher PCI bus speeds.
- The speed of the EIDE interface included with the chipset is not only
determined by the PIO or DMA modes, but is also highly dependent on the PCI clock. This is
one reason why the EIDE interface is always slower in systems with 60 MHz bus speed or
less. This is also valid in the other direction, meaning your interface will be faster
when you are running at 75 or 83 MHz bus speeds than at 66 MHz. At first this sounds fine,
but often either the interface or, in more cases, the hard disk isn't up to the faster bus
speeds. My HDDs work fine at 75 MHz bus speed, but at 83 MHz I have to reduce the PIO down
to 2. The same is valid for EIDE CD-ROM drives. This could be the cause if you are running
into strange lock-ups in windows.
- The Asus P/I-P55T2P4 is one example of a board that does not allow
you to adjust the ISA bus speed. It seems to be a fixed divider from the PCI clock. This
can cause sound cards to run into trouble if they don't like the higher ISA bus speed. I
haven't come across this problem myself yet, but I've heard of one fellow whose AWE 32
produced strange sounds when running at a faster speed. If you run into this, increase the
ISA wait states in the BIOS setup to try to remedy the problem.
- Let me say again that the RAM type and quality is of great
importance. Most 60ns EDO will run fine at 75 MHz bus speed, but for 83 MHz you'll need
high-end EDO or SDRAM (as long as the motherboard supports it).
|
|
|
Intel's Pentium processor is the most successful CPU ever. Because
of that it has to come first here. The nice thing about this CPU is that it is also the
most overclockable CPU ever. This can be attributed to Intel's increased quality demands
put in place after the floating-point flaw disaster. The new Pentium MMX is just as
overclockable as the Pentium Classic, maybe even better. This CPU normally runs at 2.8V.
Most motherboards that offer this voltage also offer 2.9 or 2.93V. This is only 0.1 V more
than 2.8 V and probably just the right thing for overclock-unwilling Pentium MMX CPUs. My
own Pentium MMX 200 runs fine with 2.8 V at 208/83 and 225/75 MHz. For 250/83 I have to
increase the voltage to 2.9 V and everything works fine. BTW mine's a 'SL23S'.
Please let me put a short note about the so much feared 'overclock
protection' from Intel. So far the only thing Intel has done once was disabling
the CPU pin 'BF1', which is responsible for the multiplier settings x2.5 and x3. My survey
could reveal, that the affected CPUs are about 50% of the 'SY022' and maybe 10% of the
'SU073', both P133 CPUs. You still can run those P133 at 166 MHz, but only with 83 MHz bus
speed.
Now since the message about thousands of remarked or counterfeit Pentium CPUs has
eventually reached the US, Intel has to at least make some announcements to this
counterfeit and overclock protection again, which they did. However already 2 years back,
Intel was claiming to soon ship their chips with a overclock protection, which never took
place. I still doubt that Intel will invest any money in producing this feature at all.
Before they do that, they should and will have to do something against the remarking
opportunity of their chips. Hence I wouldn't worry at all. Intel will probably never avoid
that their chips can be clocked higher, but they may push the motherboard industry to
produce motherboards that don't offer any overclocking abilities.
If you've read all the paragraphs above, you'll remember the
following things:
- Always try to increase the bus speed first if you can.
- Don't increase the multiplier while decreasing the bus speed - you
won't gain anything.
- Try higher voltages and don't be afraid of it!
- Avoid the P133 'SY022' and 'SU073' if you can.
- Don't buy a remarked Pentium - there are loads of them around!
The most overclockable Pentium CPUs:
- P150 is the absolute winner - it's most likely
nothing else than a P166 in disguise!
- P166 Classic & MMX (and hence the P150) is super
for 187.5 @ 2.5 x 75 MHz and in most term runs fine at 200 @ 3 x 66 MHz.
- P133 great for 150 @ 2 x 75 MHz or 166 @ 2 x 83 MHz
- forget about higher multiplier settings with that CPU.
- P75 most of them run at least flawlessly at 90 @ 1.5
x 60 MHz, many of them at 100 @ 1.5 x 66 MHz.
- P200 Classic & MMX superb at 208 @ 2.5 x 83 MHz,
great at 225 @ 3 x 75 MHz, amazing at 250 @ 83 MHz - the CPU for the real speed freaks!
Where do you want to overclock today?
| Pentium at |
1st choice |
2nd choice |
3rd choice |
4th choice |
| 75 MHz |
112.5 MHz @ 1.5 x 75 MHz |
100 MHz @ 1.5 x 66 MHz |
90 MHz @ 1.5 x 60 MHz |
83 MHz @ 1.5 x 55 MHz |
| 90 MHz |
125 MHz @ 1.5 x 83 MHz |
112.5 MHz @ 1.5 x 75 MHz |
100 MHz @ 1.5 x 66 MHz |
|
| 100 MHz |
125 MHz @ 1.5 x 83 MHz |
112.5 MHz @ 1.5 x 75 MHz |
|
|
| 120 MHz |
125 MHz @ 1.5 x 83 MHz |
133 MHz @ 2 x 66 MHz |
112.5 MHz @ 1.5 x 75 MHz |
|
| 133 MHz |
166 MHz @ 2 x 83 MHz |
150 MHz @ 2 x 75 MHz |
166 MHz @ 2.5 x 66 MHz |
|
| 150 MHz |
166 MHz @ 2 x 83 MHz |
187.5 MHz @ 2.5 x 75 MHz |
200 MHz @ 3 x 66 MHz |
150 MHz @ 2 x 75 MHz |
| 166 MHz |
208 Mhz @ 2.5 x 83 Mhz |
166 MHz @ 2 x 83 MHz |
187.5 MHz @ 2.5 x 75 MHz |
200 MHz @ 3 x 66 MHz |
| 166 MHz MMX |
266 Mhz @ 3.5 x 75 Mhz |
250 MHz @ 3 x 83 MHz |
225 MHz @ 3 x 75 MHz |
208 MHz @ 2.5 x 83 MHz |
| 200 MHz |
250 MHz @ 3 x 83 MHz |
225 MHz @ 3 x 75 MHz |
208 MHz @ 2.5 x 83 MHz |
|
| 200 MHz MMX |
290 MHz @ 3.5 x 83 MHz |
266 MHz @ 3.5 x 75 MHz |
250 MHz @ 3 x 83 MHz |
225 MHz @ 3 x 75 MHz |
| 233 MHz MMX |
290 MHz @ 3.5 x 83 MHz |
266 MHz @ 3.5 x 75 MHz |
250 MHz @ 3 x 83 MHz |
|
To get a P166 running at 208 MHz is a tough thing and requires high
quality hardware - I hope I'll succeed with SDRAM and the R-534, if I ever should receive
it.
There is no excuse for running a P150 at 2.5 x 60 MHz as intended -
this CPU definitely runs at least at 166 @ 2.5 x 66 MHz or 150 @ 2 x 75 MHz, which is even
better!!
Read Important
News for Overclockers! |
|
|
The Intel Pentium Pro still doesn't seem to be a popular CPU for
overclocking. This is really not fair to the powerful PPro since this CPU works just as
well as the Pentium for overclocking.
The first letdown seems to be the lack of any available motherboard
for this CPU that runs at a 75 or 83 MHz bus speed. Hence you'll have to stick to the good
old 50, 60 and 66 MHz settings. For the PPro the same methods are valid as for the Pentium
- first try improving your bus speed.
A PPro150 or a PPro 180 doesn't deserve to run at the puny bus speed
of 50 or 60 MHz. They are screaming to be overclocked to the 66 MHz bus speed and will
reward you with a considerable performance increase and a high reliability.
| Pentium Pro at |
1st choice |
2nd choice |
| 150 MHz |
166 MHz @ 2.5 x 66 MHz |
|
| 180 MHz |
233 MHz @ 3.5 x 66 MHz |
200 MHz @ 3 x 66 MHz |
| 200 MHz |
266 MHz @ 4 x 66 MHz |
233 MHz @ 3.5 x 66 MHz |
I admit, that it will be tough, to get a PPro180 to 233 MHz and a
PPro 200 to 266 MHz, but it's worth a try. The second choices however work out in most of
the cases and the PPro 150 runs great at 166 MHz. The problem with the PPro is the
difficulty to adjust things. Here the SoftMenuTM
technology is a let down, because it wouldn't let you choose 233 or 266 MHz. You can't
adjust the voltage either, because the motherboards adjusts it by itself or more by a VRM
module. Nevertheless the PPro is a worthy overclockable CPU! |
|
|
I have mixed feelings about writing this paragraph. One one hand, I
promised you information on this subject ages ago, but on the other hand, much of the
information regarding the 6x86 M1 and the upcoming M2 is about to change. Nevertheless, I
will refer to the good old, well known 6x86 and it's later stepping versions.
Due to the massive heat production of the older versions (steppings
of less than 2.7) and the overly high heat production of even the latest versions, this
CPU is not as flexible as the Pentium for overclocking. The first important thing to
remember is that you are indeed able to kill your 6x86 with overclocking. I've never heard
of an Intel CPU with this problem, not even a 486, but I have heard several stories of
fried 6x86 CPUs. Hence I DO NOT recommend you overclock this CPU at all. The only reason I
don't refrain from writing about the 6x86 is that I've been promising this information for
a long time.
Overclocking the 6x86 is quite a bit more restrictive than
overclocking a Pentium. This is mainly due to its heat production but also can be
attributed to it's limited multiplier settings of x2 and x3. You can more or less forget
about the x3 multiplier because the only scenario where it makes sense to use it is at 3 x
50 Mhz. Due to the pathetically low bus speed, this is completely uninteresting in the way
of performance. Hence this only leaves the x2 multiplier.
If you really want to overclock your 6x86, think small! Think in
small steps!! It is worth considering just one step up. This means P120+ (100 MHz) to
P133+ (110 MHz), P133+ (110 MHz) to P150+ (120 MHz) and P150+ (120 MHz) to P166+ (133
MHz). The step from the P166+ (133 MHz) to P200+ (150 MHz) seems to be too big and has a
fairly low success rate with quite a high risk of losing the CPU.
You'll achieve the highest success rate with 2.7 or 3.7 stepping
6x86 CPUs because they run more stable and produce less heat.
Cooling is paramount for the overclocking of a 6x86, so don't even
think about overclocking this CPU without a HUGE heat sink or a power peltier.
I hope all this will all change with the release of the split
voltage 6x86. This chip will be run at 2.8 V and should result in much less heat
production. Maybe the 6x86 will suddenly turn into a really great overclocking CPU. |
|
|
The K5 has been late in arriving, but is proving to be very powerful
indeed. The latest performance figures show fairly nice results compared to the Pentium
and the pricing of the K5 is just wonderful! 'How about the overclocking then?' I ask,
'after all AMD is the manufacturer of the most overclocked 486 CPUs in history'. My first
overclocked CPU was an AMD 486/40 (to 50 MHz) then an AMD 486/100 (to 120 MHz) and
everybody knows about the beautiful and cheap AMD 5x86-133, which just runs great at 160
MHz and still is faster than many P100 systems.
In my initial experience, the older PR75, PR90 and PR100 K5 CPUs are
not great fellows for overclocking. Most of the time, the system would remain dead after
moving up only one step and these versions share the same serious heat problem with the
6x86.
The PR120 and PR133, however, seem to be much better candidates.
These two guys run at the same external and internal speed as the PR90 and PR100!! This
proves that the chip inside has been improved and altered. Obviously these CPUs don't
produce as much heat as the first incarnations, which is one of the first requirements for
overclocking. The new PR150 and PR166 (at 120 and 133 MHz) seem to be of the same design.
Unfortunately I haven't got enough information to tell you anything
decent about successful overclocking of the K5, but I hope that'll change. My Overclocking
Survey has already got quite a lot of K5 entries. |
|
| 1 |
Turn off computer, open it up, get your motherboard manual |
| 2 |
Check your CPU markings on top and bottom of CPU, write them down
and put your CPU back in again |
| 3 |
Check the current clock speed and multiplier jumper settings on
your motherboard, compare them with your manual, write them down |
| 4 |
Check the supply voltage jumper settings on your motherboard,
compare them with manual and your CPU marking, write it down |
| 5 |
Have you thought of a decent cooling for your CPU ? Apply it ! |
| 6 |
Change the jumper settings for clock speed and/or multiplier
according to your manual |
| 7 |
Check if everything is ok, no jumper forgotten or put wrongly |
| 8 |
Start computer |
| 9 |
Does it reach BIOS setup ? |
| 10 |
If yes, go to 13 |
| 11 |
Turn off computer and change jumper to higher supply voltage
according to manual, if possible |
| 12 |
If you still shouldn't reach BIOS setup, forget about overclocking
to this speed |
| 13 |
Change BIOS setup settings to the right values |
| 14 |
Does it reach full working operation system ? |
| 15 |
If yes, start testing (I recommend Winstone or the BAPCo Suite.
Don't take this job too easily! It's better to occur crashes or lock ups now, than coming
across them when it counts!) |
| 16 |
If no, you should try 11 or check your cooling, you also can try
some more conservative memory timings in the BIOS setup. This means increasing the wait
states or the read/write cycles; but don't forget to check later if you gained speed by
trying some benchmarks, cause there's no point in overclocking if your memory access is
getting slower. |
| 17 |
If everything works well - congratulations, if not, try 11, check
cooling |
| 18 |
Don't change supply voltage unless you have to. It only makes the
chip hotter. |
| 19 |
Never forget: cooling is of main importance ! |
One additional thing: Windows 95 is much more picky
about overclocking. Therefore it might be that you can't run Windows 95 properly with
your o/c CPU though it was working just fine with DOS and Windows 3.11. You'll either have
to improve the cooling or you'll have to forget about overclocking your CPU with Windows
95.
|
|
|