Re: [OT] Water or refrigeration PC cooling

["Musashi" <simonr@xxxxxxx>]

Having being involved in the OCing scene for 5 - 6 years now, my humble=
opinion is neither. The reason being these:
There are 2 ways to go with overclocking - Pure Multipliers or FSB.
Although multiplier clocking will give you improve the speed at which your
=
processor runs, any clocker will tell you the FSB is where the real gains
a=
re to be made. Now if you want to push the fsb of your processor there are
=
several things to take into account:
First and foremost, is whether the particular core of the proccessor, AMD
o=
r Intel has already been pushed up by the manufacturers (266 -333 as this
w=
ill limit your potential to clock it harder)
Secondly is the memory that you intend to run it with suitable or stable
en=
ough to handle those faster FSBs (Crucial used to be as it was very
comfort=
able running quite hot, not so much these days)
Thirdly, if you are clocking the FSB this will impact on the other
componen=
ts within your PC, the PCI and ISA bus both run at multiples of 33mhz, eg
p=
c100 (x3) pc133 (x4) etc. The further you move from that 33 devisor on
your=
FSB speed the more unstable your PCI components are likey to become (to th=
e point that if you increase your FSB for arguements sake from 100 to 117,
=
and your system wont post you are smack bang in the middle of a devisor
and=
stressing your ancilliary components sometimes to the point of failiure) a=
nd I have blown 200 and 300 graphics and sound cards doing this.
All things considered you decide to go down the overclocking route, you
buy=
the most expensive motherboard you can buy, a new vapochill system, Radeon=
8700 fire it up, cool the core down to -90oc and it all starts working, bu=
t the system becomes unstable - why because once you start overclocking
the=
processor it needs more power (core voltage) and unless you have a nice sh=
iny PSU with a very clean +5v line then, you will find problems, so
120quid=
for a new sparkle psu (these are reconed to have the most stable 5 volt li=
ne).
You finally finish that rig =A32000 should just about do it, your
performan=
ce gain is likely to be ??
15 - 20% if you are lucky !!

In the good old days of clocking my 1 gig AXIA ran day in day out at 1500
(=
well just under) a 50% increase in speed for the cost of a bigger heat
sink=
whereas the cost of an xp1500 would have been over 300 quid. Today even wi=
th a vapochill i couldnt get 4500 out of a barton 3000 and even if I could
=
the rest of the components in my system probably couldnt keep up.

The rate of processing power is growing at a phenomenal rate these days,
an=
d 2 months after a new processor comes out its successor is here. A retail
=
Barton xp3000 is =A3273 a barton xp2800 is =A3187an xp2500 is =A3109, in 2
=
months that 3000 will be the same price as a 2800 and in a another 2
months=
probably the same price as a 2500. Even with using my GIS at work on my p4=
workstation my bottleneck isnt my processor (although this gets hit pretty=
hard) it is my hard Drives as my cheapskate boss specced ata133 drives and=
not SCSI 160s. The point that I am trying to make is that gone are the day=
s when bolstering your processor by overclocking it are going to give you
t=
hat shiny PC that you always dreamed of, the rest of the architecture in
th=
ere is so antiquated that it is struggling to keep up. We lightened the
loa=
d on the graphics by putting GPUs and huge amounts of ddr memory on
graphic=
s cards, we put expensive synthesizer chips on our sound cards and we made
=
our memory run twice as fast but we are still stuck with 33mhz bus speeds
t=
hat date back to the original IBM AT architecture and no matter how cool a
=
vapochill rig runs it cannot overcome this.

My advice, the relative cost of processing power is cheap today, and that
s=
hiny freon cooled sytem may give you the edge for a month or two until a
ne=
w processor comes out that is faster than your already stressed rig.
Better=
spend your money where you will see serious advantages, serial 150 hard dr=
ives are just round the corner, but scsi 160 is here now (and it is VERY
FA=
ST). Buy the fastest memory you can pc3700 with a low CAS latency (2.5) or
=
even better buy it as matched pairs as something like pc3500 CAS2.
If you want to do AV work buy a dual processor rig (much better than a
Vapo=
chill rig and still cheaper) with at least a gig of matched RAM. I dont
kno=
w what you want to do with your new toy but unless all you want to do is
ch=
ase your tail trying to break the fastest 3d mark (which is what most
clock=
ers do these days) I suggest getting someone to spec and build a rig for
yo=
ur application.

If you want to talk to someone about the futility of clocking and how to
br=
eak numerous world records (on air cooled sysytems !!!) give Andy Gibson a
=
bell at Overclockers.co.uk (he sold his vapochill and still trounces most
o=
f the freon cooled clockers).

Sorry for my rant,=20

Simon
----- Original Message -----=20
From: Paul=20
To: ukha_d@xxxxxxx=20
Sent: Wednesday, April 30, 2003 1:16 PM
Subject: [ukha_d] [OT] Water or refrigeration PC cooling


Sorry for another OT - I'm thinking about upgrading one of my PCs to the
latest, fastest, shiniest and (most expensive) PC around and was
thinking about water or refrigeration cooling. Anyone have experience
with this?



Paul.





[Non-text portions of this message have been removed]