What are the baest manufacturers/models of PSU?

[kony]

Hello again. I have a similar set up to the poster above, on a KT333 chipset
XP2500, currently two PATA drives in a raid, plus a DVD-ROM. I couldn't get
the Fortron/Sparkle 400W PSU in the end. Instead I got an Antec Truepower
430W. How far can I push this? I too have two case fans and a GeForce 4 MX.
My RAM CAS is stable at 2. I plan to slower increase the FSB above 166, and
later get a much better graphics card. And I must get round to adding
another PATA drive to copy the RAID drives. What will that do to my power
consumption?

Both are good power supplies... although I mentioned the Sparkle as a
good value the Truepower should be better for your particular
configuration due to how close the CPU is situated to the power
supply.

Adding another drive, upgrading the video card, and oveclocking to the
highest level possible still should not exceed the capacity of that
power supply. You will reach the limits of the motherboard, memory or
CPU long before that happens. All the changes proposed might
increase the total consumption by 30-100W (as a rough guess, we don't
know yet how far you'll overclock or which video card), but as your
system sits right now it would run from a good name-brand 250W PSU, so
you have a lot of breathing room as far as power is concerned.

The larger issue is cooling... not knowing your case or actual
effective airflow I can only suggest monitoring temperatures after
each significant upgrade or change. Two fans can handle what you're
suggesting with a good case and minimal obstructions to airflow, like
no stamped-in-metal fan guards.


Dave

 

[larrymoencurly]

<dsads> wrote in message news:[email protected]...

http://www.pcpowercooling.com/products/power_supplies/index.htm

Are they really worth the extra cost, compared to Fortron/Sparkle
PSUs, which are only $20-80 from www.directron.com or www.newegg.com ?
Because in every aspect that really matters (i.e., not trivial
bragging poitns), they're just as good, and they're cheap enough that
a person can buy a spare and still have change left over from what a
PCP&C would cost. I could understand paying extra for a PCP&C PSU if
it would shut down in case of overheating, but such a feature isn't
available, not even as a special-order option, according to PCP&C
e-mail.

 

[dummy]

I like to use Antec Power Supplies.


Additional Info:
Many things can cause a computer to crash. Sometimes it is
intermittent power and often it is just bad memory. This can be
caused by overheating, bad drivers, corrupted files, memory leaks in
programs, Sound Cards, Video Cards, Network cards, any other cards,
IRQ Conflicts. MS XP and Win2k are both hard on the hardware and are
affected by inadequate hardware. Normal procedure would be to remove
all IO cards except the video card and see if the comptuer still
crashes. Then add other cards you use to see what is causing the
computer to creash. I have seen a specific program cause the crashes
because it was causing a memory leak.

Normally with a power supply there is an allowed 5% variance. There
will be some variance on all power supplies. This is to be expected.

 

[Nicolas The Great]

This means that if the 12 V rail is showing a 12.5 V then it is
actually too much power there.

On the 3V &5 V rails I've noticed around 2.8V & 4.7V respectively.

Is this fine?

thanks

I got an Eneramx 350W and my lines look like this
core 1.73
3.3 goes 3.26 to 3.28 never above or below.
+5V stays at 5.00
12V stays at 11.64

Nick

 

[Lane Lewis]

"Gary" <[email protected]> wrote in message news:[email protected]...


Are they really worth the extra cost, compared to Fortron/Sparkle
PSUs, which are only $20-80 from www.directron.com or www.newegg.com ?
Because in every aspect that really matters (i.e., not trivial
bragging poitns), they're just as good, and they're cheap enough that
a person can buy a spare and still have change left over from what a
PCP&C would cost. I could understand paying extra for a PCP&C PSU if
it would shut down in case of overheating, but such a feature isn't
available, not even as a special-order option, according to PCP&C
e-mail.

Note that 20-80 is probably OEM (brown bag) not retail, but for home use
the Fortron/Sparkle are fine. Zalman's doesn't sell OEM and will come with
Manual and retail box and full guarantee which is one of the reasons I
recommend Zalman over the other two. If in a server situation 6-8 hard
drives go with PCP&C for the additional support.

Lane

 

[S.Boardman]

snip!

Both are good power supplies... although I mentioned the Sparkle as a
good value the Truepower should be better for your particular
configuration due to how close the CPU is situated to the power
supply.

I remembered you'd said this, which is why I went for the Antec Although
I did have to bend the bit of metal the PSU sits on, as the fan grill got in
the way :-(

Adding another drive, upgrading the video card, and oveclocking to the
highest level possible still should not exceed the capacity of that
power supply. You will reach the limits of the motherboard, memory or
CPU long before that happens. All the changes proposed might
increase the total consumption by 30-100W (as a rough guess, we don't
know yet how far you'll overclock or which video card), but as your
system sits right now it would run from a good name-brand 250W PSU, so
you have a lot of breathing room as far as power is concerned.
Good.

The larger issue is cooling... not knowing your case or actual
effective airflow I can only suggest monitoring temperatures after
each significant upgrade or change. Two fans can handle what you're
suggesting with a good case and minimal obstructions to airflow, like
no stamped-in-metal fan guards.

The midi-tower case is a bit claustrophobic. Not sure about the stamped in
metal fan guards. I have two flat RAID cables between the intake fan and
main 'space' though. There is an output fan in the case below the PSU fan,
and the slot is removed next to the graphics card. The case side is leaning
against the braided mobo power cable (much longer than the previous one and
it 'loops out') leaving a 1 cm gap between the side and the case. I would
have the side off but the cat takes far too much interest! The max temp the
CPU has been is 39 deg. It's a few degrees cooler with the side off. Current
CPU fan speed 2556 RPM at 24/7 100% CPU capacity.

 

[kony]

I remembered you'd said this, which is why I went for the Antec Although
I did have to bend the bit of metal the PSU sits on, as the fan grill got in
the way :-(

not surprising, as power supplies keep getting higher in capacity they
end up with either smaller heatsinks to provide room for the 2nd fan
or the fan and grill are push towards the the exterior as much as
possible. Back when the typical ATX power supply was 200-250W the PS2
PSU casing seemed a good size but today with good high-wattage power
supplies really cramming the parts into a PS2 sized case it looks a
bit like we need a new, slightly larger power supply standard. Rarely
there are some that are longer, perhaps the best short-term solution
since "most" people can spare an extra inch between the power supply
and the top (optical drive) external 5.25 bays.

The midi-tower case is a bit claustrophobic. Not sure about the stamped in
metal fan guards.

Everywhere you have a fan, the consideration is how much the airflow
is restricted. Either there is NO grill at all, only an open hole the
fan sits in, or one of two most common types of grills- Either round
wire welded together, most often chrome-plated, or the casing itself
had holes punched or stamped out of it during production, hence
"stamped-in-metal" as i referred to it. Some stamped-in-metal grills
are fairly open, restrict airlflow only a little, but more often they
reduce airflow by over 50%.. in other words, most often it requires
over twice as many fans to achieve same airflow because of a poor
budget manufacturing choice, and they cause more turbulent noise too.

The biggest problem with the worst of the stamped-in-metal grills,
those with a circular hold pattern instead of slots, is the difficulty
to cut them out.. not only does it require metalwork tools, but
there's no easy way to cut out the hole without leaving lots of metal
particles behind so it's prudent to disassemble the system first. On
the other hand the metal slot type can more easily be cut out with
tin-snips... not the best looking result but is fast, easy, and most
people aren't looking at the back of their PC too often.

Myself I always preplan a system before even putting parts in a case,
do all that cutting and finishwork beforehand. It's much better to
have too much airflow and need to reduce the fan speed (and noise) or
unplug a fan rather than have to add a fan later.

I have two flat RAID cables between the intake fan and
main 'space' though.

That might be significant or might not, depends on the amount of open
area around those cables... the air doesn't necessarily need a staight
path but too much impedance is of course less efficient.
Stamped-in-metal grills are "usually" a bigger problem.

There is an output fan in the case below the PSU fan,
and the slot is removed next to the graphics card. The case side is leaning
against the braided mobo power cable (much longer than the previous one and
it 'loops out') leaving a 1 cm gap between the side and the case.

It does sound like a pretty cramped case. It might help to use
plastic wire-ties to secure the cable to the rear of the power supply
or put a knot in the cable to "use up" the extra length.

I would have the side off but the cat takes far too much interest!
The max temp the CPU has been is 39 deg. It's a few degrees cooler with the
side off. Current CPU fan speed 2556 RPM at 24/7 100% CPU capacity.

Obviously the goal is to keep the system closed, but some cases just
aren't suited for high-performance = high-heat components. The 39C
CPU temp is fine but you'll have to see how much it rises once you
start overclocking it... ultimately that may be what limits your
overclocking the most, as beyond a certain point (which can vary)
you'll need to raise the CPU voltage to get higher overclock if you're
not satisfied with the speed it can hit at default voltage. That
voltage increase usually does more to raise CPU power consumption,
temp, than the frequency increase does, though they have to be
considered together since one necessitates the other. Generally I try
to avoid going over 1.75V but most of the Athlon boxes here are below
1.7V.



Dave

 

[kony]

Note that 20-80 is probably OEM (brown bag) not retail, but for home use
the Fortron/Sparkle are fine. Zalman's doesn't sell OEM and will come with
Manual and retail box and full guarantee which is one of the reasons I
recommend Zalman over the other two. If in a server situation 6-8 hard
drives go with PCP&C for the additional support.

Lane

Some of the same retail units are sold as OEM, but still retail
pricing is pretty reasonable. The 400W below seems disproportionately
priced but is the only one of the 4 with active PFC.

Retail 300W $27
http://www.newegg.com/app/ViewProductDesc.asp?description=17-104-979

Retail 350W $36
http://www.newegg.com/app/ViewProductDesc.asp?description=17-104-966

Retail 400W $67
http://www.newegg.com/app/ViewProductDesc.asp?description=17-104-974

Retail 530W $69
http://www.newegg.com/app/ViewProductDesc.asp?description=17-104-968



Dave

 

[S.Boardman]

snip!

Everywhere you have a fan, the consideration is how much the airflow
is restricted. Either there is NO grill at all, only an open hole the
fan sits in, or one of two most common types of grills- Either round
wire welded together, most often chrome-plated, or the casing itself
had holes punched or stamped out of it during production, hence
"stamped-in-metal" as i referred to it.

snip fan stuff!

I have the holes punched out of the casing. For the outtake fan, the fan
sits directly on the casing. The input fan at the front sits against a piece
of metal about an inch from punched holes, and is encased in a plastic
protector.

That might be significant or might not, depends on the amount of open
area around those cables... the air doesn't necessarily need a staight
path but too much impedance is of course less efficient.
Stamped-in-metal grills are "usually" a bigger problem.

About an inch of space on either side on the edges. I suppose I ought to get
round ones. Are they interchangeable between IDE and RAID - I don't need
anything special? They have to be UDMA 133 ones?

It does sound like a pretty cramped case. It might help to use
plastic wire-ties to secure the cable to the rear of the power supply
or put a knot in the cable to "use up" the extra length.

A knot - that's a good idea! The rest of the cables are stuffed in above my
DVD drive, in the space at the top of my case.

Obviously the goal is to keep the system closed, but some cases just
aren't suited for high-performance = high-heat components. The 39C
CPU temp is fine but you'll have to see how much it rises once you
start overclocking it... ultimately that may be what limits your
overclocking the most, as beyond a certain point (which can vary)
you'll need to raise the CPU voltage to get higher overclock if you're
not satisfied with the speed it can hit at default voltage. That
voltage increase usually does more to raise CPU power consumption,
temp, than the frequency increase does, though they have to be
considered together since one necessitates the other. Generally I try
to avoid going over 1.75V but most of the Athlon boxes here are below
1.7V

How will I know when I need to raise the voltage?
The heatsink is rated for a XP3200.

 

[kony]

I have the holes punched out of the casing. For the outtake fan, the fan
sits directly on the casing. The input fan at the front sits against a piece
of metal about an inch from punched holes, and is encased in a plastic
protector.

It sounds like your case is working against you. Sometimes it's
easier to just add an intake fan to the side-panel, since it just pops
off and doesn't require any system downtime to take all the bits out
of the whole case.

About an inch of space on either side on the edges. I suppose I ought to get
round ones. Are they interchangeable between IDE and RAID - I don't need
anything special? They have to be UDMA 133 ones?

RAID uses the same cables... ATA33 are 40 conductor and very rarely
round. ATA66/100/133 are all the same, 80 conductor cable. If you
didn't see specs just a picture of a round cable the odds are very
high that it's 80-conductor. Another clue can be that 80-conductor
cable are usually with colored connectors, blue to motherboard, grey
in the middle for slave and black on the other end for master.

How will I know when I need to raise the voltage?
The heatsink is rated for a XP3200.

It helps to have an idea of the average ceiling for your particular
model of CPU. There are several places to find that info but the
simplest in this forum would be to just go here and select the CPU
from the drop-dwon list: http://www.cpudatabase.com/ . You can pretty
much ignore any results with water-cooling or even more exotic
measures like the Prometia. You can usually ignore the lowest scores
too for a variety of reasons like incompetence, no desire to push the
CPU, or bad power supply, motherboard, or just a "dud" CPU... you'll
have to assume for the time being that you don't have a dud CPU, and
hope for typical results.

Providing you get your case airflow improved you should be able to run
at least 1.7V with that heatsink, but you may not even be able to
choose >1.85V with your motherboard, and more-often-than-not the heat
resulting from voltages over 1.8V will limit overclocking... you might
be able to hit a higher speed at 1.8V than 1.9V with an air-cooler.
So anyway from that list you'll see most people are hitting the
2.2-2.3GHz range, so that's a good target. Write down your BIOS
settings since there's a pretty good chance you'll be clearing the
CMOS to get the system booting again after it's set too fast.

With the FSB left unchanged and manipulating the multiplier you'll be
isolating the CPU (if we can assume the motherboard is capable of
supplying enough power, which it should since there are 2.2GHz stock
speed CPUs supported by that board. You can either raise the
multiplier till it won't POST, then clear the CMOS (or check the
manual for a keyboard key to revert to CPU defaults at power-on), then
raise the voltage a notch till it runs stable, or go backwards- Raise
the voltage to the max you think is acceptible and then change the
multiplier, working backwards till the reduction in voltage introduces
instability... at some point you'll have to decide how much work you
want to put into it, just settling for an easy overclock or finding
the max or the best voltage to speed tradeoff... keeping track of the
voltage vs speed results may help you see a "curve", where each minor
increase in speed requries increasing the voltage more and more.

I suggest you try 1.7V and see how high it'll run stable, putting the
system through it's paces, and do a Google search for some CPU stress
tests. If it's stable then you decide to leave it there, decrease
voltage for testing same speed again, or raise speed with same
voltage. FOr a margin of error you might want to increase the voltage
0.025 over the lowest that tests stable in use.

After you get an idea of the voltage and total MHz speed for the CPU
then you can drop the multiplier to default again, leave the voltage
raised, and test the FSB, see how high that will go... you don't want
to change the voltage for that since you need a "constant", you know
the CPU can run higher than the resulting frequency at the voltage
used so the CPU is then removed from the equation and focus is then on
memory, PCI bus, AGP, etc. This part of the testing is a lot more
difficult, time-consuming... this being your first exercise, you might
consider leaving the FSB under 174MHz, being sure to test with
http://www.memtest86.com for several hours BEFORE loading and running
windows... we don't want the memory causing any data corruption. I
usually leave the memtest floopy in the system so if i accidentally
forget it'll boot to that instead of windows.

At this point I feel like I only skimmed the surface, didn't really
give a comprehensive overview at all... do a Google search for some
overclocking tutorials, 'tis better to be safe than sorry. Also be
sure to backup data on the hard drives prior to trying any FSB over
~174MHz, as higher resulting PCI bus speeds could corrupt the hard
drive(s). All overclocking is done at your own risk of course.


Dave

 

[S.Boardman]

RAID uses the same cables... ATA33 are 40 conductor and very rarely
round. ATA66/100/133 are all the same, 80 conductor cable. If you
didn't see specs just a picture of a round cable the odds are very
high that it's 80-conductor. Another clue can be that 80-conductor
cable are usually with colored connectors, blue to motherboard, grey
in the middle for slave and black on the other end for master.

Thanks. Any better than others? One for each RAID drive and one spare
<snip o/c advice>
It's all so complicated! I don't supposed I could just up the FSB until it
falls over then try and up the voltage a bit and see if it then works???
BTW there are lots of V number... is it Vcore I want? The windows utility I
have from MSI (only reliable for temp and CPU fan speed) says 1.65.

At this point I feel like I only skimmed the surface, didn't really
give a comprehensive overview at all... do a Google search for some
overclocking tutorials, 'tis better to be safe than sorry. Also be
sure to backup data on the hard drives prior to trying any FSB over
~174MHz, as higher resulting PCI bus speeds could corrupt the hard
drive(s). All overclocking is done at your own risk of course.

The memory is Corsair PC 3200.
I really must bite the bullet and change the RAID over (and then backup the
entire drive to the spare one on a regular basis). I've a box of three 80Gb
drives sat here, just waiting for their chance to shine...

 

[kony]

Thanks. Any better than others? One for each RAID drive and one spare

Too often there's little to no tech specs provided on cables... They
should be twisted pairs but almost all are, is obvious when you look
at them. It can help to have pull-tabs on the ends, reduces strain
when unplugging them, but some people consider the pull-tabs unsightly
so to each their own preference. Primarily care is needed when
unplugging multiple times to avoid stressing them too much, but to a
lesser extent it's true of old ribbon cables too.

<snip o/c advice>
It's all so complicated! I don't supposed I could just up the FSB until it
falls over then try and up the voltage a bit and see if it then works???
BTW there are lots of V number... is it Vcore I want? The windows utility I
have from MSI (only reliable for temp and CPU fan speed) says 1.65.

The thing about raising FSB is that to get much of an o'c you'd need a
large FSB increase, which may put the PCI, AGP, etc, out of spec.
Typically you might see USB, network adapter, or hard drive problems
at overly high bus speeds. It's better to find the CPU's ceiling
first by isolating it, and regardless you'll probably need a 12-12.5X
multiplier to get most of the o'c anyway on that board, I don't recall
it supporting 200MHxz FSB with 1/5 PCI divider.

it sounds like a lot of work, but it may not take a long longer to do
than describe, once it's not working you have an idea of where it
fails, what speed vs voltage and can work from there, considering
whether to decrease speed or increase voltage.
For a generic answer that may or may not work, try 174MHz FSB, 12.5X,
1.7V. If it works you might then lower voltage back to 1.65 default.

I don't know if your board supports multipliers over 12.5X with that
CPU... likely it will list them but choosing one would result in
(that) multiplier minus 8... to get the +8 multipliers a bridge on the
CPU needs burnt through, IIRC.

The memory is Corsair PC 3200.
I really must bite the bullet and change the RAID over (and then backup the
entire drive to the spare one on a regular basis). I've a box of three 80Gb
drives sat here, just waiting for their chance to shine...

When you've made the backup, ready to change the array over, that
might be a good time to try the FSB overclocking while the backup
drive(s) are unplugged.


Dave