Conflicting Video Card Current Requirements and PS Ratings Systems

[JB]

It seems like both industries operate in a vacuum.

I looked at a video card whose requirement was "a 350 W PS with 22 A
available on the +12V rail".

Most PS ratings that I checked are rated for a maximum of around 18A on the
+12V rail regardless of the advertised power rating.

Some have dual or triple +12V rails whose combined ratings may be 36A or
higher.

So how do we make a determination as to what the video card really needs and
how to provide it?

Some would, I suppose, put the rails in parallel and hope for the best. But
if the PS is not designed for that mode, the result could be instability
with unpredictable effects.

If the actual load requirement for the video card were disclosed, we could
make some guestimates as to all other loads and choose a power supply that
would probably work. IMO the video card requirement is ridiculous and shows
a lack of concern for the customer.

Thanks for any insights.

Mike

 

[kony]

It seems like both industries operate in a vacuum.

I looked at a video card whose requirement was "a 350 W PS with 22 A
available on the +12V rail".

.... but you didn't mention the card nor the rest of the
system. Since a video card manufacturer can't know other
system details, they just have to generalize because some
psu aren't honestly rated and they can't know the 12V
current the rest of the system uses.

Most PS ratings that I checked are rated for a maximum of around 18A on the
+12V rail regardless of the advertised power rating.

You haven't checked many online then? Newer generations of
higher wattage PSU typically have more than 18A rating,
though sometimes it is for current on multiple 12V rails
which is ok, since one rail on modern systems powers the CPU
which is part of the consumption the video card manufacturer
factors in when spec'ing 22A.

Some have dual or triple +12V rails whose combined ratings may be 36A or
higher.

So how do we make a determination as to what the video card really needs and
how to provide it?

Start with details. Video card, CPU, # of hard drives, PSU
make/model/rating.

Some would, I suppose, put the rails in parallel and hope for the best. But
if the PS is not designed for that mode, the result could be instability
with unpredictable effects.

Not necessarily, depends on the specfics. Generally if the
rails are truely in parallel it is more versatile, not less.

If the actual load requirement for the video card were disclosed, we could
make some guestimates as to all other loads and choose a power supply that
would probably work. IMO the video card requirement is ridiculous and shows
a lack of concern for the customer.

I agree the video card manufacturers should plainly state
current:rail consumption, it would make things a lot easier
for those who know the other variables in their system. It
is a shame the engineers and marketing departments can't
(dont') provide these specs. You can have some idea of
certain models of psu that are shown acceptible with your
card by seeking fellow users of same card with equivalent
systems, but cold hard numbers are always the more accurate
way to go.


On the other hand, it is seldom a good idea to buy a PSU
with barely enough capacity rather than one with quite a bit
of margin, not just for stability reasons but to achieve
longer life. Granted, sometimes even that is not enough,
certain PSU cost corners like poor fans or capacitors might
reduce life regardless of having a capacity margin, but
generally if one has the budget for a power hungry video
card it just makes sense to put a bit more of the budget
into a higher quality and capacity PSU as well.

There are some online comparisons of video card power
consumption, if you're lucky a Google search might find one
that has tested your card or at least same GPU, memory,
frequencies... even if a different brand of card it should
be very close to same power consumption as they're generally
built on the same reference design or at least close enough
to that in consumption even when certain manufacturers like
MSI or Asus decide to change the design a bit.

 

[larry moe 'n curly]

It seems like both industries operate in a vacuum.

I looked at a video card whose requirement was "a 350 W PS with 22 A
available on the +12V rail".

Most PS ratings that I checked are rated for a maximum of around 18A on the
+12V rail regardless of the advertised power rating.

Some have dual or triple +12V rails whose combined ratings may be 36A or
higher.

So how do we make a determination as to what the video card really needs and
how to provide it?

The only reliable ways are by finding somebody who's taken actual
measurements, or buy/steal a DC clamp-on ammeter and measure for
yourself. Estimated or stated ratings can be way off, but fortunately
they're almost always higher than the actual current consumption.

Is it possible that the 22A @ +12V rating is for the whole PSU and not
just for the rail that plugs into the graphics card? Because I
thought that multi-rail +12V was introduced as a safety measure, to
limit each +12V rail to 20A. Also experts like JonnyGuru.com have
found that some PSUs, such as the Antec Trio (made by Seasonic), that
are claimed to have double or triple +12V rails actually have only
one, so there's no need to worry about the per-rail amp ratings.

 

[Ken Maltby]

It seems like both industries operate in a vacuum.

I looked at a video card whose requirement was "a 350 W PS with 22 A
available on the +12V rail".

Most PS ratings that I checked are rated for a maximum of around 18A on
the +12V rail regardless of the advertised power rating.

Some have dual or triple +12V rails whose combined ratings may be 36A or
higher.

So how do we make a determination as to what the video card really needs
and how to provide it?

Some would, I suppose, put the rails in parallel and hope for the best.
But if the PS is not designed for that mode, the result could be
instability with unpredictable effects.

If the actual load requirement for the video card were disclosed, we could
make some guestimates as to all other loads and choose a power supply that
would probably work. IMO the video card requirement is ridiculous and
shows a lack of concern for the customer.

Thanks for any insights.

Mike

http://www.overclock.net/faqs/88626-info-do-you-need-multiple-12v.html

Luck;
Ken

 

[JB]

... but you didn't mention the card nor the rest of the
system. Since a video card manufacturer can't know other
system details, they just have to generalize because some
psu aren't honestly rated and they can't know the 12V
current the rest of the system uses.


You haven't checked many online then? Newer generations of
higher wattage PSU typically have more than 18A rating,
though sometimes it is for current on multiple 12V rails
which is ok, since one rail on modern systems powers the CPU
which is part of the consumption the video card manufacturer
factors in when spec'ing 22A.




Start with details. Video card, CPU, # of hard drives, PSU
make/model/rating.



Not necessarily, depends on the specfics. Generally if the
rails are truely in parallel it is more versatile, not less.




I agree the video card manufacturers should plainly state
current:rail consumption, it would make things a lot easier
for those who know the other variables in their system. It
is a shame the engineers and marketing departments can't
(dont') provide these specs. You can have some idea of
certain models of psu that are shown acceptible with your
card by seeking fellow users of same card with equivalent
systems, but cold hard numbers are always the more accurate
way to go.


On the other hand, it is seldom a good idea to buy a PSU
with barely enough capacity rather than one with quite a bit
of margin, not just for stability reasons but to achieve
longer life. Granted, sometimes even that is not enough,
certain PSU cost corners like poor fans or capacitors might
reduce life regardless of having a capacity margin, but
generally if one has the budget for a power hungry video
card it just makes sense to put a bit more of the budget
into a higher quality and capacity PSU as well.

There are some online comparisons of video card power
consumption, if you're lucky a Google search might find one
that has tested your card or at least same GPU, memory,
frequencies... even if a different brand of card it should
be very close to same power consumption as they're generally
built on the same reference design or at least close enough
to that in consumption even when certain manufacturers like
MSI or Asus decide to change the design a bit.<

Thank you for your input Kony. I am sure you can do a power budget and
allocate it per rail but that is not what I need.

I am savvy on the engineering considerations but my gripe is with the
ribbon- clerk type disclosure of power requirements.

For example, the particular card whose requirement I mentioned as being 22A
on the 12 V rail is an EVGA 8600GT. My gut feel is that the card does not
need anywhere near that much current. In fact, IFhe link below gives the
power dissipation for that card as being 30W, not the 264W implied by the
requirement on the box. So, IF the numbers at the site are correct, the
card itself needs 2.5 A.

http://www.atomicmpc.com.au/forums.asp?s=2&c=7&t=9354

Obviously EVGVA's intent is to tell the customer his total system needs that
much capability on the 12V rail...a prognistication only a true seer should
make. But that is not what they say so it's worse than no information at
all AFAIC.

BTW, that card is just one example, it's not one I intend to buy. I know
other companies put similar requirements on the customer and think it's just
sloppy work across the board. I don't think a user should have to buy a
clamp-on ammeter and verify the ratings.

When I get my components selected I will select a PS and then see what you
folks think.

Thank you.

Mike

 

[JB]

The only reliable ways are by finding somebody who's taken actual
measurements, or buy/steal a DC clamp-on ammeter and measure for
yourself. Estimated or stated ratings can be way off, but fortunately
they're almost always higher than the actual current consumption.

Is it possible that the 22A @ +12V rating is for the whole PSU and not
just for the rail that plugs into the graphics card? Because I
thought that multi-rail +12V was introduced as a safety measure, to
limit each +12V rail to 20A. Also experts like JonnyGuru.com have
found that some PSUs, such as the Antec Trio (made by Seasonic), that
are claimed to have double or triple +12V rails actually have only
one, so there's no need to worry about the per-rail amp ratings.<

Yep..like I was saying to Kony, the disclosure is pretty sloppy across the
board.. I tried to get a schematic on an Antec years ago and it turned out
to be too secret for the unwashed masses. lol I can understand that but
would even a block diagram would tell you if each rail had separate feedback
control or whatever awithout disclosing anything proprietary.

Thanks for your input.

 

[Paul]

Thank you for your input Kony. I am sure you can do a power budget and
allocate it per rail but that is not what I need.

I am savvy on the engineering considerations but my gripe is with the
ribbon- clerk type disclosure of power requirements.

For example, the particular card whose requirement I mentioned as being
22A on the 12 V rail is an EVGA 8600GT. My gut feel is that the card
does not need anywhere near that much current. In fact, IFhe link below
gives the power dissipation for that card as being 30W, not the 264W
implied by the requirement on the box. So, IF the numbers at the site
are correct, the card itself needs 2.5 A.

http://www.atomicmpc.com.au/forums.asp?s=2&c=7&t=9354

Obviously EVGVA's intent is to tell the customer his total system needs
that much capability on the 12V rail...a prognistication only a true
seer should make. But that is not what they say so it's worse than no
information at all AFAIC.

BTW, that card is just one example, it's not one I intend to buy. I
know other companies put similar requirements on the customer and think
it's just sloppy work across the board. I don't think a user should
have to buy a clamp-on ammeter and verify the ratings.

When I get my components selected I will select a PS and then see what
you folks think.

Thank you.

Mike

Xbitlabs measures video card power consumption, by using current shunts
on the slot. The 8600GTS is about 12V @ 3.75 amps total, so the 8600GT must
be a bit less than that. The card draws that current, partially through
the slot connector, and partially through the Aux connector on the end
of the video card. At this low current level, if they had wanted, they
could have drawn it all through the slot. Using an Aux connector is quite
conservative.

http://www.xbitlabs.com/articles/video/display/geforce8600gts_7.html

Even if the companies making the products gave out power estimates, they'd
probably be wrong anyway

Paul

 

[kony]

Thank you for your input Kony. I am sure you can do a power budget and
allocate it per rail but that is not what I need.

I am savvy on the engineering considerations but my gripe is with the
ribbon- clerk type disclosure of power requirements.

For example, the particular card whose requirement I mentioned as being 22A
on the 12 V rail is an EVGA 8600GT. My gut feel is that the card does not
need anywhere near that much current. In fact, IFhe link below gives the
power dissipation for that card as being 30W, not the 264W implied by the
requirement on the box. So, IF the numbers at the site are correct, the
card itself needs 2.5 A.

http://www.atomicmpc.com.au/forums.asp?s=2&c=7&t=9354

I agree a 22A rating is arbitrarily high for that card, it
is much closer to 30W than even 60W.

Obviously EVGVA's intent is to tell the customer his total system needs that
much capability on the 12V rail...a prognistication only a true seer should
make. But that is not what they say so it's worse than no information at
all AFAIC.

BTW, that card is just one example, it's not one I intend to buy. I know
other companies put similar requirements on the customer and think it's just
sloppy work across the board. I don't think a user should have to buy a
clamp-on ammeter and verify the ratings.

When I get my components selected I will select a PS and then see what you
folks think.

The problem seems to be that once PSU start rating current
from (supposed) split rails, it makes the current rating
seem higher than it really is. Suppose a PSU really had
split rails, suppose also it had 22A split 11A/11A. I doubt
any are labeled like this as they would tend to imply one
rail has more capability at such a low current and indeed
one might support more but getting down around this level
the question arises as to how they split the rail and how
much current remains per rail including powering other
devices.

It means there is a secondary reason why the current might
need be overspec'd to cover all scenarios when they use the
ill-conceived total 12V capacity rating as a guide, but
unfortunately some customers' eyes are bound to glaze over
if we attempted to tell them the real specifics they needed
to know instead of a simple number, and this doesn't even
consider yet that many PSU are overrated, not true ~ 22A (or
whatever the reating) capacity regardless of the shared rail
factor, are instead rated for peak instead of sustained (for
MTBF life rating) current and often at an unrealistically
low temperature like 25C instead of the real operating
state.