26 Aug 2007,Paul <nospam@[EMAIL PROTECTED]
> in news:fatm30$ukb$1@[EMAIL PROTECTED]
"wow wow wow". This is exactly why i need to ask questions. thank you. I
was
nowhere near discovering some of the aspects you outline below. And of
other
aspects, i'd only noticed the phrases, but not (yet?) realized their
significance to
me. Also, (not buttering you up, just noticing..) the way you typed this
up, you
appear super familar.
Summary: my PS almost certainly limits me to old cards. My PS is strong on
+5v
and weak on +12v.
My revised selection criteria: low-cost, reliable (new/warranteed/RMA-able
or
fanless used). Very secondary criteria: bandwidth, vram, dual-crt.
I'd like to know whether I'm stuck choosing from used 16mb-era gfx cards
(eg, tnt2.
And of course I'd research these myself). AFAIK any fan card old enough to
use +
5v, will have worn out fans.
Maybe the low-end agp cards (such as fx-5500oc, see below) draw low enough
power for my +12v line to supply?
(data from www.sparklepower.com/proPCPS_ATX.html)
My PS looks fine for summed wattages:
Sparkle's summed (+5v & +12v) ratings are:
175W for my ATX-300GT (site's values match my label's)
120W for ATX-300PA and ATX-300PN (two newer 300w models with +12v1
+12v2 lines)
Sparkle total of all "+" lines are:
280W for all three models.
i found posts which summed +12v demand (including a recent post by you
(Paul
'nospam needed') :-) )
mobo + 1hd + usb hd + fd + 2cd + gfx card = amps, +12v
??? + 0.5 + 0.2 + 5 + 2 + ???gfx = ???Amps [my PS label rsays +12v==13A]
Details:
cpu. I strongly suspect AZ-11 mobo feeds Duron 800Mhz either 3.3v or 5v.
800Mhz
maxes 35.4w at 1.6-1.7v. I can't find or further decipher AMD spec info.
usb hd. i guessed at USB signal watts. Probably negligible. Might be 3.3v
or 5v.
The external usb hd has own PS in power cord. I have no usb mouse,
keyboard, or
"usb hub".
floppy. Generic 5 watts from googling
Zero amp stuff:
nic. Dlink pdf on cd says 0.8w(240mA)max 3.3v.
ram. Runs off 3.3v?
0fans. No case fans because it's a 110v a/c deskfan strapped to the open
side of
case. Seems very good flow.
??? amps:
gfx card. Depends on card, but assuming the fx5500oc agp, i will probably
phone
BFG, asking for link to published specs (with power vs volt lines). fx5500
lacks hd
power plug, so should need less than agp (2.0 or 3.0) spec's max?
mobo. can't find info yet. az-11 manual warns only about min 200W PS,
supplying
20A, +5v and 10mA, +5Vsb. (Maybe acronyms in the manual would imply
likely
approximate amps?)
i thought comparing amps (+12v?) of this Rage iic gfx card might be
useful, but
can't find that.
I googled for VIA KT133 conflicts. So far, I've found no explosive looking
info.
the power question is probably all that's puzzling me...
___________
but for anyone interested (google archaeologists) the following is my
~2nd sweep
thru the research process
> http://www.xbitlabs.com/articles/video/display/ati-powercons.html
> http://www.xbitlabs.com/articles/video/display/ati-vs-nv-power.html
>
> Xbitlabs has been measuring power, in their video card reviews, on
> and off since those two articles above. To find power, go to this
> page:
>
> http://www.xbitlabs.com/search/
[snip]
sampling:
http://www.xbitlabs.com/images/video/graphics-cards-2003/3dmark03_1024
_pure.gif
implies that bang-for-buck, 9500 pro does well against 9600 xt.
But this is a speed benchmark. And not only do we minority 'workstation'
users
have to filter thru gaming-oriented info, but various cad/modeling
software probably
need differing benchmarks (all of which are obscure?)
Bad for me: xbitlabs never ran an fx-5500. Though possibly they ran
another NV34
based card.
> They like to use "contem****ary" when reviewing the current crop of
> cards. This article is from July 2006.
>
> http://www.xbitlabs.com/articles/video/display/power-noise.html
>
> There should not be a big difference in power draw if using two
> monitors. If the card is in 3D mode, it should already be drawing
> the power stated in the Xbitlabs measurement. Running another DAC
> or running a TMDS interface won't make that much difference.
> Running two monitors is more demanding of the video card memory
> bandwidth, and so selecting the cheapest card might not help in
> that regard.
ok. Data bits is larger when dual-display, while data watts is negligibly
different.
> What will make a difference, is trying to run two monitors with
> a Duron 800. The processor could well be the rate limiting thing.
i haven't googled this yet, but i wonder: is CRT power demand *of the
data* roughly
a factor of total pixels? or of display area? (or neither?)
> To work through a reasonably modern example, I picked this card:
>
> ASUS N7600GS SILENT/HTD/256M GeForce 7600GS 256MB 128-bit GDDR2
> AGP 4X/8X $87
>
> http://www.newegg.com/Product/ProductReview.aspx?Item=N82E168141210
> 64
>
> The card has 256MB of onboard RAM. The interface to that RAM is
> 128 bits wide, whereas a lot of low end cards only use 64 bits.
> The really expensive cards have 256 bits or more in width. Width
> helps with GPU to video memory bandwidth.
http://en.wikipedia.org/wiki/Comparison_of_ATI_Graphics_Processing_Units#Rade
on_R300_series
bit width
Radeon 9500 Pro 128
Radeon 9600 XT 128 or 256
(later implies both are native-agp)
en.wikipedia.org/wiki/GeForce_FX_Series#GeForce_FX_models
FX 5500 NV34 64 or 128 bit native-agp
BFG site's info is skimpy. Can't find bit-width, but bfg fx5500oc is
probably 64-bit
> That card memory is probably "true memory". The low end cards
> sometimes use Turbocache or Hypermemory, and you have to read more
> than one version of advertising copy,
yes
[snip]
> http://en.wikipedia.org/wiki/Turbocache
> http://en.wikipedia.org/wiki/HyperMemory
>
> The above Asus example card, has an aluminum heatsink. I would
> estimate the power of the card at around 35W or so. (Allowing a
> couple watts for the HSI bridge chip on the back of the card.)
[snip clarification on output & power connectors]
those used ati's are fully featured. but lack of HD plugin for fx5500oc
agp suggests
it needs less power. but the fx5500 lacks dvi out.
According to those wikipedia pages, all agp have "true" vram/memory.
Implies that
it doesn't matter whether cards may be native pci-e , because running thru
agp-
spec slot means they can't cache.
> So there is room to cut corners on video cards, but you may not
> end up getting all the features you were looking for.
>
> More info here. Non-quadro cards are listed as fine, as long as
> the video card has sufficient memory for the number of open
> windows. (click "tested" button.)
>
> http://www.solidworks.com/pages/services/videocardtesting.html
[perhaps appropriate disclaimer: i didn't intend to mislead anyone to
thinking i
expect to be using solidworks. this ng was most common in results from
googling.]
I chose "ATI"," winxp". SW chart shows:
ATI 9500 Pro p***** w/limitations (see popup). Note that 9600 pro is in
same
group, so possibly SW never tested 9600xt, yet it would be similar.
For "ATI", clicking "show all" button gives the same chart of cards.
SW2008 is likely not yet tested.
______________
boredom alert. leftover thoughts:
NV34, last of native-agp?
Anisotropic Filtering apparently slows the FX cards, but this may be an
example of
a tradeoff that benefits modelers.
bfg fx5500oc is probably 64-bit:
www.google.com/search?q=FX-5500+bfg+%7C+oc+Bits+Bit-width
www.nvnews.net/vbulletin/showthread.php?t=37880 2004
www.nvnews.net/vbulletin/showpost.php?p=439430&postcount=13
"as verified by a phone call to BFG....well hell, they don't even know...
all the
engineers had left for the day."
www.nvnews.net/vbulletin/showpost.php?p=443252&postcount=18
"it was 64-bit One of the main reasons I took it back to Best Buy."
lack of 'turbocache' is actually performance disadvantage, but *should*
force
advertisers to list vram correctly. OTOH, why not also 'turbocache' the
fastest
cards?
gfg4 are last of +5v powered cards (don't know which ati are last +5v)
|
