Video Card UpGrade

[BaldEagl]

Let's take a closer look at Lusche's PSU specifically:

It is rated at 450W and 70% efficiency.  This gives us a real wattage rating of 315W.

Before doing this though, we have to reduce the total rated wattage by the rated wattage on the 3.3 and 5V rails.  450W-145W x 70% = 213.5W on the combined 12V rails.

Since we know that Amps x Volts = Watts then his rated 29 Amps should show 348W on the 12V rails but if we add the 145W from the 3.3 and 5V rails to this it would give us 493W which is higher than the PSU's rating.  Alternately, we can take the 213.5W that we calculated were available on the 12V rails and arrive at real amps available of 18 (213.5/12), far under the PSU's rating of 29 Amps.

So what's going on here?  His PSU is likely rated at peak power meaning the numbers that they show in the specs are optimistic and available only for very brief periods.

So, will this PSU run a 9600 GT (Minimum of a 300 Watt power supply with +12 Volt current rating of 18 Amps.)?  Yes, because it meets the minimum required specs under continuous load.  Will it run a 9800 GT (Minimum of a 400 Watt power supply with +12 Volt current rating of 26 Amps.)?  Maybe but any continuous draw in those ranges will fry either the PSU or the video card.

I for one think the card manufacturers understand this and consider it when making their recommendations.  Do they err on the side of safety?  Most likely because look what just happened to a PSU rated at 450W with 29 Amps on the 12V rails.

And this is why running a PSU less than the card manufacturer's specs is not a good idea for the novice and is simply bad advice in a forum such as this.


BTW, If you do this exercise with a PCP&C or a certified Seasonic PSU the numbers come out very close to the rated power.

[Cajunn]

I'm upgrading too, I'm looking at this card any advise:

  NVIDIA GeForce GTS 250

    * Core Clock Speed 738 MHz

    * Graphics Card Performance This card is compatible with the new rendering features of DirectX 10, which is featured in the Windows Vista operating system.

    * Video Memory Installed 1 GB

    * Technology DDR3 SDRAM 256-bit

    * Memory Clock Speed 2.2 GHz

    * Features:
 Nvidia PhysX-Ready , Nvidia Lumenex Engine , Nvidia 3-way SLI ready , Nvidia CUDA technology , Nvidia GigaThread technology , Nvidia SLI Multi-monitor support , Nvidia GeForce 3D Stereo technology

[Lusche]

Cajunn, as you may see from this thread, it's hard to judge if a certain card makes sense for without knowing the rest of the system.

[SPKmes]

sorry for this intrusion....but with what has been said here about power I have a question. Does this all mean that if your power supply is over working other than the obvious burnout issues... it would starve different areas (cpu, graphics card, soundcard, etc) of needed power to run effectively?.

[Cajunn]

Cajunn, as you may see from this thread, it's hard to judge if a certain card makes sense for without knowing the rest of the system.

 I just had my system built, and I know my power supply is 550 or maybe 650 because I requested it. And it was built on a gaming platform, I'm running a Asus EAH4670 series graphics card right now which is 512/MD3 and I'm just going to a 1G setup. I'm running Intel dual core 2.66GHz 4 gig's of ram which I'm going to 8 gig's in the near future.

Note: I'm running Windows XP64, and I'm sticking with that as long as I can!

[Strip]

Usually (with power supplies) if a power supply is rated for 300 watts this number takes into account efficiency. So you do not need to correct output figures....

[humble]

Thats not entirely correct. Unless specified as continuous power that number almost always refers to "peak" power. You can expect a minimum of 10% less clean continuous power. Additionally unless you know the actual conditions at test you don't know how sound the rated output is. A unit rated at "300W" at 25C will generate significantly less at 50C and might even shut down entirely at that temp. As an example the 910W silencer delivers 910W of continuous power at 50C and 1000W peak. Now if we drop down to the turbo cool (both PC power and Cooling) it delivers 510W continuous/650 Peak at 50C. So only 78% of maximum power in a top of the line mid range unit. Realize that at the tested environment this PS delivers up to twice as much power as a typical 500W generic power supply.

Take a look at the temp your system is generating and compare it to the chart below


without specific 411 multiplying efficiency X "rated output" will ballpark you reasonably well under normal operating conditions. Your "300W" power supply probably delivers 180-210 total Watts in normal use...
 

[flatiron1]

there is a low power 9600 gt and 9800 gt out now made by evga and galaxy.

[BaldEagl]

sorry for this intrusion....but with what has been said here about power I have a question. Does this all mean that if your power supply is over working other than the obvious burnout issues... it would starve different areas (cpu, graphics card, soundcard, etc) of needed power to run effectively?.

Yes.  You'll typically experience a grey screen, black-out, BSOD or other anomalies.

[BaldEagl]

I just had my system built, and I know my power supply is 550 or maybe 650 because I requested it. And it was built on a gaming platform, I'm running a Asus EAH4670 series graphics card right now which is 512/MD3 and I'm just going to a 1G setup. I'm running Intel dual core 2.66GHz 4 gig's of ram which I'm going to 8 gig's in the near future.

Note: I'm running Windows XP64, and I'm sticking with that as long as I can!

Need to know how many amps it has on the 12V rails.  If you open the case there should be a sticker on the PSU.  You'll also need to make sure it has the right connectors (6 pin, 8 pin or 6 + 8 pin for the card).

[Boozeman]

Hehe, just as I thought, the silence is deafening.

I must say it's rather funny that you are still quoting manufaturers recommendations on what a certain video card draws, despite clear evidence from numerous hardware review sites, that the power consumption of a whole system (and that are rather top notch ones) including the card in question, does acutally draw less power than the minimum recommendation for the card alone.

So my point still stands: If one of you guys can prove that a 9800GT draws significantly more than 120 W (just the card, not the system!) in any application, I'm all ears. It just has to be a independent 3rd party test, no manufacturer claim. 

 

[Masherbrum]

See Rule #4

[Skuzzy]

The manufacturer knows their hardware better than someone at some WEB site does. 

The truth is, video cards will draw different levels of power based on what they are doing.  The manufacturer specs the power draw based on an absolute worst case scenario.  Just as Intel and AMD specs the CPU power draw at worst case levels as well.

They have no reason to lie about such things.  Quite the opposite, they have every reason to keep that advertised power draw level as low as possible.  Taking the word of someone at a WEB site (show me their credentials.  Are they EE's?  Systems programmers? Have they actually done circuit and/or chip design?  How long?  Why did they quit?  How long have then been doing systems software design?) over the word of the people who designed and built the parts is a recipe for failure.

I do check out WEB site reviewers and for the most part, they do not know the difference between a sine wave and a square wave.  It is quite apparent, in many cases, they read the instruction manual on using some piece of equipment, then read the numbers it spewed forth without really understanding the cause of those numbers.

If you (collectively) wish to trust them over the engineers who actually designed the parts and are fully qualified to speak to how much power those parts will use, that is your choice.  It does not make it the right thing to do.  It just makes it your choice.


Oh, and the name calling stops right now.

[humble]

Hehe, just as I thought, the silence is deafening.

I must say it's rather funny that you are still quoting manufaturers recommendations on what a certain video card draws, despite clear evidence from numerous hardware review sites, that the power consumption of a whole system (and that are rather top notch ones) including the card in question, does acutally draw less power than the minimum recommendation for the card alone.

So my point still stands: If one of you guys can prove that a 9800GT draws significantly more than 120 W (just the card, not the system!) in any application, I'm all ears. It just has to be a independent 3rd party test, no manufacturer claim. 

 

I'm sorry but your mistaken, a 9800 series card at full load will draw 230W or so of power....just that simple. A generic 450W PS doesn't even generate that much total power. As to what a specific card will draw playing AH in a sustained manner....I'm not sure. Put I can guarantee you that the PS in question does not put out anywhere near the "oomph" you think it does.

[humble]

Skuzzy,

"Name calling" isn't an issue, some poor guy wasting some hard earned coin due to well meant but mistaken advice is the issue. Anyone with significant time around computers knows that heat is the #1 system killer. Not only does performance degrade but system life can be cut dramatically. Forgetting the likely reality that the PS in question cant generate enough power to start even if it can its taxed to its limits and generating to much heat....degrading its own performance and the entire systems life span. You do not need an uber PS....all you need for most systems is 265-330 total watts of good power. The key is understanding PS realities, as the chart I posted illustrates most new out of the box generic PS's only generate 320-350W at normal operating temps. The heat they generate doing so degrades that "out of the box" performance pretty quickly. So dropping a high demand card into a PC with a marginal PS creates a scenario where even if it appears to be OK the added system heat will kill not only the VC but the everything else within a year or so.

None of this is my thought, every quality power supply and independent review site recognizes heat and marginal PS output as the two leading "PC killers".