Tetrium wrote:

Why would you want a 350W/400W PSU for a Pentium 3 btw?

The build will have lot of things to power up, including 12v molex for graphics, 12v water pumps, 3-5 chassis fans, lot of extension cards, drives etc

alexanrs wrote:

I like that Enermax was mentioned, since I just bought a nifty 350W Enermax PSU with 32A on the 5V rails. Should be fine for a Duron+AGP videocard without an auxiliary power conector. Did Enermax use good caps?

What model of Enermax and motherboard ?

Tetrium wrote:

RacoonRider wrote:

Tetrium wrote:

I think the 5v:12v power draw of old graphics cards is still somewhat unexplored territory and I'd like to learn more about this 😀

By the fact that both my 9800Pros have yellowed plastic around 5V pin, I'd say they use 5V line a lot 😀 It's not hard to measure though, a few resistors and a voltmeter is all it takes when it comes to high-end videocards with Molex and Berg connectors.

Wow, why didn't I think of that, cheers 😁
Would still be a lot of work though 🤣!

You would need an ammeter. Ideally with an inductive pick-up.
Even with the dedicated power cable(s) they still simultaneously pull power from the slot so unless you are going to assume that part you need a more complicated test fixture.
Wasn't there a thread about that a week or so ago?
.

Inductive pickups only work with AC voltages. In a perfect world you would measure all the + voltage lines going into the motherboard in addition to the power supply. Not exactly an easy thing to setup for the home lab.

Godlike wrote:

What model of Enermax and motherboard ?

Enermax EG365P-VE and the motherboard is a humble PCChips M810LU rev 8.0

luckybob wrote:

Inductive pickups only work with AC voltages. In a perfect world you would measure all the + voltage lines going into the motherboard in addition to the power supply. Not exactly an easy thing to setup for the home lab.

That's only CT type pickups. Hall sensors detect DC.

alexanrs wrote:

Enermax EG365P-VE ...

Hitachi bulk caps and PCE-Tur secondaries, the latter shouldn't be trusted after a decade. Once recapped that would be an excellent PSU for retrocomps.

BTW right now I'm tinkering with with a brand new Enermax NAXN ENP500AGT, overall it's just a hair above the 'passable' level but it comes stuffed full of Jun Fu and Su'scon shit 😵 that needs to be extirpated and replaced with real caps.

. . .

The label says 19A on +5V but in fact it has a 30A rectifier on +5V.

PCBONEZ wrote:

You would need an ammeter. Ideally with an inductive pick-up. Even with the dedicated power cable(s) they still simultaneously p […]
Show full quote

You would need an ammeter. Ideally with an inductive pick-up.
Even with the dedicated power cable(s) they still simultaneously pull power from the slot so unless you are going to assume that part you need a more complicated test fixture.
Wasn't there a thread about that a week or so ago?
.

Not an ammeter, a voltmeter and a shunt resistor. An amperemeter would be most clumsy and unreliable, since you would have all the power go through it. You could use an amperemeter with a shunt resistor as well for pretty much the same result as a voltmeter and a shunt resistor.

RacoonRider wrote:

Tetrium wrote:

I think the 5v:12v power draw of old graphics cards is still somewhat unexplored territory and I'd like to learn more about this 😀

By the fact that both my 9800Pros have yellowed plastic around 5V pin, I'd say they use 5V line a lot 😀 It's not hard to measure though, a few resistors and a voltmeter is all it takes when it comes to high-end videocards with Molex and Berg connectors.

There was an article on xbitlabs which tested power draw of a bunch of video cards around that time period. The 9800 Pro page is here:
http://www.xbitlabs.com/articles/graphics/dis … owercons_8.html

It showed that the 9800 Pro was the last generation of ATI cards that relies most heavily on the 5v rail.
The next generation cards (X800/etc) shift much more of the consumption to the 12v rail, and draw less current through the AGP slot, favoring the molex connection instead.
The difference in power usage of those generations of cards could be an important thing to consider with some system builds.

RacoonRider wrote:

PCBONEZ wrote:

You would need an ammeter. Ideally with an inductive pick-up. Even with the dedicated power cable(s) they still simultaneously p […]
Show full quote

You would need an ammeter. Ideally with an inductive pick-up.
Even with the dedicated power cable(s) they still simultaneously pull power from the slot so unless you are going to assume that part you need a more complicated test fixture.
Wasn't there a thread about that a week or so ago?
.

Not an ammeter, a voltmeter and a shunt resistor. An amperemeter would be most clumsy and unreliable, since you would have all the power go through it. You could use an amperemeter with a shunt resistor as well for pretty much the same result as a voltmeter and a shunt resistor.

No no no no no. A quality clamp ammeter equipped with a Hall sensor rather than a CT is what you want here. The clamp allows you to insert the sensor without dismantling any of the PSU leads.

TELVM wrote:

Hitachi bulk caps and PCE-Tur secondaries, the latter shouldn't be trusted after a decade. Once recapped that would be an excell […]
Show full quote

alexanrs wrote:

Enermax EG365P-VE ...

Hitachi bulk caps and PCE-Tur secondaries, the latter shouldn't be trusted after a decade. Once recapped that would be an excellent PSU for retrocomps.

BTW right now I'm tinkering with with a brand new Enermax NAXN ENP500AGT, overall it's just a hair above the 'passable' level but it comes stuffed full of Jun Fu and Su'scon shit 😵 that needs to be extirpated and replaced with real caps.

. . .

The label says 19A on +5V but in fact it has a 30A rectifier on +5V.

Indeed, pretty much almost everything from the time period has shit for caps. Pity that OS-CON caps were hardly ever used at all, at least re-capping is easy but just inconvenient.

Good luck on the re-cap, maybe put up a quick guide so that people in the future find it.

gdjacobs wrote:

RacoonRider wrote:

PCBONEZ wrote:

You would need an ammeter. Ideally with an inductive pick-up. Even with the dedicated power cable(s) they still simultaneously p […]
Show full quote

You would need an ammeter. Ideally with an inductive pick-up.
Even with the dedicated power cable(s) they still simultaneously pull power from the slot so unless you are going to assume that part you need a more complicated test fixture.
Wasn't there a thread about that a week or so ago?
.

Not an ammeter, a voltmeter and a shunt resistor. An amperemeter would be most clumsy and unreliable, since you would have all the power go through it. You could use an amperemeter with a shunt resistor as well for pretty much the same result as a voltmeter and a shunt resistor.

No no no no no. A quality clamp ammeter equipped with a Hall sensor rather than a CT is what you want here. The clamp allows you to insert the sensor without dismantling any of the PSU leads.

What's wrong with this method? http://www.xbitlabs.com/articles/graphics/dis … owercons_2.html

shamino wrote:

There was an article on xbitlabs which tested power draw of a bunch of video cards around that time period. The 9800 Pro page i […]
Show full quote

RacoonRider wrote:

Tetrium wrote:

I think the 5v:12v power draw of old graphics cards is still somewhat unexplored territory and I'd like to learn more about this 😀

By the fact that both my 9800Pros have yellowed plastic around 5V pin, I'd say they use 5V line a lot 😀 It's not hard to measure though, a few resistors and a voltmeter is all it takes when it comes to high-end videocards with Molex and Berg connectors.

There was an article on xbitlabs which tested power draw of a bunch of video cards around that time period. The 9800 Pro page is here:
http://www.xbitlabs.com/articles/graphics/dis … owercons_8.html

It showed that the 9800 Pro was the last generation of ATI cards that relies most heavily on the 5v rail.
The next generation cards (X800/etc) shift much more of the consumption to the 12v rail, and draw less current through the AGP slot, favoring the molex connection instead.
The difference in power usage of those generations of cards could be an important thing to consider with some system builds.

whoops...I had actually made a small ascii table with the 5v/12v power draws of all significant graphics cards before they stopped using the 5v line heavily, but forgot to "Submit" it before closing my browser down last night 😊
I actually used a couple of the xbitlabs.com articles as the source of this little table and, while not setting these power draws in stone, could be used as a rule of thumb or a quick reference.

Will do it again and edit this reply 🤣

edit:
R9600Pro----450/760-------------3.3v: 3.32A-------------5v: 0.54A----------------12v: 0.54A (yes, both 0.54 it's not my typo 😜)
R9600XT----600/850-------------3.3v: 3.52A-------------5v: 1.17A----------------12v: 0.61A
R9800Pro----440/760-------------3.3v: 5.08A-------------5v: 6.08A----------------12v: 0.62A
R9800XT----470/840-------------3.3v: 3.84A-------------5v: 4.80A----------------12v: 2.71A

X800Pro----475/900--------------3.3v: 2.38A-------------5v: 2.33A----------------12v: 2.45A
X800XT----550/1250-------------3.3v: 2.71A-------------5v: 3.50A----------------12v: 3.42A

FX5700-----470/660-------------3.3v: 4.06A-------------5v: 1.12A----------------12v: 0.63A
FX5700u----530/1000------------3.3v: 2.90A-------------5v: 1.88A----------------12v: 2.52A
FX5900u----455/880-------------------------------------5v: 2.84A----------------12v: 3.84A
FX5950u----500/1050------------3.3v: 6.05A-------------5v: 5.51A----------------12v: 2.88A

GF6800-----375/1150------------------------------------5v: 4.26A----------------12v: 3.67A
GF6800U----440/1250------------------------------------5v: 4.48A----------------12v: 4.62A

Source: a few xbitlabs articles but I had forgotten to save all links yesterday
I left out the 3.3v rail if it was insignificant to me (think 0.5A or so)

RacoonRider wrote:

What's wrong with this method? http://www.xbitlabs.com/articles/graphics/dis … owercons_2.html

Sensitivity will be quite poor (they're using a low resistance to prevent excessive voltage drop). Hall sensors have very little parasitic impact on the circuit and are selected based on the desired current measurement range for the instrument resulting in superior sensitivity. With the low voltages that will be measured across the shunts, calibration will be an issue.

Even with the low resistance of the shunts, voltage drop across 3.3V lines could potentially throw delivered voltages out of ATX spec. Shunt resistors work well in the power supply where they're installed before the voltage feedback to the switching controller which allows for compensation for any voltage drop.

A shunt adapter would have to be made for every type of connector to be measured. ATX, 12V aux, PCIe, SATA, etc. A clamp multimeter can be clipped the 12V, 5V, or 3.3V conductors on any unwrapped cable.

Also, Uni-T as a "professional multimeter". That's a laugh.

gdjacobs,

Oh, you mean this thingy?

I didn't understand you at first. Yes, that's much better 😀 I just would not buy one specifically for the matter. Shunting is a cheapo way of doing this 😀

Or via adapter:
http://www.extech.com/instruments/product.asp … d=50&prodid=547

I disagree about shunting being the cheapo way. You can use inline Hall sensors in place of shunt resistors if you want to mod the wiring or build adapters. They start at 1.64 USD (one supplier) for an SMT part.
http://www.digikey.com/product-search/en/sens … =0&pageSize=500

RacoonRider wrote:

RacoonRider wrote:

PCBONEZ wrote:

You would need an ammeter. Ideally with an inductive pick-up. Even with the dedicated power cable(s) they still simultaneously p […]
Show full quote

You would need an ammeter. Ideally with an inductive pick-up.
Even with the dedicated power cable(s) they still simultaneously pull power from the slot so unless you are going to assume that part you need a more complicated test fixture.
Wasn't there a thread about that a week or so ago?
.

Not an ammeter, a voltmeter and a shunt resistor. An amperemeter would be most clumsy and unreliable, since you would have all the power go through it. You could use an amperemeter with a shunt resistor as well for pretty much the same result as a voltmeter and a shunt resistor.

What's wrong with this method? http://www.xbitlabs.com/articles/graphics/dis … owercons_2.html

Sorry I'm slow guys. Real Life keeps getting in the way lately.

RacoonRider wrote:

What's wrong with this method? http://www.xbitlabs.com/articles/graphics/dis … owercons_2.html

If I had a 6' board and a shovel with a 6' handle I would use the shovel to load the muck on the truck and they are using the board.
The board will get the job done but it's not the right tool for the job.

BTW: One name for a circuit where you measure voltage across a shunt is "ammeter". They are just doing it the hard way.
Rp is the shunt. Rg and Ig are the equivalent circuit for the meter movement when an analog meter is used..
http://hyperphysics.phy-astr.gsu.edu/hbase/ma … etic/ammet.html

I will admit that is the best go at it I have seen on a review site so far, but accuracy is going to be an issue.
The method could be made to work fairly well but their implementation as it is could only be called ghetto.

They have a lot of little problems when that all stacked up probably put their accuracy at like +/-20%, maybe less.
- They aren't using a high accuracy meter or parts to start with. Or high precision.
- Their ghetto shunts are too high of resistance. Proper shunts that can handle enough amps are like 0.000x ohms or less. [And their accuracy/tolerance is typically +/-0.5 % ] Their 0.03 ohms is 2 orders of magnitude too high for what is used in metering circuits.
- It looks like they calculated the 0.03 ohms vice measuring it. If they are going to use general parts that [probably] have a tolerance of +/-5 % they should have checked the actual value with a high accuracy ohm meter rather than just taking the number off the box (so to speak).
- Assuming the PSU is dead on-spec at 12.0v, the 12v the drop across 0.03 ohms is 0.36 volts so the video card would only actually be getting 11.64 volts. (The low side of ATX spec for 12v is 11.4v) So if the PSU 12v is even a little low the card will be undervolted. As overclockers can tell you undervolted parts use less watts.
- Similar situation for the other voltages.
- Their so called shunts use 4x 5 watt resistors. Total watt rating is 20 watts. Anymore watts and the resistors are overheated. Unlike true shunts standard resistors aren't particularly temperature stable so their resistance will be different making all the calculations wrong.
- Proper shunts are huge an bulky and why you think ammeters are too big but the thing is they are in a round-about way making an ammeter and if they want any accuracy they need to use the proper kind of shunts.
- The taping off pins and soldering wires thing works so long as the tape doesn't slide off and screw up a slot. A better way would be use a ribbon riser and tap into those wires. It would also avoid having to solder on every card you want to test.

I could probably think of more but that should be enough and I'm hungry.
.

PCBONEZ wrote:

BTW: One name for a circuit where you measure voltage across a shunt is "ammeter".

Good to know! In Russia we only have one word for everything that measures Amps, so the dictionnary left me kind of pointless as to what the difference between amperemeter and ammeter is.

Everything else, that I'm not going to quote, seems reasonable as well 😀

Another little tidbit.
Most DMMs I've had can measure amps directly so long as you keep it under 10 amps.
.

RacoonRider wrote:

Good to know! In Russia we only have one word for everything that measures Amps, so the dictionnary left me kind of pointless as to what the difference between amperemeter and ammeter is.

Basically a case of old vs new words for the same thing. I've seen ampmeter used as well.

Years ago capacitors were called condensers.
That is why the names of so many capacitor companies end in "con". ["con" is short for condenser.]
.

PCBONEZ wrote:

As overclockers can tell you undervolted parts use less watts.

Here you are wrong. A video card is not a simple ohmic power consumer at the 12V, but includes at least one switching regulator. When the input voltage gets lower the regulator increases the input current to be able to supply the same output voltage and current. This can go on until the input current becomes too high for the switching regulator and the protection kicks in.