VOGONS


Reply 20 of 21, by Tetrium

User metadata
Rank l33t++
Rank
l33t++
dulu wrote on 2021-02-22, 15:55:
Tetrium wrote on 2021-02-22, 06:33:

The majority of sA motherboards don't have the P4 connector on the motherboard, so once PSUs started to really move away from using 5v mainly and the supply of PSUs that could deliver the needed amount of 5v started to slowly dry up and age, this problem became more and more prominent.
If no decent 5v PSU is available, it will effectively rule out the vast majority of sA boards, including (afaic at least) all sA boards with universal AGP.

A question of curiosity - how many minimum amps on 5V rail should the power supply have, if we do not plan to overclock socket A?

Personally I always wanted to use a PSU with at the least 30A on the 5v rails for any Athlon, but with the lower TDP ones I guess one could get (again, my opinion and the way I personally picked my parts) away with a 25A one. This is with decent PSUs. The most used PSU brand I used is FSP because back then these PSUs were pretty solid and easy to find because lots of OEMs favored them.

Others may give you a different advice though, but especially with the many A7V133 and A7V boards me and my colleagues noticed that these boards really favored a PSU with heavy 5v lines or the system 'seemed' to be less stable. And all OEM/prebuild systems I saw the guts for myself and that had an FSP always seemed to have at least 30A (usually 300W PSUs with 250W and 235W PSUs of this brand being mostly confined to P2 and P3).

Whats missing in your collections?
My retro rigs (old topic)
Interesting Vogons threads (links to Vogonswiki)
Report spammers here!

Reply 21 of 21, by dulu

User metadata
Rank Newbie
Rank
Newbie

Palit 9800 pro 128bit has been installed for several days. After some time, I noticed that the power supply was clearly louder than usual. I figured the reason was the increased use of the 5V line over the ti4200 8x which I used before. I also got interested in the data on energy consumption on individual lines of the power supply provided in this thread

PSU for Athlon XP 3200+, Radeon 9800 XT?

I have a multimeter. I have never measured the amperage, and the topic of sA energy consumption is perfect for learning. In order not to extend the topic, I will only describe the final successful measurement attempts. Previously, I had a problem with computer reboots due to too small contact surface between the wires and the multimeter sockets.

The first idea was to measure the current separately on each of the four 5V lines in a 20pin plug. Effect - each line (during the measurement) consumed about 0.4A. Assuming that all 5V cables are connected in the power supply, it would mean that the overclocked platform uses 4 * 5V * 0.4 = 8W, which of course is not possible. Another thought - maybe this board used 12V line. (20pin plug has only one 12V wire) Measurement read 0.09A regardless of load, so no. Bad contact surfaces created additional resistance through which more current flowed through the remaining lines (i think so). Please do not call the fire department - I have finished my measurements and no building has burnt down.

20210227_162948.jpg
Filename
20210227_162948.jpg
File size
1.6 MiB
Views
44 views
File license
Fair use/fair dealing exception

Since all 5V wires in the molex plug come from one point in the power supply (I know it is so, I checked), they need to be connected, run through a multimeter and then lead to the 20pin plug. Theoretically, I could only connect one wire from the meter to the mainboard (the board started),or only one wire from PSU, but it would theoretically cause four times more current on the tracks, which could end up damaging the board. And the problem with bad contact of individual wires. Finally, for successful measurements it looked like this.

20210227_180308.jpg
Filename
20210227_180308.jpg
File size
1.49 MiB
Views
44 views
File license
Fair use/fair dealing exception

Here are the measurement results. It turned out that the multimeter is not suitable for what I do because ... I exceeded its measurement range.

Current on the 12V line in 20pin (unchanging, regardless of the load) - 0,09A

Def: AXP T-bred 2700+ 166x13 1,65V; ram 166 2:2:2:9 2,7V
OC: 187x13 1,8V; ram 187 2:3:2:9 2,7V

Q3A - Def: 12,6A ; OC: 19,3A ; Delta: 6,7A
SuperPI - Def: 12,4A ; OC: 18,5A ; Delta: 6,1
Doom3 - Def: 13,4A ; OC: out of scale
Prime95 - Def: 15,1A ; OC: out of scale

20210227_171028(1).jpg
Filename
20210227_171028(1).jpg
File size
1.28 MiB
Views
44 views
File license
Fair use/fair dealing exception

At least the energy measurements for the graphics card were easy. Everything is clear here. Desktop - 3.6A, games - 4.2A. 12V desktop - 0.75A

Final thoughts:
Assuming, with a slight margin of safety, that the difference between Def and OC in Prime is 7A, after OC - the platform takes 22.1A from the 5V line. By adding 4.2A from the GPU, we get a current of 26.3A, which means that for several days I have been exceeding the maximum current for my power supply 😀 Ofc assumed delta could be lower but discussed results are made for 1.8V, and I used 1.825V on a daily basis ... 😀