Which PSU rail is more import for CPU?

[hasu]

My AMD 6000+ machine with Antec 430W, sometimes reboots by itself when stressed, a typical behavior of electrically under powered system. What power rails are more important for CPUs? 3.3v or 12v? Does that differ among AMD/Intel CPUs or between older generation and newer AMDs?

Interestingly my AMD 6-Core 1100T used to run fine with 4 sticks of RAM and two hard drives with 280W small form factor PSU.

Edit: Both 6000+ and 1100T have the same TDP of 125W

 

[cbn]

12v.

P.S. Here is a good thread written by jonny guru on the topic of how multiple 12v rails are configured (if you wanted to look into this further):

http://www.jonnyguru.com/forums/showthread.php?t=3990

 

[Shehriazad]

Somewhat off-topic: Do you still game with that 6000+? I'd be curious how it does in modern games...

 

[VirtualLarry]

I would check the 6000+ mobo for swollen caps. Other than that, replace the PSU. And if that doesn't work, junk it, it's old, and you'll have a shiny new PSU left over for a new rig.

 

[hasu]

Somewhat off-topic: Do you still game with that 6000+? I'd be curious how it does in modern games...

No more gaming!
I use it to VPN to my office PC. I rarely use it but I still like it when I use !!

 

[hasu]

I would check the 6000+ mobo for swollen caps. Other than that, replace the PSU. And if that doesn't work, junk it, it's old, and you'll have a shiny new PSU left over for a new rig.

I tried with another Antec 450W and the problem persists. PSU may not be the culprit..
The machines shuts down by itself only when it is stressed (with HeavyLoad).

No more new rigs in the near future. I already have one of the fastest machine in the world i7-4790K! I never built a machine based on such a new CPU. I used to go for yesteryear's favorite. But this time I am glad that I opted for the latest and greatest!

But still AMD is my favorite like many other under dogs!

 

[Denithor]

Hmm, shutting down under load. What kind of temps you running while under load? If high, the first thing I would check is that the heatsink isn't clogged with dust (have seen thermal throttling/shutting down several times due to dusty HSF on customer computers). Blow it out with compressed air if dusty, see if that fixes it.

I would assume it's been quite a while since you mounted the HSF on the CPU (likely several years), sometimes the thermal transfer paste can "dry out" (not true drying, but close enough) or even get squeezed out from under the HSF, leaving an insulating layer of air instead of paste. If dusting doesn't fix the problem, time to remove the HSF and remount properly with fresh paste.

 

[tweakboy]

Depends good gear jasu.. gb

 

[hasu]

Hmm, shutting down under load. What kind of temps you running while under load? If high, the first thing I would check is that the heatsink isn't clogged with dust (have seen thermal throttling/shutting down several times due to dusty HSF on customer computers). Blow it out with compressed air if dusty, see if that fixes it.

I would assume it's been quite a while since you mounted the HSF on the CPU (likely several years), sometimes the thermal transfer paste can "dry out" (not true drying, but close enough) or even get squeezed out from under the HSF, leaving an insulating layer of air instead of paste. If dusting doesn't fix the problem, time to remove the HSF and remount properly with fresh paste.

Spot on! I just installed Tiger ONE utility for the WinFast motherboad and found that the temperature goes to 70 deg C under HeavyLoad. I wiped the heat sink clean and applied some arctic cooling that I had in my junk box and still CPU temperature goes to 70+ deg C. I guess the real solution might be to get a better HSF!

 

[VirtualLarry]

I've seen that a lot with OEM AM2 6000+ boxes. I cleaned the heatsink and re-pasted it, no big improvement. So I pulled out a 125W copper-cored 4-heatpipe AMD heatsink and used that. Surprisingly, no change either. Those CPUs just run hot, and seem to degrade (run hotter!) as they age, regardless of cooling.

Needless to say, not my favorite CPUs.

 

[flexy]

Are you overclocking the CPU?

While a sudden shut-off or freeze during stress testing MAY point to a problem with the PSU I wouldn't come to this conclusion too fast, especially if you're overclocking.

But if you are NOT overclocking and the system shuts down at stock speeds, yes I would strongly assume that something is off with the PSU.

I am sorry, 450W and 430W does not sound "too convincing" to me and the problem is in addition that the Antec is likely a multi-rail PSU, means it has two 12V rails which are made to draw like 15A each (18A peak) which is about 180W-ish per rail.

NOW - if the PSUs are older it can be that they degraded to a point where they cannot supply the CPU anymore, even NON overclocked. With a weak 430W PSU 10% degradation might be all that's needed....

TDP for a A6000+ is 125W, this does NOT leave a lot of headroom on a rail where 180W would be the maximum and of course not if the PSU is older. Throw in a few drives, fans etc...and you are VERY close to what this PSU can handle.

And by the way, your PSU not only has to supply your graphics card (which is LIKELY on the second rail)....but also your board, on-board devices, pcie slots, sound card etc. and as already mentioned fans, drives and what not. If there is only one rail rated for 15A constant current and 180-ish watts.....uhm....DO THE MATH.... (especially if you happen to share the one rail where the CPU is connected to with the board or whatever else).

This is *interesting* since I just in the recent days read a lot about multi-rail and single-rail PSUs and how *allegedly* multi-rail PSUs are not sufficient anymore for "modern systems".

(This is nonsense, by the way, I am running a Haswell system on a 8 year old Toughpower 750 with FOUR RAILS, best PSU on Earth. I gave a brand-new XFX 750W Core Pro to my wife since I prefer my old Toughpower

BUT: Here is the deal: A weaker PSU like yours has not four rails where ATX board power, PCIE, 12V EPS, power for drives etc. are spread over four rails (which would be fine)...but ONLY TWO rails. And this is where the problem is, IMHO.

 

[hasu]

Are you overclocking the CPU?

While a sudden shut-off or freeze during stress testing MAY point to a problem with the PSU I wouldn't come to this conclusion too fast, especially if you're overclocking.

But if you are NOT overclocking and the system shuts down at stock speeds, yes I would strongly assume that something is off with the PSU.

I am sorry, 450W and 430W does not sound "too convincing" to me and the problem is in addition that the Antec is likely a multi-rail PSU, means it has two 12V rails which are made to draw like 15A each (18A peak) which is about 180W-ish per rail.

NOW - if the PSUs are older it can be that they degraded to a point where they cannot supply the CPU anymore, even NON overclocked. With a weak 430W PSU 10% degradation might be all that's needed....

TDP for a A6000+ is 125W, this does NOT leave a lot of headroom on a rail where 180W would be the maximum and of course not if the PSU is older. Throw in a few drives, fans etc...and you are VERY close to what this PSU can handle.

And by the way, your PSU not only has to supply your graphics card (which is LIKELY on the second rail)....but also your board, on-board devices, pcie slots, sound card etc. and as already mentioned fans, drives and what not. If there is only one rail rated for 15A constant current and 180-ish watts.....uhm....DO THE MATH.... (especially if you happen to share the one rail where the CPU is connected to with the board or whatever else).

This is *interesting* since I just in the recent days read a lot about multi-rail and single-rail PSUs and how *allegedly* multi-rail PSUs are not sufficient anymore for "modern systems".

(This is nonsense, by the way, I am running a Haswell system on a 8 year old Toughpower 750 with FOUR RAILS, best PSU on Earth. I gave a brand-new XFX 750W Core Pro to my wife since I prefer my old Toughpower

BUT: Here is the deal: A weaker PSU like yours has not four rails where ATX board power, PCIE, 12V EPS, power for drives etc. are spread over four rails (which would be fine)...but ONLY TWO rails. And this is where the problem is, IMHO.

I have a new high wattage PSU bought during Christmas sale. I think it is 750W or so. But its per rail current for 12v is only 18A. That PSU has more than two rails for +12v though. IIRC, Antec 450W has 17A per +12v line. In fact, I already put the old PSU (430W) back into the system. Even though this machine itself is old, the actual running time might have been much less. It was sitting on a shelf for the last 5 years or so!

I think the real reason for the shutdown is the heat (it just shuts down, not a restart and no BSOD), and the stock fan (with heat pipes) is unable to cool down the CPU as fast as it should under load. Now, with the Tiger monitoring program (WinFast) installed, it starts beeping once it crosses 70 deg C.

 

[Xpage]

Can you try to undervolt the CPU, even a small amount would make a large difference in temp. Or drop the clocks slightly.

AMD CPUs usually have a lot of extra voltage that they do not need and underclock well.

 

[VirtualLarry]

I think the real reason for the shutdown is the heat (it just shuts down, not a restart and no BSOD), and the stock fan (with heat pipes) is unable to cool down the CPU as fast as it should under load. Now, with the Tiger monitoring program (WinFast) installed, it starts beeping once it crosses 70 deg C.

Like I tried to indicate in my prior post, these CPUs (A64 6000+, especially the 125W model), were heat monsters. And they seemingly get worse with age. (More leakage?)

Get an Artic Cooler Freezer 64 Pro on there, or something like it. That might be your only solution.

 

[flexy]

I have a new high wattage PSU bought during Christmas sale. I think it is 750W or so. But its per rail current for 12v is only 18A. That PSU has more than two rails for +12v though. IIRC, Antec 450W has 17A per +12v line. In fact, I already put the old PSU (430W) back into the system. Even though this machine itself is old, the actual running time might have been much less. It was sitting on a shelf for the last 5 years or so!

I think the real reason for the shutdown is the heat (it just shuts down, not a restart and no BSOD), and the stock fan (with heat pipes) is unable to cool down the CPU as fast as it should under load. Now, with the Tiger monitoring program (WinFast) installed, it starts beeping once it crosses 70 deg C.

"Only" 18A is normal, MOST multi-rail PSUs have "only" 18A per rail which is totally fine, the 17A or the 18A per se are NOT the problem. (The problem would be if the PSU has only 2 rails 18A each and the system draws close to this power). So does the problem also occur with the 750W PSU with the 4 rails?

 

[hasu]

"Only" 18A is normal, MOST multi-rail PSUs have "only" 18A per rail which is totally fine, the 17A or the 18A per se are NOT the problem. (The problem would be if the PSU has only 2 rails 18A each and the system draws close to this power). So does the problem also occur with the 750W PSU with the 4 rails?

Well.. this particular machine has 4 SSD's, 4 sticks of DDR2 and a fan-less video card. So it is extremely unlikely that it would take more than 430W. I had a 1100T which ran fine on 280W SFF PSU with 16GB DDR3, 2 desktop hard drives and a DVD drive.

I also suspected PSU initially but now I think it is the heat that causes the CPU thermal shutdown. I think VirtualLarry is correct that this CPU is aging differently compared to other CPUs. May be AMD X2-6000+ has properties of a Germanium transistor.

Edit: Actually the new PSU is 650W with 4 rails for +12v each with only 15A. So the best bet was Antec 450 with 18A per rail. The problem existed with Antec 450 as I mentioned.