Overclocking and PSU

[lopri]

I recently found out that my PSU is the bottleneck of my overclocking, and it brought me a question that I wanted answered. Frankly I am happy with my current overclock and do not intend to take apart my rig for a PSU swap, so consider this issue purely theoretic. Without a further ado, my rig consists of:

Xeon 3060 @3.6GHz/3.7Ghz (depends on my mood )
P5W-DH
Team DDR2-667
OCZ GameXtream 700W (Four 12V rails with 18A max individually, and 600W combined)

The PSU handles up to 3.70GHz (9x411) excellently. But once I raise the FSB to 422, making the CPU frequency 3.8GHz, it starts to show its limit. I first thought I hit the limit of either my CPU or RAM but the growing evidence is now evident. Prime examples are spontaneous shutoff/reboot while under heavy load. With 1.60Vcore and 2.40Vdimm, applications never crash to desktop indicating that the other components are able to handle the load, but PSU inevitably fails, usually 10~20 mins into the 100% load on both cores. (Temps are well taken care of) Similar symptom occurs when memory is clocked very high, too. (>1.05GHz)

My question is: Is this a good enough reason to warrant an RMA? My reasoning is that in no way the CPU/memory load can max out the wattages my PSU specifies. Then again I'm pushing the CPU/memory way beyond their specs. But then again people buy way bigger PSUs for overclocking.

If anyone works in PSU industry I'd appreciate your enlightenment!

 

[n7]

How do you come to the conclusion this is a PSU issue?

Everything you've stated indicates a CPU/RAM being maxed out beyond limits issue, not PSU.

I'd love to blame OCZ, but if you are trying to say an OCed Xeon/RAM is drawing more than 18A...let's just nicely say i strongly disagree with your analysis.

 

[cmdrdredd]

Your memory is rated at 667 and likely needs more vots than you're giving. Try 2.3v and see what happens. Also loosen the timings slightly to say 4-4-4-10 and work from there.

 

[n7]

Hehe, he already runs them over 2.3V no worries

I still cannot understand how he attributes the instability to a PSU.

When i clock too high for CPU/RAM, i get random reboots too.

So i lower OC...not blame my PSU

Who knowz though.

 

[Baked]

You have DDR2-667 RAM. Enough said. Being Team memory doesn't mean a damn thing. You're running it way above it's rated spec, what else do you think would happen? 700W PSU ain't enough? Please.

 

[lopri]

OK Guys. You guys got me totally wrong! Ugh.. Let's just say it's the PSU. (Are you saying it's not possible for PSU to blame?) When PSU becomes a bottleneck of overclock, is RMA request justifiable? Christ, I told you it's theoretic!

 

[n7]

I'm confused (& alchohol isn't helping me)...

I would say that in many cases, a poor PSU could indeed be an issue.

But you have an excellent PSU, which really does not appear to be the problem as far as i can tell.

So what exactly are you trying to get across in this thread?

 

[lopri]

Say your CPU/memory/board are proven @3.40GHz and you somehow diagnosed that your PSU is to blame for less-than-ideal 3.20GHz. Your PSU is rated 700W and you don't think you're maxing out the wattage. Now, the 12V rails can be a delicate issue as you pointed out, but 12x18=216 and we're not counting video card yet. (And I believe memory sucks up from 3V rails) Is this situation an acceptable excuse for a PSU RMA?

 

[n7]

Sure.

If your PSU is actually causing issues with OCing, i'd be all over RMAing it.

Key word being if.

BTW, since your PSU is actually the identical to mine, just a different color & name on it, here's how the 12V rails are distributed in case you didn't already know:

12V1 = CPU
12V2 = CPU2/PCI-e2
12V3 = mobo/peripherals
12V4 = PCI-e1

 

[lopri]

Thank you. But the real question is, "Will PSU makers regard such an RMA request as valid?", assuming you tell them as it is.

 

[n7]

Heh...that's a good question.

I suppose you'd somewhat be at their mercy then.

OCZ seems pretty decent though; i imagine they'd take care of you.

 

[Raider1284]

I would really doubt you could RMA due to overclocking, same reason you cant rma a cpu if it doesnt overclock well. Overclocking requires alot more power, but your PSU only has to work at the normal default speeds. Unless the psu is failing when your not overclocking it's not defective. Some companies, like EVGA for example still cover their stuff even if you modify it/run it past default specs, so if OCZ is like them, you could probally get an RMA.

 

[Howard]

That PSU has a lot of ripple (which MAY affect OCing) on the +12V at high loads, but I doubt your rig is actually drawing that much power.

 

[lopri]

The real question is whether PSU vendors will issue an RMA. I think they should, as long as the power usage is under the specified capacity regardless of whatever the hell a user does within the given limit. In this case, the PSU is supposed to supply 218W from the 12V1 rail, and no ammount of OC'ing will achieve such high wattage consumption? I'd really like an explanation from someone who's involved in the PSU industry.

 

[Duvie]

Originally posted by: lopri
Thank you. But the real question is, "Will PSU makers regard such an RMA request as valid?", assuming you tell them as it is.

I would say NO if you put it in those terms....The best you can do is check the rails with a Multimeter and if you find it doesn't deliver its ratings you have a better chance.

The fact is 700watts according to PSU charts is more then enough for any dual core setup today if not for a quad core. You would need far better proof.

This ocing you have should not overtax this PSU unit unless it was just plain faulty. In that case checking the rails would result in you having the proof you need to RMA...without saying OCING....

The fact is ocing isn't a factor. You are not overlclocking the PSU. You are just increasing the necessary power needed from the components. If you can attribute to not enough power is being delivered even though 700watts should be ample then you have a case. Proving that is tougher, since the motherboard could be the reason for what it is doing with the power it receives form the PSU...

Remember cooling can effect PSUs as well...The hotter they get causes less efficiency. The lower the efficiency the less power that is actually delivered...Most PSU are 80% over 40c...Therfore meaning the most your PSU could deliver would be 560watts. Howevere I still think your PSU should be ample for this setup unless it was faulty...NOW PROVE IT...


I have more power hungry systems on 550watt and 600watt PSU's

 

[Duvie]

I just looked~!~!...The 3060 is a conroe derivative or basically a E6600....3.6-3.7ghz is far higher then most get...

Quit whining....I dont think you have a case or can prove it is the PSU....

 

[Thor86]

LOL, any component that runs fine at spec cannot be RMA'd because it doesn't run out of specifications.

 

[Mr Fox]

Originally posted by: lopri
The real question is whether PSU vendors will issue an RMA. I think they should, as long as the power usage is under the specified capacity regardless of whatever the hell a user does within the given limit. In this case, the PSU is supposed to supply 218W from the 12V1 rail, and no ammount of OC'ing will achieve such high wattage consumption? I'd really like an explanation from someone who's involved in the PSU industry.

Industry Standard is +/- 10 % If you can show higher variation....they should RMA it.


The Ripple that you are seeing is the wall.... and it might just take a bump.... or maybe not... you will know because it will get ugly thermally......


if you don't already have here is Intel Thermal Analysis Tool : http://www.techpowerup.com/downloads/392/mirrors.php

A Great tool to Allow CPU Temp Logging (Both Cores)..... it will dump Data Into an Excel File... that can be graphed and broke down.

One other Thing.... What have you done with the North and South Bridge ? The stock solutions suck.... They maybe a part of your issue....

My 3060 is good at 3.4 & 3.6 .... I have not had the Gonadz to go Higher

 

[Sunrise089]

OP - what makes you think it is your PS?...since it probably isn't IMHO.

 

[lopri]

Originally posted by: Duvie
I just looked~!~!...The 3060 is a conroe derivative or basically a E6600....3.6-3.7ghz is far higher then most get...

Quit whining....I dont think you have a case or can prove it is the PSU....

Hey I'm not whining. This is more about a fundamental. I want to know what's the prerequisite for a PSU RMA.

 

[lopri]

Originally posted by: Duvie
The fact is ocing isn't a factor. You are not overlclocking the PSU. You are just increasing the necessary power needed from the components. If you can attribute to not enough power is being delivered even though 700watts should be ample then you have a case. Proving that is tougher, since the motherboard could be the reason for what it is doing with the power it receives form the PSU...

I think this summarizes my understanding as well. Forget about how high my OC is, but what if I can achieve 3.8GHz by swapping the PSU? Then do I have a case?

 

[lopri]

Originally posted by: Sunrise089
OP - what makes you think it is your PS?...since it probably isn't IMHO.

• 1. System is stable @3.7GHz. (VCore 1.55V set in BIOS but it drops to 1.52~1.53V under load)
  2. Temps are well checked. At 3.6GHz/1.5V, CoreTemps reading rarely goes above 60C under dual Prime95. Under 70C @3.8GHz.
  3. CPU responses to the Vcore increase (up to 3.8GHz) almost linearly, which gives me an impression that the CPU has room to grow as long as temps are in check.
  4. While benching @3.8Ghz, application never gives me errors nor crashes to desktop. For instance, if I run Orthos it runs without erroring out until the problem occurs.
  5. Then suddenly within 10~20 minutes into heavy loads, the system shuts off / reboots. This also happens while running 3DMarks @3.8GHz.
  6. Memtest (both under DOS and Windows) run fine forever @3.8GHz, as long as CPU isn't under 100% load.

Above described is the symptom I'm experiencing. But again, I was more curious about how one can assure the quality of PSU, not how to get a new PSU.

 

[Duvie]

Originally posted by: lopri

Originally posted by: Sunrise089
OP - what makes you think it is your PS?...since it probably isn't IMHO.

• 1. System is stable @3.7GHz. (VCore 1.55V set in BIOS but it drops to 1.52~1.53V under load)
  2. Temps are well checked. At 3.6GHz/1.5V, CoreTemps reading rarely goes above 60C under dual Prime95. Under 70C @3.8GHz.
  3. CPU responses to the Vcore increase (up to 3.8GHz) almost linearly, which gives me an impression that the CPU has room to grow as long as temps are in check.
  4. While benching @3.8Ghz, application never gives me errors nor crashes to desktop. For instance, if I run Orthos it runs without erroring out until the problem occurs.
  5. Then suddenly within 10~20 minutes into heavy loads, the system shuts off / reboots. This also happens while running 3DMarks @3.8GHz.
  6. Memtest (both under DOS and Windows) run fine forever @3.8GHz, as long as CPU isn't under 100% load.

Above described is the symptom I'm experiencing. But again, I was more curious about how one can assure the quality of PSU, not how to get a new PSU.

1) So...Most boards and Asus especially have had "vcore droop" under load for years....This is the mobo not the PSU. My system sets bios andundervolts .02-.03v at idle and load drives it down .01-.02v...pretty standard

2) temps are not reliable...sorry to say...70c is too high IMO anyways and 60c is different per board. Too many variation in temperature monitoring from manufacturer to manfacturer.

3) Nothing in computer hardware works linear...throw that notion right out the window...If anything vcore increases to "stable" clock speed is more like an inverse square graph...It has been this way for years...First .025v gives you multile fsb...next is less...next is even less...until you get to ap point where .025v gives you like 1-2mhz of fsb....Nothing new here

4) Hence while ppl run orthos or stress testers...Components dont always fail instantaneously...some take a period of time to be stressed...hence memory is a big contributor to symptoms like this...

5) Reboots can be attributed to system errors...do you have the default setting of automatic restarts turned off in windows XP? This is on by default and can mask BSOD...which are often related to memory or chipset issues resulting from out of spec...
In the past for me, reboots from a PSU is from one failing hard, and often times will keep doing this even at idle and sort of looping...Most power issues from pushing it to far in OCing has always resulted in just a straight up freezing or hard lock. memory is usually BSOD, and actually temperature is usually reboots....

6) Again how does that prove PSU???? It doesn't...



Dont mean to pick on you but you haven't shown anything that definitely points to or even helps make your claim that the PSU is the culprit...

I will give you a more likely culprit then the PSU based on your symptoms...

MOtherboard chipset!!! Have you given it more voltage? Are you actively cooling it?

 

[bjp999]

Theoretically speaking, I doubt that a manufacturer would refuse to give you an RMA if you reported that the PSU was flakey and causing spontaneous reboots. Doing it a second time would be a challenge! I think this is a moot point since the expert concensus is that your PSU is not the cause.

 

[GuitarDaddy]

I would say the answer to your theoretical question is no. Failure to produce an expected or comparible(to another PSU) overclock would not be a valid reason for an RMA.

Manufacturers guarnantee performance at a rated level, if you can prove that the PSU becomes unstable within the rated wattage and amperage levels then you have a valid RMA.

In a case like you describe it could well be that the PSU's failure to supply stable power at it's rated level is the cause of poor overclocking, but justifing an RMA would require proving (or claiming) that the PSU failed at its rated level. Just saying that it is bottlenecking your overclock, or doesn't overclock as well as brand X will not fly as a valid justification IMO