Psu

Since no1 can answer my saoftware question i'll have a go at the hardware & work back :confused

Can anyone suggest a way to test a PSU? Basically at gaming load (not 3DMark 06 load :confused) my PC has a tendacy to freeze or restart.

I suppose the only reason for it to cut out is a lack of amps on the 12V rails, how would I go about finding out & what will I need to do so?

:major2-Bit-Geek:major

 

Use a MultiMeter you can buy one at local hardware store, I prefer a digital
http://www.youtube.com/watch?v=KzjMIcER4EU
http://www.sears.com/shc/s/s_10153_...d&site=&creative=3514039735&OVKEY=multi-meter

 

Yeah, the digital multimeter is a good way to go. Also, most PSUs have the specs printed on the side of the unit. A minimum of 18a or 19a per rail is recommended in a multi-railed system. If there's only 2 rails, at least 22a is nice. If you only have a single rail, I'd recommend replacing the unit unless it's one of the modern higher-end units with 50a+ on the single rail. For example: I'm running an Antec 650w with 3 rails: 22a, 22a, and 25a. I have zero issues with 2 SATA HD, 2 optical drives, AMD X2, 4gb DDR2, GTX260 c216.

 

I didn't have a problem with the PSU until I put more RAM in it I think, but when I got a new CPU it didn't get worse :confused

Old Syetem:
ASUS M3A: http://usa.asus.com/products.aspx?l1=3&l2=149&l3=592&l4=0&model=1934&modelmenu=1
AMD Phenom X4 9750 2.4GHz @ stock:http://products.amd.com/pages/DesktopCPUDetail.aspx?id=397
ATI Radeon HD 4870 1GB @ 800MHz & 1000MHz: http://www.sapphiretech.com/presentation/product/?psn=000101&pid=179
2GB DDR2 Kingston Hyper X 1066MHz in Dual Channel (KHX8500D2/1G): http://www.kingston.com/hyperx/products/khx_ddr2.asp
SATAII Samsung 320GB HDD
SATAII Samsung 160GB HDD
CoolerMaster Extreme Power Duo 650W: http://www.coolermaster.com/products/product.php?language=en&act=detail&tbcate=22&id=5924
X3 12CM & 1 8CM FAN
Typically 1 USB 2.0 decive installed at all times

I simply got another 2GB of the Kingston Hyper X RAM & I started getting random shut downs during games, but not 3DMark :confused. Then I added an AMD Phenom II 940 & overclcoked it to 3.4GHz but it didn't cause the crashes to be anymore frequent: http://products.amd.com/en-us/DesktopCPUDetail.aspx?id=509

Any idea why it would stop working & how I might go about fixing it?
The attached pic is of the PSU's vital stats :-D, not quite the same as the one in the link becasue Coolermaster don't have the Exteme Power (DUO) 650W on their website :confused

Thanx for you're help so far, it's much apreciated

:major2-Bit-Geek:major

 

Well, you have plenty of power and if the PSU is good then the next obvious is the memory, run memtest86+ to check your RAM or pull out the new sticks and try it that way. Preferably overnight. About the only other thing I can think of is excessive heat on the GPU.

 

Thanks augiedoggie,
I'll give memtest a go overnight and tomorrow i'll post... the results, i've never used it so lets find out what it does exactly :-D

I will put everything back to stock clocks for the test (CPU & GPU) but leve the RAM as that is supposed to be able to run at 1066MHz.

As for the GPU overheating, it's been an excellent card nice 'n' cool :cool
It runs at 40C idle & it is rated to run at 100C just fine

*I wish I had a way of finding out the load temps*

:major2-Bit-Geek:major

 

You can not properly test a PSU with a "multimeter".
All you are going to test is output at a idle.
In order to be tested PROPERLY a PSU has to be stress tested under a full load with proper equipment.

A PSU is under it's fullest load & strongest power consumption during bootup .

A PSU tested with a multimeter can give the false appearance of being OK & be dropping power under a load.

 

Actually, you can test a PSU under load with a multimeter. You can test primarily the 5v and 12v lines, but it can be done....

 

ok I ran mem test overnight on te 2GB of RAM I know to be good & it is good :-D

I then tested the other 2 sticks individually, the older one was fine but the 4th stick had some errors I have pics as I have no idea what any of it means :confused

So the newest stick of RAM is probably the culprit, how could a stick of RAM be bad yet work fine until stressed?

Sorry about the quality of the photo's :-o

:major2-Bit-Geek:major

 

All I can tell you is that 'Zero" errors is what you want. When you stress components you unearth flaws which is why people run things like Prime95 etc... when they buy/build a new system to see if there are weak components so that after a 24hr. period of testing then one can be pretty sure of having a good rig. Oh, since you mentioned OC'ing then that might be what sent the RAM to the RMA pile.

 

I was going to leve the newest stick in for longer but since it showed errors after about an hour of running I took it out :confused

Oh no, what I meant by overclocking the RAM is that it comes as standard as 800MHz you need to setup 1066MHz the BIOS, it is in fact DDR2 8500 though.

As for the CPU & GPU I only increased the multiplier to get a higer clock 3.5GHz with 17.50 so far X18 won't boot :confused & the GPU was overclocked using Catylist Control Centre by ATI, I also upped the GPU fan speed to 60% to keep it cool :cool

If I take the RAM back to the shop & they say it's ok (which they usually do when I take stuff back :banghead) what then?

I also thought maybe the motherboard may be the cultprit as alot of people have support problems with the ASUS M3A just getting CPU's like my PII 940 to work. Is their any way my mobo could be at fault?

:major2-Bit-Geek:major

 

Ya, that RAM is bad or perhaps the socket itself is bad so try it in another slot but if another stick worked in the same slot then the socket should be OK.

I'm not an OC'er so I can't help you there but OC'ing does stress stuff especially if you up voltages.

Well, I haven't used memtest86+ for a long time but maybe there's a way to print it out IDK, anyways tell the store that you used the memory test and show them a pic of the results and they should give you a new stick.

As to the mobo, run Prime 95 and take pics of that too with your rig @ stock speeds or OC'd for your own comfort level, your OC might boot properly but that doesn't mean it's really stable. Good luck.

BTW, don't tell the store that you OC'd.

 

I need to do more testing but I definately have a problem with either the slot or the stick itself, I will have a go with just 2GB installed (the good ones), I will try them in both the first & second channel to see if I have good slots :-D

Voltage wise with the ASUS M3A it is set the Auto becuase the board only overclcoks when set to auto becuase the available voltages are too low if you know what I mean :confused, so no voltages have been altered.

I had a quick go with Prime 95 & when my CPU was at 3.5GHz up from 3.0GHz it crashed instantly LOL but after returning my CPU to stock I got some results, I ran it twice & have attached the "results.txt".

Oh I know OC'ing invalidates any garentee LOL, so does replacing the stock cooler for both AMD & Intel :confused

Thanx for the help so far 1st class stuff :highfive, I am actually getting results rather than relying on paranoia & frustration :-D

:major2-Bit-Geek:major

 

Ok so I did a few tests with Prime 95's torture test & here is what I have.

Attached I have some more details I wrote in notepad during & after the tests, bearing in mind I could only give 20mins to each test or until it reported errors which was around 5mins.

So according to the results all 4 slots work fine but the first two stick of RAM I got don't work in the second channel (A2 & B2). I think I am even more confused now LOL

Any ideas anyone?

:major2-Bit-Geek:major

 

Do as augie suggests, and if they are still reluctant, most RAM these days has a lifetime factory warranty. Visit the manufacturer's web site and submit an RMA request and they will replace it. On this same note, the shop should know this and should have no problems exchanging it since they can also replace it under factory warranty.

 

The memory...
Many times manually setting your memory timings & voltages will correct errors, mobos don't always make the best choices went set to auto especially when using multiple sticks and brands that support different timings...
It helps to know exactly what chips your men sticks are made of, so you know what voltage a timings they support, even memory sticks of the same brand can be make from different chips supporting different timmings and voltages

If you having issues with your PSU like fluctuating voltages this can make your memory error

I'd suggest testing your PSU with a multimeter, I wouldn't trust software but you could run Everest or similar to monitor voltage fluctuations just take the values with a grain of salt...

FYI in the past... Coolmaster has been at the bottom of the food chain quality wise

The PSU is the heart of any PC, Twice as important when OCing, funny how much some people will spend on componets only to cheap out on the PSU:confused

Good Luck...:major

 

The problem is I have no idea which sticks are faulty :confused

I would normally agree but as soon as I remove all my settings from the BIOS (timings etc) the errors seem to have dissapeared, I ran prime 95 one last time with all 4GB in at once & it ran perfectly for 9 hours according to the logs :confused

The name of the RAM is further down the thread except there is 4GB of it being used all the time rather than 2, X4 1GB DDR2 Kingston Hyper X 1066MHz (runs error free at 800MHz). They have timings suggested here at MG (5-5-5-15) also suggested by Kingston is to set the voltage to 2.2V but I didn't need to because it ran at 1066MHz using auto voltage.

I think my Grandad has a multimeter both analog & digital versions

Again I was suggested that exact CoolerMaster here at MG :-D
I needed a new one because my old PSU didn't have good enough rails to support my ATI 4870 1GB.

I'd say it's because PSU's are more complicated than people realise, many people simply think how many watt's do I need, some don't even think about that :-D I didn't use to give a thought to the PSU only the wattage, thankfully I know better now though :-D

:major2-Bit-Geek:major

 

As long as you ignore the few who actually provided a useful solution, then you will remain confused.

You were advised to use a multimeter. Normal is for a defective supply to boot and run a computer. Until you know the supply 'system' (which is more than just a power supply) is good, then everything else done can only be 'it might be good'.

You are confused because you do not have definitive answers. Even Prime95 will not report what you needed to know - definitively. First establish the power supply 'system' is good - definitively - without doubt. That means the multimeter - and ignore the many who do not understand how a power supply system works.

For example, power supply does not draw maximum power when first powered. But those who never designed power supplies believe that myth. I designed and built power supplies. Get the meter. List to the few who suggested up front you best advise.

 

I am not aware acting on advice is classed as ingoring as :confused

I have systematically follwed & acted on all advice, except tesing the PSU itself with a multimeter because I have not had a chance to borrow one yet. I would buy one but I only have... 4p in my poket atm LOL

I'm wondering why you joined MG to reply to this particular post :confused

:major2-Bit-Geek:major

 

Getting (borrowing) is where all efforts should be directed. Executing Prime95, for example, does nothing useful. As the very first posters noted, you need the meter, and you need the meter now.

If the power supply 'system' is not confirmed good, then anything else can act defective while really good. Without a stable power 'system', then other tests cannot provide a reliable answer. What value is other testing when even good components can report failure? Until voltages are established, other tests an only report 'maybe'.

 

If you dio get a multimeter you still aren't going to be able to fully test a power supply under a load.
The best way is to take the PSU to a shop & have it tested properly.

Or, simply try a known good PSU in the problem unit.

 

A defective power supply can boot a computer. How does that shop tell anything more? Do they have the many thousands of dollars in test equipment we had to design and manufacturer supplies? Of course not. Most computer repair techs have no electrical knowledge. Electrical knowledge is not required to pass the A+ Certification tests.

Will the shop use a power supply tester? Routine is for those testers to report a defective supply as good. Those testers can identify some defective supplies. But cannot report a supply as good. Many computer repairmen do not know that nor know why.

By far, VDC numbers from that multimeter with system under maximum load will provide the only definitive - no doubt - answer. Best when those numbers are posted here so that additional information hidden in those numbers can be extracted. Most don't know how much a multimeter reports.

Power supplies are complex items. But are also some of the most robust devices in the building.

Take a typical computer grade UPS as an example. In battery backup mode, its 120 volt output can be two 200 volt square waves with a spike of up to 270 volts between those square waves. That would be some of the 'dirtiest' electricity a computer will ever see. So 'dirty' as to be destructive to small electric motors or power strip protectors. But completely acceptable to any computer power supply.

Meanwhile, numbers from the multimeter when supply is under heavy load will report more than most computer repair shops can do. After all, the best computer repair shop will only do what we engineers have taught them - use a 3.5 digit multimeter when supply is under maximum load.

If additional load from memory causes power supply failures, then the always defective power supply could have been identified long ago using the multimeter.

 

I have no idea how to put my PSU at max load, but I think a good place to start would be to recreate the situation that caused the PC to restart. Play a game while having someone takes reading every couple of mins, once an hour comes & goes without any hicup I can safely say it's probably fine.

I say an hour because thats longer than I could play high end games before I got a restart (thank god for autosave LOL).

It seems to be fine sice I swapped around the RAM though :confused
I will still see if I can get my hands on a multimeter tomorrow to test the PSU under load.

Thanks for you're help guys I will report my findings when... I have some :-D

:major2-Bit-Geek:major

 

That's the right idea. Maximum load does not exist during startup. Maximum load can be obtained by multitasking to all peripherals simultaneously. For example, play complex video graphics (ie a movie), while downloading something from the Internet, while a camera is powered from a USB port, while playing sound loudly, while searching the hard drive, while reading a DVD, while ... Now that is a maximum load.

Moving on. Is memory defective? Don't recall if anyone mentioned Memtst86. That test can be executed in room temperature. Then repeated using a diagnostic tool. Execute Memtst86 in a 100 degree F room or with memory heated by a hair dryer on highest heat setting. Heat is a diagnostic tool. An ideal temperature to any semiconductor. But weak and getting worse memory may fail Memtst at this higher and otherwise ideal temperature. Now you have numbers from Memtst that ID a memory location, a data line, or an address line - often to only one memory board. Not just what failed. Also important is why and what makes the intermittent failure hard.

If your problem went away by swapping, well that is classic shotgunning. The problem remains. Symptom simply cured or masked.

For example, how does memory fail? A memory location only changes when unique values are loaded in adjacent memory cells. Move that memory and those unique values no longer exist in adjacent cells. Did you cure it? Of course not. Swapping (shotgunning) only masked a failure which will probably get worse with age.

If voltages are bad, some memory locations may become marginal - cannot save or read data at full speed. In this case, the memory appears bad because something is defective in the power supply system.

Essential to tracing any such failure are numbers - to follow the evidence.

 

I got myself a digital multimeter & have been messing around with it to see what I can find out

But the first real test was Pime95's tirture test 2 (max heat, power consumption, some RAM tested). I got a molex & pulgged in & got some odd readings :confused

*Red & Blk wires 5.17V (shouldn't that be 5.5V?)
*Yellow & Blk wires 11.89/90V (that sould be 12V?)

Can anyone shead some light on these results & give me any pointers on how to test my PSU properly (what to plug the multimeter into).

I am a noob with multimeters LOL

:major2-Bit-Geek:major

 

Connect black lead to the chassis. Red probe wires in the nylon connector between supply and motherboard.

Touch the purple wire. It should read a three digit number around 5 volts.

Now, Prime95 does nothing useful. To create a load, multitask to many or all peripherals. For example, play complex graphics (ie a movie), while downloading from the internet, while playing sound loudly, while talking to something on USB ports, while searching the hard drive, while, ...

Now you are read to read voltages. Again read the purple wire. Then repeat by reading voltages on any one of red, orange, and yellow wires. All readings to three significant digits.

 

Ok I found some time to do some proper testing on my PSU

To load up the PSU I ran alot of things all at once like you suggested:

*A movie in VLC player
*A movie in windows media centre
*A music video in WMP
*A music video in Real Player
*Downloaded a catylist driver
*Started a Deep Optimise & defrag in Smart Defrag
*Started Burnout Paradise (no disk required & it's a legal copy :confused)


I couldn't think of anything else to do that would run for long enough. Here are my results:

*Purple wire: 5.11V
*Yellow wire: 11.89V
*Grey Wire: 5.17V
*Red wire: 5.16V
*Orange: 3.38V

If I have missed anything I could do it again I guess. I used a black molex wire as a ground.

Everything has been running just fine since I turned the RAM down from 1066MHz down to 800MHz (stock). Could it be the motherboard that is at fault? I know the M3A is stretched past what ASUS supports CPU wise :-D but many posts I have read have pointed out alot of compatability issues with lower CPU's than mine & with high end RAM like mine, I know it seems like a obvious answer but my computer only started acting up once I installed another 2GB of RAM added to the 2GB that was already there :confused

I suppose we can get to that next, once we rule out the PSU :-D

:major2-Bit-Geek:major

 

You will also find the green wire at near zero when power is applied. Also useful would have been the purple wire when computer is off.

However you power supply is fully functional.

You might look at the current numbers for 3.3, 5, and 12 volts. Your 12 volts is more heavily loaded - closer to its maximum numbers than the other voltages. But those voltages say your power supply is perfectly acceptable.

Forgot what was to be answered. If I remember, the only question was whether the power supply was sufficient. Quite sufficient.

 

Memory - memory has internal timing requirements. In its simplest form, motherboard puts up and address and waits for memory to set internally. Then memory read signal clocks that address into logic. The content of that memory comes out some nanoseconds later. After waiting enough time for data to get stable, motherboard reads that data.

Memory has timing numbers. Same size memory can have different timing parameters. Some memory is faster than others. Normally the memory provider will provide memory that is fast enough - that meets motherboard timing requirements.

And then some will counterfeit memory. Or remark memory with a faster number.

Testing memory can be at room temperature. But many memory defects can be found with temperature. For example, the memory diagnostic should also execute in a 100 degree F room or with memory heated by a hairdryer on highest heat settings. Also useful is to selectively heat the motherboard interface chip (ie Northbridge) to see how timing changes when it is heated affect that diagnostic test.

Temperature - especially heat - are diagnostic tools because temperature can change the timing response of digital circuits. All ICs (such as memory) have parameters that can changer across the entire temperature span and remain in spec. But marginal, slowly failing, and getting worse semiconductors that pass in room temperatures can also fail at what is perfectly ideal but warmer temperatures.

Use the memory tester in conjunction with temperature to confirm memory or its interface circuits are OK. Selectively heat to determine which IC gets unstable.

 

So basically test my RAM again using Prime95 and/or memtest but heat it up to about 38C/100F?

Also is there any tests I can do to make sure the mobo is functioning correctly as it's the only old/crappy part leaft in my PC. Put it this way I got it for £46/$75 about a year ago, now you can't even buy them as far as I know.

As you can see it's got all the fetures I need but it was the base model: http://usa.asus.com/products.aspx?l1=3&l2=149&l3=592&l4=0&model=1934&modelmenu=1

The RAM has already passed all tests at room temp (around 30C case ambiant), it's just when I set the RAM manually to 1066MHz from the default 800MHHz, thats when I get all sorts of odd issues & what with all the complaints about my mobo all over other forums (including the ASUS forum) I just wanna make sure :-D

:major2-Bit-Geek:major