Power Supply Problem?

[AlexScherr]

I've just finished my first home built machine, and am having stability problems that seem to relate to the power supply.

Here are the specs:

PSU: Antec NEO HE 550W

running

Case: Antec P180
Mobo: Asus P5N32-SLI SE Deluxe
CPU: Intel Core 2 Duo E6400
CPU Fan: Zalman CNPS9500
RAM: 2GB Corsair DDR2 800 (PC 6400)
GPU: Geforce 7950GX2 (Gigabyte)
Monitor: Dell 2407FPW (24" LCD)
Sound: Creative X-Fi Platinum (Card and 5.25" front)
HDD: Seagate Barracuda 7200.10 (320 MB)
DVD: Nec 16x DVD/CD R/W
Floppy: NEC 3.5"
Fans: 4 x 120mm

Everything booted on the first run, and all drivers installed without difficulty. On the mobo, I have disabled onboard audio (because of the XFi), onboard gameport, and the onboard LAN (I'm still on dialup . . . don't ask!) Device manager looks smooth, and there are no apparent problems with the configuration.

Everything works fine in normal, non-gaming usage: fast, precise, very good. My CPU temp at idle is 35C; my GPU temp at idle is 55C - 60C.

However, in games, after a short period of time, the machine shuts down: slowing, then a frozen screen with looping sound. Oblivion shuts down after 2 - 3 minutes in the first location; Pacific Fighters takes longer to shut down, but will do so fairly quickly; even Galciv2 shuts down after about 5 minutes of play. Setting video options lower extends the times before freezing, but it still freezes.

I did some careful temperature checking during the short periods that I can use a game. My CPU temp rises to 37C, and my GPU to 65 - 70, nothing alarming, but in truth I can't run the games for too long before freezing: these are readings after only 2 - 3 minutes and an exit. Still, I don't think that it's a temperature problem.

Rebooting after a freeze usually works fine, although I will occasionally get a BSOD during Windows logon, with a "MACHINE_CHECK_EXCEPTION" error. I have occasionally had to boot into safe mode, then restore, but in general it will go to where I want it to go.

I've been scanning these boards, and this sounds like a power supply problem. So my question:

Is 550W simply not enough for this rig? Are my symptoms the result of inadequate power? Or is there something else that I should check? If it is the PSU, I'm ready to spring for a more powerful unit: but how much more should I go? and do you have any recommendations?

Thanks for your help, I really appreciate it,

Alex

 

[waylander]

This sounds more like either a ram or temperature issue than a psu one. Try taking the side case panel off and putting a household fan in the opening and see if you are more stable to completely discount the temperature as a problem.

 

[AlexScherr]

Thanks for the reply. I actually tried that last night: ran a small household fan directly on the machine with the side panel off, in a room cooled down to 65F. The GPU temperature readings (off the nvidia control panel) dropper noticeably, into the low 50C's, but still the same instability issues.

I'll try some RAM testing tonight: Prime95, SuperPi 32M, and/or Memtest86+, just to see if it's a memory issue. Would that cover the memory problems? I'll post my results,

Alex

 

[waylander]

Run memtest86 first and run it for around 4 hours.

I have trouble believing that the neo550 wouldn't be able to power your system but your video card IS a power hog... You might need more amps on your 12v rail but the neo has 42amps so again I can't see that being the issue. One way to rule out the psu is to get a psu tester, they only cost $15 from newegg.

 

[AlexScherr]

I'll do that with Memtester, and post results.

I'm a bit surprised about the PSU myself. I did ask around a fair amount, and was reassured that 550W would be plenty. We'll see, however.

As to testing the power output, what I'm confused about (and can't seem to get a straight answer to) is how much power I need for this configuration. I'm new at this (building from scratch) and have always bought retail before. Since having this problem, I've tried to research the power requirements using the online calculators; and I've gotten ranges all the way from 384W up to 580W.

FWIW, the ASUS mobo manual says that it requires a minimum of 450W for safe operation; and posters at the ASUS website say that the board itself is a power hog. I've also read that the 7950 sucks up power, and the same about any of the Conroe chips. Adding on the SATA drive, the other disc drives, the sound card, the fans . . .

Without knowing exactly how much I need, I'm not sure that knowing what my Antec will produce will help. But it can't hurt to know, and I'll check out the testers at Newegg.

Thanks again, and I'll post in a bit.

A---

 

[Mondoman]

1st: Major kudos for the detailed list of specs! 8)
2nd: I'm inclined to think your PS has enough power. The Conroes are not quite as bad power hogs as the worst of the Pentium Ds. I also have a hard time believing that "good" (non-value-line) Corsair RAM could be the problem.
3rd: However, I'm a bit concerned about the Zalman. The stock Intel hsf is designed to blow air down onto the MB outward in all directions, in order to cool the components surrounding the CPU. In the case of your MB, that is not just the PWMs (which get VERY hot under high-stress apps such as gaming), but also the radiators for the north and south bridge heat pipes, and even some fins on the northbridge's hs. No matter which direction you point the Zalman, at least two other sides will go unserved. Even in the direction the fan is blowing, components low-down to the MB (such as the PWMs) will likely not be properly cooled.
Thus, my suggestion is to install the stock Intel hsf and see if that fixes your problem. The "flower-type" Zalmans (e.g. 7000 and 7700) are examples of aftermarket cpu hsfs that should do fine in cooling adjacent MB components.

 

[phreejak]

"MACHINE_CHECK_EXCEPTION" can mean a few things:
- bad cpu
- a cpu that was too aggressively overclocked
- underpowered or faulty PSU

There is a whole section on BSODs in this months issue of MaximumPC. I remembered reading about this particular BSOD when I saw your post

 

[AlexScherr]

Thanks for both of the suggestions.

Just to update you all, I was able to run a series of CPU / RAM based tests last night. I ran SuperPi through all of the routines, including the 32M run, with no flaws. I ran Prime 95 for two hours, and no errors. Finally, I set Memtest86+ to run off a bootabe CD overnight. It ran for 9 hours without an error. I checked my CPU temp quickly at the end of the 9 hour test, and it was at 38C (up 3C from its idle temp of 35C). I would think that this eliminates the CPU or the memory as the problem.

However, without having done any thorough vid card testing, I am noticing some bad results on the 7950. I did run a quick test of it using something called the "Video Card Stability Test" freeware package. Two results are interesting. First, the test is supposed to take 30 minutes to run, but it froze the machine within 3 minutes (much like the games).

Second, it has a benchmarking function. I ran it when the machine was cool and got a 1500 score; I ran it when the machine was hot, and instead got a 550 score. While the scores don't mean anything in absolute terms, surely this shows a change in this particular vid card under hot conditions. The radical difference between cool and hot scores may even suggest that one of the two GPU's on the 7950GX2 is simply shutting off under stress. But perhaps I'm wrong about that.

So now I have to figure out what to do about the vid card. I plan to run some GPU intensive tests tonight: any suggestions? I'll do it both under hot (closed case) and cool (open case, external fan) conditions, and see if there's a difference.

I still think that there may be a power problem with the machine. Phreejak's post about the "MACHINE_CHECK_EXCEPTION" error matches what I've read on the net about the error: either it's a CPU problem (bad or overclocked) or a power supply problem. Given the results of my CPU and memory tests, it looks like the power supply is the most likely cause. So, I've decided to RMA my current PSU, to replace it with a Silverstone Zeus 750W, which has been getting good to glowing reviews. If you all are right, and the power supply isn't the problem, I'll still have a stronger power supply, for about $50 more than my original cost.

If the power supply doesn't solve the stability problem, I'll then have to focus on the vid card. I think the right plan is to buy a cheap card for the PCI-E x16 slot, and do the RMA dance with Newegg to see if it's just a bad unit or some incompability between the card and the mobo. If the former, I'll then have that wonderful 7950GX2 power, plus a backup; if the latter, I can get a different card in the same price range.

Finally, Mondoman, that's interesting about the Zalman. I haven't even thought about the cooling of components near the processor. You're absolutely right that the Zalman 9500 pulls heat up and away from the processor only, doing nothing special for the surrounding area. I'll probably test this after seeing what the power supply does. Do you think that the results of my CPU / memory testing (coupled with the relatively cool CPU temps) make this necessary?

Again, thanks to all for helping out with this. I'll post results later.

A---

 

[phreejak]

You'll be happy with that new PSU. It's the same one I use and my rig has a tremendous energy demand on it. It's stable, quiet and powerful (especially for future proofing).

 

[AlexScherr]

Well, it looks like it has been resolved, but not by the new power supply (which works just fine, and which I will keep.) Once I installed the new PSU, I continued to get freezes. I then went out and purchased a lower-grade vid card (Geforce 7600GT), and installed it: still the same freezes.

I then called Asus. The first person I got was rude and unhelpful, but the second person actually worked me through a test regime, and we discovered . . . .

that I'd bought the wrong memory for the motherboard. I had installed Corsair DDR2 800 memory, on the assurance from Corsair that the mobo would support it. The Asus tech rep said that it was not, and proceeded to prove it, by asking me to remove one of the two memory cards from the mobo, the one closest to the CPU. With one memory card remaining, the machine performed much better, and only froze during 3dMark05 testing and in Oblivion. I can play Pacific Fighters without any trouble. I have purchased new memory that Asus specifically lists as compatible (in the DDR2 667 range), and will RMS the original purchase.

So, the lessons for me:

-- DDR2 800 memory does not work with the Asus P5N32-SLI SE Deluxe mobo.

-- a Geforce 7950 GX2 will work.

Thanks all for the help.

 

[waylander]

This sounds more like either a ram or temperature issue than a psu one. Try taking the side case panel off and putting a household fan in the opening and see if you are more stable to completely discount the temperature as a problem.

That was from my first post... Ram is almost always the first thing I check for stability or POST problems...

Glad you got it working though and have fun with it.

 

[AlexScherr]

You were right, Waylander! I should have thought of this earlier. However, in my own defense, I did call Corsair before I purchased the mobo and RAM and they ASSURED me that it would be compatible. Oh well: live and learn!

Thanks for your help,

Alex

 

[AlexScherr]

Further developments:

When I first installed Windows XP (Home SP2), I screwed up the SATA installation, resulting in my hard drives being recognized as G: and H: drives, and a series of disk errors. Last night, I reinstalled Windows, and succeeded in having those drives recognized as C: and D:. No more disk errors.

I've been running on a single Corsair 1 MP stick in the the third RAM slot until my new RAM arrives (on the advice of the tech support person.) In reinstalling windows, I also took care to install the Asus mobo drivers correctly, specifically the Nvidia driver.

Now I'm getting nearly full, and completely freeze free performance from my machine, but ONLY with the single RAM chip in the 3rd slot. In that configuration, the machine completes 3dMark05, and any other video test I give it. It runs Oblivion at 1920 x 1200 with 2xAA without a single pause. Cool! I thought: maybe I've solved the problem.

So I installed the other Corsair stick in the first ram slot (the other yellow one). The machine now recognizes 2MB . . . but it also freezes in 3dMark and almost instantly in Oblivion. I'm not sure what this means: full performance with a single 1MB stick in the 3rd slot; freezing with both sticks in the proper slots. I need to do the following tests:

-- install the second Corsair stick again and REINSTALL the Asus and NVIDIA drivers from the disk. Don't know of that will make a difference, but seems worth a try.

-- remove the Corsair DDR2 800 memory, and replace it with OCZ DDR 667 memory (specifically listed as compatible in the mobo manual). Reinstall drivers. Test in both 1 stick and 2 stick configurations.

If I can get 2 sticks of approved DDR2 667 RAM to work, that would seem to confirm that the problem is the mobo's compatibility with DDR2 800. On the other hand, if I have the same 1 stick / 2 stick problem with the slower RAM, then the problem is something else.

I will say that when I removed the 2nd Corsair stick, it was very hot: not burning but uncomfortable. I don't know how to assess this: RAM may always run hot for all I know. But if I do get the same problem with the new and approved RAM, I'll try some different cooling configurations, to see if that matters.

So: on we go. At least I know I can run the machine with one stick of the Corsair. As to the rest, well, more later.

Alex