PSU rails question

Hello, I have a question regarding PSU's and videocards.

I recently bought a 680W PSU in order to prepare for the upgrade in my videocard in my computer (AMD 64X2 4200, 2GB memory, 250GB HD, DVDRW, DVD reader) and unfortunately after a seconds of wanting to play some games on it, the card (a XFX 8600GT 512 DDR2) seems to die out (the system will either reboot, crash the video only or crash the video and lock up the rest of the system).

I looked into it and was told that 680w is not enough for a card of that power, that the rails aren't enough for it (the specs on my PSU are here) and that I'd need 900w at least. The confusing thing is that a little later I ran into someone with a similar system to mine running SLI-linked dual 8600 cards using 600w, which I'm pretty sure would use a bit more juice than mine.

Could someone explain this for me? Getting tired of the video constantly cutting out and while I'm loathe to buy another new PSU, better that than frying the card or worse. Thanks for any assistance.

 

Your power supply is plenty strong for your video card. The 8600GT is not a powerful enough of a card to be causing enough power drain to crash your PC. You'd need two 8800s or two 9600s in SLI to even get close. While the 8600 is OK for some of the lighter games, it simply does not have the power to play some of the more high end games unless all the in-game settings are cranked down to 'low'. For example: playing Crysis on an 8600GT will require all the video settings to be at the lowest possible for smooth game play. You might be able to gradually turn up some of them settings a little bit at a time, but the 8600 is not really a high end gaming card, and therefore doesn't require a high end power supply. Anything over 450w should be fine for an 8600. Your 680w is plenty strong. It has two 12v rails; one at 22a, the other at 24a. These are good numbers. Whoever said you needed 900w is insane, or grossly unaware, or just wanted to sound like they knew what they were talking about (which they don't). You can easily run 2 of the 8600GT in SLI on 550w, so, again, your PSU isn't the problem here. I think the problem is either the video card is overheating, or the CPU is overheating, or both. Before you buy anything, download and run a temp monitoring program. Let it run in the background while gaming to get an idea of your temps. Some of these programs will record your voltages too. Check with your motherboard maker; most of them have these types of programs as free downloads (Asus, Biostar, Gigabyte, and MSI have 'em). You can also find some here at Major Geeks. Here's one I like:
HWMonitor http://www.majorgeeks.com/HWMonitor_d5842.html (monitors temps, voltages, fan speeds, etc)
...and here's some others. I've tried some of these, but not all of 'em.
CoreTemp http://www.majorgeeks.com/Core_Temp_d5665.html (for CPU temps)
MBProbe http://www.majorgeeks.com/MBProbe_d283.html (temps; voltages; etc; kind of old but should be OK)
NextSensor http://www.majorgeeks.com/NextSensor_d4988.html (temps; voltages)
Many other programs can be found here http://www.majorgeeks.com/downloads11.html and here http://www.majorgeeks.com/downloads7.html I know nVidia has a program called "nTune" for monitoring video card temps. More info is available here http://www.nvidia.com/object/ntune_5.05.54.00.html
It could also be that the card is just faulty But lets hope that isn't the case. XFX is good company, with a lifetime warranty on all their products, and they do returns pretty quick. Have you uninstalled your video drivers, and installed the newest ones available? You can get 'em here (you didn't say what version of Windows you're running, so I posted 2 links for you ):
Windows XP (32 bit) http://www.nvidia.com/object/winxp_175.19_whql.html
Windows Vista (32 bit) http://www.nvidia.com/object/winvista_x86_175.19_whql.html
I hope all this helps! Good luck!

[dlb]

 

The only thing that dlb may have overlooked in his comprehensive review is the possibilty of ram failure also. However, this item would be lower down on my list of remedial action but it may be wise to run memtest to double check.

Why not take off the side cover of your pc and see if the same problem continues. If it does get a domestic fan and point it at the case with the cover off.

Good Luck

 

OOOOPS! My bad... rolleyes I got on a roll and it slipped my mind. Good call RR...

As mentioned, a RAM diagnostic is a good idea too. As mentioned, memtest is a good Windows based tester, but I personally prefer the bootable ones. For example:
Windows Memory Diagnostic http://www.majorgeeks.com/Microsoft_Windows_Memory_Diagnostic_d3955.html (will create a bootable floppy or ISO file to burn as an image to CD. Boot to it and press T for the extended tests. Run at least 3-4 passes. Be patient; it will take some time)
Memtest http://www.majorgeeks.com/MemTest_d350.html (Windows based)
There are many more (and other versions of Memtest) available here http://www.majorgeeks.com/download.php?id=26

RR- thanks for the heads up!

 

The only thing I would like to add here would be power connections. I have a similar setup as you and although you don't mention your motherboard, mine is an Abit AN8 32X and along with the extra connectors for power to the pci-e graphics cards (you do know they require their own power right?) there is a connector on the motherboard thats recommended to recieve a regular 4 pin molex power connector for extra stability.

Also, although your power supply is adequate you have to make some wise choices when utilizing your 12 volt rails, don't pile everything on one long connection wire set, spread the weath around.

 

Thanks for all the help. As for the type of mobo I've got, it's an ASUS M2V. MS memtest for Vista says the memory's okay. As for the temperature specs, I've attached the file that HW Monitor came up with.

As for updating the drivers, did that yesterday and now it's added a new quirk: flickering just before it crashes. Tempted to roll back the drivers, as it wasn't an issue previously.

I'm going to try the fan into the open system to see if that helps as well.

 

How long did you run the test for?

 

Ran the memory test again last night, four passes (wasn't paying attention to the time, but it was well over two hours and the gauge read 457.2% coverage). It came up with 0 errors, so I don't think it's the memory.

 

afraid that is nowhere near long enough. You should run it for 24hrs. Whether you do or not is up to you

 

Ah. I misunderstood and thought it was to run for four cycles. I'll run it for 24 hours when I get home. Thanks.

 

I was suggesting this for a different memory diagnostic. When using this one:
http://www.majorgeeks.com/Microsoft_Windows_Memory_Diagnostic_d3955.html press T for the extended tests, and run at least 3 or 4 passes. It's best to let it run overnight, but 3 or 4 passes can give you a good idea of whether the RAM is OK. When you launch the downloaded file, you'll be given the option of creating a bootable floppy disk, or creating an .ISO image that you'd burn to CD to create a bootable CD. Then boot to it, and press T, and let it run. When burning ISO files, they must[/b] be burned as images, and not as a regular data file. Most (if not all) CD burning apps have options for burning CDs from image or ISO files. You cannot burn images using the Windows built-in burning software (unless you install a 3rd party add-on or plug-in).

 

Well, I ran the diagnostic several times over the weekend, and it kept crashing after about four hours. I also ran the other one while in Windows and it came up with much the same thing, so it looks like it might be the memory. I think I'm going to replace them just to be on the safe side.