Odd behavior after upgrade, PSU problem?

[squigish]

I recently upgraded my computer, and it seems to be much less stable than it was.

Specifically, with the new configuration, when I boot the computer, it will often fail to POST, reboot, and then POST successfully. Occasionally, it has refused to POST at all. I'm wondering if my PSU might perhaps be insufficient for my setup.

I also had a situation where my WDC1002FAEX Caviar Black 1TB hard drive caused the BIOS to hang during POST whenever it got to the part where it identifies the device connected to whichever SATA port the caviar black drive is plugged in to. However, I hotplugged the drive, and now it works fine, both hotplugged and during boot-up. I'm doing an RMA with western digital (the drive was empty when it failed, how often does that happen?)

Unchanged from old to new config:
Mobo: Gigabyte GA-H55M-USB3 v 2.0, latest BIOS firmware (F11 I think)
CPU: Intel core i3 560 Clarkdale 3.33 GHz
PSU: Antec Earthwatts Green EA-430D 430W continuous rated
Optical Drive: Lite-ON Black 12x Blu-Ray Burner IHBS112-29
Hard drives: Hitachi 1TB 7200 RPM SATA
Maxtor 200GB IDE
Western Digital Caviar Green 2TB WD20EARS
Card Reader: AFT XM5U All-in-one USB 2.0 Card Reader
2x 120mm fan, 1x 80mm fan


Parts in old config that were removed:
RAM: G.SKILL 4GB (2 x 2GB) DDR3 1333 (PC3 10666) F3-10600CL9D-4GBNT
G.SKILL 2GB (2 x 1GB) DDR3 1333 (PC3 10666) F3-10666CL9D-2GBNQ
GPU: Asus Nvidia Geforce 210 Silent
SSD: Kingston SSDNow V+ Series SNVP325-S2/64GB

Parts added to the new Config:
RAM: CORSAIR XMS3 16GB (4 x 4GB) DDR3 1600 (PC3 12800) CMX16GX3M4A1600C9
GPU: ZOTAC ZT-20202-10L GeForce GT 220 512MB 128-bit DDR2
SSD: Corsair Force Series 3 CSSD-F90GB3-BK 2.5" 90GB SATA III
Hard Drive: Western Digital Caviar Black 1TB WDC1002FAEX

The system had been running stable for about 8 months. I run linux Ubuntu 10.04 Lucid, and my computer runs 24/7. The only overclocking I did was to enable the XMP profile for the Corsair ram, enabling it to run at 1600 MHz.

How much additional power does having lots of ram consume? Which rail does it come from? Neither the old nor the new graphics card had an external power connector, so I assumed that neither drew much power.

I was able to get the system stable again by removing two sticks of the new corsair ram, so now I'm running on 8GB of ram, but I'd like to get the full 16GB working. The Mobo is speced to support a max of 16 GB of ram, although this specific set of ram isn't on the QVL (the largest config on the QVL is only 4GB), the CM3X8GX3M4A1333C9 is, which is a similar model number to mine.

I've run memtest86+ on the ram, and everything comes out clean. The video card I pulled from another working system. Is my mobo bad? Or the PSU? any ideas?

 

[TollhouseFrank]

I was able to get the system stable again by removing two sticks of the new corsair ram

your issue solved!

However, the problem comes down to

A: is it the RAM not stable at 1600 and needing run at 1333

B: 1, possibly 2 bad slots

C: RAM timing issues between the 2 different sets of RAM you have

Have you run memtest on both sets in the known 'good' slots? If so, try in the other slots (if the board supports it and boots) to see if possibly those slots are bad.

 

[squigish]

A: is it the RAM not stable at 1600 and needing run at 1333

This is a possibility, as it is currently running stable at 1333. If it doesn't work at 1600, should I RMA it? I guess next step is to clock the two working sticks up to 1600 and see what happens

B: 1, possibly 2 bad slots

Meaning RAM slots on the motherboard? I was previously running 4 sticks of ram (2x2GB+2x1GB), so this seems unlikely, at least with my (very basic) knowledge of how RAM works.

C: RAM timing issues between the 2 different sets of RAM you have

How would I determine if this is the case? I'm afraid I don't really know what you mean by timing issues. Are you talking about differences between sticks 1+2 vs 3+4 of the new ram, or differences between old ram (6gb) and new ram (8gb or 16gb)?

Have you run memtest on both sets in the known 'good' slots? If so, try in the other slots (if the board supports it and boots) to see if possibly those slots are bad.

I have only run memtest on the full 16gb set. I don't remember which sticks were in which slots.

The manual says that the ram must be placed in slots 1 and/or 3, although I haven't tried otherwise, so it might work. I am currently using slots 1 and 3.

Regarding the power issues, what determines how much power the RAM uses? Capacity? Number of sticks? Speed? All of the above? Is this unlikely to be the problem?

My plan: Clock the current (working) 8GB of ram from 1333 up to 1600 and see if it's still stable. Then swap the two known working sticks for the other two sticks, and run memtest. Should I run memtest with them at 1333 or 1600? Is the 90 min. fade test useful in this situation, or should I just run the standard tests? Is it worth it to run the test multiple times? If so, how many?

I realize I'm asking a lot of questions. Please respond even if you can only answer a few of them.

Thanks a bunch!

 

[Oulse]

In your situation, where it fails to boot, a quick insert may do the trick. When testing RAM for a problem that happens occasionally, it's best to run the rests for as long as possible. I leave an extensive test running 12+ hours, so it goes through at least a few times. The longer the better, as a good stick will (virtually) never fail. Bad RAM may fail only 1/1000 accesses, but that still means you're going to get crashes more than once a month.

 

[TollhouseFrank]

This is a possibility, as it is currently running stable at 1333. If it doesn't work at 1600, should I RMA it? I guess next step is to clock the two working sticks up to 1600 and see what happens

If the sticks were advertised to run at 1600 and don't, then either your board isn't supplying the proper voltage (check RAM manufacturer's suggested voltage settings) or the sticks are not compatible with the board (it does happen)

Meaning RAM slots on the motherboard? I was previously running 4 sticks of ram (2x2GB+2x1GB), so this seems unlikely, at least with my (very basic) knowledge of how RAM works.

It is quite possible. It may be that there is a bug and one set of slots can't run faster than 1333. One of the sets may have gone bad. Right now, this is just a suggestion that likely isn't the proper answer, but it is a possibility.

How would I determine if this is the case? I'm afraid I don't really know what you mean by timing issues. Are you talking about differences between sticks 1+2 vs 3+4 of the new ram, or differences between old ram (6gb) and new ram (8gb or 16gb)?

If you go into your RAM settings in BIOS/UEFI (depends on which your board has), you will see settings for timings as possibly something like AUTO, and then manually set things like TRAS, CAS, etc.... You need to check your RAM Manufacturer's suggested settings to make sure what the settings on each pair of sticks are. It is possible one set is tweaked a bit tighter than the other set and that is causing issues that would cause you to have to manually set looser settings.

I have only run memtest on the full 16gb set. I don't remember which sticks were in which slots.

Do it one pair at a time In slots 1&3. Let the test run say, overnight while you are asleep. That will give it time to find if errors are likely going to happen or not with that pair.

Regarding the power issues, what determines how much power the RAM uses? Capacity? Number of sticks? Speed? All of the above? Is this unlikely to be the problem?

Check the RAM manufacturer's recommended settings. Though all RAM has a chip on it that tells the BIOS/UEFI what it's auto settings are, sometimes that may not be detected properly, or you may have 2 different sets with different auto settings that are causing conflicts.


As for the rest of your questions, Let's get through these first. Mismatched RAM sets can cause issues, and what these suggestions will do is find out if it is mismatched settings (like timings or voltage) causing the issues or not.