It seems that my system is in need of some extra RAM. With the dockers I have running, the size of my backups (500k+ files, 2.1TB) is causing CrashPlan to use too much memory and my server is crashing.

 

I figure while I'm at it, I may as well go bigger 'cause bigger's always better, right? So I'm looking at adding 2x8GB to my existing 2x4GB ram for a total of 24GB.

 

My current memory, as specified in my .sig, is a Corsair Vengeance 2x4GB kit and is on the Gigabyte memory compatibility list for my GA-F2a88XM-D3H. I'm looking at either the CORSAIR Vengeance Pro 2x8GB kit with exactly matching speed & timings (1600MHz, 9-9-9-24) to what I currently have, or a CORSAIR Vengeance Pro 2x8GB kit at a faster speed, but slower timings (2400MHz, 11-13-13-31). My reason for considering this memory, is if I need to add additional in the future, I've already got a faster set in there and I can buy a matching faster set sometime down the road.

 

My understanding is that the first set I'm looking at with the matching specs would be better because all the specs match.

My first question: would the second set run well (I realize the BIOS will run it at 1600, not 2400) in combination, or will it end up forcing all 4 sticks to run at the slower timings of the 2400 set as well? If it does, would that make a noticeable difference? I'm not planning on adding any VMs to the setup, though I'll probably end up adding a few more dockers as things peak my interest.

 

My second question: is that neither of those sets is on the memory compatibility list for the mobo. However, the date on that list is Sept 2013, so these sticks may not even have been manufactured back then. How critical is it, for unRAID server purposes, that the memory be on the mobo compatibility list?

 

At this point, I'm not really looking to replace the existing memory with 4x8GB (total 32GB), I'm just looking at a smaller step. So, while that's an option, it's not one that I'm open to at the moment.

My advice remains unchanged: http://lime-technology.com/forum/index.php?topic=56023.msg534454#msg534454

 

The higher-speed modules will work fine ... as you've noted, they'll just be clocked at the lower speed of your current modules.

 

HOWEVER ... with unbuffered memory it's best to not populate more than 2 slots.  The bus loading from 4 modules will significantly degrade the data and address signaling, and can result in a much less reliable memory subsystem.

 

I'd either just install 2 8GB modules (for 16GB total) or buy a pair of 16GB modules (which your system supports -- but they're appreciably more expensive than the lower capacity modules).

 

Not sure if this recommendation from Asus is pertinent across the board on all motherboards,

 

Always install DIMMS with the same CAS latency

 

They don't care about the rest of the timings, but Asus does recommend all the same CAS latencies across the DIMMS

This is interesting.

 

I ended up getting a G.Skill RipjawsX 2x8GB set clocking in at PC2133 with 9-11-11-31 timings.

 

Does anyone have any idea why MemTest 86 (the one built into the boot image on my unRAID USB stick) is reporting:

RAM: 630 MHz (DDR3-1261) - BCLK: 94
Tmings: CAS 9-9-10-24 @ 128-bit Mode

 

At first, I added the 2 sticks in with the existing PC1600 memory that was in there. Then, after a few minutes, it hit me just how slow that was, so I shut down the machine and pulled the older, slower memory out, and it still reports the same DD3-1261 speed and odd timings.

 

I went into the BIOS settings and set the AMP (AMD's equivalent of Intel's XMP) to Auto. As I understand it, that's supposed to take maximum advantage of the memory's clock speed.

HOWEVER ... with unbuffered memory it's best to not populate more than 2 slots.  The bus loading from 4 modules will significantly degrade the data and address signaling, and can result in a much less reliable memory subsystem.

Hi Gary - I have recently repurposed an old desktop system as an unRAID server.  The CPU is an i5-2500K and the mobo is an Asus P8H67M-EVO.  The system had 2x4GB sticks of RAM and I added 2x8GB to have a total of 24GB of RAM.

 

The system has been up and running for about three weeks but I have had one kernel panic and one or two other strange occurrences.

 

Should I go down to just two sticks of RAM?  Or should I first run Memtest with all four sticks and then if it shows issues go down to two sticks and run Memtest again?

You would probably be better off starting your own topic fur this question. Reference back to Gary's comment, maybe, by let your question stands on its own.

... Should I go down to just two sticks of RAM?  Or should I first run Memtest with all four sticks and then if it shows issues go down to two sticks and run Memtest again?

 

I'd drop to 2 8GB modules.  Memtest won't necessarily find any errors with 4 modules unless you run a LONG time, as the random bit errors you'll get from excessive bus loading are NOT because the memory is bad, but because the bus signaling is degraded and will occasionally result in bit errors as a result.

 

You would probably be better off starting your own topic fur this question. Reference back to Gary's comment, maybe, by let your question stands on its own.

 

... no need so split this off -- it's clearly a related topic, and is a one-off question anyway.

 

This is interesting.

 

I ended up getting a G.Skill RipjawsX 2x8GB set clocking in at PC2133 with 9-11-11-31 timings.

 

Does anyone have any idea why MemTest 86 (the one built into the boot image on my unRAID USB stick) is reporting:

RAM: 630 MHz (DDR3-1261) - BCLK: 94
Tmings: CAS 9-9-10-24 @ 128-bit Mode

 

At first, I added the 2 sticks in with the existing PC1600 memory that was in there. Then, after a few minutes, it hit me just how slow that was, so I shut down the machine and pulled the older, slower memory out, and it still reports the same DD3-1261 speed and odd timings.

 

I went into the BIOS settings and set the AMP (AMD's equivalent of Intel's XMP) to Auto. As I understand it, that's supposed to take maximum advantage of the memory's clock speed.

 

Because that's what the AUTO setting set it to. If you want to pump it up to mem spec you'll have to do it manually, but i wouldnt bother as you're likely to sacrifice stability for 0.001% faster memory access.