Mismatched dual channel?

[DDY]

Can someone explain this to me?

I was under the impression that dual channel only worked with pairs of RAM modules with the same capacity.

I've gone from 2x 1GB to 1GB+4GB but the machine still tells me it's running in dual, is my rig lying to me or is 1GB of the 4GB stick running in dual with the remaining 3GB in single?



Background: since DDR3 prices have rocketed I decided it's now or never to upgrade the RAM in my machine, which, has 1GB soldered in plus a single slot previously occupied with a 1GB stick. This explains the weird 1GB 800MHz + 4GB 1600MHz combo I've got going on, just installed the 4GB stick I bought used which I'm testing for errors.

[DanceswithUnix]

Depending on chipset & bios it is either putting the 4GB on the end of the 1GB area ("dual channel asymmetric") or as you suggest some systems seem to be able to interleave lines on the first 1GB and then run the last 3GB of the second dimm after. That seems to be called "flex", or whatever Intel feel like branding it today.

I don't think it makes that much difference in reality. The chipset can access both slots independently and will try and pull any tricks at its disposal to make best use of that. So whilst it isn't ideal to have unbalanced memory, both slots can be accessed at the same time so it is still dual channel.

So if you thought it would be single channel, well it isn't anywhere near that bad. Just enjoy your extra ram

[DDY]

Thanks for the info, turns out there's more to my upgrade story.

Since the upgrade I noticed graphical corruption, which on my laptop with integrated graphics pointed to a bad stick of memory, however, Memtest had run several loops, all of which had passed.

I did a bit of poking and it turns out the Intel mobile 4 series chipset has a known bug which causes graphical corruption if the memory is in "Flex" configuration and VT-d is enabled.

Although I do run VMs on this machine, VT-d isn't essential and with it disabled I no longer had graphics problems.

The bug is mentioned in this Intel doc: (errata number 10)
http://www.intel.com/assets/PDF/specupdate/320123.pdf

[Bambooz]

however, Memtest had run several loops, all of which had passed.

Even though the RAM itself apparently wasn't the issue..: http://forum.canardpc.com/threads/68...d-betatesters-!

Avoid V4.xx for more modern systems if you can. I've wasted countless hours with 4.xx not detecting anything in a dozen or so loops, while the 5.0 betas found problems within minutes.

[DDY]

Even though the RAM itself apparently wasn't the issue..: http://forum.canardpc.com/threads/68...d-betatesters-!

Avoid V4.xx for more modern systems if you can. I've wasted countless hours with 4.xx not detecting anything in a dozen or so loops, while the 5.0 betas found problems within minutes.

That's good to know thanks.

As for modern, does that include my Dell XT2?