Running single channel RAM on a z68 motherboard

[rm55]

Strange question maybe, but I'm considering running single channel RAM on my new setup (z68 + 3570k). Reason being that I've got 3x 2GB DDR3 1600 left over from my old i7 920 setup. So my choices are:-

1) Run 2x2GB dual channel = 4GB total
2) Run 3x2GB single channel = 6GB total
3) Buy either a 4th 2GB DIMM or maybe 2x4GB and run a total of 12GB at full speed.

Option 3) is obviously better but is too expensive for now.

Now I know from the i7 920 days that triple channel RAM is no better in the real world. So will the same hold true that while I'll only get half the RAM bandwidth will I actually notice a performance drop using real programs rather than benchmarks?

[CCODEZ]

IMO, go with either option one or option 3, cost permitting. You can only have either a dual channel mode OR single channel mode, not both at the same time. 3 Sticks will set your MB to single channel which would hamper your performance more than the extra 2GB is worth.

[rm55]

Thanks for the reply. The MB and CPU arrived yesterday but it appears the z68 MB BIOS was an older version that didn't support Ivybridge. It wouldn't boot so I couldn't update the BIOS anyway.

I was in 2 minds about this purchase anyway so that was all it took for me to decide to DSR and stick with what I have for now.

[Devastater6194]

Stick with option one until you can afford option 3.

[rm55]

Thanks guys. I've still not bought the MB actually having had second thoughts if it was worth the money!

[rm55]

Thought I'd post back if anyone was interested. I finally took the plunge and replaced my Q6600 with a 3570k and MSI Z77a-GD45. I'm very pleased, especially with the MSI motherboard - £75 for an SLI capable MB beats the completion by at least £30. I first tried using my old RAM as mentioned above, then grabbed some Samsung Green 30nm 2x4GB when OCUK had it on special offer last week.

Anyway I did a few speed tests, which are pretty unscientific but the upshot is that the difference between dual and single channel is not noticeable unless you're timing on a stopwatch. Please see the small print below the graphs but basically the results are as follows:-

Default speeds
(1) 2x 2GB Corsair 1333MHz 9-9-9-24 1T 1.5v
(2) 3x 2GB Corsair 1333MHz 9-9-9-24 2T 1.5v
(3) 2x 4GB Samsung 1600MHZ 9-9-9-24 1T 1.35v

Overclocked RAM and CPU to 3.9GHz
(4) 2x 4GB 2133MHZ 10-10-10-28 1T 1.5v

**************************************************

Cinebench 11.5 (higher is better)


Lightwave Rendering (lower is better)
Scene1 = 1,000,000 polygon car model with HDRI and radiosity
Scene2 = 430,000 polgons, volumetric render


AIDA64 (MB/sec) - higher is better


Sisoft Sandra (GB/sec) - higher is better


Intel Burn Test (GFlops/sec) - higher is better


Metro 2033 Benchmark (fps) 1920x1080 DX11


Heaven 3 benchmark (higher is better)


Notes:
1. The first 3 non-overclocked runs were all at BIOS default speeds which is why 3x2gb is at 2T not 1T.
2. The Samsung 2x4GB defaulted to 1600MHZ 1T, but was still noticeably SLOWER than the old Corsair 1333MHZ stuff; am guessing some of the timings are slacker? *
3. Very little difference between dual channel and single. CPU-Z actually reported 3x2GB as running dual channel but I don't see how this can be correct. Memory benchmarks show a 10%-30% improvement with matched pairs, but in real world stuff it make no difference.
4. Based on these results I think I might mix my old Corsair stuff with the new Samsung and run the lot at 1.5v 9-9-9-24.

* based on this, I'd say if like me you have an old DDR3 triple channel 1.65v kit left over from an x58 system, think twice before buying new RAM, as your old stuff should run at 1333MHZ or even 1600MHZ at 1.5v - and either way will be faster!