Memoery sped question

[isochrony]

I've finally gotten past the "not running with 64Gb of RAM" problems on both of my MBs. So now I'm running memtest86 v4.20 and trying to understand what it's telling me. The setup is two MBs as follows:

Gigabyte X79S-UP5 with 64GB of G.Skill F3-12800CL10Q-32GBZL (2 packs in alternate channels). 10-10-10-30 timing
ASUS Sabertooth X79 with 64GB of Crucial Ballistix BLS4K8G3D1609ES2LX0.16FER (2 packs in alternate channels) 9-9-9-24 timing
Both processors are running at 3200 MHz according to memtest; I let the BIOS default the value and set the XMP profile to 1 for each build. Nothing overclocked that I can find. The other values from memtest make me pause.

The Gigabyte is showing numbers of:
L1 Cache: 36782 MHz
L2 Cache: 16931 MHz
L3 Cache: 11721 MHz
Memory: 7111 MHz

The Asus is showing numbers of:
L1 Cache: 106669 MHz
L2 Cache:45071 MHz
L3 Cache:31373 MHz
Memory:17391 MHz

And the Asus is running memtest through all 64Gb of tests at about 2.5x the speed of the Gigabyte. Of course, 17391/7111 ~2.5, so this makes some sense.

My question is why? Does the difference in the CAS level really make this much difference or should I be looking for some other setting fouled up on the Gigabyte?

Before I commit both these boards to cases, I want the best/fastest one for running my video editor box...

Thanks for any opinions, pointers, etc.

Isochrony

 

[PassMark]

> Does the difference in the CAS level really make this much difference

Unlikely.
What CPU is in each box?
Did you try moving the RAM between the boxes and see if the performance is due to the RAM or the box (CPU/MB)

 

[isochrony]

I've got a different I7-3930K in each socket. Two head-scratching observations:

1) The beefier cooler is on the slower MB. I find this odd because the smaller cooler has trouble keeping the CPU in the faster MB cool; a beefier cooler is on order. I haven't noticed either CPU slowing down due to heat though.

2) The Ballistix RAM in the Asus MB was clocking at 1600 MHz. I replaced it with a set of G.Skill F3-17000CL11Q2-64GBZLD clocking at 2133Mhz, and I'm getting almost the exact same performance numbers - different in the last digit from what I posted.

I tried setting the Gigabyte with Channel and Rank interleaving to no apparent effect.

What else should I be looking for?

 

[isochrony]

I finally got time to fiddle some more. I put the Ballistix ram into the Gigabyte, loaded optimized, then made sure XMP Profile 1 was being used (clocks at 1600MHz). Now I'm getting similar speeds as measured by wall-clock time compared to the ASUS with the 2133MHz ram installed. However, the numbers memtest coughed-up as noted above are unchanged for the Gigabyte.

Net: I don't believe the memtest figures are telling me anything useful. And since it's running memtest fine, I'm going to go ahead and fire up Windows and see what happens.

Anyone have any insight, I'm still curious...