Help with 2666 MHz, 4930K

[Maban]

My girlfriend got a 2666 4x4GB kit for her 4930K/X79-Deluxe. It seems to run the XMP profile stable but running an AIDA bandwidth test shows the bandwidth isn't where it's supposed to be. It's roughly 32GB/s at the rated speeds but when it's dropped down to 2400 or less it goes up to 50-60GB/s.

 

[cadaveca]

what kit exactly?

 

[Maban]

F3-2666C11Q-16GTXD

 

[cadaveca]

It's the board's BIOS screwing with you. You need to go into BIOS, go to "TOOLS" section, and check the listing there for timing settings and input both primary and secondary manually.

 

[Maban]

Manually set the XMP timings. Used the same timings for both. Less latency for 2666 but far higher bandwidth for 2333.

 

[cadaveca]

you have to set more than just the first four timings manually.

uh, can for get FAT32 USB drive, and capture screenshots of memory timings page, top to bottom, by pressing F12? Let me take a look, I'll tell ya what to set. I'll capture some screenshots of my own board to show ya.

Even 2333 is wonky, look at write performance.

BIOS:

performance:

So, two possible problems, maybe three. The BIOS isn't setting timings and voltages correctly, the CPU isn't capable at the given speeds and voltages, or a problem with the memory.

As far as I am concerned, 2666 should be no big deal, so let's blame BIOS, compare some settings, and see what works, or doesn't.

 

[Maban]

CAS Latency (CL) 11T
RAS To CAS Delay (tRCD) 13T
RAS Precharge (tRP) 13T
RAS Active Time (tRAS) 35T
Row Refresh Cycle Time (tRFC) 214T
Command Rate (CR) 2T
CAS To CAS Delay (tCCD) 4T
RAS To RAS Delay (tRRD) 7T
Write Recovery Time (tWR) 16T
Read To Read Delay (tRTR) Different Rank: 5T, Different DIMM: 7T
Read To Write Delay (tRTW) Same Rank: 9T, Different Rank: 9T, Different DIMM: 9T
Write To Read Delay (tWTR) 10T, Same Rank: 22T, Different Rank: 7T, Different DIMM: 7T
Write To Write Delay (tWTW) Different Rank: 8T, Different DIMM: 8T
Read To Precharge Delay (tRTP) 10T
Four Activate Window Delay (tFAW) 27T
Write CAS Latency (tWCL) 7T
Additive Latency (tAL) 0T
CKE Min. Pulse Width (tCKE) 6T
Refresh Period (tREF) 9360T
Burst Length (BL) 8

These are the settings that were used according to AIDA. (no FAT32 flash) I changed them to what you have in your screenshots and got the same bad results: http://puu.sh/9AasX/10b58aa87f.jpg

 

[Maban]

I think I fixed it. Maybe not fully though, still lower than your result, but I can tinker more I guess. tCCD and tCCDWR were set rather high. I changed tCCD to 1 and tCCDWR to 0.

 

[cadaveca]

Looking good. And yeah, it's usually tertiary timings that cause this sort of issue. TCCD @ 0, if it works will help pull things in a bit too. Keep working at it!

 

[Maban]

Any suggestions on how to get tCCD to 0? It won't boot at 0.

 

[cadaveca]

try boost VCCSA. Might not work, I am running Samsung-based sticks that can be pushed a bit harder than what you got now.

 

[bubbleawsome]

I've seen a few guys on other forums have issues with the 49xx series memory/PCI/SATA. Seems its links to the rest of the mobo are a bit odd sometimes. While 2666 should be ok it just might not be possible. (I hope it is though)