GA-F2A88XM-D3H + 7850K + DDR3-2400 possible? Need reliable settings...

[MaxGadgetGuy]

OK so I'm trying to find a way to successfully run DDR3 memory @ 2400 MHz with this motherboard  with RELIABILITY (Prime95 for 1 hour). This board seems to love 2133 MHz RAM but hate 2400 (becomes unstable all else equal)!

I previously was using G.Skill Ripjaws 2133 ram using 11-11-11-30 1.6v and it's setup using the BIOS detected XMP Profile 1. This worked just fine. I needed another set of RAM and found a deal on G.Skill Trident 2400 ram with 10-12-12-31 1.65v settings. Interestingly, I wanted to try it in the same system. First off, I could NOT successfully get the new 2400 RAM to be detected using XMP profiles. It kept showing the timings in XMP from the previous memory. In desperation, I decided to try re-flashing the F6 BIOS (same version) dated 5/28/14 and *Voila* it finally was able to read the new XMP timings! This leads one to wonder is there no other way to get this mobo to re-detect XMP timings when changing RAM (same 2 of 4 sockets) without a re-flash of the BIOS???

Secondly, now that I can use the XMP timings for 2400 (in the BIOS as XMP profile #1), I can see it correctly applying the new timings and slightly higher voltage. But no matter what voltages I try (stock, +0.05v to NB core, +0.10v to NB core) I can't run this build stable in Windows 7 x64 to even get 20 minutes in Prime95 (blended). However, if i simply change the memory multiplier to 21.33 (to have it run at 2133 MHz) the BIOS picks 9-11-12-28 1.65v (seems reasonable) and the system is perfectly stable! I can run Prime95 for over 1.5 hours no problem. NOTE: I was able to run Memtest 5.01 from a USB stick and find no errors with the new memory at 2400 MHz. Memtest defaults to only one CPU core for the test. So I know the RAM is solid.

So the question is, with the latest F6 BIOS, is there some voltage changes, Northbridge freq/voltage adjustment, etc. I need to make to get this reliable @ 2400? I saw a G.Skill rep state that sometimes adding 0.1v to the CPU-NB helps AMD run memory stable as if it needs the boost at higher memory speeds. I've tried setting CPU Load and NB Load to Extreme and Standard, with no difference. I've also tried adding up to 0.11v to the NB core and no luck. Does anyone have any other suggestions. I know that the difference between 2133 (CL9) and 2400 (CL10) is small, but it would be nice to run on auto-detect settings reliably since the board seems to detect the XMP profile now for the G.Skill Trident 2400 RAM. 

[Tiger]

You wrote, "needed another set of RAM". If you are combining two sets then you won't get reliable settings at rated speeds. Just use one kit.

[MaxGadgetGuy]

The original memory (2133) was completely removed and used in another computer build. The newer DDR3-2400 replaced the older 2133 as stated above "in the same sockets (2 of 4)." 

The question remains can this board which is a newer model from Gigabyte ever run ANY set of DDR3-2400 reliably? My firm answer after time invested is clearly NO - unless they make another BIOS revision to address it.
 

[-Andy-]

I have the Kingston Hyperx Beast KHX21C11T3K2/8X kit 2 x 4GB 2133 MHz running @ 2400 without any issues for 7 months now. Below my BIOS settings with the GB GA-F2A88XM-D3H:

Memory freq: 2400
Voltage: 1.67
Timing mode: advanced manual
Timings:
cas: 11
trcd: 13
trp: 13
tras: 30
tcmd: 2

Intel XMP: disabled
AMD AMP: disabled

Hope this helps

Andy

[MaxGadgetGuy]

Thanks for the reply. My issue is reaching pure stability with 2400 RAM via passing Prime95!

I'll detail a bit more... I have had NO problems booting to Windows 7 x64 with the mobo set to XMP Profile 1 (DDR3-2400) and the timings and voltage are all set exactly to the G.Skill specs automatically (10-12-12-31 @ 1.65v).

BUT...I can not get Prime95 to run even 5 minutes without seeing a rounding error from one of the CPU cores. Not even increasing the RAM voltage by 0.05v and the Northbridge by .10-.15v makes any difference. However, simply changing the memory multiplier to 21.33 instead of 24 allows me to have NO problems in Prime95 effectively at DDR3-2133 (improved timings of 9-11-11-28). Interestingly, these settings still come from XMP profile 1. The only thing I changed in the BIOS was the RAM multiplier from AUTO (24) to 21.33.

I'm curious if ANYONE has been able to run this motherboard at DDR3-2400 and successfully pass with Prime95 (Blended test) for 1 or more hours???! I really feel that Prime95 should pass since it works at 21.33 without fail.

It's my belief the F6 release BIOS needs an update or this board can't reliably handle 2400 speed. Either that or I'm missing some key BIOS memory settings but XMP should just work!?

[klh591]

After reading everything you have stated I would think your problem is the memory NOT the motherboard.

I have used hundreds of G.Skill memory sticks and the only G.Skill memory I had problems in Gigabyte
motherboards was 2400 or above.  All of my 1600, 1866, 2133 sticks worked, no matter what board.

I usually had to downspeed anything over 2133 to get it to work,  no matter what chipset.

[MaxGadgetGuy]

Well I thought of the memory as well but it has passed in Memtest 5.01+ for 4 passes at the 2400 speed with no errors detected. This was done using the stock XMP Profile #1 timings & voltage. The Memtest information screen shows the right timings as well. So I have a hard time thinking its really the Gskill memory at fault. I suspect the BIOS since the Gigabyte F6 BIOS (and beta versions) were the first people got XMP to work with 2400 speed RAM.

If I can convince my buddy, maybe I can test the memory in an Intel Z87 or above mobo and see what happens at 2400 MHz in Prime 95...