Correct settings needed!

[oldtech]

Hello and thanks for activating.
I think I need help configuring my timing for optimal stability and performance, I can't seem to accomplish both.

CPU
AMD FX-8320 "Vishera"
MOBO
M5A97 LE R2.0
RAM
F3-2133C11-8GXL

[emissary42]

Does your board not feature the ASUS D.O.C.P. function?

That should give you a good starting point utilizing the kits XMP which probably only needs very few adjustments for AMD like for example upping the tRFC by one step.

[oldtech]

Hello and thanks for the reply.

Yes it does, I have two DOCP profiles. Right now I'm running the 2133 profile. What adjustments should I make?
Also this enabled an option to adjust my pcix bus speed, and my gpu has a core speed of 150 but it's running at 100... should I boost that to match?

[emissary42]

To be honest i am not sure what you are talking about. Your board has no PCI-X slots, so there should only be a PCI and a PCIe clock and both should be kept at their standard values for stability reasons. Also the GPU and it's power states with their according frequencies have not to do anything with the XMP or DOCP. So if in doubt leave those settings alone.

Other than that only make adjustments if they are needed for stability. There is no universal optimal setting to use XMP based kits on AMD systems. Sometimes the workarounds just work, other times you have to adjust timings like the tRFC and/or voltages e.g. Vnb other times. That not only depends on the memory, but also on the CPU and the MB. To maximize performance you would have to fine tune all timings manually, but that can be quite time consuming and is only really worth it if you are 2D benching (SuperPi, wPrime etc).

[oldtech]

So the problem is the kit is XMP which I can safely then assume based on process of elimination from your reply, is for intel based setups.
Now it all makes sense.

Is there any chance you know of a kit designed for AMD Chips that operates at the same specs as this kit and the same size, designed for dual channel setup; or the next closest to this?

EDIT:
I am not sure if this was what you meant but I edited the ReFresh Cycle time ? from 300 to one step up which was 350. I noticed that since doing that an ongoing memory leak I noticed while gaming (adding major system resource drain, instability, and when there was stability exceptionally poor performance and overheating) where the application would just steadily increase how much memory it was using. Now it's holding steady at 1.4 GB, and 60fps (though I still have to keep graphics down but my card is an HD6770 from 6 years ago)

[emissary42]

The only kits with AMD memory profiles are those from AMD themselves, but i am not sure if they are sold any more because availability is pretty bad here in europe.

However ASUS D.O.C.P., Gigabyte E.O.C.P. and others manufacturers workarounds were made to use XMP based kits on AMD and they usually do a decent job with most settings. The issue with tRFC is, that on AMD systems there are no values between 90/110/160/300/350 and sometimes the board mechanisms will chose a lower value than specified because it is closer then the next bigger one and that can lead to instabilities. So then increasing it by one step usually helps to stabilize the system.

The other common thing is the voltage for the memory controller in the CPU, that has to be increased for higher memory ratios. Some mainboards do it themselves and nail it right away, with others you have to raise it yourself (to be fair this regularly happens with Intel based systems as well).

[oldtech]

Well, I put the Refresh Cycle Time from 300 to 350 and it worked for awhile but there is still instability.

Can you explain what you meant by tRFC? Was I correct in editing the refresh cycle time?
Right now its back at 300 which was the default on the DOCP profile, and I've tried using the other profile which was 2134 MhZ but the system is once again unstable. I can't make it through 20 minutes of a video before it crashes and I have to hard boot.

There is no cooling problems, this is all related to the ram and started when I bought it. I can't believe there is no manufacturer making AMD ram kits.. its like AMD is going under.
I feel like a fool for switching from intel when I did now, because ever since I have I've been facing non stop instability and been trying to fix it by throwing more money at new hardware--and being disappointed every step of the way.

How can I make this ramkit run stable on my system? Should I go back to the other DOCP profile of 2133 now or stick with 2134? Should I change any other settings? Please don't use acronyms for any settings I should change as my bios doesn't list them that way.

[emissary42]

Both profiles should be DDR3-2133. Can you show me a screenshot of Aida64 > Motherboard > SPD?

Have you already tried manually rising the voltage for the memory controller a bit?

[oldtech]

Both profiles should be DDR3-2133.

Have you already tried manually rising the voltage for the memory controller a bit?

They technically are, but one is labelled as 2134 in the bios and I can't see any other differences (using it still tests / indicates 2133 during bootup)

I use the 2133 profile, and all the timings are set as indicated on the package (nothing left to auto)

Are you referring to the DRAM voltage? I can up that from 1.5 to 1.50625 and in increments of 0.00625

EDIT:
Can I get that screenshot out of the same section of CPU-Z to save myself an install?


HTPC.txt

[emissary42]

No i was not, the memory controller has its own voltage. It should be mentioned in your MB user manual somewhere.

CPU-Z does not show the whole XMP, only the primary timings. Differences between the two profiles probably are in the secondary timing section. You can copy the secondary timings from the Aida64 output if you wanna try manual timings (the command rate for both profiles is 2T).

[oldtech]

My apologies for taking so long to post.

having stability issues, even with adjusted voltage on the controller.
Am forced to resort to auto (defaults to 1600MhZ) to keep the system up long enough to post.
I'm including the requested info from Aida.
Hope you're still following this @emissary42

The freezing / forced boot is happening (when set to either DOCP, with or without the adjusted voltage in place, no other settings messed with) typically during gpu intensive stuff. I was ready to blame the gpu and in the past I have when I had it first happen, but when running at the slower ram speed there is no problem. Gpu was also over heating on the DOCP prof a lot, which I suspect was causing the freeze.

I've got little experience troubleshooting performance bottlenecks. I appreciate your input and anyone elses who might want to chime in.



Field Value
Memory Module Properties
Module Name G Skill RipjawsX F3-2133C11-8GXL
Serial Number None
Module Size 8 GB (2 ranks, 8 banks)
Module Type Unbuffered DIMM
Memory Type DDR3 SDRAM
Memory Speed DDR3-1600 (800 MHz)
Module Width 64 bit
Module Voltage 1.5 V
Error Detection Method None
Refresh Rate Normal (7.8 us)
DRAM Manufacturer G Skill

Memory Timings
@ 800 MHz 11-11-11-28 (CL-RCD-RP-RAS) / 39-208-5-12-6-6-24 (RC-RFC-RRD-WR-WTR-RTP-FAW)
@ 761 MHz 10-10-10-27 (CL-RCD-RP-RAS) / 37-199-5-12-6-6-23 (RC-RFC-RRD-WR-WTR-RTP-FAW)
@ 685 MHz 9-9-9-24 (CL-RCD-RP-RAS) / 33-179-5-11-6-6-21 (RC-RFC-RRD-WR-WTR-RTP-FAW)
@ 609 MHz 8-8-8-22 (CL-RCD-RP-RAS) / 30-159-4-10-5-5-19 (RC-RFC-RRD-WR-WTR-RTP-FAW)
@ 533 MHz 7-7-7-19 (CL-RCD-RP-RAS) / 26-139-4-8-4-4-16 (RC-RFC-RRD-WR-WTR-RTP-FAW)
@ 457 MHz 6-6-6-16 (CL-RCD-RP-RAS) / 22-119-3-7-4-4-14 (RC-RFC-RRD-WR-WTR-RTP-FAW)

Extreme Memory Profile v1.3
Profile Name Enthusiast (Certified)
Memory Speed DDR3-2133 (1066 MHz)
Voltage 1.50 V
Refresh Period (tREF) 7.9 us
Recommended DIMMs Per Channel 2
@ 1067 MHz 11-13-13-31 (CL-RCD-RP-RAS) / 44-278-3-6-16-9-9-27-7 (RC-RFC-CR-RRD-WR-WTR-RTP-FAW-WCL)

Extreme Memory Profile v1.3
Profile Name Extreme
Memory Speed DDR3-2133 (1066 MHz)
Voltage 1.50 V
Refresh Period (tREF) 7.8 us
Recommended DIMMs Per Channel 2
@ 1066 MHz 11-13-13-31 (CL-RCD-RP-RAS) / 44-278-2-6-16-8-8-27-7 (RC-RFC-CR-RRD-WR-WTR-RTP-FAW-WCL)

Memory Module Features
Auto Self Refresh (ASR) Not Supported
DLL-Off Mode Supported
Extended Temperature Range Supported
Extended Temperature 1X Refresh Rate Not Supported
Module Thermal Sensor Not Supported
On-Die Thermal Sensor Readout (ODTS) Not Supported
Partial Array Self Refresh (PASR) Not Supported
RZQ/6 Supported
RZQ/7 Supported

Memory Module Manufacturer
Company Name G.Skill International Enterprise
Product Information http://en.gskill.com/en/series/desktop-memory

[emissary42]

Gpu was also over heating on the DOCP prof a lot, which I suspect was causing the freeze.

I don't know what settings you changed in the BIOS, but in general there is not direct connection between memory ratio, timings and voltage to dGPU temperatures. So if you actually saw toasty GPU temperatures under load and not just the GPU driver crashing, you might have an issue with your GPU.

Extreme Memory Profile v1.3
Profile Name Enthusiast (Certified)
Memory Speed DDR3-2133 (1066 MHz)
Voltage 1.50 V
Refresh Period (tREF) 7.9 us
Recommended DIMMs Per Channel 2
@ 1067 MHz 11-13-13-31 (CL-RCD-RP-RAS) / 44-278-3-6-16-9-9-27-7 (RC-RFC-CR-RRD-WR-WTR-RTP-FAW-WCL)

That is basically all the info you need to manually configure your memory:

Frequency = 1066 MHz or DDR3-2133
Vdimm = 1.50V or AUTO
tCL = 11
tRCD = 13
tRP = 13
tRAS = 31
Command Rate = 2T / 2N
tRC = 44
tRFC = 300
tRRD = 6 (7)
tWR = 16
tWTR = 9
tRTP = 9
tFAW = 27
tWCL / tWL = 7 (8)

Raise controller voltage in small steps for stability, you probably will need less than +100mV.

[oldtech]

I don't really know what to make of the crashing, maybe the temps aren't that high 44C isn't that bad right?
All I know is they happen when I run on 2133, but tbh I'm sure its happened while the gpu wasn't under any load, just in a desktop environment.

Not sure what else to do at this point. Any suggestions for tools I can use to test the ram, cpu, gpu?

[GSKILL TECH]

Try a lower frequency like DDR3-1866, if that works then you can work up.

[oldtech]

My bios does not use acronyms, it uses full labels for timings and every label begins with 'DRAM' so some of this stuff isn't syncing up with fields in my bios.
I'll post my bios fields as they're labelled and hopefully I can get everything matched up right.