Help Needed Overclocking FX8350 on Asus Sabretooth 990FX R2.0

Man the cpu voltage looks low at 4.6ghz. (230 bus) and the memory has gotta be gosh awfully bad for performance at 754mhz.

Cas 9-10-9-27-36 2Tcr in that order should be no issue 1.55v

OR bump the memory divider back up one, and keep the loose timings look for 1000mhz at 250 bus is a sweet spot I've found in most cases. Ram should be ok for the overclock, IMC likes a little extra voltage.

NB/HT links, 2600mhz and beyond is Big OC for the NB but easy clocks for HT being native since FX-8100 at 2600mhz. It's an un-necessary clock speed unless your running multiple processors or multiple GPU's, performance gains are small.

Ram for AMD absolutely loves tight tight timings. Your performance will go up and up here as the Cpu access the Ram quite often. If your sticks could pull off Cas 10-12-12 or tighter at 1100mhz effective, you got a happy IMC.

Sweet spots for dividers are 225, 250, 275, and 300 reference clocks, or just a click under like 273 to better ensure daily overclock stability (273 x 11 = 3000mhz cpu in example)

Since you have super loose timings, I say go for the big RAM speed in raw Mhz. The more the better. But you never know, sometimes the big speed ram can do some pretty tight timings at lower speeds. But you have much trial and error ahead.

Sorry for no further updates but the 7 day work-week has left me very tired and I just lounged on the couch all day yesterday. BBQ today and up at 4am tomorrow so will post with further results once I actually get some time on my PC (kids are currently playing LEGO on it).
Please bear with me .....

Hey Z_M, it is your life man and you have to live it with your family. NO problemO. Luck to you and get some rest.
RGone...

I am back. Family gone back to the uk after being here on holidays for 2 weeks.
Back to the grind..

First off, silly question --
RAM timings ..can see the various options under DRAM timing control, but which ones relate to which ...?
All set to auto at the moment and want to follow RGone instructions.

This and more is in my info_tips FX in mobo section...

...Most ram is shown when bought with the first four timings listed as outlined below. CPUz normally shows the first four timings in order down the page and adds one more which is tREF and has the CMD rate listed lastly most of the time.

The operations that these numbers indicate are the following: CL-tRCD-tRP-tRAS-CMD. To understand them, bear in mind that the memory is internally organized as a matrix, where the data are stored at the intersection of the lines and columns.

For instance 8, 8, 8, 24, 2T >> CL-tRCD-tRP-tRAS-CMD.

CL: CAS Latency. The time it takes between a command having been sent to the memory and when it begins to reply to it. It is the time it takes between the processor asking for some data from the memory and then returning it.
tRCD: RAS to CAS Delay. The time it takes between the activation of the line (RAS) and the column (CAS) where the data are stored in the matrix.
tRP: RAS Precharge. The time it takes between disabling the access to a line of data and the beginning of the access to another line of data.
tRAS: Active to Precharge Delay. How long the memory has to wait until the next access to the memory can be initiated.
CMD: Command Rate. The time it takes between the memory chip having been activated and when the first command may be sent to the memory. Sometimes this value is not announced. It usually is T1 (1 clock cycle) or T2
(2 clock cycles).
RGone...

Hey RGone --

Thank you kindly for the quick response RGone.
I had just about figured most of it out using the CPU-Z Memory tab to correlate headings and timings against what's in the BIOS.

Ago, so over the last few hours I did the following.
1. Set the Memory Frequeney to 2145MHz.
2. Changed the Timings manually via DRAM Timing Control to 11-12-11-29-40 (per instructions) - was unsure where to change to CAS 2 so omitted that.
3. Up'd the DRAM Voltage from 1.491 to 1.55
4. Up'd the CPU_NB Voltage from 1.187 to 1.25.
5. Changed CPU_NB Frequency to 2530MHz
6. Changed HT Link Speed to 2530MHz

Obviously I changed little by little abnd tested via P95 to see if failures would occur.
With all the above changes made, temps are fine but when running Blend test, get failures on Worker 7 & 8 after 6 minutes.

Will have to leave it at that for now as I need to hit the hay and have been messing with the BIOS since 9pm(ish).
Not sure where to go next - should I reduce the Mem Freq to "1840" and maybe up the CPU Voltage ??

Tempted to go back down to 4.5GHz and tweak there as had no failures ??

The joy of it all. Will try and do some research tomorrow day and get back on tomorrow night for another few hours !!
(Determined) ;-)

RGone / Johan45 -
Should I get new RAM ?? I am sure I would be able to sell of my HyperX RAM easily enough and only take a small hit.
I was enjoying my OC until I hit this glitch and if it is down to the Kingston and AMD not playing well together, I will get new RAM.
What is the best RAM at the moment for the FX8350 and Sabretooth combination in relation to OCing ??
Will be back on tonight ...

1. Set the Memory Frequeney to 2145MHz.<< That is probably too high for most any DDR3-1866 ram unless you are very lucky,. Need to drop memory multplier by one factor.

If you bring the ram speed down one notch, the ram probably will not be a problem. So other than just overclocking the ram higher than it will go being DDR3-1866 ram, I really don't see anything wrong with the voltages you set in 2 posts above.

Now just as an FYI, my G Skill Ares DDR3-1866 is rated to run that speed with CAS9 whereas yours is rated that speed at CAS10. So my ram should be just a shade more powerful to do work. But just to swap ram when it might have been a little too greatly overlocked is not really worth it until you see if it will work just less overclocked.
RGone...ster.

Afternoon All,

Well I haven't given up but due to my limited intellect it's moving rather slowly ;-)

I made all necessary changes as advised, I have played with various DRAM Timings and eventually settled on "10-11-10-30-42 (1Tcr)".
Have up'd the Voltage by small increments of .003125 and it's currently at "1.4125".

Anyhoo, ran another P95 test just a few minutes ago and failed on worker 7 after 15 minutes. Let it run for 30 minutes and all other workers running fine and fully processing each set of instructions...

Thought I would dl some of my BIOS settings in case I am missing anything ?
Also, have uploaded latest temps with current setting whilst running P95 for 30 minutes and the failure error on Worker 7..

Help

Two things that might fix that a bit more V_Core and I would also try setting the CPU_NB to 1.28v. One thing I have seen with that kingston ram is it tends to needmore CPU_NB voltage than others tend to.

Core C6 State = disabled
HPC Mode = enabled
APM = disabled

Cheers Johan,
Will up the CPU_NB to 1.28 and run P95 again. If it fails again, will start to increase VCore bit-by-bit.

@Mr. Scott- from reading, my understanding was that you didn't need to adjust HPC or APM unless going for really high OC ?
Can add to more heat being produced ?

APM and HPC try to keep you inside the 'envelope' if you will, thus limiting your OC.
If you're overclocking, you're already outside that envelope, so why would you want to keep these items working against you?

Really high OC is subjective to the user. The masses would say anything over 4.5 is really high. I personally don't think like that.

Mr.Scott said:

Core C6 State = disabled
HPC Mode = enabled
APM = disabled

Click to expand...

I FULLY concur with that by Mr. Scott. IF you begin to overclock and the motherboard has those settings in bios they should be adjusted as Mr. Scott suggests.

When I posted this statement and had images of the 4 captures we needed, I 'found' a typO and changed it. I post from 3 or 4 computers and one of my saved txt files had a typO in it. This is what will from now on show up in my get on the same page post and suggest what to do in gettting a baseline for testing.

Normally during setup and testing we disable C1/E, C6, Cool N Quiet, APM, TurboCore and in windows performance manager itself we set to "performance" mode. ALSO if you have HPC in bios you would ENABLE it. That way there are not "other" settings messing with the overclocking process. Some of those settings are not available on all models of cpu but where in evidence we disable for setup of overclock process.

The typO was APC instead of the APM that is there now. So from the first it is suggested to Disable APM not if you think you will overclock to or beyond some speed and I have now added to Enable HPC if it is in the mobo bios.
RGone...ster.

If I may make a suggestion that could possibly help alleviate a whole lotta headbanging. Drop the multiplier on the Cpu to 17.5 so the Cpu is at 4.0, then try and run prime blend or memtest, I'd like to see if the memory is stable at 1800+. I had a really difficult time trying to get a certain set of DDR3 1866 sticks to run at 1866 or above. It took me raising the Cpu NB Voltage to 1.35 to get the sticks stable. You may find it easier to push the Cpu OC if you know the memory is stable at 1800+ first.

Ok -
1. Changed settings per Mr.Scotts recommendations.

2. Then as per Mandrake, I left DRAM timings (and all other settings) as is and dropped CPU Multipler from 20 down to 17.5 (4600 to 4000).

3. Then back into Windows and ran P95 Blend for 20+ mins and it was Successful (for the 1st time in a long time ) !!

4. Also ran Memtest and have all results on screen shoots as follows --

Just set CPU Mul back to 20 and have up'd the CPU_NB Voltage from 1.25v to 1.28v as per Johan.
Currently running Prime95 Blend test again to see if this helps in resolving failure on worker after 15 mins ??

Well, all was looking well and was about to kill P95 Blend test after it had ran for 20+ mins but as I went to stop the test, Worker 7 failed again - this time on 22mins+.
On CPU_NB at 1.25v it failed at 15mins.
On CPU_NB at 1.28v it failed at 22mins.

Going to call it a night as its just 00.05 here and have finished working nights.
Will try again tomorrow evening.

Just to mention - I did change the Performance Option in Windows as well to High Performance.
Also, the temps are getting a bit higher with the socket going to 61 and core hitting 51 (at the highest whilst running the test).

Thanks for the help this evening guys.
Much appreciated !!

I feel you have two choices, I'll explain why I keep going back to the memory.

I had my system 2 hours Prime blend stable @ 4.7 Ghz with a set of Rip Jaw DDR3 2x4g 1600's in it. I then went out and and purchased a set of G Skill Sniper 2x4g 1866's thinking I would be able to plug them in and have no issues, wrong! I couldn't for the life of me get them to run at 1866 with the rig Oced to the previously stable 4.7 Ghz. Thinking that it was a ram issue I RMA'ed them received a new set and had the same issues. In that time I purchased another set of Ram G Skill Ripjaw X 2133's. which I was able to plug in and run at 1866 without an issue, unlike the snipers. I played with it for a bit but just chalked it up to the Fx 8350 or CHV mobo not liking the Snipers and I stuck them in my Intel rig. About a year later I got the bug to try and get then to run on the Fx 8350 and decided that I was going to get the Snipers stable if it killed me. Which I did, what I learned is, for what ever reason I could run other sticks at 1866 with only 1.25 Cpu Nb voltage but the Snipers required 1.35 Nb Cpu Voltage. I only was able to finally get them stable by returning the 8350 to stock clocks and test the sticks. To make a long story short, I feel that your best bet is to either set the Cpu back to the 17.5 multiplier with the Fsb at 230 and run Prime Blend for 2 hours and see if it will pass. If it doesn't then it's most likely the Cpu Nb Voltage that needs to be raised. The other option is set the Dram multiplier back one and see if you can get the Cpu 2 hours prime stable. If it doesn't pass then it's likely the Cpu voltage that needs to be raised. Once you find true stability you can always raise the Dram multiplier back up and then test again.

Additionally, just an FYI "Maxmem" is not "Memtest" To run memtest you would have to load it on a usb flash drive and then run it, preferably overnight.

Morning,
Thanks for the Info Mandrake. I will follow your advice and start with running a 2hr P95 Blend test with CPU Multiplier back to 17.5 and see if it passes.

In the interim, just downloading Memtest now (apologies for the confusion - newbie error)

Going to kick it off now so fingers crossed