• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Help w/ FX-8370 Overclocking and Memory Speeds

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Kundahli

Registered
Joined
Aug 15, 2015
Hello everyone,

I am a new member here, but have found the content of this forum extremely valuable for getting myself up and running with my FX overclocking.

I need a bit of help as I seem to be encountering instability from my memory speed.

Ive been following the guide posted on this site. I began by overclocking using just the multiplier and worked up to a 22.5 multiplier while keeping the host clock at 200, this was at 1.38V and LLC set to Medium. This was stable for a 20min P95 run with temps of CPU: 61C / Package: 59C

I then followed the guide and dropped the multiplier back down to 20 and worked to stabilize an OC using the FSB. I was able to get this to 220 while holding the multiplier at 20 and keeping voltage at 1.38 with LLC again at Medium. I stressed this for 2 hours in P95 and got the temps in the screen shot below. I would like to note that the 64C high on the package was only a transient spike while trying to screen grab during testing.

Anyway, a few things Ive noticed is that even when I manually set the HTT Link Frequency to 2600 and the NB Frequency to 2200 (both default speeds) they will change when looking at CPUZ. If I raise them any higher than default I fail to post. How important are these settings for establishing a stable OC

Also, my memory speed with the FSB at 220 is 733 although my RAM is rated at 800. Is there anyway I can get my RAM back up to 800Mhz while using an FSB clock?

Right now I would really like to establish as efficient, Vcore wise, of an overclock as possible using both the FSB and multiplier. I think pushing my Vcore any higher than 1.4V will max the temps. The Cooler Master 212 Evo is struggling at 1.38 as it is.

Any help or advice you could offer is greatly appreciated as Im very new. Reading the guide was helpful, but it seems like OCing is one of those things that is best learned by doing

Some screenshots of temps after 2hr of P95 with the FSB OC are below.

Screen Cap of 2 hours Prime95 Stress Test at 4.4ghz  -Multipler 20- -FSB 220-.PNG

Screen Cap of 2 hours Prime95 Stress Test at 4.4ghz  -Multipler 20- -FSB 220-  SPD SETTINGS ONLY.PNG
 
I wanted to attach some baseline temperature data in case that will help with the advice. This is after 30min of Prime95 Blend @ 1.4 with no changes in clock speed or multiplier and all AMD power save options turned off in the bios.

Screen Cap of 30min Prime95 Stress Test at 4.0ghz  -Multipler 20- -FSB 200- Vcore 1.4.PNG
 
Anyway, I think I figured out my issue. My RAM seems to not like to go over 1600Mhz. If I use the FSB to overclock that will change my my RAM speed. If I go over 1600mhz I get instability and can't post.

What ends up happening is I can either choose to gimp my RAM speed and use a 6.67 memory multiplier when overclocking with the FSB or I can just use the normal 8.00x memory multiplier and avoid overclocking with the FSB using just the clock ratio.

I seem to get better temps when I overclock with just the clock ratio (22 x 200) vs the FSB (20 x 220). Is this common?
 
Try setting Ram manually to:

1866 (933) with timings 9-10-9-27-36 2T (rest of timings on auto) (2T is command rate)

Run Ram 1.65v

All this with Cpu multiplier only overclock (keep to minimum while testing Ram stability with OCCT linpack) Increase cpu clocks after you've found Ram stability and continue testing Cpu overclock if ram is stable.
 
Try setting Ram manually to:

1866 (933) with timings 9-10-9-27-36 2T (rest of timings on auto) (2T is command rate)

Run Ram 1.65v

All this with Cpu multiplier only overclock (keep to minimum while testing Ram stability with OCCT linpack) Increase cpu clocks after you've found Ram stability and continue testing Cpu overclock if ram is stable.

Thanks for the suggestion, ShrimpBrime.

I followed your instructions this morning, but didnt have much luck. This Mushkin RAM is a little strange in that its default voltage at 1600mhz is 1.35. I tried the settings you mentioned with 1.65V, 1.5V, 1.4V, and 1.32V and failed to post each time.

I followed this up by trying to keep the XMP Profile set to "Profile 1" while using a 9.33x multiplier on the memory. I tried this with 1.35 and 1.5V and failed to post each time.

I'm open for trying anything else.
 
What speed are you at with the ram 1866 and 220 reference clock?
 
@ShrimpBrime, I just had a bit of a breakthrough. I set the XMP Profile to "Profile 1" then set all timings to "Auto". I used a 933 multiplier and was able to boot into windows, but I crashed within 5 minutes.

@Johan, I haven't been able to get stability with my ram at 1866 yet. As far as the CPU+memory speeds I have gotten, I seem to be pretty stable at 4.4 using both 22 x 200 w/ memory at 1600 or 20 x 220 w/ memory at 1433 (6.67 memory multipler)
 
Setting XMP then setting timings to auto kind of defeats the purpose TBH but that's minor ATM. My guess is that at 1866 with timings as S_B suggested you're going to need a bit more CPU_NB voltage, since it's now at 2400 and 1866 puts an additional load on it. I would set it manually to 1.25v. It might be NB core on that board Gigabyte isn't what I use but I have looked at a few BIOS shots.
 
@Johan Ah, alright. I was under the impression that the XMP Profiles automatically adjust RAM timings based on the memory multiplier.

When I use ShrimpBrimes suggested timings should I have the XMP Profile enabled or disabled?

The stock NBcore on this motherboard is 1.1625, for reference.

So I've tried the following combination:
S_B's suggested memory timings
XMP Disabled
Memory Multiplier: 9.33
NBcore: 1.25
DRAM Voltage 1.5 (I tried again with 1.65)

With this I fail to post. I can try to get some screenshots of my BIOS settings up later today if that would be helpful?
 
You can always up the NBcore to 1.27 or 1.3v and give that a try. If that doesn't work then we'll have to try something else. I know that some of these CPUs have a hard time with the faster ram.
 
You can always up the NBcore to 1.27 or 1.3v and give that a try. If that doesn't work then we'll have to try something else. I know that some of these CPUs have a hard time with the faster ram.

@Johan, thanks for the continued help.

Here is what I have tried following up on your quoted suggestions:

CPU Clock: 22
FSB: 200
Vcore: 1.3875
NBCore: 1.25 and 1.3

XMP Profile: Disabled
Memory Timings: as per S_B's suggestions
Memory Multiplier: 9.33 (1866Mhz)
DRAM Voltage: 1.5 and 1.65

So four combinations in total with the different variations of NBcore and DRAM Voltage. All four of them failed to post. Any ideas?



Just to make sure we are all on the same page, my memory is only spec'd at 1600mhz. I've been able to get stability (1 hour P95 Blend) with this memory frequency using the following:

CPU Clock: 22
FSB: 200
Vcore: 1.3875
NBCore: 1.1625

XMP Profile: Profile 1
Memory Multiplier 8.00
Memory Timings: Auto
DRAM Voltage: 1.35
 
OK well if the memory wants to run at 1600 then you can up the reference clock until you get 1600 speed again. Drop a multi for the CPU for 4.4 and also for the NB to keep it around 2.2. Keep the NBCore at stock volts as well as the ram and us stock timings
 
Would you mind elaborating on OCing the memory frequency with the XMP Profile on and the timings set to "Auto". This was the closest I came to becoming stable at 1866. Although OCing the memory isnt a priority for me, it would be fun to see if I could do it.

These were the settings I used when I was able to post with the memory multiplier at 9.33:

CPU Clock: 22
FSB: 200
Vcore: 1.3875
NBCore: 1.1625

XMP Profile: Profile 1
Memory Timings: Auto
Memory Multiplier: 9.33 (1866Mhz)
DRAM Voltage: 1.404


Perhaps bumping the NBcore with these settings will help stabalize the speed with the timings on "auto". I know you said this defeats the purpose of OCing the memory, but I dont really understand why? If you wouldn't mind explaining I would appreciate your insight
 
Last edited:
I said it defeats the XMP since you set all the timings back to auto.
Either way if you want to get 1866 working. Leave it all on auto I'd set the Dram voltage to 1.5, make sure the NB speed is only 2200 or less. Post the machine and then show us some shots of CPUz- main and memory tabs
 
I said it defeats the XMP since you set all the timings back to auto.
Either way if you want to get 1866 working. Leave it all on auto I'd set the Dram voltage to 1.5, make sure the NB speed is only 2200 or less. Post the machine and then show us some shots of CPUz- main and memory tabs

I think I understand what you are saying: So if I enable XMP and manually adjust the memory multiplier to 1866, but leave the timings to auto, the timings will stay at 1600mhz timings (something like 9-9-9-24 35 2)?

I will get those images up later tonight.

Thanks again for your help.
 
AFter a bit of research, I'm having trouble finding blackline 1866 memory with similar part numbers. However while looking at their list, I've found plenty of 2133 blackline memory with some timings you can try at the 1866 speeds.

Some 800 also with loose timings. Just a matter of finding which timings will work best.

800mhz timing part number 992055 is 10-10-10-27 1.5v

2133mhz part number 994164R timings of 10-12-12-28 1.6v

I'm more familiar with their RedLine parts and have had great luck with them. I'll continue research but 9-9-9-24 is a really common timing for most blacklines, it may take some loosening up to get over that 800mhz frequency.
 
AFter a bit of research, I'm having trouble finding blackline 1866 memory with similar part numbers. However while looking at their list, I've found plenty of 2133 blackline memory with some timings you can try at the 1866 speeds.

Some 800 also with loose timings. Just a matter of finding which timings will work best.

800mhz timing part number 992055 is 10-10-10-27 1.5v

2133mhz part number 994164R timings of 10-12-12-28 1.6v

I'm more familiar with their RedLine parts and have had great luck with them. I'll continue research but 9-9-9-24 is a really common timing for most blacklines, it may take some loosening up to get over that 800mhz frequency.

Thanks, I will play around a bit when I get a chance and post back (will also get those images of CPU-Z up as well).

In the meantime, for those interested, I emailed Mushkin directly and asked them about overclocking this particular model of RAM. This was their response:

"If you are going to overclock the memory to 1866, then yes you would need to set the timings manually in the BIOS.
There isn't really an 1866 spec that you could use, well there is for basic plug and play type memory(13-13-13-32 @ 1866) , but that is not performance memory.

Typically what you would do when overclocking memory is loosen the timings when you go up a step in frequency. For example with this memory if you set the frequency to 1866 you would set the timings to 10-10-10-27 1.35V
Then you should test the new settings with MemTest to make sure you are error free. http://memtest.org/
With the memory at CL10 @ 1866 the performance is going to be practically the same as CL9 @ 1600, you gain with the higher frequency but lose out on the looser timings. So you will want to see if you can get the timings tighter.

The correct way to tighten the timings would be to tighten one timing at a time and test with MemTest after each change, like you would try 9-10-10-27 and that was MemTest error free then go 9-9-10-27 and so on. If you got errors then you increase the memory voltage a bit to see if the errors go away with more voltage. By small increase I mean 0.05V
However with this kit you can probably get away with setting 9-9-9-24 @ 1866 with 1.50V

I have to stress that if you go beyond rated specs you need to test with MemTest and be error free for at least one pass of MemTest before you try to boot into the O/S
Unstable memory can and does corrupt operating systems. If you are error free in MemTest the chance of O/S corruption is very low.
 
Here is the screenshot of CPU-Z using 1866 speed

CPU Clock: 22
FSB: 200
Vcore: 1.3875
NBCore: 1.1875

XMP Profile: Profile 1
Memory Timings: Auto
Memory Multiplier: 9.33 (1866Mhz)
DRAM Voltage: 1.354

RAM at 1866 - XMP Auto - NBVcore 1.1875 -.PNG
 
I want to imagine that these timings are loose for the low voltage.

You could try a little more Ram voltrage 1.5'ish range and tighten timings like this IE: 11-11-10 and test 11-10-10 and test.

You'd have a more enjoyable overclocking experience with Mushkin Redlines often times their 933 Ram will do 1200mhz in most cases stability between 1100 - 1200 effective speeds.

Glad you got something working though!
 
Back