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

FX-8350 F@h unstable

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
Most current stable one week F@h 24/7 is the following:

BIOS settings:
Current CPU Speed: 4500mhz
CPU Ratio: 18
CPU Bus: 250
Mem Freq.: DDR3-2000mhz
CPU/NB Freq.: 2250mhz
HT Link Speed: 2250mhz
DRAM Timings: 8-10-9-27-1T
CPU Voltage: 1.46875v
CPU/NB Voltage: 1.2v

Next I'm going to experiment here:

BIOS settings:
Current CPU Speed: 4500mhz
CPU Ratio: 18
CPU Bus: 250
Mem Freq.: DDR3-2000mhz
CPU/NB Freq.: 2500mhz
HT Link Speed: 2500mhz
DRAM Timings: 8-10-9-27-1T
CPU Voltage: 1.46875v
CPU/NB Voltage: 1.3v
 
Ok done!

BIOS settings:
Current CPU Speed: 4500mhz
CPU Ratio: 18
CPU Bus: 250
Mem Freq.: DDR3-2000mhz
CPU/NB Freq.: 2500mhz
HT Link Speed: 2500mhz
DRAM Timings: 8-10-9-27-1T
CPU Voltage: 1.46875v
CPU/NB Voltage: 1.25v
 
"csimon", AMD even suggests in the overclocking guide for the FX processors that 1.35V CPU_NB is not too high for use with the FX using air cooling. That said with the CPU_NB still at a default voltage of around 1.1 and I have seen some at 1.185 for some reason, it covers a lot of ground to go from that default to 1.35Volts.

The wisdom has been to raise the CPU_NB speed to more readily accommodate the use of faster ram speeds. When the CPU_NB is speeded up the CPU_NB often needs an increase in voltage to be stable with a higher frequency.

Now what voltage does the sped up CPU_NB need for it to function without error with the increase? That is determined system by system since there are variables. One thing to know though is that increased CPU_NB voltage does bring some increase in operating temperature within the cpu itself since that is where the CPU_NB aKa the IMC is located.

In my own particular instance I use a 1.2815 CPU_NB voltage with the CPU_NB LLC set to HIGH. It is enough CPU_NB voltage on my CHV with FX-8350 or the FX-8120 I had in use before. However my same CPU on the Asrock Fatal1ty mobo seemed to require 1.3125V CPU_NB. This with ram speeds in the range of DDR1900. I have to admit though that my CPU_NB and HT Frequency are normally on the order of +2550Mhz and depending on the exact CPU clock they both may exceed 2600Mhz.

Using 1.2825V with my CHV and 1.3125V in bios for the CHV and Fatal1ty and leaving the voltages set at that setting even when slowing my overclock has no adverse effects on my configuration. I have fully adequete cooling to ignore any temp rise for CPU_NB speed or voltage increase. That condition though is not necessarily so for all users.

I did however arrive at the CPU_NB voltage by trial and error beginning at 1.225V and moving upward until I found my needed voltage and after arriving at the needed voltage and having very adequete cooling; I do not have to constantly adjust that voltage as I might if running just on air cooling.

There is a difference in my use of the computer and your use though. There are times when I am doing photo editting and or little more than surfing the web. You are likely running F@H in the background and may wish the 4.5Ghz speed for that for use in the background. I throttle back nowdays to 4.0Ghz when doing little actual work on the computer. My actual video editting is done at 4.8Ghz with ram speed as close as I can get it to DDR1900 since the DDR1900 is using two different kits of ram. One rated to run DDR1600 and the other at DDR1866 but with the both I can use 16gigs of ram. An increased amount of ram can also require that the CPU_NB voltage rise to facilitate stability. My use of 1.2815 seems to accommodate the 16gigs of ram as well, so my CPU_NB stays set always to that voltage since I now use 16Gigs of ram with two kits of mis-matched speed ram.

The overall point is that when the system is reaching its limits and is very close to stable at all times, the voltages should require hands-on testing. Undervolted normally will bring an error. Slightly Overvolted is seldom noticed, but the idea of just too much voltage is not generally what most like to see. So test your system. Test it thoroughly and as completely as you can. We are not the final say but your system itself is and only you can dial it in. Test and test again and that will bring your own system to overclocking fruition. Luck man.
RGone...
 
Even though I don't have a FX-8350, my highest temp with my systems (2x PH II 1090T, 1xFX-8120 and 1xFX-8150) is 39-41c, and that's on the FX-8120 at 3.6GHZ. All using the V6GT cooler and running F@H.

8120-3.6GHZ-c.JPG
 
Why not try a little less Cpu NB V first like 1.25?
Oh well at least I tried. It took a dump on my during the night so I had to raise the cpu/nb volts to 1.3v again.

Current settings:

BIOS settings:
Current CPU Speed: 4500mhz
CPU Ratio: 18
CPU Bus: 250
Mem Freq.: DDR3-2000mhz
CPU/NB Freq.: 2500mhz
HT Link Speed: 2500mhz
DRAM Timings: 8-10-9-27-1T
CPU Voltage: 1.46875v
CPU/NB Voltage: 1.3v

Thanks for that reassurement RGone, I feel better about higher cpu/nb voltages & frequency.
 
Good to see you're feeling a bit more comfortable tweaking these settings
 
"csminon", a couple of questions:

1. Did you get CPU_NB and HT Freq okay when running them faster than their defaults as "Joni" asked?

2. Is this really important?
Help Stanford University scientists studying Alzheimer's, Huntington's, Parkinson’s, and many cancers by simply running a piece of software on your computer.

The problems we are trying to solve require so many calculations, we ask people to donate their unused computer power to crunch some of the numbers.

I ask that question because as I look at your buss speeds and ram timings it looks as if you were adjusting to try and run benchmarks for HWBot or similar. Not actually like you had important applications and data to be run on the system.

I went to Mushkin site and looked at that ram. I had noticed your signature said you were running it at 8-10-8-27-1T and I wanted to see what the ram was actually rated to run. I found this at Mushkin site >>
Redline 996997 - 8GB (2x4GB) DDR3 PC3-17000 9-11-10-28 Voltage:1.65V.

You have that ram screwed down mighty tight below specs for only about 100Mhz reduction in the rated ram speed. Plus I see you running the ram at 1T and not 2T. The Mushkin site did not show a 1T/2T rating so most ram makers that do not show which as a recommendation have it understood it is 2T. In fact generally across the board from the ram manufacturers, the majority of their ram is rated at 2T for AMD systems.

Many of us tweak our systems to the Nth degree when we benchmark and I would think that is fine since benchmarks do not take nearly the precision that many other applications take to run accurately. I do not run applications on my computer with everything clocked as high as I can run benches, nor do I screw the ram timings down as tight as I do for benching. This allows me to run without any odd happenings occurring for one thing and also to run lesser voltages in general for most of my 24/7 type computer usage.

I am just beginning to ponder which maybe the most important? Running F@H in a manner that does not crash and finishes Units for the cause. Or is there really a reliable increase in the number of units finished with just a tightening of ram timings that require the speed up of the other busses and increase of voltage to the other busses? I really have begun to ponder the process.
RGone...
 
+1 I mentioned the ram a while back. It's basically an OC'd set of 1333.
 
"csminon", a couple of questions:

1. Did you get CPU_NB and HT Freq okay when running them faster than their defaults as "Joni" asked?

2. Is this really important?
Help Stanford University scientists studying Alzheimer's, Huntington's, Parkinson’s, and many cancers by simply running a piece of software on your computer.

The problems we are trying to solve require so many calculations, we ask people to donate their unused computer power to crunch some of the numbers.

I ask that question because as I look at your buss speeds and ram timings it looks as if you were adjusting to try and run benchmarks for HWBot or similar. Not actually like you had important applications and data to be run on the system.

I went to Mushkin site and looked at that ram. I had noticed your signature said you were running it at 8-10-8-27-1T and I wanted to see what the ram was actually rated to run. I found this at Mushkin site >>
Redline 996997 - 8GB (2x4GB) DDR3 PC3-17000 9-11-10-28 Voltage:1.65V.

You have that ram screwed down mighty tight below specs for only about 100Mhz reduction in the rated ram speed. Plus I see you running the ram at 1T and not 2T. The Mushkin site did not show a 1T/2T rating so most ram makers that do not show which as a recommendation have it understood it is 2T. In fact generally across the board from the ram manufacturers, the majority of their ram is rated at 2T for AMD systems.

Many of us tweak our systems to the Nth degree when we benchmark and I would think that is fine since benchmarks do not take nearly the precision that many other applications take to run accurately. I do not run applications on my computer with everything clocked as high as I can run benches, nor do I screw the ram timings down as tight as I do for benching. This allows me to run without any odd happenings occurring for one thing and also to run lesser voltages in general for most of my 24/7 type computer usage.

I am just beginning to ponder which maybe the most important? Running F@H in a manner that does not crash and finishes Units for the cause. Or is there really a reliable increase in the number of units finished with just a tightening of ram timings that require the speed up of the other busses and increase of voltage to the other busses? I really have begun to ponder the process.
RGone...

RGone,

1. The CPU_NB and HT Freq is fine at defaults and overclocked.

2. It is important to me in day-to-day usage. This is not a dedicated folding machine by any means, I also work and play on here. However, I am current at 30,000,000+ points for Stanford.

I think I understand where you are headed about the ram, cpu/nb & ht thing.

I believe that a stable overclock on the cpu is best for maximum smp folding. However, I think that when overclocking the cpu/nb & ht in conjunction with this it may actually hamper smp production because of the excess strain to the cpu, not to mention the extra higher temps.

So, I think I may be answering your question when I say that I am just pushing the limits of the ram at this point since I can go no further on the cpu at current temps.

Perhaps I should re-emphasize the following:

I am stable at these settings-

BIOS settings:
Current CPU Speed: 4500mhz
CPU Ratio: 18
CPU Bus: 250
Mem Freq.: DDR3-2000mhz
CPU/NB Freq.: 2250mhz
HT Link Speed: 2250mhz
DRAM Timings: 8-10-9-27-1T
CPU Voltage: 1.46875v
CPU/NB Voltage: 1.1875v

And I am currently testing these settings-

BIOS settings:
Current CPU Speed: 4500mhz
CPU Ratio: 18
CPU Bus: 250
Mem Freq.: DDR3-2000mhz
CPU/NB Freq.: 2500mhz
HT Link Speed: 2500mhz
DRAM Timings: 8-10-9-27-1T
CPU Voltage: 1.46875v
CPU/NB Voltage: 1.3v
 
Last edited:
Tried to quick find the exact answer but couldn't- what are you using for cooling out of curiousity? With the CC coming up I need to get my 8350 to run a little faster without overheating and it looks like you've got that down...
 
The reason I mentioned the ram in the first place was because of how sporadic the error was. 48 hours straight then poop. Sounded like a memory issue to me. Perhaps speeding up the NB will cure that or maybe loosening the timings a tad.
 
Tried to quick find the exact answer but couldn't- what are you using for cooling out of curiousity? With the CC coming up I need to get my 8350 to run a little faster without overheating and it looks like you've got that down...

I'm watercooling the cpu with a Enzotech Sapphire Series SCW-REV.A CPU waterblock, Thermochill PA160 radiator & shroud (designed by Cathar) w/ Cougar Vortex PWM 120 fan, and a Laing D5 pump. The case is an old-school Lian Li PC-V2000B. Temps are fair, I'm replacing the case and cooling systerm this summer since this one is dated.

The reason I mentioned the ram in the first place was because of how sporadic the error was. 48 hours straight then poop. Sounded like a memory issue to me. Perhaps speeding up the NB will cure that or maybe loosening the timings a tad.

Yep, poop it went. However, cpu/nb & ht were stable at defaults so now I will attempt to return cpu/nb w/ voltage to (near) default and leave ht at 2500. Temps were high anyway.

Now testing here:

BIOS settings:
Current CPU Speed: 4500mhz
CPU Ratio: 18
CPU Bus: 250
Mem Freq.: DDR3-2000mhz
CPU/NB Freq.: 2250mhz
HT Link Speed: 2500mhz
DRAM Timings: 8-10-9-27-1T
CPU Voltage: 1.46875v
CPU/NB Voltage: 1.2v
 
Last edited:
Ah, okay.... that is still some pretty heavy-duty cooling compared to just the air that I'm working with. Definitely gives you some more flexibility to have fun with it looks like.

Just curious... what is a typical PPD range for you at the top stable speed you've reached compared to say, 4.0 or 4.2ghz? For benching I can see a huge reason to chase down the higher clocks obviously, but I haven't seen one way or the other yet how much it helps boost folding output...
 
OC depending on the wu I have gotten 20-23k ppd at 4.7 on just the cpu. If I use both I lose about 1-1.5k ppd on the cpu but gain on the gpu. Here's the pic I posted for you on the temps. This was after 4 days straight temps around 70f. Mind you I don't know if it makes much of a difference but I'm only running ddr3 @1600, my 1866 sticks pooped their pants.
 

Attachments

  • Capture1.PNG
    Capture1.PNG
    28.9 KB · Views: 49
I've been folding 24 hours now with 8350 at 4.6 GHz, on air, fans at minimum. Here's a ss of the temps after 24 hours. I also get around 20000 ppd, and lose around 2k if I use the GPU. Still trying to get this to 4.7 for Chimp Challenge.

4.6.png
 
Back