• 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.
I agree with the above comments, also Joni if you don't already, you may want to stick a fan at the back of the motherboards. That is if the case allows you to stick a fan on the right side panel and/or you could always modify the case for one.
 
Wow you have one golden piece of silicon there Joni.

Dunno, I doubt it's anything special. I didn't do Prime for 12 hours, only did quick couple of hours of blend and now folding for the 2nd day. Seems to be f@h stable though.

I agree with the above comments, also Joni if you don't already, you may want to stick a fan at the back of the motherboards. That is if the case allows you to stick a fan on the right side panel and/or you could always modify the case for one.

I did try that a week ago, but it didn't seem to have much effect on my setup.

I had case open and extra 2000 rpm fan ready in my hand. I then ran Prime until I hit 72/62 C and pointed the fan directly in the back of the socket. Socket temp went from 72 to 64 but it didn't seem to have no effect on core temperature, it was 64/62 with the fan. It only seemed to cool the socket sensor.

Maybe it's different with the case closed?
 
That is what I I am referring to, cooling the socket area a bit more. See post 97 and 98 the difference between your cpu temp and mine?
 
That is what I I am referring to, cooling the socket area a bit more. See post 97 and 98 the difference between your cpu temp and mine?

Oh, I see. But are you sure we aren't cooling just the socket sensor, while the CPU temperature actually stays the same? I mean if we are cooling the socket area, it should also lower the core temperature, or should it?

When I pointed that 2000rpm fan directly to the CPU (between the heatsink and the ram) it lowered both socket and core temperatures, but on the backside of the mobo (pointed directly to the sensor) it only lowered socket temperature.
 
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.

View attachment 126262

I said golden cause you said youwere at 4.6G and HW monitor says 1.416 vcore. For that speed I need almost 1.47v.
 
Oh, I see. But are you sure we aren't cooling just the socket sensor, while the CPU temperature actually stays the same? I mean if we are cooling the socket area, it should also lower the core temperature, or should it?

When I pointed that 2000rpm fan directly to the CPU (between the heatsink and the ram) it lowered both socket and core temperatures, but on the backside of the mobo (pointed directly to the sensor) it only lowered socket temperature.
This topic is up to dispute and I don't know the real answer per se. I can tell you from experimenting I had my rear fan intaking air and blowing it on my motherboard socket area at first and I was having instability issues when I was really trying to push my chip around 5.4. My temps weren't showing any issues, but I have heard that blowing air directly on that area can cause the sensors to be fooled, thinking that the socket is cooler then it really was. I shut the fan down to see my socket temps were really high. So I reversed the fan to exhaust sucking air out. I also do not have the fan running high, it's running about 800rpm. When I reversed the fan I was no longer having issues and my socket temps were about 6-7 c cooler. Then they were without a fan. With all that said I don't know if reversing the fan was what helped my instability because it isn't the only parameter i changed.
 
I said golden cause you said youwere at 4.6G and HW monitor says 1.416 vcore. For that speed I need almost 1.47v.

Yeah, but it's tested just with two hours of blend. I'm pretty sure it fails if I leave prime running for the night. I'm curious, is that 1.47 what you set on bios, vdrop or vdroop? Also did you use LLC?

I might have pretty different methods than what people usually suggest. I only disable APM Master Mode and Turbo then enable HPC mode. I like to leave all the power saving options in the bios on as I haven't found them to cause any instability. I only up the CPU multiplier and leave everything else untouched. For the voltages, I set LLC to get rid of the huge vdroop I have (Ultra High for 4.6), then leave voltage on auto and add positive or negative offset when needed. If it's not on auto, it wont idle on nice 0,9V and also, if I use a manual voltage instead of auto and offset, for some reason I can't seem to get this running as low voltages than with auto & offset.

This topic is up to dispute and I don't know the real answer per se. I can tell you from experimenting I had my rear fan intaking air and blowing it on my motherboard socket area at first and I was having instability issues when I was really trying to push my chip around 5.4. My temps weren't showing any issues, but I have heard that blowing air directly on that area can cause the sensors to be fooled, thinking that the socket is cooler then it really was. I shut the fan down to see my socket temps were really high. So I reversed the fan to exhaust sucking air out. I also do not have the fan running high, it's running about 800rpm. When I reversed the fan I was no longer having issues and my socket temps were about 6-7 c cooler. Then they were without a fan. With all that said I don't know if reversing the fan was what helped my instability because it isn't the only parameter i changed.

I'll try to fix a fan there tomorrow. Would be great to get those temps lower even for couple of Celcius.
 
Last edited:
Joni , I have it on manual with the LLC set to extreme. My Vcore is set at 1.45 and jumps to 1.47 under load. But different boards work differently that way. This one only jumps .02v where my M5A99FX PRO jumps .07v on extreme.
 
Joni , I have it on manual with the LLC set to extreme. My Vcore is set at 1.45 and jumps to 1.47 under load. But different boards work differently that way. This one only jumps .02v where my M5A99FX PRO jumps .07v on extreme.

I remeber my M5A99X on extreme gave the Cpu V a good boot in the behind, it may have been similar to your pro Johan. I don't think I've used Extreme on the CHV.
 
Had a bit of discussion a while back with RGne and Trents, I'm using it like an offset so I can run a lower volt when it's not working. RGone suggested the using the offset voltages instead of manual , I gave it a shot for a few hours one day and I'm back to this. I find it more predictable. On my 965 it runs ~ 1.4 , when LLC kicks in it jumps to 1.47 , never any issues with it.
 
Had a bit of discussion a while back with RGne and Trents, I'm using it like an offset so I can run a lower volt when it's not working. RGone suggested the using the offset voltages instead of manual , I gave it a shot for a few hours one day and I'm back to this. I find it more predictable. On my 965 it runs ~ 1.4 , when LLC kicks in it jumps to 1.47 , never any issues with it.

Could you high voltage be because of NB or RAM? Try to lower the voltage, to something that you know fails on blend, then do a small fft to see if it passes that.

I did this on 4.4. small fft passed 12 hours with 1.356 V on load but fails blend and large fft. I lowered ram to 1600 and all tests passed. Also by increasing the cpu vcore considerable all passed with ram at 1866.
 
That makes sense Joni but I'm willing to put more volts into it for the faster RAM. Right now it's at just a tick over 2000 and I get good results with it. I also have my NB at 2500 and 1.3v The IMC is working a lot harder and will need more power.
But my CPU was a dog right from the get go
 
Last edited:
But my CPU was a dog right from the get go

I remember that "Johan45". You came into the forum and it seemed yours was the worst FX-8350 we had ever heard of. There was one other of them about that same time that did not seem much better. Otherwise most FX-8350 seemed better than any other Vishera.

I know at first I gave my FX-8350 a 5 on the 10 scale. Then as I used and benched it I felt it was at least an 8, I thought. And then at 4.8 and video edittig, I felt it was a 10. So much so did I like the 4.8Ghz video editting that I just about refused to push it really high again just for the heck of it and risk a pretty good cpu.

Then we tried to talk "Joni" out of using his FX-8350 but to get his money back and then even though it was not a 1.272 Volt Cpu his FX-8350 started to look like a pretty good piece of silicon for sure.

I did the short thread on using Cpu offset voltage just because it was something that was used by some on the Deneb processors and seemed to work to keep the Vcore lower most of the time unless just pushed. But on my CHV board it did not seem to do exactly what I wanted it to do. I went back to manual and Ultra-High on the CPU LLC and though it will not lower the voltage unloaded like runnng cpu offset voltage will, it seems more predictable. I think it is because the CHV Cpu LLC seems to work backwards from the other lower tiered boards of even Asus but most of the other brands as well.

We spend so much time trying to get all these users with tell me how to overclock when there are thousands of How-Tos out there that we seldom do what this threads is doing and take the time to hammer on what is happening with a board and a cpu. Or similar situations.

I always thought motherboards of a brand and model would all be alike. I think I was vrey wrong. If cellphones or radios varied as much as motherboards seem to do, then we would never get a connection to a dialed number or get a specific radio station. Hehehe.

A few times I got a couple of DFI motherboards with consecutive serial numbers. They were sent me that way from the assembly line. They would do pretty much the same if I stuck the same cpu and ram in them. I have gotten the same model motherboard with serial numbers very far apart and they may well be pretty much alike or vary big time. I never found out 'the' why of such actions. I know how the robots assemble the boards and the container after container of parts are placed on the boards and you would think they would all be alike.

C_D has more than one CHV of the early type with no -z in the model number. He has changed out the board that he has used as HIS bench board that he writes about in here and no matter which one of the CHV boards he uses...his stuff does not act like mine. So is it the board that makes for odd stuff or is it the fact he swapped the boards but not the cpus and ram and power supply and all of that which remains the same after swapping thru a couple of the same model of motherboard and the same bios on each?

All of his CHV boards run better than most of the cheaper boards we see all the time. But they never run just like mine. We have swapped settings between us and still his multiple CHV boards never run just like mine or my CHV never seems to run just like his.

And then I put the Asrock Fatal1ty board together with an FX-6300 and got a real eye-opener on cpus. No matter what, I could not get the FX-6300 to impress me. Too much Vcore needed, I thought. With my cooling I was not really able to recognize the FX-6300 being much cooler. AND THEN, I put my FX-8350 in the Asrock FatLady board as C_D and I affectionately call her and she did well with the FX-8350 until I was testing P95 Blend up the ladder. Some how and I have not found the answer yet, I lost my way and after a testing everything I could, I thought my FX-8350 had LOST a step. You know slowed! I was sure my FX-8350 had slowed.

Then I put the FX-8350 back on my CHV and she took off like a rocket again. I am certain now that I know what happened. The FX-8350 was not using exactly the same software monitored voltages and the CPU LLC works backwards from the CHV and I was trying to HOP from one speed to another in large jumps and I just got lost. Could not seem to find my way again until I got back to my darling CHV board. Hehehe. What a dunce I am.

So what the heck is all this text about? Learn the principles. When you know the principles, test; then test and after that test our stuff again. Only working with the board and cpu one plans to use can he find out what his system will really do. Each of us that knows the principles very well can try and offer suggestions for go fast, but only the one with the keyboard hooked up to the system can prove out each system. Even after all that you can swap boards and be a klutz like I think I was with my FX-8350 and the Fatal1ty board.
RGone...
 
Last edited:
I was trying to HOP from one speed to another in large jumps and I just got lost.
Rgone well said, I can't count the times I've lost my way being lazy and just trying to get to a number. Taking good notes and writing down every change has been most beneficial for me when I'm really pushing.


Only working with the board and cpu one plans to use can he find out what his system will really do. Each of us that knows the principles very well can try and offer suggestions for go fast, but only the one with the keyboard hooked up to the system can prove out each system.
Learning what your combo will do is the best way to overclock!
 
yeah... its the principles that will steer you right. as long as you understand how the chip works, what does what, what your errors mean, and the safety limits, then aproach overclocking methodically, you'll be able to overclock

there is a great reason hidden in all of this text that you aren't seeing a lot of specifics... its because overclocking is different from cpu to cpu, and different from motherboard to motherboard, and different ram to ram, and again, different psu to psu, even different based on cooling solutions (cases, cpu coolers, fans) used... you need to learn how your system works. not how mine or rgone's or anyone else's system works.

learn the basics, play with it yourself, familiarize yourself with how it behaves, soon you'll start to learn what YOUR chip/mb/ram/psu can do, and what you need to do to get it stable.
 
That makes sense Joni but I'm willing to put more volts into it for the faster RAM. Right now it's at just a tick over 2000 and I get good results with it. I also have my NB at 2500 and 1.3v The IMC is working a lot harder and will need more power.
But my CPU was a dog right from the get go

Just curious what VID does your chip have?

By the way, you are coming pretty fast on F@h stats. What do you use?

I saw there is Mandrake, You and I pretty close :)
 
I honestly don't remember what my chips VID is. I do know that as soon as I went over 4.5 my voltage skyrocketed. I'm fairly certain that my baby is a leaker judging by the volts and heat output.
I'm using my main PC in my sig and after OC'ing the cards a bit last night it was estimating almost 82000 ppd. I just set up my HTPC to join in this morning before work but still have some bugs. Have to optimize the Radeon. So with luck between the 2 I'll be around 100k
 
I honestly don't remember what my chips VID is. I do know that as soon as I went over 4.5 my voltage skyrocketed. I'm fairly certain that my baby is a leaker judging by the volts and heat output.
I'm using my main PC in my sig and after OC'ing the cards a bit last night it was estimating almost 82000 ppd. I just set up my HTPC to join in this morning before work but still have some bugs. Have to optimize the Radeon. So with luck between the 2 I'll be around 100k

CoreTemp should show the default VID if Turbo is off. It's amazing what GPU's can do, I have 6 processors running and get only about 100k. I need to unbox fifth 3570k today :/
 
I'll check tonight when I get home for you it might be late though.
Ya just 1 of my 580's was hitting 36000 ppd at one point yesterday.
 
Just curious what VID does your chip have?

By the way, you are coming pretty fast on F@h stats. What do you use?

I saw there is Mandrake, You and I pretty close :)
Close ha, it's not going to be close. Johans 1 rig w the 2 580's is producing more then my 3 combined. Joni if you're going to be doing 100k I'll be limping along at 70-75k at best. Maybe I'll crank up my Athlon II I should be able to get at least another 1k a day out of that ;).
 
Back