Overclocking fx 8350

[Martinez87]

So I recently upgrade my gpu card and noticed something very weird due to overclock fx 8350

Specs
Motherboard: sabertooth 990fx r2.0
CPU: FX-8350
Cooler: Dark rock pro 3
GPU: Club 3D Radeon HD 7870 / gigabyte gtx 970 g1
RAM: Crucial Ballistix Tactical 1866 4GBx2
HDD: Crucial SSD MX100 256GB
PSU: Thermaltake smart se 630w

max overclock with radeon gpu
Core speed: 4,6ghz
multiplier & bus: 23x 200
Core voltage: 1,416v idle and 1,428v load (llc ultra high)
NB & HT clock: 2600Mhz
after 1h prime95 blend test max core temp was 61deg with no errors
No matter that I do with multiplier, bus and voltage couldn't reach stable 4,7ghz

max overclock with geforce gpu
Core speed: 4,7ghz
multiplier & bus: 20x 235
Core voltage: 1,44v idle and 1,464v load (llc ultra high)
NB & HT clock: 2600Mhz
after 1h prime95 blend test max core temp was 61deg with no errors
Couldn't get it stable on these settings when I switch back to radeon card so how is it possible?

And then I try to reach 4,8ghz
Core speed: 4,8ghz
multiplier & bus: 20.5x 234
Core voltage: 1,464v idle and 1,488v load (llc ultra high)
NB & HT clock: 2600Mhz
after 1h prime95 blend test max core temp was 64deg with no errors
when I switch back to radeon card after 3min I get errors on these settings so that is very weird

Some saying 62deg is max temp for fx 8350 and others 65deg so I decided to stay with 4,7ghz
but gpu card shouldn't affect cpu overclock - could you explain this for me?

 

[newtekie1]

The max temp is the TJMax temp set by AMD, and that is 70°C. Beyond that the CPU will throttle to protect itself. Some people say anything beyond 62°C will damage the CPU, I don't believe them, I believe AMD who actually made the CPU. However, there is truth to the fact that FX CPUs can become unstable if they get too hot. So an overclock that is stable at 55°C might be unstable at 65°C. Though I don't think that is what is happening to you.

I think your issue is the AMD graphics card is pulling more power than the GTX970. This is causing your power supply to provide more clean/stable power with the 970, and that in turn allows you to get a slightly better overclock.

 

[de.das.dude]

The max temp is the TJMax temp set by AMD, and that is 70°C. Beyond that the CPU will throttle to protect itself. Some people say anything beyond 62°C will damage the CPU, I don't believe them, I believe AMD who actually made the CPU. However, there is truth to the fact that FX CPUs can become unstable if they get too hot. So an overclock that is stable at 55°C might be unstable at 65°C. Though I don't think that is what is happening to you.

I think your issue is the AMD graphics card is pulling more power than the GTX970. This is causing your power supply to provide more clean/stable power with the 970, and that in turn allows you to get a slightly better overclock.

this is the only sane explanation i can find. Radeon cards have higher idle draws if i am not wrong.

 

[xfia]

this is the only sane explanation i can find. Radeon cards have higher idle draws if i am not wrong.

i see more.. we dont have the whole picture. the fsb is being changed.

 

[Martinez87]

Thanks for Yours opinions

GPU Power Requirements from official sites:
gtx970 - 145w
hd7870 - 175w
it's not to much difference but maybe you are right
but I overclocked slighty gtx970 to 1460Mhz on core and to 7400Mhz on memory without touching the voltage
overclocked gtx970 power consumption should be more than 145w
CPU 4,8ghz power consumption from some benchmark:
fx8350 215w idle and 365w load
so 630w psu should be enough but maybe it's not enough

What do Yuo mean "the fsb is being changed"
I tested both cards on the same settings so fsb should be the same
FSB:RAM 1:4 cpu/nb voltage 1,275v
I think it's extremely hard or even impossible to go higher than 4,6ghz with NB&HT set to 2600Mhz
only using multiplier and being stable with safe temps
of course on air cooler

 

[newtekie1]

That Thermaltake, from what I can find on it. doesn't look like that good of a unit. So more draw on the 12v rail, even a little, could bring it down to the point where the CPU becomes unstable.

Also, those are max numbers, the 7870 draws a lot more than the GTX970 at idle. The GPU is idle when you are running Prime95. I'm willing to bet if you ran Prime95 with GPU-Z's Render Test open, which will load both the CPU and GPU, you'd find the CPU overclock isn't stable regardless of which graphics card you have in the system.

 

[Martinez87]

Thanks for Your replie newtekie1
Your argumentation was reasonable to me "running Prime95 with GPU-Z's Render Test"
I did this test now but only for 40min
So I load both overclocked CPU & GPU without any errors with Fx 8350 4,7ghz & gtx970 1,46ghz
I thought Cpu temp will be be higher because gpu will provide extra heat but no
Gpu-z shows during test 50% of power consumption
So the psu is handling my performance quite good I think
But maybe i'm wrong

 

[eidairaman1]

The max temp is the TJMax temp set by AMD, and that is 70°C. Beyond that the CPU will throttle to protect itself. Some people say anything beyond 62°C will damage the CPU, I don't believe them, I believe AMD who actually made the CPU. However, there is truth to the fact that FX CPUs can become unstable if they get too hot. So an overclock that is stable at 55°C might be unstable at 65°C. Though I don't think that is what is happening to you.

I think your issue is the AMD graphics card is pulling more power than the GTX970. This is causing your power supply to provide more clean/stable power with the 970, and that in turn allows you to get a slightly better overclock.

Yeah considering looking at my specs my CPU is at 5.1GHz on a Scythe Ashura (Smaller than a Dark Rock 3) and im at 1.476Vcore, at 5.2 It hits a Thermal wall when Vcore is set higher, 5.3 it locks at Windows Welcome Screen.

 

[Martinez87]

safe temp from official site...
http://products.amd.com/en-us/search/CPU/AMD-FX-Series/AMD-FX-8-Core-Black-Edition/FX-8350/92

 

[xfia]

@Martinez87 your fsb!!! http://www.amd.com/Documents/AMD_FX_Performance_Tuning_Guide.pdf

you got a rockn mobo so you can change the fsb but it would make more sense with your ram set at 1600mhz with a fsb of like 230-233

amd does not set throttle temps on mobo's.. it is the manufacturers that do. can you run it over 61c? yes but keeping it under 62c will keep you at max lifespan.

 

[Martinez87]

xfia You are probably right but what the fsb has in common with different gpu on board? With one prime test can be passed but with other can not
I don't undarstand that
Tried it with corsair 750w and the same effect

 

[Vayra86]

What you are likely suffering from is what LLC (load line calibration) actually counters to a degree.

Basically, every time you put a load on CPU or GPU, you take power from the 12V+ line. That means the current on that line will drop a bit (vDroop) and LLC 'evens that out'.

Hotter components take more power, again putting more stress on the 12V line. At some point that effect strengthens itself, and your 12V current can drop very briefly 'below required'. That's when your OC goes boom. As GPU also takes power and AMD idle GPU takes a bit more than Nvidia (also at load, btw), that can also incur a penalty on the 12V current.

You can load up OCCT and look at the 12V current at load, and see if you have the correct LLC setting to get a stable, straight line in the graphs you get when benching on OCCT. For LLC, higher is not always better.

Disclaimer: I have no experience with FX overclocking, but vDroop is always an issue. Besides PSU wattage, more expensive PSU's generally have more stability on the 12V and are better capable of keeping it straight when the load changes.

 

[Martinez87]

tried overclock (idle radeon gpu) with high llc and many combinations of multiplier, bus and voltage and can't reach 4,7ghz
than I run first GPU-Z's Render Test and while running it I start Prime95 blend test (load both radeon gpu and fx 8350)
and insane thing happened - I can pass prime95 test (1 hour) with 4,7ghz and 4,8ghz too (llc ultra high and 630w psu) - settings and temps are similar as in post one
but I'm unstable then radeon gpu is not stressed = idle
so wtf is happening?
I've never seen something like this

 

[eidairaman1]

turn off LLC, Cool n Quiet, spread spectrum, turbo core

 

[Martinez87]

Completly turning of llc will provide higher voltage and lower oc possibility - ok, end of testing, this wierd card is going for sale, thanks for opinions. End of thread.

 

[PCGamerDR]

Try increasing NB Volt and see how that goes, fx8350's support ram up to 1600, i remember had to increase my nb volt a bit on my cpu to be completely stable with my 2400mhz kits. Also i hate prime95, try aida64 or occt .

 

[chuck216]

Question is why are you upping the buss speed with an 8350? They run a lot more stably if you simply change the multipier 23.5 for 3.7 ghz, 24 for 3.8 etc.

 

[cdawall]

Question is why are you upping the buss speed with an 8350? They run a lot more stably if you simply change the multipier 23.5 for 3.7 ghz, 24 for 3.8 etc.

No its not. It makes no difference at all if you keep the settings tied to the bus within check. Keep the NB and HTT link speed at sane levels and stability is the same whether you run 200mhz bus or 300mhz bus.

 

[xfia]

turn off LLC, Cool n Quiet, spread spectrum, turbo core

that is really friendly to mother earth to be telling everyone to turn off QnC.. you clearly need to question what you think you know and leave it up to @cdawall and others with more experience. this is why we need a new guide from amd..
this might have been someone from amd.. microsoft.. intel.. hell maybe the cia. crazy to think they wouldnt do something like this.

 

[Jism]

I see some enourmous bullshit going on here...

LLC is menth to cover the voltage-drop when load is applied. It's electrons-nature. When you OC, you need to have LLC to prevent the system from going unstable after load is applied.

Second; there is a difference in between reading the sensors and the actual CPU core temp. The TJ-max should never be exceeded when oc'ing or stresstesting. If the TJ-max (Check "Coretemp") is exceeded you are burning out your CPU from the inside, and limiting any good OC's. So make sure your cooling is allright and stays well under the CPU TJ-max limit.

now, as for your OC; you should know that both PCI-express bus and CPU 12V are sharing the same line. Radeons tend to draw a higher current from the PCI-express bus then the standard 75W. For most systems, this is no real issue, but when hunting for the CPU limits, your also seeking the power limits your Board or PSU can deliver. You should lower your OC of your CPU if you are getting problems with 3D testing. Then your complete OC is just not stable.

 

[eidairaman1]

Explain why im at 5.1GHz Stable then. No LLC, im on air where most have to be water cooled, by the way xfia, i do talk with cdawall often on this stuff. I used the FX Ocers club for help even.

 

[Jism]

As i said before, some Radeons tend to draw a higher current from the PCI-express bus, from the very same line as your CPU 12V feed. Your OC is'nt 24/7 stable then. Your motherboard or PSU is'nt able to provide the extra current that's drawn. Or your PCI-express bus is OC'ed as well, you need to put that on 100MHz only.


Try something like this to exclude, your Radeon is drawing a higher current then allowed:

It feeds the PCI-express bus with an extra 12V which makes it possible to share the load between the CPU & GPU. There is no harm caused in using such a wire. 5.1GHz OC does not mean a thing, by the looks of it your system is just not stable at 4.8GHz. Yes you can run 5.1 but for how long again? I call a system thats running 48 hours any blend / prime / OCCT test stable.

 

[NC37]

Voltages shouldn't increase with LLC. I've been running without LLC for years. With LLC on I had to run at close to 1.45v to get this 8320 stable at 4.3Ghz. With LLC off I run at 1.4v @ 4.5Ghz. I had it booting into 4.7 but temps got high and I backed off.

Heat increases too because of that. More voltage, more heat.

 

[Jism]

Motherboards with a crappy VRM / capacitors tend to have a higher VDROP when load is applied. That's where LLC (Load line calibration) is coming in. It simply compensates the moment load is applied, and thus the voltage drops, basicly increasing the CPU voltage in order to keep it operating at a certain voltage. However, when LLC is at max setup, it will simply put voltage way over into the CPU, as seen on some older AMD cpu's where the base CPU voltage is set at 1.45V and LLC is kicking in with almost 1.6V. LLC works in most cases but it also generates more heat and making the CPU unstable when getting too hot.

 

[xfia]

Explain why im at 5.1GHz Stable then. No LLC, im on air where most have to be water cooled, by the way xfia, i do talk with cdawall often on this stuff. I used the FX Ocers club for help even.

its all right here for you now.. you can turn off qnc and c6 for higher oc's but why not turn them back on? i would say there is a certain stability that your at a lose if you cant turn them back on. i know people that have to touch every single thing in the bios like the engineers left them a virus on auto to kill lol.. need to help get away from thinking microsoft doesnt know what they are doing.. after all they did skip windows 9 just to make everyone happy.