Stablizing fx 9590 and ran with voltage

[Bajantechnician]

@Benji_w

My FX-8350 @ 5.0 GHz is stable, but my Load-Line-Calibration kicks the Core Voltage up to ~1.47V.

Temperature wise, when all cores are on full load (i.e. Prime95), the temperature gets quite high...high 60*C's nearing 70*C.

I had a Swiftech H220...but both (original and replacement unit) each failed within 24 hours of actual operation.

 

@Bajantechnician

Load-Line-Calibration counter-acts voltage drooping when the CPU is under load / stress.

For an example, you sent your Core Voltage to 1.5V. During idle and low loads it may be at 1.5V, but once the CPU is under heavy load, the voltage droops down to say...1.48V, making your overclock / CPU unstable.

 

Adjust the Load-Line-Calibration enough so the voltage is stable.

If you use a too high LLC, then you voltage is be higher than necessary.

 

 

As mentioned before, for the sake of troubleshooting

• Disable Cool N' Quiet
• Disable C1E
• Disable Core C6 State
• Enable HPC Mode (High Performance Computing)
• Disable APM Master Mode (Application Power Management)

You are running your DRAM at 1866 MHz, and with a CPU-NB of 2200 MHz. Drop your CPU-NB down to 1.25V ~ 1.3V. Even 1.3V is not necessary -- 1.4V is just ridiculous...so extreme cpu llc is not necessary ?

It's set on ultra high

[-rascal-]

so extreme cpu llc is not necessary ?

It's set on ultra high

 

Each chip is different, just like overclocking and voltages, so use whatever is necessary.

If you have any program that can monitor Core Voltage (i.e. HWMonitor or CPU-Z), you should be able to see it if it drops, or increases too high.

 

I personally use CPU-Z and HWMonitor (HWMonitor shows the maximums and minimums)...since I won't be always sitting in front of my computer, watching every second, as it's stress testing.

 

Typically, High or Ultra High should be okay (High may cause it to dip just slightly lower - again, this can vary).

 

Also, like I said before, set the correct timings for your DRAM -- to eliminate possible memory related issues.

The timings and voltages should be found on the sticker on the RAM itself, or on the packaging.

 

Adjusting your Over-Current protection settings may help as well. Play around with it, but typically, 110% or 120% should be more than enough.

[Bajantechnician]

Each chip is different, just like overclocking and voltages, so use whatever is necessary.

If you have any program that can monitor Core Voltage (i.e. HWMonitor or CPU-Z), you should be able to see it if it drops, or increases too high.

 

I personally use CPU-Z and HWMonitor (HWMonitor shows the maximums and minimums)...since I won't be always sitting in front of my computer, watching every second, as it's stress testing.

 

Typically, High or Ultra High should be okay (High may cause it to dip just slightly lower - again, this can vary).

 

Also, like I said before, set the correct timings for your DRAM -- to eliminate possible memory related issues.

The timings and voltages should be found on the sticker on the RAM itself, or on the packaging.

 

Adjusting your Over-Current protection settings may help as well. Play around with it, but typically, 110% or 120% should be more than enough.

overcurent is set on auto and timings are fixed in bios. I used cpuz to check if the timings are right. They are 9-10-9-27.

Voltages go from ~1.42 to ~1.54

[vitor_cut]

i have done a little test in my rig and if you put the LLC to ultrahigh and the cpu power to 120%, you can keep the power saving features and have a higher incidence of boost clock, and i run aida 64 stress test for 30h and socket temp goes to 60ºC and package(cpu) to 52ºC ( with a 240 AIO, in a NZXT H440)

 

C1E- disable

EPU saving- disable

[Benji_w]

@Benji_w

My FX-8350 @ 5.0 GHz is stable, but my Load-Line-Calibration kicks the Core Voltage up to ~1.47V.

Temperature wise, when all cores are on full load (i.e. Prime95), the temperature gets quite high...high 60*C's nearing 70*C.

I had a Swiftech H220...but both (original and replacement unit) each failed within 24 hours of actual operation.

 

@Bajantechnician

Load-Line-Calibration counter-acts voltage drooping when the CPU is under load / stress.

For an example, you sent your Core Voltage to 1.5V. During idle and low loads it may be at 1.5V, but once the CPU is under heavy load, the voltage droops down to say...1.48V, making your overclock / CPU unstable.

 

Adjust the Load-Line-Calibration enough so the voltage is stable.

If you use a too high LLC, then you voltage is be higher than necessary.

 

 

As mentioned before, for the sake of troubleshooting

• Disable Cool N' Quiet
• Disable C1E
• Disable Core C6 State
• Enable HPC Mode (High Performance Computing)
• Disable APM Master Mode (Application Power Management)

You are running your DRAM at 1866 MHz, and with a CPU-NB of 2200 MHz. Drop your CPU-NB down to 1.25V ~ 1.3V. Even 1.3V is not necessary -- 1.4V is just ridiculous...

If I recall, I am able to run 2400 MHz CPU-NB with 1.23V (?). Will have to check.

 

Increasing your CPU-VDDA voltage can help stabilize overclocks. DO NOT increase this voltage too high; only a very small amount is needed.

The stock CPU-VDDA voltage should be 2.5V; try increasing it to 2.5V ~ 2.55V. DO NOT EXCEED 2.6V.

 

Tweaking the NB or SB voltage can also help, but those are the finer bits of tuning. I'd recommend that you deal with these voltages in the end.

 

Do me a favour, download IBT AVX and run it at maximum for 10 runs?

 

5ghz here with all power saving features enabled to reduce heat and power usage and pretty damn stable too.

 

SB voltage helps stabilize multiplier,

VDDA helps FSB overclocking

CPU/NB, depending on your speed, the sweet spot is just under 1.3v so 1.275v.

 

LLC to high and power to 130% will enable Cool N Quiet without hanging and all the other features can also be enabled.

[-rascal-]

Do me a favour, download IBT AVX and run it at maximum for 10 runs?

 

Will do.

I'll report back later today...hopefully.

[Benji_w]

Will do.

I'll report back later today...hopefully.

 

You didn't report back, I hope everything is ok?

[Theminecraftaddict555]

image.jpeg

My profile pic is frightened and will need some cat therapy

[Theminecraftaddict555]

You didn't report back, I hope everything is ok?

Probably busy with some stuff, life can get in the way sometimes

[Benji_w]

Probably busy with some stuff, life can get in the way sometimes

 

Fair point, need to build my gf's minecraft system tomorrow.

 

6600k

z170 Sabretooth 

m.2 ssd

3000mhz ddr4

 

It's not fair, her system alone beats the crap out of both of mine... 

[-rascal-]

You didn't report back, I hope everything is ok?

 

I had a good ol' Monday night hockey game I had to attend. I am running the test now.

 

EDIT:

 

Well I'll be damned (but not too surprised), my 5.0 GHz overclock isn't 100% stable.

The highest number of passes I can get is only either four or five before the system locks up, and I have to do a soft reset.

The CPU temperature actually climbs up pretty high;  in the picture that I took, the chip actually touches 81*C.

 

Interesting how day-to-day for the last few months of use doesn't cause the system to crash / lock up.

 

Spoiler

 

I am able to run AIDA64 stress test okay for about ~6 hours. I'm actually amazed IBT just puts it to its knees so quick.

Well, I might just have to settle with 4.8 GHz / 4.9 GHz for the time being   ...at least until I get a new CPU cooler.

...a little hesitant with water coolers now since I had two Swiftech H220's die on me within 24 hours of actual operation. This is actual a common problem with the older generation H220 (not the H220X).

 

http://forums.swiftech.com/viewtopic.php?f=2&t=2548&sid=705ffdb60d027aad17d11fed980385e4 (Talking to Bryan over at Swiftech)

 

http://linustechtips.com/main/topic/148175-new-swiftech-h220-just-failed/ (high res pictures of the ceramic bearing, apparently, turned to powder in the coolant on page 2)

[Benji_w]

I had a good ol' Monday night hockey game I had to attend. I am running the test now.

 

EDIT:

 

Well I'll be damned (but not too surprised), my 5.0 GHz overclock isn't 100% stable.

The highest number of passes I can get is only either four or five before the system locks up, and I have to do a soft reset.

The CPU temperature actually climbs up pretty high;  in the picture that I took, the chip actually touches 81*C.

 

Interesting how day-to-day for the last few months of use doesn't cause the system to crash / lock up.

 

Spoiler

 

I am able to run AIDA64 stress test okay for about ~6 hours. I'm actually amazed IBT just puts it to its knees so quick.

Well, I might just have to settle with 4.8 GHz / 4.9 GHz for the time being   ...at least until I get a new CPU cooler.

...a little hesitant with water coolers now since I had two Swiftech H220's die on me within 24 hours of actual operation. This is actual a common problem with the older generation H220 (not the H220X).

 

http://forums.swiftech.com/viewtopic.php?f=2&t=2548&sid=705ffdb60d027aad17d11fed980385e4 (Talking to Bryan over at Swiftech)

 

http://linustechtips.com/main/topic/148175-new-swiftech-h220-just-failed/ (high res pictures of the ceramic bearing, apparently, turned to powder in the coolant on page 2)

 

 

Thought so, Your volts are too low, and to make it worse standard IBT is easier to pass, you need this version And standard tests are piss easy to pass, Very high and Maximum should be your goal.

 

http://www.overclock.net/attachments/13202 IBT AVX... Have fun!   

 

[Guest]

I am so confused, doesn't the 9590 turbo to 5 ghz stock? 

 

Only recommendations are air cooling > water cooling because of the VRM's. 

 

Monitor VRM temperatures. 

 

To much voltage can cause instability just the same as under volting. 

[Benji_w]

I am so confused, doesn't the 9590 turbo to 5 ghz stock? 

 

Only recommendations are air cooling > water cooling because of the VRM's. 

 

Monitor VRM temperatures. 

 

To much voltage can cause instability just the same as under volting. 

 

Can stick a fan over the VRM heatsink while using an AIO, Don't monitor VRM temps, to do that you need to use AI suite and it's a bag of shit.