Overclock.net banner
1 - 20 of 23 Posts

·
Registered
Joined
·
10 Posts
Discussion Starter · #1 ·
Hi there, i am trying to over clock my 4100 but seem to be hitting some trouble.

Specs

AMD FX 4100 (All Cores Unlocked in Bios)
CPU Cooler: Notchua NH D14
Motherboard: Asus M5A78L-M/USB3
RAM: 16 GB (4x4GB) Corsair Vengeance 1600Mhz 9-9-9-24
Case: NZXT Phantom Enthusiast Full Tower

Currently I am unstable using the following

CPU Freq: 205
Multiplier: x22
Core Speed: 4.5 Ghz
VCore: 1.45
LLC: Auto
(Everything Else Auto)

RAM Freq: 1600Mhz
Voltage: 1.5
Timings: 9-9-9-24 CR2
(Everything Else Auto)

Prime Shows a fatal error after a hour on blend and PC restarts but Ive read the CPU Vcore should not exceed 1.45V for 24hr use. Also under 100% load in prime the cpu does not exceed 42c.

Any help on getting a good overclock from this machine would be appreciated.
 

·
Premium Member
Joined
·
4,265 Posts
Quote:
Currently I am unstable using the following

CPU Freq: 205
Multiplier: x22
Core Speed: 4.5 Ghz
VCore: 1.45
LLC: Auto
(Everything Else Auto)
If APM was still enabled (assuming this based on everything else being left alone here), the chip would not maintain 4.5ghz under a parallel workload, it would have throttled down to attempt to stay close to TDP limits. Hate to say it, but this means that you are unstable, and may not have even been at 4.5ghz during the test.

My theory, is that after an hour, the VRMs have heat-soaked the motherboard to the point that the voltage starts dipping, which causes the instability. That motherboard is going to be a limiting factor to having an overclock that can handle continuous workloads like that. I realize it's "only" a 4 core, but zambezi is still a total power pig when overclocked.

You've got a 300W capable HSF strapped to a motherboard that will tap out at 140W or less. With the reduced air-flow on the VRMs compared to the "stock" AMD heatsinks, I expect you've probably reduced the motherboards continuous available stable power to ~100W or less.

Don't use the reference clock to overclock at this stage. That should be experimented with for fine tuning later on. When you adjust your reference clock (called "CPU freq"), you are also adjusting RAM/NB/HT bus speeds (many boards don't report the change in speed to these other buses, so you just have to "know" that it is being adjusted, and do the math yourself), any of which could be the source of an instability. OCing with the reference clock without compensating for these changes on those buses, will make it impossible to isolate the source of the instability.
 

·
Registered
Joined
·
10 Posts
Discussion Starter · #4 ·
Thanks for the replays. I've tried a different tact today so here goes. I can get to 4.5Ghz with the following:
CPU freq: 200
Mem clock: 1600
Pcie clock: 100

CPU ratio: 22.5
And turbo: disabled
CPU/NB freq: 2400

CPU over voltage: 1.440625
(Bios and CPU-z show 1.488v?)
Vddnb over voltage: 1.25000
Llc: 22.575%

Ht link speed: 2400
Ht over voltage: auto

Memory over voltage: 1.5000

PCI/PCIe CLK status: enabled

Gart error reporting: disabled
Microcode updation: enabled
Secure virtual machine: disabled
Cool n quiet: disabled
Advance clock calibration: auto
Unleashing mode: disabled
Active CPU cores: auto

APU disabled using amgmsgtweaker every time windows starts.

I ran prime on small ftt for 3 hours fine, ran blend for 3 hours fine.

Problem is if I up the multi from 22.5 to 23 so the cou goes from 4.5 to 4.6 then it fails in prime no matter how much extra voltage I give the CPU. Any ideas?
 

·
Registered
Joined
·
10 Posts
Discussion Starter · #10 ·
So I've ran prime for 4 hours fft at 200 x 22.5 with 1.44v vcore and it's stable. If I up the multiplier by .5 to 23 then it fails in prime within seconds even at 1.55v vcore. Any ideas?
 

·
Registered
Joined
·
716 Posts
Well, you could try upping the fsb as explained in the guide. Start with small increments and work your way up, paying attention to the north bridge and HTT speeds. You may have to loosen your memory timings too after say 210 fsb. Try 9-10-10-30 or 10-10-10-30 (or upping the ram voltage, but I wouldnt go more than 1.6volts on 1.5volt ram).

Are you using the onboard video ? I believe that would get overclocked too, so that has to be watched because that would cause system instability. You can change that so that it doesnt go far over stock settings in your GPU Overclocking options.

I would definately put a fan blowing on the VRMs. (better yet, custom vrm heatsinks plus a fan)

You do have a fantastic cpu cooler on a motherboard not really made for overclocking.
 

·
Registered
Joined
·
1,669 Posts
Get a better motherboard or modify the VRM section.

Talking from experience and plenty of knowledge, that motherboard has the weakest, non-cooled and highest throttling phase design ive known.

Google voltage regulation module and AM3+ phase lists.

You'll soon see why, and to others yes i do know that the 4100 is low wattage but i also know that motherboard throttles that cpu under full load even at stock.
 

·
Premium Member
Joined
·
4,265 Posts
Running torture tests on that motherboard is just asking for a burned out motherboard.

A voltage setting and the actual voltage that winds up being delivered can vary pretty wildly. The "setting" is just a goal, but the regulation system often fails to achieve the goal. Beefier VRM designs do a better job, but they all require experimentation with LLC levels to get tighter regulation for a given power demand.

Watch CPU voltage WHILE the CPU is under a load and see what happens. It is VERY unlikely that the motherboard is HOLDing 1.55V under a load.
 

·
Registered
Joined
·
716 Posts
The cpu temps look fine, as expected with that cpu cooler. In your case, its the vrm temps you have to look out for. While running prime95 for a while, put your finger on a vrm and see how long you can hold it there. Of course, if you have a temp probe or infrared thermometer, you could just read it.
 

·
Premium Member
Joined
·
4,265 Posts
Yea, but those mosfets could be 100C runnning hard. I'm not so worried about the finger (a very mild burn will undoubtedly heal, and certainly won't be the only hot thing you've ever touched (on purpose or by accident), but the damage someone might do upon the "surprise" of feeling the burn sensation; A quick snap of a wrist (reflex) to escape the danger next to a huge HSF in a computer case with potentially sharp edges could break something or cause a larger injury.

I use the "touch" method too, just be aware that things in there can be hotter than you might expect.
 

·
Registered
Joined
·
10 Posts
Discussion Starter · #19 ·
Just ran IntelBurnTest and about 5 mins in i touched the mosfets and could keep my finger there for as long as i wanted too.

Might help that my rear exhaust is a noctua 120mm fan and I have a 200mm exhaust on the top of the case right bove the d14 heatsink, plus the middle fan on the d14 under hangs the heatsink slightly pushing air over the mosfets.

I think with a good motherboard my cpu would hit 5ghz with ease, i just dont think the motherboard i have is stable enough for a 4.6 overclock.

I mean the cpu never reaches 50c the mobo temp is 27c all the time and the cooling of the case is fantastic.
 
1 - 20 of 23 Posts
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top