AsRock Fatality 990FX (CPU Throttling)

Hey everyone,

Edited by Atomic - 15 Jan 2016 at 1:15pm

Hey Wardog,

Atomic wrote: Hey Wardog, If I am honest, that is a pretty big oversight by AsRock. To design a motherboard titled as gaming and for this particular product not being able to handle a pre-set over clock included in the motherboard as one of it's features. Just because the heat sink which covers the VRM's (The one with the Killer logo), is so poorly designed it is critically overheating and causing the system to loose stability.

Those preset overclock settings are not really guaranteed to work, since each CPU and combination of components is basically unique. The system those settings were determined on may he able to handle them, but your particular CPU may, for example, need more voltage to stay stable. The same goes for cooling; if your cooling is not arranged in the exact same manner that was used for obtaining those settings, you may need to put in some extra effort.

Am I wrong in saying that Asus or Gigabyte DO NOT have these same problems?

Yes. Some of their boards may have stronger VRM sections, but water cooling combined with an overclock pretty much always necessitates the use of a spot fan - depending on the CPU and OC, of course.

3) Have I made a mistake by going with AMD instead of INTEL for a gaming system? If not, then why? :-(

Not really. Generally speaking your GPU is the most important component in your rig when it comes to gaming. A stock FX-6300 should be plenty for most games and GPUs, although some games (like Crysis 3) could benefit from an FX-8xxx CPU or a more powerful Intel CPU, and if you had an SLI setup you might actually see the CPU bottleneck the GPUs at 1080p.

Intel's CPUs do have an advantage in performance (and power usage) at the moment, but that doesn't directly make AMD a bad choice for gaming. They provide decent bang for buck, especially when ovwrclocked. We just need to find you the best settings for your system.

Atomic, while you figure out how to keep your VRM cool, you could maybe try to find out what you can get out of your CPU at stock voltages. Increasing the clock speed alone has a smaller effect on power draw than increasing voltage, which should make life easier for the VRM. Just take your time with it and you should see warning signs before your rig is in any real danger of getting damaged:

Note: You may want to check you have Turbo Core, Cool'n'Quiet etc. disabled. Since we are just finding the maximum stock voltage clock speed, I personally don't think having those enabled matters much. It's when you go for maximum OC that you need maximum control over your system parameters, but for now all you really need is CPU-Z, HWMonitor and Prime95.

1. Start with stock settings and run Prime95 blend on all cores for about 20-30 minutes with HWMonitor running as well. This will give you baseline temperatures and show you how much Vcore and the clock speeds fluctuate (if at all) when loaded.
2. Increase the CPU clock speed multiplier by 0.5 (or the clock speed by 100 MHz), check with CPU-Z that the setting has stuck and run Prime95 again on all cores for 20-30 minutes while monitoring everything on HWMonitor. If temperatures stay below 60 C (CPU/package) and 70 C (socket/CPUTIN most likely) and no workers are dropped, repeat this step.
3a. At some point one or more workers will stop before 20-30 minutes has passed. This means the system is unstable and will (usually) need more Vcore to function at this clock speed. Lower the clock speed by 100 Mhz and run Prime95 blend on all cores for two hours. If workers are dropped, you need to lower the clock speed again. If no workers are dropped, you have reached what some people consider reasonable stability. Personally I would suggest running Prime95 for at least 4 hours, preferably 8.
3b. If the temperatures reach 60 C/70 C, you've pretty much reached your limits with the current cooling. If Prime95 is still running and the temperatures are running, stop the test immediately. I'm not expecting heat to be an issue for the CPU itself, but your socket temps may be high due that water cooling system. Also, if you notice clock speeds dropping or cores not being under 100% load while running Prime95, that's likely your VRM throttling and you need to stop the test and back off with the OC. Note that there will be short periods of time between tests that the CPU may not be under 100% load, so before you panic, make sure that the core(s) in question are actually in the middle of a test.

All of this at your own risk, naturally. ;) It's hard to say what your maximum OC at stock Vcore (1.35 V, I think) will be. The 3.8 GHz p state has a Vcore of 1.4125 V seems high considering the increase in clock speed is under 10%. I wouldn't be surprised if you could hit that at stock voltage, especially if the water cooler is keeping the CPU cool.