Great to see there is still a fanbase around the AM3+ socket and FX-9000 CPUs!
I'm currently using two FX-9590, a FX-9370 and several FX-8000/6000 CPUs. Firstly, I would like to contribute my findings and experiences while digging further into OC capabilities and difficulties.9370 vs 9590:
.) The 9370 consumes a litte bit more Watts than the 9590 and gets about 5°C warmer at stock speed, while using the same motherboard and cooler (ASUS TUF Sabertooth 990FX R2.0 and Noctua NH-D14).
.) The 9370 CPU-NB needs more Volts to reach 2400 and is unstable at 2600 MHz, while my 9590 is basically capable of that.Higher than DDR3-1866 support? No!:
Some shops and sites claim the FX-9000 chips would officially support 2133 or even 2400 MHz, because it's based on the Centurion core. But it actually is still the same old Vishera core with the same stepping!
Evidence: Visit AMD, scroll down and look at "Features" http://www.amd.com/en-gb/products/processors/desktop/fx#Inofficial DDR3-2133 support:
Never had problems, as the CPU-NB speed is 2200 MHz.Inofficial DDR3-2400 support:
Very problematic, as you have to raise CPU-NB speed to 2400 MHz. The RAM speed itselt surely isn't the problem (at least with my AMD labeled RAMs), but I've never found a reliable scenario regarding the CPU-NB.
ASUS automatically raises CPU-NB Voltage to 1.4V, which is a bit higher than you should go in any 24/7 case. But the system boots and seems to be stable at first look. Even Prime95 seems to do fine. But whole multi-jumps above 2200 MHz lead to very strange phenomenons: IE11 doesn't show webpages correctly after a while and randomly freezes, Win7 Aero began to deactivate itself and couldn't be reactivated until a reboot and games began to show flashing textures, black screens or crash. System crashes or BSODs appeared at very long runtimes but were seldom, as long as I didn't go below 1.35V.
So I don't think this is an option in any case. Unlike the Phenom-IIs, the non-existing speed advantage does not justify the worries you get and I wouldn't trust a machine in that condition with my data!220W myths and facts:
The 220W TDP and actual consumption depends, of course, completely on its usage. Using it for simple tasks won't be much of a penalty, compared to FX-8000 CPUs.
I've experimented with compact water cooling kits, state-of-the-art at that time, but they were up to 15°(!) warmer than my NH-D14 at half-load and no chance to stay below 80° at full load, besindes the acoustic drawback of the pump. Exception is (should be) an open water cooling solution of course, but that never was an option for me. So it is no problem to cool it, even with air cooling!
That's why people should keep in mind that OC'ing a FX-8000 to the levels of a FX-9000 may and pretty sure is more power consuming and often not even possible, as the FX-9000 are some extraordinary good samples of the same chip. So, FX-8000 OC'ing may draw more power than using a stock FX-9000!
On the other side, by OC'ing the FX-9000, disabling CnQ, disabling APM (AMD Power Management), or activating HPC (High Performance Computing), you may need way more than 220W! The 220W TDP is a maximum desired value that ends in slowing down CPU speed and V. if reached. E.g.: At stock speed 2 threads of Prime95 prevents Turbo Core to work and 8 threads slows the clock to 4 GHz or even below. After messing with some BIOS options, it is possible to stay at full clock speed and use a lot more power that way.9590 V. & LLC settings @ 5 GHz:
I too never had success in reaching 5 GHz for the base clock. 4,75-4,80 GHz is realistic. Anything higher is not stable at full load, even if I gave it 1.55V or more.Some considerations about that
1.) The same boards that were 125/140W capable, suddenly are now 220W capable, without visible changes. The headroom (8+ Phases) from the past has obviously been used up to support a 220W CPU out of the box.
2.) By OC'ing that CPU, you reach power needs, the board can hardly handle.
3.) Because of that, it seems most likely, the CPU is not in need of more V., but the board is the limiting factor.
4.) Has anyone reliable information of the actual V. a core gets if running in Turbo Mode? That V. has to be sufficient and I suspect it not to be 1.55V, but something between 1.475 and 1.525 Volts. Find that V. and you have stable 5 GHz at reasonable V! But just theoretically, because it is not designed to have that kind of V. on every core at full load, which leads to enormous power needs.
5.) Regarding LLC, most boards have their CPU LLC at AUTO. Some even offer CPU-NB LLC. At CPU stock speed, the system keeps absolutely stable if LLC is set to Standard, Regular or Normal. Therefore it is just a method for reaching redicolously high V. settings (>1.55V) and shouldn't be needed at 5 GHz, right?
6.) Setting LLC it to High, Higher or Extreme gave me far to high temperatures at load (base V. at default), leading to shutdown after a short time of Prime and did nothing to stabilize clocks above 4.8 GHz.
7.) There is no difference between multi OC and HT OC.
8.) I've tried CPU, CPU-NB and DRAM overcurrent to help out. Even CPU @ 130% didn't do any good. It this is the key to a successful OC, as many reviewers hype it, I don't see it.
I hope this isn't too much information at once
and maybe it helps to clear someones questions or helps me understanding it better, if I am wrong in some cases.
What do you think of setting HTT clock equal to CPU-NB clock? Is it better equally or leave HTT at default of 2600 MHz for no (dis)advantage?