phenom II 545 overclock/unlock
I have been itching to play with Phenoms but was not able to justify the expense ... until now. I could not resist the ASUS M4A77D + Phenom II X2 545 combo for $145 (plus $10 MIR). This will be replacing a C2D-4500 (3GHz) system and I will recycle 4GB DDR2-800 RAM, thus AM2+ MB. I think it will be faster than the C2D4500 but if I can unlock to X4, it will be the jackpot. The cooler will be an old TT Big Typhoon for now. Let's see the initial overclock and unlock. If this puppy proves worthy, it might wear a new shiny Cogage True Sprit. I will post the results here, let me know if you have any suggestions.
0
Comments
Overclock is at 3.75GHz right now. I think there is hope for 4GHz since there is no heating problem and vcore is at 1.375v. I think I can go upto 1.45v.
System 1: Intel E4500@3.0GHz, 273 FSB, 4GB DDR2-546, Nvidia 9600GT
System 2: Intel E5200@3.5GHz, 280 FSB, 4GB DDR2-560, ATI HD4850
System 3: AMD PII-545@3.75GHz, 250 FSB, 4GB DDR2-833, Nvidia 9600GT
Speedfan and CPU-Z show too much fluctuation of vcore on this ASUS M4A77D motherboard; vcore is going up and down between 1.325 and 1.4 under load during stability testing. I checked the Load Line Calibration option and enabling it did not improve anything. Actually I had to disable for overclock stability. I am not sure if the vcore fluctuation is due to the motherboard or PSU. Any suggestions?
Noticed that on most if not all of my systems (all AMD) some just do it more than others. Have an Abit NF-M2 nView that I love but under load it is almost like it'll jump up to the next value that is selected in bios. Know that isn't an answer, but an observation.
Anyway, 545 is running at 3.75GHz completely stable and cool right now. And 4 GHz seems like too much strain on the CPU and motherboard. Load temperatures running IntelBurntest are unbelievably low with a TT Big Typhoon; max 40C !!! I checked the core temperature with Coretemp and Speedfan and I still think this is too low to be true.
Good idea! I will try HWMonitor too.
Edit:
I tried now and the temperatures are the same with HWMonitor. Motherboard has a CPU temperature reading too besides the K10 CPU diode. The temperature reading of the motherboard is following K10 CPU diode with 1-2C offset. I am not sure if the motherboard reading is based on the CPU diode or an independent diode on the socket. I am feeling uneasy overclocking a CPU while reading <40C core temperatures under load.
I did the followings all at once since I did not have much time to try step by step.
- upgraded the BIOS to the latest one
- replaced the crappy Antec SP500 PSU with Antec EA650
- reseated the heatsink. I had to do this anyway to remove the PSU
- reduced overclocking to 3.5 GHz
Now the CPU is unlocking all cores and running stable at 3.5 GHz. If i had a better cooler I think it would overclock to 3.75. Maybe later, I have to leave it at 3.5 for now.
By the way, unlocking disables core temperature sensor and I don't have the previous problem anymore.
Thanks Jedi. I have increased FSB to 240, decreased the HT and RAM multipliers one step down without changing vcore. I am now at 3.6 GHz stable after 12 cycles of Burntest. RAM speed is at DDR2-800 and HT at 1920 MHz. It looks like this is the perfect spot to stop and get some sleep now.
Yeah don't push it if you're getting tired as I ruined meaning an oc from being tired and going
Think you've done a great job and now time to "enjoy" it. :thumbup
Jedi, thank you very much for your company.
That's more than mutual and look forward to your input when I get off my duff and do some more overclocking!
Why, do you need more than 40 GB/s of bidirectional bandwidth? (I think maxing PCIE 2.0 bandwidth is 16 GB/s, and thats not even saturated)
Once upon a time tweaking your bus got you improved IO performance, these days, its minimal, or any at all, all your doing is heating up your motherboard. Thats why unlocked chips are so prevalent in the AMD camp. AMD got rid of the ugly FSB limitations years ago with Hyper-Transport. Your IO bandwidth is plenty at 1.8 or 2.0, whatever the default is, your system is going to bottleneck it other places, USB 2.0, your spinning drive, even a good SSD will likely be the limit, your optical drive obviously, your RAM, everything. Hyper-transport 3.0 in any flavor is never your bottleneck.
Enjoy your nice overclock
Yes, NB can be run faster than HT.
These 2 articles use DDR3 CPU's however the principles apply to DDR2 as well.
http://www.ocztechnologyforum.com/forum/showthread.php?63271-AMD-Phenom-II-mild-overclock-wild-performance
http://www.ocztechnologyforum.com/forum/showthread.php?56715-AMD-AM3-CPU-s-which-ram-speed-is-faster-and-which-CPU-NB-clock-is-best
Especially Tony's article about ram timings/speed compared to NB speed.
Keep in mind just about all OEM NB chip coolers, are Meh at best. I mean, the one on your Asus board looks like it might be okay as passive NB heatsinks go, but once you start pushing the NB you might want to consider some after market cooling for your NB chip. I recently had a board fail, I think due to a NB that was not cooled sufficently, and keep in mind, I was not even over clocking it. If not a new sink, get some air flow on that thing and monitor your temps.
But doesn't the rise in frequency still effect the NB chip heat output? I know I see a voltage adjustment for that in my BIOS.
Yes, some boards have 2 NB voltages, one for the CPU and one for the chipset. No need to adjust the Chipset NB voltage, we are not changing its speed.
You want to change the CPU-NB voltage, may be called CPU-NB VID, not the NB Voltage or Chipset voltage.
Again, you are not raising the frequency of the chipset.
See, I did not know that, so there are two independent voltage regulators on the CPU?