Pyrofeed
New Member
- Joined
- Jun 10, 2009
- Messages
- 78 (0.01/day)
| System Name | Sylar |
|---|---|
| Processor | Phenom II x4 955 BE |
| Motherboard | ASUS M4A78-E |
| Cooling | Cooler Master 120mm x2 |
| Memory | 4gb GSKILL |
| Video Card(s) | XFX 4850 |
| Storage | 500gb Samsung |
| Case | Antec 300 |
| Power Supply | OCZ ModxStream Pro 600W |
| Software | Windows 7 RC |
I had another undervolting thread a looong time ago and that was more me asking for advice. In this one I am just going to be showing my results and hopefully getting some opinions on what I did. I am not a big time overclocker or undervolter or any sort of computer playmate. What happened was I got bored again, and after putting together a thuban system for my dad I got jealous of his temps, so I decided I would once again venture forth and try to undervolt some more.
Specs:
Asus m4a78e
Phenom II X4 955 BE (C2)
Process:
The stock vcore on this cpu is a whopping 1.35. My idle temps at stock would be about 40-42 idle and 60 at load. I managed to drop my vcore down to 1.2625 stable, I haven't tried further but I have a feeling I will either only get to 1.25, or bsod at that stage. Now at this point I was stuck and I came here to ask if there was a way to drop it more, some way to make it more stable. I asked about ACC and a few other things and the consensus was basically the only way to go further would be to underclock it. My goal however, was to keep the stock clock without the stock voltages and temps. So I wondered what an overclocker would do if he needed to increase stability but he already played with his cpu to the max. He would mess with other voltages! So then (this is part two of my journey) I went and dropped the cpu/nb voltage from 1.1 to 1.025. I'm stable!
Temps:
Like i said, I idled at 40-42 and loaded at around 65. Now I idle at 34 (these are core temps) and load at 52. So overall my idle temp had a 8 deg drop, and the load had a 13 deg drop.
Notes:
I don't know how common it is for stress testing to be unreliable for overclockers, so I figured I would put this in here just in case. It seems when undervolting, the stress tests don't rule out instability, even after a LOOONG time of running a sequence of prime, occt, S&M, and wprime. For example, I managed to drop my cpu/nb down to 1 volt flat. Completely "stable". Then I noticed that due to a recent cmos clear my clock was wrong, so I pressed the windows 7 option to sync it with internet time. It synced, froze, and proceeded to give me what I dubbed the FZPSOD (or Flashing Zebra Print Screen Of Death), and shut off. The FZPSOD happened a few times all in similar circumstances, so I would just recommend that during your stress testing and after, do basic tasks such as web browsing and listening to music, because sometimes those are the things that destabilize the zebra hiding in the back of you monitor. I also realized the beauty of using software to fiddle around at first, because if it shuts off, you just boot back up and you don't have to worry about any saved settings!
Conclusion:
I may try to see if I can eek any sort of OC, however small, out of any headroom I have, just for fun, but I would not be upping the voltages. I may also try to drop my vcore lower, and try to find other voltages to mess with that would help temps and power draw. Basically, if you are not looking to go any faster than what you purchased, it is almost a certainty that your components are using more volts than they need to, so you can get a pretty reliable drop in temps by shaving these off. I am guessing that in theory better motherboards will be able to more readily due this, and hopefully I will make enough money this summer to get a crosshair 4. Any and all comments and criticisms are appreciated. Just to note, I due not expect any performance gains out of what I have done, I am not a serious benchmarker and like the performance of my components at stock, hence why instead of making them perform better, I have opted to make them more efficient. Also note that drastic drops in voltage and temp are possible by lowering the clock, something I experimented with only a tiny bit, however I did not want the coolest and most efficient 955, I wanted the coolest and most efficient 955 3.2 ghz.
Thank you for gracing these words with your experienced eyes, have a nice day!
Specs:
Asus m4a78e
Phenom II X4 955 BE (C2)
Process:
The stock vcore on this cpu is a whopping 1.35. My idle temps at stock would be about 40-42 idle and 60 at load. I managed to drop my vcore down to 1.2625 stable, I haven't tried further but I have a feeling I will either only get to 1.25, or bsod at that stage. Now at this point I was stuck and I came here to ask if there was a way to drop it more, some way to make it more stable. I asked about ACC and a few other things and the consensus was basically the only way to go further would be to underclock it. My goal however, was to keep the stock clock without the stock voltages and temps. So I wondered what an overclocker would do if he needed to increase stability but he already played with his cpu to the max. He would mess with other voltages! So then (this is part two of my journey) I went and dropped the cpu/nb voltage from 1.1 to 1.025. I'm stable!
Temps:
Like i said, I idled at 40-42 and loaded at around 65. Now I idle at 34 (these are core temps) and load at 52. So overall my idle temp had a 8 deg drop, and the load had a 13 deg drop.
Notes:
I don't know how common it is for stress testing to be unreliable for overclockers, so I figured I would put this in here just in case. It seems when undervolting, the stress tests don't rule out instability, even after a LOOONG time of running a sequence of prime, occt, S&M, and wprime. For example, I managed to drop my cpu/nb down to 1 volt flat. Completely "stable". Then I noticed that due to a recent cmos clear my clock was wrong, so I pressed the windows 7 option to sync it with internet time. It synced, froze, and proceeded to give me what I dubbed the FZPSOD (or Flashing Zebra Print Screen Of Death), and shut off. The FZPSOD happened a few times all in similar circumstances, so I would just recommend that during your stress testing and after, do basic tasks such as web browsing and listening to music, because sometimes those are the things that destabilize the zebra hiding in the back of you monitor. I also realized the beauty of using software to fiddle around at first, because if it shuts off, you just boot back up and you don't have to worry about any saved settings!
Conclusion:
I may try to see if I can eek any sort of OC, however small, out of any headroom I have, just for fun, but I would not be upping the voltages. I may also try to drop my vcore lower, and try to find other voltages to mess with that would help temps and power draw. Basically, if you are not looking to go any faster than what you purchased, it is almost a certainty that your components are using more volts than they need to, so you can get a pretty reliable drop in temps by shaving these off. I am guessing that in theory better motherboards will be able to more readily due this, and hopefully I will make enough money this summer to get a crosshair 4. Any and all comments and criticisms are appreciated. Just to note, I due not expect any performance gains out of what I have done, I am not a serious benchmarker and like the performance of my components at stock, hence why instead of making them perform better, I have opted to make them more efficient. Also note that drastic drops in voltage and temp are possible by lowering the clock, something I experimented with only a tiny bit, however I did not want the coolest and most efficient 955, I wanted the coolest and most efficient 955 3.2 ghz.
Thank you for gracing these words with your experienced eyes, have a nice day!