n00b Overclock diary: Phenom X4 955BE/Asus M4A78T-E
I've decided to make a serious first attempt at CPU overclocking, having just completed my first build. I remember reading primesuspect's journal of his early experiences with overclocking and thought it might be useful to start my own. Hopefully it won't just benefit me, but also others using similar components.
Speaking of which, my setup is:
CPU: Phenom II X4 955 (black edition)
Motherboard: Asus M4A78T-E; BIOS 2105
PSU: Coolermaster 700W Silent Pro M
Memory: G-Skill 4GB DDR3 PC12800
Graphics: XFX ATI Radeon HD 5770
SSD: Intel X25-M 80GB (for OS and Apps)
HDD: Western Digital 500GB Caviar (storage)
Chassis: Xclio case Windtunnel extreme
With everything at stock, I've been very happy with the machine's performance. Windows 7 runs very snappily and with the SSD, will load from the boot screen in ~15sec. Crysis and BFBC2 both run very well at high or very high video settings. Performance is also very good when running Ableton Live and Reason. There's always room for improvement though, and I'd like to find a stable overclock - at stock voltages for now - that I can use intermittently, when running highly demanding applications. I intent to keep the system at stock settings for day-to-day tasks.
Re. components, I based my choices mainly on specifications, cost and availability and in retrospect should perhaps have spent a bit more time researching, as I gather there are better AM3 motherboards out there. Still, can't do much about that now, so there's no point crying over spilt milk. Besides, the board performed pretty well in my initial set of tests.
I spent a lot of time reading and re-reading a number of MB reviews and AM3 OCing how-tos this morning after a dissapointing first foray into OCing last night. Using AMD Over Drive (AOD), I was not able to get past 245Mhz HT ref. clock, with all the multipliers at their lowest settings. A little extra research meant I had a much better idea of what I was doing and got much better results this afternoon.
The most helpful OCing guide I found was probably this one at tomshardware. This MB review had some useful information about the motherboard HT ref. clock performance (specifically, that the motherboard has an "HT/FSB hole" in about the range where I'd had problems last night).
At time of writing,the only testing I've done has been to establish the max HT ref. frequency of the MB. I will post updates in this thread as I continue, and list updates at the top of this post.
On to the fun part:
1. Prep
Step one was a fresh install of Windows7. I wanted to keep everything as neat as possible so I followed this procedure for installing drivers, OCing and monitoring utilities, benchmarks, games and browser:
1) install MB chipset & gfx drivers
2) reboot, defrag' C:
3) install AOD, CPU-Z, CoreTemp, OCCT, 3DMark05, CINEBENCH
4) repeat (2).
5) install Crysis, Firefox
6) repeat (2)
In the BIOS, I ensured that the onboard gfx, Cool ‘n’ Quiet, C1E and Spread Spectrum were disabled. I set the overclocking features to manual and set about finding the boards maximum HT ref. clock.
2.Establishing max. HT ref. frq
I started by setting the CPU, NB and HT multipliers to their lowest settings (8, 4 and 1, respectively) in the BIOS. I set the HT ref : DRAM ratio to 1:2 and booted to windows. From there, I loaded AOD and started increasing the HTref in 15MHz up to increments 245MHz after which I dropped to 10MHz and later 5MHz increments. After each test i ran the AOD benchmark utility. I wanted to know the max HTref that the motherboard could achieve and still complete the benchmark. From there I would start ramping down the frequency and doing more extensive stability testing with AOD.
I originally set out to use AOD as much as possible at stock voltages, but when testing the HTref, I found that going from 245 to 255MHz AOD would hang when appying the settings. I thought this might be due to the reported FSB hole of this MB, so entered the BIOS and tried manually setting the HT ref to 260Mhz. Following a reboot windows loaded with no problems, and the processor was able to complete 5mins of AOD stress testing at full load. From this point on, every increase in HT ref frq was followed by a 5 min stress test.
Thinking I was out of the hole, I tried to up the HT ref. to 265MHz in AOD, but the same thing happened - the system hung and I had to reboot. The problem was fixed by setting the HT ref. to 265Mhz in the BIOS rather than AOD, so the problem may not have been a blackspot in MB peformance but with changing the ref clock in AOD. In any case, I continued using the BIOS to change the clock speed from here on in, and using AOD only for stability testing.
As I worked I kept a spreadsheet open on a laptop next to me and recorded my results, which are below. Bearing in mind that I did not carry out stability testing past 1hr today, my max "stable" HT ref. clock was 291MHz. I'll to leave it on an 8 hr run later, for a more rigourous test. Annoyingly I forgot to get CPU-Z screengrabs, but I will do tomorrow. Once I'm fairly sure what the max stable HT ref. clock is, I'll set about finding the max CPU multiplyer.
Table Legend:
POST? did the system post? ; 0 - no; 1 - yes
OS? did windows load? 0 - no; 1 - no, failed at load screen; 2 - no, failed at login; 3 - no failed at desktop; 4 - yes
CPU-Z? did CPU-Z load? 0 - no; 1 - yes
stress5m, stress15m, stress1hr? completed 5m, 15m or 1hr of AOD stability testing? 0 - no; 1 - yes
fullstop/period: not tested
Speaking of which, my setup is:
CPU: Phenom II X4 955 (black edition)
Motherboard: Asus M4A78T-E; BIOS 2105
PSU: Coolermaster 700W Silent Pro M
Memory: G-Skill 4GB DDR3 PC12800
Graphics: XFX ATI Radeon HD 5770
SSD: Intel X25-M 80GB (for OS and Apps)
HDD: Western Digital 500GB Caviar (storage)
Chassis: Xclio case Windtunnel extreme
With everything at stock, I've been very happy with the machine's performance. Windows 7 runs very snappily and with the SSD, will load from the boot screen in ~15sec. Crysis and BFBC2 both run very well at high or very high video settings. Performance is also very good when running Ableton Live and Reason. There's always room for improvement though, and I'd like to find a stable overclock - at stock voltages for now - that I can use intermittently, when running highly demanding applications. I intent to keep the system at stock settings for day-to-day tasks.
Re. components, I based my choices mainly on specifications, cost and availability and in retrospect should perhaps have spent a bit more time researching, as I gather there are better AM3 motherboards out there. Still, can't do much about that now, so there's no point crying over spilt milk. Besides, the board performed pretty well in my initial set of tests.
I spent a lot of time reading and re-reading a number of MB reviews and AM3 OCing how-tos this morning after a dissapointing first foray into OCing last night. Using AMD Over Drive (AOD), I was not able to get past 245Mhz HT ref. clock, with all the multipliers at their lowest settings. A little extra research meant I had a much better idea of what I was doing and got much better results this afternoon.
The most helpful OCing guide I found was probably this one at tomshardware. This MB review had some useful information about the motherboard HT ref. clock performance (specifically, that the motherboard has an "HT/FSB hole" in about the range where I'd had problems last night).
At time of writing,the only testing I've done has been to establish the max HT ref. frequency of the MB. I will post updates in this thread as I continue, and list updates at the top of this post.
On to the fun part:
1. Prep
Step one was a fresh install of Windows7. I wanted to keep everything as neat as possible so I followed this procedure for installing drivers, OCing and monitoring utilities, benchmarks, games and browser:
1) install MB chipset & gfx drivers
2) reboot, defrag' C:
3) install AOD, CPU-Z, CoreTemp, OCCT, 3DMark05, CINEBENCH
4) repeat (2).
5) install Crysis, Firefox
6) repeat (2)
In the BIOS, I ensured that the onboard gfx, Cool ‘n’ Quiet, C1E and Spread Spectrum were disabled. I set the overclocking features to manual and set about finding the boards maximum HT ref. clock.
2.Establishing max. HT ref. frq
I started by setting the CPU, NB and HT multipliers to their lowest settings (8, 4 and 1, respectively) in the BIOS. I set the HT ref : DRAM ratio to 1:2 and booted to windows. From there, I loaded AOD and started increasing the HTref in 15MHz up to increments 245MHz after which I dropped to 10MHz and later 5MHz increments. After each test i ran the AOD benchmark utility. I wanted to know the max HTref that the motherboard could achieve and still complete the benchmark. From there I would start ramping down the frequency and doing more extensive stability testing with AOD.
I originally set out to use AOD as much as possible at stock voltages, but when testing the HTref, I found that going from 245 to 255MHz AOD would hang when appying the settings. I thought this might be due to the reported FSB hole of this MB, so entered the BIOS and tried manually setting the HT ref to 260Mhz. Following a reboot windows loaded with no problems, and the processor was able to complete 5mins of AOD stress testing at full load. From this point on, every increase in HT ref frq was followed by a 5 min stress test.
Thinking I was out of the hole, I tried to up the HT ref. to 265MHz in AOD, but the same thing happened - the system hung and I had to reboot. The problem was fixed by setting the HT ref. to 265Mhz in the BIOS rather than AOD, so the problem may not have been a blackspot in MB peformance but with changing the ref clock in AOD. In any case, I continued using the BIOS to change the clock speed from here on in, and using AOD only for stability testing.
As I worked I kept a spreadsheet open on a laptop next to me and recorded my results, which are below. Bearing in mind that I did not carry out stability testing past 1hr today, my max "stable" HT ref. clock was 291MHz. I'll to leave it on an 8 hr run later, for a more rigourous test. Annoyingly I forgot to get CPU-Z screengrabs, but I will do tomorrow. Once I'm fairly sure what the max stable HT ref. clock is, I'll set about finding the max CPU multiplyer.
Table Legend:
POST? did the system post? ; 0 - no; 1 - yes
OS? did windows load? 0 - no; 1 - no, failed at load screen; 2 - no, failed at login; 3 - no failed at desktop; 4 - yes
CPU-Z? did CPU-Z load? 0 - no; 1 - yes
stress5m, stress15m, stress1hr? completed 5m, 15m or 1hr of AOD stability testing? 0 - no; 1 - yes
fullstop/period: not tested
0
Comments
I need sleep, so I'll post my methods either tomorrow or tuesday, but in the meantime here are some screens:
Max HTref at stock voltage, prime-95 stable for 15min (not quite 291Mhz: the AOD stability test obviously isn't as demanding as p95):
Max AOD-stable CPU multi:
3.8Ghz (19 x 200) @ stock vcore:
Almost 4Ghz (19 x 210) @ 1.45v:
My max prime95-"stable" overclock - 4.1Ghz (19 x 215) @ 1.53v. To be clear, I don't intend on clocking the chip this high, even if it does stand up to more rigourous testing:
BTW I may have missed it (not unusual) but what are you using for cooling?
Your HT values in CPU-Z look very wrong.. Your link values are 200? I am using x18 bus speed 200.7 and HT link 2006.8
I have been trying to figure out how to get HT Link to the 3000 the board claims is possible...
Also have you found a way to get the mem to use SPD Ext XMP-1600? No matter what i have tried it is still stuck at 667 MHz...
I have also found AOD BSOD's my setup.. no idea why but it started out of nowhere...
Also how is your core voltage at 1.3v? I havent touched anything except multiplier and mine is at 1.5V right now