This evening I finally decided I had collected enough information and worked up enough nerve to try a little overclocking on my X2.
The first thing I tried was using Easytune 5. I am trying to think whether I had a rationale for this. I hadn't yet messed with the advanced settings in the BIOS, and I guess I hoped ET would wrap it all up and make it easy on me.
Baaad Easytune, baaad. Lesson learned: do not poke the clocks while Windows is running. I used ET to poke the processor up to 2410 mhz, and Windows went boom. Eventually got it stable again. The one lasting effect was that the OS now claims my hardware has changed and wants me to reactivate! When I try to reactivate it tells me I need to buy a new key (this is a licensed oem version of Windows bought for this system).
Sigh. Guess I will have to call them. Good thing I have three days. And it's Sunday. And I have to leave for Vegas 5:30 Monday morning. Thanks Bill.
After I got things cleaned up I used msconfig to set Windows to start in diagnostic mode. Figured the less stuff running the better. Then I dove into the BIOS.
The hardest part of the ramp-up was matching the terms everyone uses on the boards with the terms in the Award BIOS in my board. Someone here had recommended I start with 230 HTT, LDT 4x, 166 mult, and 1.425 vCore. When I went into the tweak section of the BIOS to change the HTT freq. I misread it, thinking that 230 was the actual frequency. Instead it was the overclock! When I rebooted the bios mentioned politely that a 430 mhz freq. was bloody lunatic, and froze. Rebooted again and hit delete and it had restored the old settings.
Figured it out the next time around and set up the first pass. At 1.425 vCore I was getting restarts and BSODs trying to boot Windows. Slowly increased the voltage to 1.475 vCore and it stabilized and was able to boot the operating system. At this point I had 230 HTT, 2530 mhz core speed, mult 166 for 180 mhz dimm speed, 3-3-3-8 (spd), and 2T.
Ran a few rounds of Prime95 and Super-pi, just long enough to make sure it wasn't going to blow up. Interesting that with vCore set to 1.475 in the BIOS, the CPUs appeared to be drawing 1.50 - 1.52 v under full load. Then went back into the BIOS to mess with the memory.
First I took it back to 200 (1:1), for a dimm speed of 230 mhz. 3-3-3-7 at 2T. Booted Windows and ran two copies of memtest through 100% coverage. Memtest saturated both cores to 100 %. Temperatures started at 36 c idle and ramped up to 54 c. The memtests covered 100% without error and I shut them down. Back into the BIOS.
Took the timing to 1T and memtest reported errors. Upped the vDimm by .2v and still got errors. So I backed off the mult to 166, 1T timing, 3-3-3-7, and ran memtest again without errors.
So that's where I stand at the end of the first session:
230 HTT
4x LDT
11 CPU mult
2530 Mhz core speed
166 memory mult.
180 Mhz dimm speed
3-3-3-7 1T
Crystalmark went from roughly 44k to 51k.
This is a moderate overclock, but it's already faster than a stock 4800+, and I don't know that I want to push it further. I have two remaining questions:
- Which is better? Running the DIMMs at 230 Mhz with 2T timing, or 180 Mhz with 1T timing? According to Sandra the memory does very well on the bandwidth bench, so I don't really need to push it, but I am curious.
- Should I worry about peaking at 54 c under full load? This is the stock AMD cooler.
I will reboot back into diagnostic mode and leave Prime running all night to make sure it is stable.
Thanks to Anandtech for providing such a great resource.
Update: I ran two copies of Prime95 with affinity set to unique cores. Ran the long FFT torture test for 40 minutes. No errors, but the CPU temp topped out at 62 deg. c. I was too chicken to leave it running that warm all night. So I shut it down. Probably look at a better cooler when I get back from my trip.
Update 2: The nice Microsoft people were able to activate my OS on the phone this afternoon. It was a fairly seamless process that involved my reading them 263,000 numbers, and then typing in a further 10^32 numbers they read to me. After which, the software was activated again.
The first thing I tried was using Easytune 5. I am trying to think whether I had a rationale for this. I hadn't yet messed with the advanced settings in the BIOS, and I guess I hoped ET would wrap it all up and make it easy on me.
Baaad Easytune, baaad. Lesson learned: do not poke the clocks while Windows is running. I used ET to poke the processor up to 2410 mhz, and Windows went boom. Eventually got it stable again. The one lasting effect was that the OS now claims my hardware has changed and wants me to reactivate! When I try to reactivate it tells me I need to buy a new key (this is a licensed oem version of Windows bought for this system).
Sigh. Guess I will have to call them. Good thing I have three days. And it's Sunday. And I have to leave for Vegas 5:30 Monday morning. Thanks Bill.
After I got things cleaned up I used msconfig to set Windows to start in diagnostic mode. Figured the less stuff running the better. Then I dove into the BIOS.
The hardest part of the ramp-up was matching the terms everyone uses on the boards with the terms in the Award BIOS in my board. Someone here had recommended I start with 230 HTT, LDT 4x, 166 mult, and 1.425 vCore. When I went into the tweak section of the BIOS to change the HTT freq. I misread it, thinking that 230 was the actual frequency. Instead it was the overclock! When I rebooted the bios mentioned politely that a 430 mhz freq. was bloody lunatic, and froze. Rebooted again and hit delete and it had restored the old settings.
Figured it out the next time around and set up the first pass. At 1.425 vCore I was getting restarts and BSODs trying to boot Windows. Slowly increased the voltage to 1.475 vCore and it stabilized and was able to boot the operating system. At this point I had 230 HTT, 2530 mhz core speed, mult 166 for 180 mhz dimm speed, 3-3-3-8 (spd), and 2T.
Ran a few rounds of Prime95 and Super-pi, just long enough to make sure it wasn't going to blow up. Interesting that with vCore set to 1.475 in the BIOS, the CPUs appeared to be drawing 1.50 - 1.52 v under full load. Then went back into the BIOS to mess with the memory.
First I took it back to 200 (1:1), for a dimm speed of 230 mhz. 3-3-3-7 at 2T. Booted Windows and ran two copies of memtest through 100% coverage. Memtest saturated both cores to 100 %. Temperatures started at 36 c idle and ramped up to 54 c. The memtests covered 100% without error and I shut them down. Back into the BIOS.
Took the timing to 1T and memtest reported errors. Upped the vDimm by .2v and still got errors. So I backed off the mult to 166, 1T timing, 3-3-3-7, and ran memtest again without errors.
So that's where I stand at the end of the first session:
230 HTT
4x LDT
11 CPU mult
2530 Mhz core speed
166 memory mult.
180 Mhz dimm speed
3-3-3-7 1T
Crystalmark went from roughly 44k to 51k.
This is a moderate overclock, but it's already faster than a stock 4800+, and I don't know that I want to push it further. I have two remaining questions:
- Which is better? Running the DIMMs at 230 Mhz with 2T timing, or 180 Mhz with 1T timing? According to Sandra the memory does very well on the bandwidth bench, so I don't really need to push it, but I am curious.
- Should I worry about peaking at 54 c under full load? This is the stock AMD cooler.
I will reboot back into diagnostic mode and leave Prime running all night to make sure it is stable.
Thanks to Anandtech for providing such a great resource.
Update: I ran two copies of Prime95 with affinity set to unique cores. Ran the long FFT torture test for 40 minutes. No errors, but the CPU temp topped out at 62 deg. c. I was too chicken to leave it running that warm all night. So I shut it down. Probably look at a better cooler when I get back from my trip.
Update 2: The nice Microsoft people were able to activate my OS on the phone this afternoon. It was a fairly seamless process that involved my reading them 263,000 numbers, and then typing in a further 10^32 numbers they read to me. After which, the software was activated again.
