ASRock & Opteron: Voltage, Temp, and OC'ing

GargGarg Purveyor of Lincoln Nightmares Icrontian
edited March 2007 in Hardware
I've just finished getting my new rig set up and running. It's an Opteron 165 on an ASRock 939Dual-SATA2.

My question is about how to interpret these voltage and temperature readings:

attachment.php?attachmentid=22599&stc=1&d=1171840303

First off: temperature. We can see that Core Temp is reporting different temperatures than Motherboard Monitor (MBM5). My question is which program, if either, I can trust. Core Temp changes the temps so fast that I have a hard time believing that a processor can fluctuate 2*C so quickly (the temp fluctuates at least 1*C on at least one core every second). And I'm not sure what each sensor MBM5 displays is actually reading on the board. Are temp2 and temp3 supposed to be each core, or is one on-chip and is another elsewhere? I don't think temp2 and temp3 are each core, because before I started Folding and the temperatures leveled out, they were reading farther apart. I could be wrong, though.

Also of note - Core Temp has crashed my system twice, so I'd like not to use it if I didn't have to. Any suggestions for other programs for reading temperatures? MBM5 has nice data logging abilities, so I'd like to use it or something like it, if it's accurate enough.

Next up: voltages. MBM5 is the only program I have that supposedly reports voltages for each core. Is that what it's actually doing, or is it making something up for the second core? If it's not, I don't like my second core being so far off the first. Might make overclocking a bit flaky.

Lastly: Assuming that either Core Temp or MBM's temperature readings are close enough (probably are), do you think I have any overclocking headroom? Right now, I'm using the AMD retail HSF with the Opteron, with the included TIM.

I've never used a HSF's included TIM before (I feel kind of dirty, now), but it looked better than the average TIM, and I'm planning to replace it soon anyhow. I'm going to install my Zalman Reserator 1, possibly with a newer water block, and definitely with a fixed pump, in a couple of weeks. But that won't lower my temps very much, since it's low-flow.

This Opteron's stepping is LCB9E. Some people out there are saying this stepping is complete crap, but I've been reading through some threads at Guru of 3d and it seems like LCB9E can yield excellent OCs, but not as frequently as the more legendary CCBBE stepping.

My RAM hasn't overclocked well in the past, but I'll give it another go on this new board and the Opteron's memory controller. It's 2x512 Corsair ValueSelect.

Thanks for reading the essay :)

Comments

  • LeonardoLeonardo Wake up and smell the glaciers Eagle River, Alaska Icrontian
    edited February 2007
    I would ditch MBM5 and install Speedfan. MBM5 hasn't been supported for some time now. I have no experience with "Core Temp." No, you shouldn't be getting one-degree fluctuations per second, not unless your PSU is about to blow and it's output is fluctuating wildly.
  • GargGarg Purveyor of Lincoln Nightmares Icrontian
    edited February 2007
    Forgot about Speedfan. I'll give that a shot :thumbsup:
  • miragemirage
    edited February 2007
    Hey Gargoyle,

    I agree with Leo on Speedfan; that is the program I use with Opty-165/Asrock combination. Here are a couple of snapshots for you while running double Prime95's at the maximum heat option. I hope these give you an idea about what you would expect from your setup.

    Cheers\\

    PS: The mobo is Asrock 939Dual-VSTA, #1 in my sig
  • GargGarg Purveyor of Lincoln Nightmares Icrontian
    edited February 2007
    Thanks Mirage! Have you ever reached a temperature where you chip isn't stable anymore? I'm curious what my ceiling is. Right now, it peaks at around 55*C at the warmest part of the day, but it can fluctuate pretty quickly between 50 and 55.

    So far so good on the OC'ing. I've been bumping the HTT up 10mhz every 12 hours or so. I'm at 9x240 (with 5/6 memory divider) right now. I think I might leave it here until this weekend when I get my Reserator rebuilt and installed, though.

    So does anyone know if the the voltages reported for each core by MBM5 are correct (specifically - that one core is getting more than the other)? Seems weird that they'd be getting different voltages, but while MBM5 is the only program I have that reports voltages for each core, the rest of its measurements have correlated well with Speedfan, so I'm not ready to dismiss it quite yet.
  • miragemirage
    edited February 2007
    Gargoyle wrote:
    Thanks Mirage! Have you ever reached a temperature where you chip isn't stable anymore? I'm curious what my ceiling is. Right now, it peaks at around 55*C at the warmest part of the day, but it can fluctuate pretty quickly between 50 and 55.

    No, the only stability issue I had seen was related to insufficient voltage. The core temperature is ~60C when I set vcore to 1.5v and increase the frequency to 3GHz. Opty165 is still stable at 60C Tcore while running at 3GHz but I do not like to run at that temperature. So 55C Tcore should not be a problem.
    Gargoyle wrote:
    So far so good on the OC'ing. I've been bumping the HTT up 10mhz every 12 hours or so. I'm at 9x240 (with 5/6 memory divider) right now. I think I might leave it here until this weekend when I get my Reserator rebuilt and installed, though.
    Please post your oc results here, I am curious.
    Gargoyle wrote:
    So does anyone know if the the voltages reported for each core by MBM5 are correct (specifically - that one core is getting more than the other)? Seems weird that they'd be getting different voltages, but while MBM5 is the only program I have that reports voltages for each core, the rest of its measurements have correlated well with Speedfan, so I'm not ready to dismiss it quite yet.

    IMHO, what you observe as different vcores for the two cores is just an artifact of MBM5 since it is not able to handle the dual-core processors appropriately.
  • GargGarg Purveyor of Lincoln Nightmares Icrontian
    edited March 2007
    mirage wrote:
    Please post your oc results here, I am curious.

    I just finished setting up my refreshed + modded Reserator. More on the reserator mods later, but basically, I put in a stronger pump and a return line to get the warm water to the top of the tank.

    9x255 at the 166 memory setting crashed while playing a game. My assumption was that it was the RAM, since it was running at 208mhz, more than it has handled in the past.

    So then I set it to 9x260, but set the memory to 133 in the BIOS. That resulted in a memory frequency of 167. Supposedly, that takes the RAM out of the picture. But this time the game (Civ4) wouldn't even finish starting up, and Orthos failed after 3 seconds. I'm assuming the CPU needs more juice, unless the high HTT itself can cause instability. I'm using the (default) 4x HT multiplier so the HT link is running at 1040 right now.

    I'm going to try upping the core voltage a tad, but I think the BIOS only allows a small increase from 1.35 to 1.40. I'll probably need to do the vCore mod to raise it a bit higher.
  • miragemirage
    edited March 2007
    Gargoyle wrote:
    9x255 at the 166 memory setting crashed while playing a game. My assumption was that it was the RAM, since it was running at 208mhz, more than it has handled in the past.

    I would also suspect memory stability in this case. Is it possible to relax the RAM timings a little bit further or increase the RAM voltage (I can increase vdimm on 939Dual-VSTA)? 200 -> 208 is not a big jump.
    Gargoyle wrote:
    So then I set it to 9x260, but set the memory to 133 in the BIOS. That resulted in a memory frequency of 167. Supposedly, that takes the RAM out of the picture. But this time the game (Civ4) wouldn't even finish starting up, and Orthos failed after 3 seconds. I'm assuming the CPU needs more juice, unless the high HTT itself can cause instability. I'm using the (default) 4x HT multiplier so the HT link is running at 1040 right now.

    I agree it seems like you need a little more vcore. But you could try 3X HT multiplier first.
    Gargoyle wrote:
    I'm going to try upping the core voltage a tad, but I think the BIOS only allows a small increase from 1.35 to 1.40. I'll probably need to do the vCore mod to raise it a bit higher.

    Did you try 1.4? It might be sufficient. If not, I recommend using conductive pen for the voltmod instead of soldering, it is safer and reversible. You know it is not easy to find 939Duals anymore :)
  • GargGarg Purveyor of Lincoln Nightmares Icrontian
    edited March 2007
    I switched to the 3X HT multiplier and raised the vCore to 1.4. I probably shouldn't have done both at once, so I could find out what the actual cause of instability was, but I'm so impatient :D.

    Actually, to set the HT multiplier, do I need to change the CPU-NB link speed and the NB-SB link speed? That's what I did, and it worked, but I was wondering if I should only change one of those. The VSTA BIOS might be a little different.

    At any rate, those changes made it stable at 9x260 all night running Orthos. I don't know how far I'll be able to go with 1.4 vCore, but I might go pick up a conductive pen at Radio Shack today. You know, just in case :D
  • miragemirage
    edited March 2007
    Gargoyle wrote:
    I switched to the 3X HT multiplier and raised the vCore to 1.4. I probably shouldn't have done both at once, so I could find out what the actual cause of instability was, but I'm so impatient :D.

    Actually, to set the HT multiplier, do I need to change the CPU-NB link speed and the NB-SB link speed? That's what I did, and it worked, but I was wondering if I should only change one of those. The VSTA BIOS might be a little different.

    At any rate, those changes made it stable at 9x260 all night running Orthos. I don't know how far I'll be able to go with 1.4 vCore, but I might go pick up a conductive pen at Radio Shack today. You know, just in case :D

    You can set both CPU-NB and NB-SB as 3x. There is no performance penalty in doing so and it is more stable. You can also set the link width as 16bit for both bridges, 8bit was default AFAIR.

    You might be able to go pretty close to HTT=300 with 1.4 vcore. But temperatures start to increase drastically at and after 1.5v. You will see how good is your water cooler at that point:) By the way, I also bought the conductive pen from Radio Shack. Good luck!
  • GargGarg Purveyor of Lincoln Nightmares Icrontian
    edited March 2007
    It ran Orthos all afternoon at 275 HTT, but it failed after just one minute after setting it to 280. I wasn't able to find a conductive pen at the Santa Fe Radio Shack, so I'll try the one here in town tomorrow if they're open. I'm guessing the chip needs more juice :(

    I was hoping to get to at least 2.6 before needing more vCore, but I suppose that even if I get stuck at 2475 Mhz, I have little to complain about. These LCB9E cores are like, well, they're like a box of chocolates. I'll let you guys finish the movie quote. :wink:
  • miragemirage
    edited March 2007
    Gargoyle wrote:
    It ran Orthos all afternoon at 275 HTT, but it failed after just one minute after setting it to 280. I wasn't able to find a conductive pen at the Santa Fe Radio Shack, so I'll try the one here in town tomorrow if they're open. I'm guessing the chip needs more juice :(

    I was hoping to get to at least 2.6 before needing more vCore, but I suppose that even if I get stuck at 2475 Mhz, I have little to complain about. These LCB9E cores are like, well, they're like a box of chocolates. I'll let you guys finish the movie quote. :wink:

    Seems like you can hit 2600 MHz at or below 1.5v. You can also find that conductive pen online with this link

    Yeap I heard that chocolates quote from Csimon too: I have two CCBBE steppings one of them reached 2.8 GHz at default 1.35v, the other needed 1.45v for 2.6 GHz. :wink:
Sign In or Register to comment.