davidk21770

I'm a long-time AMD fanboy who's waited too long for them to catch up and need a new, fast system. I've been reading some of the forums, but in the past this has been one of the best for solid (usually) unbiased info -- SO. I'm looking to build a new system. I'm going to OC as hard as it will go, but haven't decided whether it will just be quality air or water (I have a Danger Den water system on my current setup and could move it over). I'm thinking a Q6600 or Q9300 -- even though you can't get the OC clock on the Q9300 as high as the Q6600, the Q9300 OCs seem to be faster on benchmarks. (Correct me if I'm wrong!) I'm not a gamer -- I do a lot of video processing and general number crunching and my computer doubles as a PVR in the evenings, so the Gigabyte GA-P35-DS3R which seems to be the common recommendation around here worries me because of the reported audio delay. I'm not interested in Crossfire or SLI. I tried SLI for a while and found it to be fast, but not a good video (PVR) option -- doesn't like to share with an HD TV and doesn't respect aspect ratios on the HD TV. Seems to be gaming/benchmark oriented. And, I want a wide PCIe slot for my raid card. I would like it to be compatible with my old video card (eVGA 7900 GT KO). I do want a system that I can upgrade as Intel updates it chips. Cost is a consideration, but I've learned that you can't be cheap if you want a good solid everyday OC. I really like my current DFI Expert -- complicated to OC (LOTS of knobs to learn), but solid. Ok, if you've managed to get though all of that -- Thanks! What do you recommend?

The MB should be fine. EDIT: no performance issues.

That's why I sighted down each row from each of the 4 sides. Looking down the row, each pin should hide the one behind it. If any stick out, gently bend it back. It is painful and tedious, but much less expensive than replacing the chip!

Been there, done that, survived . Heat the heatsink so that it's very warm to touch -- not hot. This will soften the thermal compound. You can do this with a heater or a hair dryer. Twist it, but don't do any more damage to the pins! It'll even take gentle prying. Take a magnifying glass and I used a small straight blade screwdriver to carefully line-up the pins. I sighted down each column from all 4 directions (turn the chip 90 degrees at a time) to get them all as straight as I could (they looked perfect and I bent 2 or 3 whole rows). When you put it back in, be very careful -- they won't be quite right and you don't want to flatten any by pushing too hard. If any break, you may be ok. A lot of them are redundant grounds, so don't panic until it won't boot... PS... In the future, it's usually a good idea to run the system to heat the chip before removing the HS...

Strip your system to the bare minimum: video, PSU. See if you can get it to boot. If you do, work on from there to identify the culprit. Otherwise, try the no-boot CMOS clear again -- longer this time -- I've seen it take 72 hours! (http://www.dfi-street.com/forum/showpost.p...04&postcount=24) And, even if you have to do some surgery on your case, a top notch PSU makes a huge difference -- get a recommended one!

Possible, but IME, this is virtually always a memory problem. Have you run memtest yet? (8+ hours starting with 20 cycles of just test 5). You can enable it in the BIOS on the genie page.

The docs for the program that you're trying to use. Search for smbus. I can't load sisoft because I use nero image drive and sisoft's incompatible, so I can't see what the setting you're looking for it. EDIT: See post: http://www.dfi-street.com/forum/showpost.p...194&postcount=3

My 1st suspect when this happens is always the PSU -- no matter how big, no matter who makes it, no matter how long you've had it. Can you borrow one from a friend to try? Have you stability tested -- prime95 for 8 hours (as above -- important that you do it like his link says for a true test!), and memtest for 8+ hours (enabled from your BIOS), starting with just test 5 for 20 cycles.

Are you using an add-in network card? If so, try removing it. If that works, try a different slot. Did you install nVidia's firewall? If so, uninstall it (NOW)! Did you install nVidia's IDE driver. If so, uninstall it unless you're using nVidia raid... so many decisions . If you're using nVidia raid, get the latest nForce drivers from the nVidia site. Uninstall all nVidia drivers from safe mode and run driver cleaner; reboot to safe mode and install the new drivers, including video if you have an nVidia video card. I reboot to safe mode one more time for nVidia video card drivers to initialize properly. EDIT: I ran at 1.33v stably with my 4800 for months before OCing. (But, it wouldn't hurt to bump as above as a check). This could also be PSU. If you can borrow a friend's PSU, I'd do that and see if it helps. (Especially if I am remembering the specs on your video card right!).

You can not let a program scan the SM bus (system management bus). The expert will reboot. This will happen with any program that scans the bus. All of the ones that I have can be told not to scan the bus, to just use the default locations. Check the docs...

That's the problem with high volts on that series a64. It kills the integrated memory controller. However, I ran 1.61 for a few months with the setup in my sig was ok. I've since dropped to 1.58 (and had to drop my OC a little also).

Try this CMOS clear: http://www.dfi-street.com/forum/showpost.p...04&postcount=24 Reload your BIOS. Load optimized defaults. Don't use any saved values. And, IMO, low 50s under load is high for everyday use. I don't want to see more than 48C under dual prime after a few hours.

If your brother discharged and killed the CPU, how is it DFI's fault? If he discharged through the CPU, could he also have damaged the MB (or other components) the same way? Can you borrow a CPU? Have you tried your old PSU to see if the new one is bad? Have you done an out of the case build with a minimum system? How long did you CMOS clear and did you do this one? (http://www.dfi-street.com/forum/showpost.p...4&postcount=24). (I saw one case recently where a guy did a 72 hour CMOS clear while waiting for an RMA and had his system work...).

I little more info... 1. 1.35v not the voltage seen by your CPU. It is a register value, not a measurement. Use MBM5 or CPUz if you want a measurement. The 1.35 is mulitplied by VIDSpecial (as a %, so divide by 100). 2. Agreed on Trc. There's a lot more to relaxing memory timings than that one value. Check the OC database for systems with your memory. Check the stickies on memory timing values. Get a64info or a64tweak to see what you're currently set at (what's under the "auto" in the BIOS). 3. Core temp is bogus on AMD chips. It's a research project from AMD. They are uncalibrated and +/- 14C. Until AMD provides a calibration procedure or calibrates them before shipping, these 2 values are meaningless. 4. Set your HT multiplier for a link bandwidth of about 1000 MHz. This is not critical to OC performance, so don't worry if you're at 900. 5. What voltage you use is a function of the quality of your cooling and how much risk you want to assume. 1.5-1.55v is considered a good everyday number with good cooling -- either air or water. Watch your temps under load -- dual prime95s (like this: http://www.dfi-street.com/forum/showthread.php?t=16445) shouldn't go above about 47/48C after hours for testing. EDIT: Expert MBs are CPU temperature challenged. They measure low. The latest BIOS has an offset that you can set on the health page. What I do is let my system idle in windows with no load for 1/2 hour or more (the BIOS doesn't execute idle cycles, so it is always under load; ie., you can't do this from the BIOS!). Then I use a digital temperature probe on my heatsink as close to my CPU as possible. I assume that my CPU is a few degrees above the heatsink temperature. (I also underclock and undervolt for this test to get the CPU heat generation at an absolute minimum). 6. Do not spreadspectrum any clocks if you overclock. This is a destabilizing factor.

Even if the rails read ok (the PSU is not dead), what I've seen is that the variations under load cause problems. You can only read static or snapshot voltages with the DMM -- see catastrophic failures. This is why I always recommend borrowing a PSU when there's any doubt. You CAN use the DMM to calibrate your MB's internal DMM and then use a program like MBM5 to create a log of the voltages over time and see how they vary. I've still seen this look ok and had the PSU not be stable enough. Bottom line is if you have stability problems, the only way to rule out a PSU that hasn't failed is to swap-in another known good PSU and see if the problem persists. How big a pig is your video card? If it's lightweight, then you can probably get by with less (if you're stable now). If it's not, then I'd be saving for a good PSU...