prodigy_

Erm. Not sure if Zalman is good for "naked" setup. Look at this one: http://www.thermaltake.com/coolers/4in1hea...on/cl-p0114.htm As much as I hate Thermaltake, I must admit that BT is actually not that bad. It's really quiet (something that you don't normally expect from Tt) and it's held in place by screws. Screws IMHO are your best bet if you gonna run naked. Much less chance to damage CPU while mounting. Basically you tighten one side a little, then another side a little and so on...

OK, let's see. "Everything goes away" means that shell (explorer) crashed and was restarted. "Error msgbox" probably said either "memory could not be read" or "memory could not be written". It's not always caused by HW (sometimes you get such errors due to memory leaks). But with a new rig I'd start making stability tests to see if everything works properly. Especially if it's OCed.

Never used DDR Booster, sorry. Only know that it allows you to feed RAM with higher VDimm w/o voltmodding your MB. Not sure if DFI NF3 allows to use 5V rail for RAM but NF4 does (and so you don't need anything else, only to switch the JP17 jumper). Prob better to PM your questions about DDR Booster to AndyOCZ. I'm sure he's familiar with the device.

Well, then higher multi is better. Thanx for quick answers!

CPU/11 for 333 x8 (not sure if these sticks can do 266 but I hope so). Probably CPU/12 for HTT 300 x9 (CPU/11 if possible, but 272 seems too high). 2-2-2-x timinigs (anything else is almost useless with BH-5). Really? I was under impression that higher HTT with lesser multi is better. But maybe it's just an Intel fanboy's prejudice... :angel:

Since my Venice doesn't really like anything higher than 2700 I'm trying to figure out how to compensate this drawback... A possible config that comes to my mind is: CPU @2664 (333 x8) RAM @266 (CPU/10) The question is would it worth the effort. Should I expect any performance gain with 333 x8 (if compared to 300 x9)?

1) "Burning-in" term implies using voltages slightly higher than those suitable for 24/7 in order to get rid of certain defects in RAM chips. These voltages should be low enough to leave main chip structure intact but high enough to destroy (literally burn) unwanted parasite links between memory cells. 2) "Properly" depends on the exact type of RAM in question. Since yours is UTT CH-5, you should use 3.4-3.5V during burn-in process. (I SERIOUSLY do not recommend using 3.5V for 24/7 with these sticks. Yes, I know that they are officially rated 3.5V by OCZ. But you should be aware that MANY of them actually DIED from being fed with such voltage for long periods of time. Safe voltage for UTT CH-5 is 3.3V or lower. ALWAYS use active cooling when your RAM is OCed. Especially while burning-in. 90mm fan blowing directly to the sticks should help to keep temps within acceptable range.) Best app for burning-in is Memtest+ (current version is 1.60) - where you should choose continuous looping of test 5. Best freq for burning-in is that which gives you some amount of errors but not thousands of them. It's easier to see if there's any progress over time this way. 3) You certainly can. But for safety reasons I recommend going step by step adding 100mV at a time. 4) First results usually show up after 24 hours burn-in. If you do not see any improvement at all, then you've probably reached the top at the given voltage.

During POST (at boot) system doesn't show you any actual values, only relative ones. "DDR400" thus only means that you're using 200/1:1 divider. (Btw, create a signature with your system specs. You can use mine as a guideline.) Anyway 623-3 BIOS doesn't allow you to use higher freq for memory than your current HTT freq. So if you're not OCed, than 1:1 (200 Mhz) is your best option. To get any higher you should either OC your CPU (increase HTT) or flash a beta (702) BIOS which allows you to set RAM freq higher than HTT.

Looks like a BIOS issue. Try flashing latest official (623-3).

Mine are Kingston's but still this seems to work. 5 hours - 400 passes - 0 errors. Previously they gave me thousands of errors @245 and some errors even @233. I'll test with Prime95/Blend now to be absolutely sure... --- [Edit - update] 245 Mhz seems to be the highest memtest stable clock. Prime/Blend fails very soon (Prime/LargeFFTs doesn't so that's certainly memory related). @241 - Prime/Blend stable. @247 - 2-4 errors in Memtest+ per pass of test 5. I'll flash some -2 BIOS and try yellow slots.

Actually yes I'm gonna make some burn-in with these sticks later tomorrow. But yesterday it was only like 2 hours which is IMHO not enough for such improvement. (Ppl on our local forums suggested 24 hours at least...) [Edit - update] That's something awesome. Now doing 245 Mhz (HTT 300x9, 166 divider) at 3.2V. Still no errors. Don't know what to think about it...

Yesterday Memtest+ showed me about 20-21 errors per pass (test 5) at 237 Mhz. Now I've done 40 passes of test 5 already w/o any errors. The only thing that changed is that I (accidentally) switched sticks around. But they're still in orange slots. I'm using the same BIOS 623-3. (VDimm=3.2V, HTT 290x9 and I didn't touch DRAM settings at all). And ambient temps are higher today. --- Is this possible at all? What do you think?

Yeah, it's average OC you can expect from Venice and TCCD RAM. Anything higher is rarely stable and will definitely cost you quite some time/effort.

Theoretically, DC gives you twice the bandwidth of SC. With K8's integrated memory controller it doesn't work this way though. I'd try and see for myself, If I were you. Off-topic: if your memory is indeed based on BH-6, then you actually got old Winbond 6ns 190nm chips rated for 4.0V Max! No kidding. Anything less that 3.8V is proven to be safe for them.

1.65V is actual. BIOS reports 1.62V. Software reports 1.6V. It's been like this for about a week already and will stay for 24/7. Sooner or later I'll take IHS off and try 1.7V. Hmm, interesting. A leaked beta maybe?