why is it reccomended (sometimes) to disable CnQ when overclocking?

(i'm asking as i'm looking to get (and sometimes OC) a sempron64 and if theres no point in having CnQ then i'll get a 2500+ for more cache, and cheaper.. rather than a 3000+)

It is either by folk who do not understand CnQ, or by folk who have no need for Quiet (and are out for maximum OC at any cost). IMHO of course!

Check out this thread about Sempron 64s as well as this one which has some more of my doodles. There is another more recent thread (IIRC) arguing the toss between more or less cache: my vote is unchanged since I bought my Sempron: I could not see a way I would benefit significantly from more cache but I could see a way to benefit from CnQ.

is this true? if so, can u use crystal thingy to do CnQ properly?
(hopefully i wont have to raise vcore.. but u never know..)

CrystalCPUID's "multiplier management" completely replaces the standard CnQ behaviour. It gives you free reign in which voltage and multiplier combination to set. Whether your system will run stable at that combination of settings is your own problem. So if you need a higher voltage for your overclock, then it will allow you to set that.

Some boards, such as the ASUS A8N SLI-Premium will let you overclock/overvolt and still use Cool n' Quiet. I kind of found this by accident on my own board. I was trying to see how high I could overclock my system without having to disable CnQ and once I pushed the BIOS FSB above 210mhz, I noticed it would apply an extra 0.15v at all times.

In other words, it would idle at 1.25v (rather than 1.1v) and the 5x CPU multiplier, then under heavy load shoot up to 1.50v and the 11x multiplier. I have a 3700+ SanDiego 2.2ghz, by the way. Anyway, by dropping the memory multiplier down I was able to clock all the way up to a 250mhz FSB, which is 25% o/c and 2.75ghz, 15-minute Prime95 stable (didn't bother running longer than that since it was just a test out of curiosity) and still have CnQ enabled.

It's hillarious to have an A64 that will do 2.75ghz and still drop back to 1.25ghz @ 1.25v when you're not stressing it. I used CrystalCPUID on my old computer and it seems like the BIOS/Windows implementation of CnQ responds faster than I was ever able to make crystal respond. For that reason, I would try to use the BIOS implementation whenever possible.

The great part for myself was that I found out that my x2 3800+ runs at stock speeds and 1.15v dual prime95 stable (1.15v is the lowest voltage that I can give with CrystalCPUID, since I am limited by the cnq voltages options + asus gives an extra 0.05V on everything ) . So now under load it will run at 2.0ghz and 1.15v instead of 1.35v like it did with cnq. My bios gives no voltage options for user, so CCPUID is the only way for me to undervolt. Too bad I cannot go lower then 1.15v . CrystalCPUID has the options, but my motherboard is setting me back on this. With a better board my cpu could probably run at 1.0v @ 1.0ghz or maybe even lower voltage.

I've been playing around with that myself, today. My BIOS has settings all the way down to 0.80v, but even if I select a lower setting (such as 1.00 or 1.05) it still comes up at 1.10v, and typically bumps up to 1.12v under load. I can select as low as 1.10 in Crystal also, but again, it seems to run at no lower than 1.10/1.12v.

Still, like you said, it seems to run just fine at that voltage and 2.2ghz. I'm going to let Prime95 run overnight and see if it gets any errors. At 2.2ghz, 1.10v, with Prime95 running for 30+ minutes the CPU temps topped out at 41*C with my Kill-A-Watt showing 116 watts A/C. At the stock 1.35v it stabilizes at 47*C, although it's worth noting the Q-Fan feature kicks the fan speed up another 300rpm at that higher temp. A/C draw at 1.35v with Prime95 running is 136 watts. A whole 20 watts more, just from running stock voltage!

The newer A64 rev.E cores have a minimum voltage (my 939 Venice and 754 Venice have 1.100v as the lowest possible voltage). If I select a lower voltage than that it will simply ignore it and use 1.100v. That's rellay a pitty, because the 754 Venice does 2.0GHz at 1.20v, so I'm pretty sure that at a low 800MHz 0.950v would have been enough.

Ohh... I didn't know they upped the minimum voltage. That would explain it then. I'm at 10 hours and 22 minutes of CPU time consumed by Prime95 right now according to Task Manager for my 2.2ghz/1.1v test. Still no errors.

I did notice last night before I started up Prime95 however that I crashed out of Battlefield 2 twice in a 15-minute period, which never happens to me. I upped it to 1.15v and it was fine for my 2 hours of gaming. I know Prime95 isn't exactly the be-all and end-all program for stability, but has anyone else experienced situations where a system will be stable in Prime95 but crash out of other high load programs like games?

If you undervolt to 1.10 Volts on a chip, is there a benefit to running Cool and quiet anyway?

Because extreme overclockers don't use the SPD timings for RAM and often run well above the DDR specs.

When CnQ kicks in it changes the dividers and can end up overclocking your RAM more than is stable. If you aren't overclocking to begin with you have some headroom with the average stick of ram. If you are pushing it before CnQ that lack of safety margin can be the straw that breaks the camels back.

Definitely not if the processor will only go down to 1.1 anyway. I noticed absolutely no difference in A/C draw or temperatures when I manually selected a voltage and switched between 5x and 11x multipliers at idle. Under load of course it's different, but CnQ will just increase the multiplier under load anyway. For that reason, I ended up disabling QnC on my PC and running at 11x 200Mhz @ 1.15v full time. It seems to be rock solid in all respects including gaming now, is as cool as before at idle and MUCH cooler under load than with CnQ since with CnQ it would ramp up to 1.35v.

I've been looking in CrystalCPUID myself and there was one thing I wondered. The interval time is seems significently higher then what CnQ use (I've read CnQ react in a matter of nano secconds) while the CrystalCPUID react in a matter of seconds. Has anyone noticed anykind of slowdown using crystal? Thanks

On my old Mobile AthlonXP system I used Crystal and manually changed the timings down to like 0.1 second for the speed-up delay and 1 second for the slow-down speed. It was rare that it ever seemed any less responsive than when left at full speed, but you are correct, I believe BIOS/OS controlled CnQ is quicker to respond. Or.. use a setup like mine; I'm currently at 1.125v @ 2.2ghz full time, CnQ disabled. Apparently although 1.10 wasn't quite enough, that next tiny step up was all it took to be solid even for gaming now. I'm happy with it.

jeez I just couldnt mess with a software program to do that on the fly

Hm. good info on how c and quiet is good to leave off while undervolting as it really isnt needed then.

I cant wait till I get a venice setup

I had a strange experience with testing voltage using Crystal. I lowered down my voltage until I reached 1.15 volt stable for 24 hours. After that I figured I could set it in the bios. Well it didn't booted at all at 1.15 volt. I had to reset the cmos to be able to boot again. So now I'm at 1.2 volt and I underclock it manually when I go to bed.

Good point; I don't think I've done a cold start at 1.125v yet, only hot or warm starts. Since I leave my PC on all the time running folding@home now, I haven't had the opportunity to do a truly cold start. I'll give it a shot maybe this weekend and see if it boots up. Some mobos will actually apply a higher starting voltage, then drop it down to the BIOS defined voltage. What mobo were you using?

I've got the same thing on my 3500+ it runs fine at 1.1v but requires 1.2v to boot.

I'm using A8N-E (damn active cooling). So Maybe I could set crystal to drop the Voltage once booted. That would work.

That was my thinking when I chose the 2500+. In retrospect, and after doing more research on the effect of L2 cache on Sempron64, I might have been slightly better off with the $8 Cdn more at the time 2600+ just for the 8x instead of 7x multiplier + 200 MHz higher stock (or base, since I haven't run much at stock) speed. On the other hand the CPUs I looked at while researching Sempron were all e3 stepping or earlier, so when I found out mine was an e6 it felt like I'd gotten the x64 and SSE3 as free bonuses... so its all good

CnQ: I don't have any regrets not getting a 3000+ or higher for this, as even OC'd the Semprons run pretty cool: mine OC'd to 1.93 GHz (modest by Sempron64 standards, but my CPU's happy place right now in this matx case with stock cooling) it spends most of its life idling between 26-28C except when I'm gaming, where it climbs into the mid 30s (it peaks @ 40C running prime95).