Trying to OC my 1100T

[TeknoBug]

My case is setup exactly like the one in trent's image, and it works best (after years of experiencing with airflow in cases) except my top fan is covered because I have cats that like to sit on top of the case and stuff from their fur and feet tend to fall in. I'm a non-believer in needing 8-10 freaking fans (like some of the "pro" gamers do).

Jakerdud, why force air DOWN when heat wants to go UP?

 

[Anonaru]

I'm a non-believer in needing 8-10 freaking fans (like some of the "pro" gamers do).

I have a couple extra fans in the intel and piledriver rigs in my sig, but they REQUIRE IT (Those bad boys put out some heat)

The deneb rig is in a pretty ordinary case.. For Jekerdud's setup..Top-mounted fans should always be blowing out, right?

Also Jekerdud, 4.2 is pretty darn good. You should consider moving onto increasing your NB frequency

 

[Jekerdud]

Yeah, I'll deal with the fans when I find my PC toolkit (moved recently, have no idea where it is).

2 hours into that three hour test, and core temps have been between 45 and 52. But, a worker stopped. The rest are still going though. Should I stop and retry?

 

[keny]

Yeah, I'll deal with the fans when I find my PC toolkit (moved recently, have no idea where it is).

2 hours into that three hour test, and core temps have been between 45 and 52. But, a worker stopped. The rest are still going though. Should I stop and retry?

A worker dropping is a fail of p95, but after 2hrs without a fail I would say that's pretty much stable and shouldn't cause you any problems in day to day computing.

 

[Jekerdud]

Yay! Do you guys need any screenshots, or am I good to start moving into fine tuning the overclock (getting the NB/HT dealt with, adjusting the FSB up and multiplier down, etc).

 

[Anonaru]

Yay! Do you guys need any screenshots, or am I good to start moving into fine tuning the overclock (getting the NB/HT dealt with, adjusting the FSB up and multiplier down, etc).

You're good to start fine tuning I think.. But I want to advise you on something--

A worker failing in P95 on Deneb / Thuban in particular usually means you'll be facing a possible random BSOD while gaming.

Even after 2 hours. Then again, it might not amount to anything at all-- Either way, I don't really think adjusting your NB is going to mess you up any.. If anything, I'd approach your same frequency by downing the multiplier by .5 and upping your FSB until they match. This might increase stability just enough

 

[Jekerdud]

Ok, I will try that (taking down the multiplier and upping the FSB).

Anonaru, about the NB, you said "a single voltage bump"... what did you mean by that? And do you think setting my NB to 2400 would be a good start? Anything with the HT yet, or will we leave that alone for the time being?

 

[RGone]

Yay! Do you guys need any screenshots, or am I good to start moving into fine tuning the overclock (getting the NB/HT dealt with, adjusting the FSB up and multiplier down, etc).

Here is my problem with fine tuning. A worker "has" quit at 205x20.5= 4.2Ghz.

The most likely reason for that 'stopped' worker is not enough voltage to the cpu. You are at the very upper edge of what should be used for Vcore at 1.52 indicated. This means to me that the 4.2Ghz really needs more Vcore to be stable but the situation seems to be that you have pretty high Vcore already.

I would drop the multiplier to 20.0 and keep the 205 FSB and that should give 4.1Ghz. Then run a couple of hours of P95 Blend and see if that gives NO worker failure. If you get nO worker failure at 4.1Ghz set as described, then it should point out that the cpu wants more Vcore to be fully stable at the 2 or 3 hour mark if the cpu speed is beyond 4.1Ghz.

If 4.1Ghz is now fully stable and No worker drops out in P95 Blend, then 4.1Ghz may well be your max speed. Only then would I start any so-called tweak of CPU/NB etc. We would tweak now because we know 4.1Ghz is in reality the max cpu speed with what is very close to the max Vcore I would use for 24/7 use.

I do not like pushing to the very edge and then expecting that to work from now on for 24/7 use. What will happen is that a month, or maybe 3 months later you will have what some come back and say and that is they are having BSODs but they declare it was P95 Blend stable some months before. There is one in this forum section right now with that problem.

I had much rather have 4.0/4.1Ghz that I do not have to push to the very edge in voltage and know that I have had zero worker stoppage. YMMV. But that is what I look for in deciding on my 24/7 speeds.

 

[Anonaru]

Here is my problem with fine tuning. A worker "has" quit at 205x20.5= 4.2Ghz.

The most likely reason for that 'stopped' worker is not enough voltage to the cpu. You are at the very upper edge of what should be used for Vcore at 1.52 indicated. This means to me that the 4.2Ghz really needs more Vcore to be stable but the situation seems to be that you have pretty high Vcore already.

I would drop the multiplier to 20.0 and keep the 205 FSB and that should give 4.1Ghz. Then run a couple of hours of P95 Blend and see if that gives NO worker failure. If you get nO worker failure at 4.1Ghz set as described, then it should point out that the cpu wants more Vcore to be fully stable at the 2 or 3 hour mark if the cpu speed is beyond 4.1Ghz.

If 4.1Ghz is now fully stable and No worker drops out in P95 Blend, then 4.1Ghz may well be your max speed. Only then would I start any so-called tweak of CPU/NB etc. We would tweak now because we know 4.1Ghz is in reality the max cpu speed with what is very close to the max Vcore I would use for 24/7 use.

I do not like pushing to the very edge and then expecting that to work from now on for 24/7 use. What will happen is that a month, or maybe 3 months later you will have what some come back and say and that is they are having BSODs but they declare it was P95 Blend stable some months before. There is one in this forum section right now with that problem.

I had much rather have 4.0/4.1Ghz that I do not have to push to the very edge in voltage and know that I have had zero worker stoppage. YMMV. But that is what I look for in deciding on my 24/7 speeds.

This is actaully very sound advice. I'm sorry, I am an experienced OC'er, and understand these sorts of things without them being said-- I keep forgetting I'm not making a reference of information for myself .

Ok, I will try that (taking down the multiplier and upping the FSB).

Anonaru, about the NB, you said "a single voltage bump"... what did you mean by that? And do you think setting my NB to 2400 would be a good start? Anything with the HT yet, or will we leave that alone for the time being?

A 2400 NB will be good for most applications. A single bump on some boards is whatever the minimum increment it will allow you to raise voltage.

I like what RGone suggested though-- Lets hold off on FSB and NB tuning / voltage until you get a fully stable OC. You may very well be able to get back up to 4.2 on the lower, safer voltage, but hold off on that just a bit longer

 

[Jekerdud]

Well, I just restarted from setting the bios to 210 x 20, looks like I'll reset it. 2 hours would be good, right?

As for the rest of it, I wrote down what bios is currently listing. I did this both so you guys could see what they were, and so I didn't have to restart just to find them:

The NB has a pull down list of frequencies for me to apply, the closest to 2400 is 2460, the next lowest is 2255, next highest is 2665.

NB voltage is still on auto, and was currently at 1.103. HT voltage was also on auto, and was at 1.203

A single bump on some boards is whatever the minimum increment it will allow you to raise voltage.

I'm not sure what a single bump would be though, .25? .025? It's manual input, so what I think would be small, might be too large

 

[RGone]

A bump most often means just that. One of whatever the least bump is allowed by the bios. Good boards make that bump about 0.025Volts.

 

[Anonaru]

Well, I just restarted from setting the bios to 210 x 20, looks like I'll reset it. 2 hours would be good, right?

As for the rest of it, I wrote down what bios is currently listing. I did this both so you guys could see what they were, and so I didn't have to restart just to find them:

The NB has a pull down list of frequencies for me to apply, the closest to 2400 is 2460, the next lowest is 2255, next highest is 2665.

NB voltage is still on auto, and was currently at 1.103. HT voltage was also on auto, and was at 1.203



I'm not sure what a single bump would be though, .25? .025? It's manual input, so what I think would be small, might be too large

If you failed in 2 hours, I'd run for 4-6 hours just to make sure you nailed the problem.
2460 is fine, and make sure you're looking at NB-CPU voltage. Don't mess with your chipset NB /HT Voltage (Or set them at 1.105 and 1.205, just in case its needed)
Your NB-CPU voltage is manual entry as well? I'd give it a .025v bump in that case, but not until you make sure you've nailed your current problem.

 

[Jekerdud]

While I am running this test, the vcore is still going up to 1.52 (due to LLC). When I brought the Vcore voltage up last time, I was at 200 x 21, which brought me to 4.2Ghz. I did an hour test after that, then I went back and increased the FSB to 205, and decreased the multiplier by .5. The tests except for the last one since then were good. Should I decrease the voltage back to 1.45? It is currently at 1.475, and the only reason it goes up is because of LLC. The only reason I ask is because of Anonaru mentioning "a lower, safer voltage."

 

[RGone]

Jekerdud get in this thing with us. I write hoping you can get to a lower Vcore by suggesting to clock back to 4.1Ghz. Anonaru is hoping you can use less voltage. Yes man try less Vcore. We do not or should not have to suggest nor really authorize an adjustment at your end. What we would need is if you adjust and have an error, we need to know what you changed and by how much. Other than that the whole idea of an overclock is the most stable speed with temperatures that are within the considered safe/stable range.

RGone...

PS: my good CHV board has more than one level of LLC. I can adjust the level of LLC so that I do n0t get an added voltage under load as you are seeing. OR yes I can choose another level of LLC and get an "over-voltage" as you see. I have no idea what LLC 'ramps' you have. Maybe you have an LLC setting that will only add the amount of voltage to the cpu that is lost because of load. Like I said my CHV allows such. End PS.

 

[Jekerdud]

Ah ok. Well, I couldn't get a lower Vcore just by going back to 4.1Ghz. I had to lower the Vcore itself.

I just finished running Prime 95 for 4 hours, and no workers stopped, and nothing went wrong!

 

[Jekerdud]

I'm starting a test after setting the CPU/NB frequency to 2460. How long should that test run for? If this one goes good, would I try going back up? Or just leave it at that and call it stable?

 

[trents]

You should run the Prime95 blend test for at least 2 hours to have reasonable assurance of stability once you get past the preliminary stages of overclocking and find your approximate max CPU coverclock. Yes, you should run the test for that long as well when you make any changes after that.

 

[Jekerdud]

Ok, will do. I will increase the CPU/NB to 2665 to see if it handles that well. Since it was earlier mentioned that Thuban cores don't like going above 2700 very much, I will leave it at that if it stabilizes.

Beyond doing that, trents, what do you mean by "get past the preliminary stages of overclocking and find your approximate max CPU clock"? I've has some successful runs, and some not, at 4.2Ghz. Would I just go back and decrease the multiplier and increase the FSB slowly (how many increments should an FSB be increased by anyways?), until I hit a new wall? We've already established that my Vcore needs to stay where it is (1.45), since LLC takes it to 1.52 at 1.475. But, would I start trying to go for higher FSB, closer to 225 as was mentioned early on?

 

[trents]

Cut back on the multiplier some (maybe to 18x) and then start to increase the FSB by 5 mhz increments. Before you do that, however, back off on your memory frequnency to the 1333 mhz divider and put your HT Link and CPUNB to stock. The reason for this is because the ram, the CPUNB and the HT Link are all tuned to the FSB and they will get faster as you increase the FSB. So you need to make a little headroom for that. When your FSB is about to 245 mhz your ram ought to be back up to around 1600 mhz. Ajust your CPU multiplier as necessary to so that the CPU is around 4.0 ghz when the FSB gets the ram back up to 1666. Then stop increasing the FSB and tweak the CPU overclock with the multiplier. Finally, after all that play with the CPU voltage to find the lowest amount that will provide 2+ hour Prime95 blend test stability.

 

[trents]

Oh, did you find out if the system was stable when the ram was at 1600 mhz and 1.65 volts? If not and you need to run it at 1333 then start the ram at 1066 before doing the FSB increases.