Trying to OC my 1100T

[keny]

Side random question: what do you guys usually do when waiting on P95 to run?

Prey that it doesn't catch fire

 

[RGone]

Prey that it doesn't catch fire

Hehehe. I hear that man.

 

[keny]

I think you have about summed it up on you're last post RGone"ster" with him hitting the max CPU multi at around 4-4.2ghz range and a small bump on the fsb will about max it out without ramping up the vcore to silly levels

 

[Jekerdud]

Well, I just had a crash while waiting on the test to get you guys some screenshots. I think I will stop there and just optimize it, unless you guys think I can do something else. The last temperature I seen was 52c, but I was watching TV at the time, so I was only checking it every 10 minutes or so. It went about 40 minutes before crashing.

Here is some pictures of CPUID so you guys can see where my system is at:

 

[RGone]

1. I would set the ram command rate to 2T since that is what CPUz is reading as the suggested Command rate for the ram at DDR1600.

2. Still no HWMonitor of P95 Blend mode which you could run for say 15 mins and post. It does not have to be at the end of a full run of P95 for hours at this time.

3. I thought you said that approx 4.2Ghz had already tested out hours of P95 blend? But then suddenly you have an issue at 4.2Ghz?

4. I see CPUz is showing 1.488 as the Vcore of the cpu. That is getting close to the 1.5V Vcore that I really like to stay slightly below for 24/7 use. Don't get me wrong, I will pour the coals to Vcore for running just benches if I have to to get to a really high cpu speed, but I do not like to run much over 1.475V Vcore day in and day out. That is me. Others may do otherwise.

5. I do not have any problem with your 1.488 indicated in CPUz. I stated what I 'tried' to do with Vcore. 1.488 Vcore is not that far out of my 'sights' for long-term use.

6. I would drop the 21 multiplier to 20.5 and try it with 205 FSB for 4.2Ghz also and see if that is 'more' stable at what appears to be very close to the cpu mhz limit. Often dropping the multiplier and getting the last bit of cpu Mhz with a little FSB and slightly lower multiplier will bring that cpu Mhz into stability with the same or maybe a little less Vcore.

7. You will have to do the testing at your end since that is what will tell the real story.

 

[keny]

Has it passed p95 at those ram speeds (1600) before, as you could drop it to 1333(jedec#4) to make sure the imc is not stressed , i doubt it is stressed on a thuban at 1600, but if you keep failing, and RGone's suggestion does'nt work , which im pretty sure it will, you can give a 1333 ram speed a crack, also are you setting the ram voltage to 1.65v at 1600?

Mark

 

[RGone]

also are you setting the ram voltage to 1.65v at 1600?

Mark

That must be older style ram if it still needs 1.65V to reach the DDR1600.

 

[Jekerdud]

No, I haven't adjusted the RAM voltage yet, as I am following advice on the first page to leave everything (voltage wise) except the Vcore voltage at stock, until I start hitting my max. The Vcore voltage is set to 1.475. It's LLC that's bumping it up according to some replies at the end of the first page.

The RAM has been at 1600 since I started testing.

To answer your #3 question Rgone, it did fine on the first test, but then when I went to get some screenshots of HWMonitor while it was running for about an hour, at 40 minutes it crashed. I will do one more 20 minute run to get you a screenshot.

 

[RGone]

1. Ram voltage should alway be set to what the manufacturer recommends for the speed you are running the ram at. Setting the correct ram voltage would not be included in leaving voltages at stock since stock for that ram at DDR1600 is 1.65Volts. You got to flex wiht what peeps say.

2. Ran an hour all okay. Start another 1 hour run and fails at 40mins. Does not sound like the rig is stable yet.

3. Yes, let us see the 20mins run with HWMonitor logging min/max volts and temps.

4. LLC can raise the voltage and is designed to work somewhat like that. What that means is the actual voltage to the cpu is higher for real than the 1.475V.

 

[Jekerdud]

So I should set it (the RAM voltage) to 1.65?

Here is the HWMonitor screenshot at the 20 minute mark:

 

[keny]

Looking at he monitor it looks like you're running 1.65v on the ram already, but best to make sure in the bios, and you're pushing over 1.5v under load on the CPU :/

 

[Jekerdud]

It has only went above 1.5 since I increased it to hit 4.2ghz. Previously, it was set to 1.45 in the bios, and LLC only made it go to 1.49 under load. This was for everything at 4.1ghz and under.

 

[keny]

If that failed then you're not going to get much more from it via the CPU multi, I would do what RGone said and run the multi at the last known good setting and clock the CPU up via the fsb from there as once they hit a wall on the CPU multi it takes a big hike in vcore to make the next step and you're not going to get there as you're on 1.5v now and that's about as much as I would run 24/7 and then that's quite high for a daily vcore.

 

[Jekerdud]

If that failed then you're not going to get much more from it via the CPU multi, I would do what RGone said and run the multi at the last known good setting and clock the CPU up via the fsb from there as once they hit a wall on the CPU multi it takes a big hike in vcore to make the next step and you're not going to get there as you're on 1.5v now and that's about as much as I would run 24/7 and then that's quite high for a daily vcore.

So would I do that from the setting where my Vcore voltage was at 1.45? Or where it is now at 1.475?

Also, I just set the command rate to 2T, the FSB to 205, and the multiplier to 20.5, here's hoping it's stable! I'll be running an hour test to confirm. Do you guys need screenshots from that too?

Edit: I just ran that test, and everything went fine, all passes! So, I am wondering what to do next, I'll sum up what has been done, and what hasn't been done (after reading through the whole post):

What has been done:

• FSB Set to 205. Multiplier set to 20.5. This brought my speed according to CPUID to 4216.6 Mhz

• Vcore voltage set to 1.45 at first, but once I hit 4.2Ghz (200x21), I had to set it to 1.475, which is the current setting

• RAM command rate set to 2T, and the RAM voltage set to 1.65 instead of auto

What has not been changed:

• NB still auto, which is 2000Mhz. NB voltage is also still default. Same with HT.

• CPUNB is also still default

So there's some changes that need done it looks like, but I wanted to wait until you guys thought things looked good enough to start changing those. Lastly, here is a screenshot of HWMonitor from my latest test. The temp hit 55 for about 30 seconds, then dropped again and stayed around 53 the rest of the test.

 

[trents]

Yes, try it with those changes. Yes, the ram would need 1.65v to run at 1600 mhz. That would be "stock" for that ram. Sometimes the newer AMD CPUs don't like that older style ram that required 1.65v to run the ram at its max rated frequency. It can put extra strain on the CPU's integrated memory controller. But you should try it anyway. Won't hurt anything. Your core temps are safe at 55c and your socket temps are safe at 65c but both temps are likely on the cusp of being unstable. I still have a suspicion your cooling isn't operating up to par. I'm running my 960T unlocked to 6 cores and 4.0 ghz using 1.6 volts of CPU voltage under load with an air cooler and my temps are not any higher than yours.

Tell us how you have your case ventilation set up. How many case fans do you have, how big are they, where are they placed and how are they set up with regard to push-pull? I'll attach a pic of what proper case ventilation should look like as far as air flow direction.

What about your H100? Is the radiator on the inside or outside of the case and how do you have it's fans set up? Are the fans drawing cool air from the outside through the radiator and into the case or the other way around? I'll attach a pic of what proper case ventilation should look like as far as air flow direction.

 

[Jekerdud]

Push/pull from what I read is having fans on both sides of the radiator, which is what I have. But, I'll attach pics of the directions. The smaller one is the case I have, the second one isn't my system exactly, it's a random pic I found on google, but it should show you how my fans are set up on my radiator, which is in between the fans. Blue arrows are intake, red are exhaust. I thinking that having the radiator fans blow in was a good thing, since I have the 140mm side fan exhausting the warm air anyways.

The RAM has been set to 1.65 for a while now. The only things I haven't adjusted are the NB and HT, when would I start changing those? And, what will I need to do to confirm 100% stability?

 

[trents]

Are the radiator fans lined up well with the case's top panel fan mount holes? The googled picture's case's fan mount holes look to be larger than 120mm and I know the radiator uses 120mm fans. Looking at the pic on the right that you googled it doesn't look like those case fans line up well with the radiator fans. Could you take a digital camera pic of your case with the side panel off?

I would hold off on overclocking the NB until you settle on the overclock of the CPU cores.

 

[Jekerdud]

I don't have a digital camera, but it's pretty much exactly like the picture, since it looks to be the same case (I am 99% positive it's the same case, cable management is in the same position, same rubber sleeving). The radiator itself is lined up with the mount holes that coolermaster put in there for a radiator of that size. The top fans in the pic are just like mine, 140mm, mounted on the mount holes provided on the case.

I haven't done any testing or haven't changed anything for about 3 hours, the list of changes above with my HWMonitor screenshot are where everything is at currently. What should I do next?

 

[trents]

Have you been running Prime95 for 3 hours stable? If so, I would call that good. My only other suggestion would be to experiment with flipping the radiator over so that it's exhausting the warmed air out the case. Heat naturally wants to rise and you are fighting against that in your current configuration. I think it's worth a try, anyway.

 

[Jekerdud]

I've only done 1 hour tests, but I'll do a three hour test.