Newbie Help - E8400, P5q-e

[Cannan]

Hi all;

 

I'm very new to overclocking so I thought I had better post what I was doing to make sure I am on the right track. On a side note, I have read a lot of posts on this forum and its great to see so many people willing to help.

 

My machine was assembled for me when I bought it, and is currently running stock:

 

Processor: E8400 Dual Core (E0 stepping)

Mobo: ASUS P5Q-E

RAM: Mushkin Ascent PC2-8000 4GB (2x2GB) DDR2-1000 CL5-5-5-12

Case: Coolermaster HAF-932

Power: Corsair TX750W

Cooling: Stock

 

I have ordered an Xigmatek HDT-S1283 heatsink with some AC MX-2 Thermal compound, so I won't be on stock cooling for long. My goal is to moderately overclock my machine under stock cooling, and then higher when I get the new heatsink in place. Gotta walk before I run though, so I'll start with seeing how what I can do with with stock cooling.

 

I installed CPU-Z, CoreTemp, RealTemp, Speedfan and Prime95 to get a feel for things. With stock cooling, my CPU cores are running around 38C idle and 55C under load. Realtemp reports 5C cooler overall, but I am playing it safe and going with the higher numbers.

 

The first thing I did was pick up the recommended timings for my RAM from the Mushkin forums (here: http://forums.mushkin.com/phpbb2/viewtopic...136&p=46826) For some reason my DRAM Voltage was set to 2.20v, so I turned that down to 2.10v to meet Mushkin's recommendation. This brings me to my first question: the timings on the Mushkin forums are listed in this manner "Write Recovery Time: 6(4-6)" with the first number being the starting point and the number in the brackets being what you should optimize to. Is optimizing the timings that critical? And how would one ensure they are optimized?

 

With the timings in place I am ready to try to overclock the CPU. If I have understood all that I have read (and there's no guarantee of that!) then I understand I should do the following:

 

- Disable C1E and SpeedStep

- Set AI overclock tuner to "Manual"

- Disable CPU Spread Spectrum

- Ensure PCIE Frequency is set to 100MHz

- Set CPU Voltage to 1.25

- Increase FSB Frequency by a small amount (NOTE: Increasing FSB in the ASUS BIOS automatically jumps the DRAM Frequency to a higher setting, so ensure DRAM Frequency is set to correct specifications according to the RAM. The available values will change as FSB changes).

- Boot to windows and run Prime95 and monitor the temperatures

- If after several hours Prime95 displays no errors, then raise the FSB more

- Once I find a satisfactory FSB, then start reducing CPU Voltage until no longer stable to reduce heat (then bring it back to where it was last stable).

- Then repeat it all again when I get the new heatsink

 

Does all of this sound correct? If I have missed something, please let me know - I'd really appreciate it!

 

Sorry for the long post - hopefully this thread will be of use to those with similar hardware.

 

-Cannan

Edited February 3, 2009 by Cannan

[Turophiliac]

it looks like you have everything under control, your overclocking should go smoothly with the steps you wrote. one thing i'd suggest is making sure you have the latest version of RealTemp, 2.84. I believe it's in beta right now though. it gives the most accurate temp for the e8400. if you can't find it, just set the tjmax to 100C and you should be fine with any temperature program. also, i'd suggest not overclocking without the aftermarket cooler, it's just not worth it... especially if you're getting 55C temps load on stock, you really shouldn't go a whole lot higher with the stock cooler.

Edited January 24, 2009 by turophiliac

[Krieg1337]

45nm TJMax is 95, not 100.

[Turophiliac]

45nm TJMax is 95, not 100.

some of the unlocked series ones are 95C, but the e7*00 and e8*00 series are officially released as 100C.

source

New Realtemp beta with correct tjmax of 100C

[werty316]

http://intel.wingateweb.com/taiwan08/publi...02_Nov_1006.pdf

Goto page 14.

[Cannan]

some of the unlocked series ones are 95C, but the e7*00 and e8*00 series are officially released as 100C.

source

New Realtemp beta with correct tjmax of 100C

 

 

Thanks for the feedback - I picked up the 2.84 beta and the readings are more inline with the other temperature programs I am using. I think I will wait until my new heatsink comes in before doing anything serious.

For now, I'll turn everything back to default EXCEPT the Ram timings (including the DRAM voltage). Any point in trying to change my RAM timings to the 'optimal' timings listed on the Mushkin forums (link above)? I am not sure how I can test what optimal is On the subject of RAM, should I be trying to set the DRAM frequency to its highest rated setting (around 1000 for my Ascents) or should I be trying to keep a 1:1 ratio with FSB?

 

Finally, when I do get the new heatsink, I am thinking I should try to keep temps (when fully overclocked) in the mid-60s at the very highest under load, that sound about right?

 

Thanks!

[Turophiliac]

I try to keep my temps maxed out at 60C, but that's just me, I wouldn't worry about maxing out at 65C but if you are going to be folding or anything, i'd keep it at 60. this is my opinion, other people have different "max safe" temps, so its you're decision.

[ebarone]

IMO, memory ratios are worthless. I've run my system at all kinds of different ratios and its never affected anything. So for memory, I would say you should try to max your mem frequency. Once you find your top frequency, see how tight you can get your timings. You can even pump more voltage in (but not much more!!!) to try and get faster speeds. You probably wont be able to go much higher than 1100MHz on the memory, but yea... if you can set that independent of the FSB, just max it out. I think my ratio right now is 10:9 or something close, and I have zero problems.

 

As far as your timings go, for the most part they are not amazingly critical. CAS 6 isnt noticeably slower than CAS 5 for example. That said, I would follow the optimization recommendations from that forum, because, well, why not? There are a couple memory timings that do make noticable impacts to performance, such as the Common Performance Number listed. On my motherboard its called Static tRead Delay, and going from 8 to 6 can in fact make a solid difference.

 

To find out if they're optimized, unfortunately, you'll have to change each setting one at a time and ensure stability =/ Time consuming, I know. To speed that process along, I'd say burn a memtest86 bootable CD and run that, a few passes should be a good indicator of stability and not take more than a couple mins. At least it beats booting to windows anyways haha.

 

You should optimize your memory frequency before you do your timings though, because imo frequency >>> timings.

 

Good luck, let us know how it all goes When is your Xigmatek due to arrive?

[Cannan]

IMO, memory ratios are worthless. I've run my system at all kinds of different ratios and its never affected anything. So for memory, I would say you should try to max your mem frequency. Once you find your top frequency, see how tight you can get your timings. You can even pump more voltage in (but not much more!!!) to try and get faster speeds. You probably wont be able to go much higher than 1100MHz on the memory, but yea... if you can set that independent of the FSB, just max it out. I think my ratio right now is 10:9 or something close, and I have zero problems.

 

As far as your timings go, for the most part they are not amazingly critical. CAS 6 isnt noticeably slower than CAS 5 for example. That said, I would follow the optimization recommendations from that forum, because, well, why not? There are a couple memory timings that do make noticable impacts to performance, such as the Common Performance Number listed. On my motherboard its called Static tRead Delay, and going from 8 to 6 can in fact make a solid difference.

 

To find out if they're optimized, unfortunately, you'll have to change each setting one at a time and ensure stability =/ Time consuming, I know. To speed that process along, I'd say burn a memtest86 bootable CD and run that, a few passes should be a good indicator of stability and not take more than a couple mins. At least it beats booting to windows anyways haha.

 

You should optimize your memory frequency before you do your timings though, because imo frequency >>> timings.

 

Good luck, let us know how it all goes When is your Xigmatek due to arrive?

 

 

Thanks! It looks like my Xigmatek will arrive in the next day or so, but I'll have to wait until the weekend to install it - work has a habit of getting in the way

In the meantime I'll pick up memTest and give that a try. I'm also debating whether I should buy a copy of Everest Advanced - so far it seems like a great utility.

 

I'll keep updating this thread as I go!

[ebarone]

Yea, doesnt RL jut suck

[Cannan]

Yea, doesnt RL jut suck

 

 

Sure does!

 

A quick update: I managed to find some time over the weekend to install my Xigmatek HDT-S1283 (using the Xigmatek ACK-I7361 Crossbow Mounting Kit and AC MX-2 cooling paste) and what a difference! I'm now running 33C idle and 42C under load - a good 13C difference under load. The installation was fairly easy too. The HAF-932's hole in the back-plate means I could set the heatsink without taking the mobo out, and the mounting kit helped a lot as well. Whole process was about 90 minutes - 2 hours if you count the 30 minutes I spent patting myself on the back

 

I hope to start the OCing process this week - I'll let you know how it goes!

[ebarone]

Thats fantastic! A quick tip that I found after getting my HDT set up, the fan moves quite a bit of air. I found I got better temps by about 2-4 degrees by ensuring the the exhaust fan on my case, which is about an inch away from the back of my cpu cooler, could keep up with the Xigmatek fan. Once I swapped out the old stock case fan with a higher CFM fan, it improved my airflow like whoa, and helped my temps. Good luck with the OCing!