CM Hyper 212 and a Intel I3. Help me out?

[momosmokey]

Noticed your QPI is at 8800... that is absolutely nuts BTW. Is there a option to bring it under 8000?? Anything over 8000 puts alot of stress on your northbridge as well as IMC of the i3. Id HIGHLY recommend bringing the QPI down before continuing with your OC as i believe this could be part of your stability issues.

 

I simply set it on AUTO. I had looked around about what people thought QPI should be set at and some people did mention keeping it lower. But I assumed AUTO would...well, take care of it; foolish me. Thanks, I'll give it a shot.

 

What i know, you can actually incuce errors by overvolting the ram. to low voltage increases risk of refresh not keeping up, but increasing voltage might make them slower to shift from 1 to 0 (more charge stored up).

Also ram speed doesn't really help performance by much..

 

OK. Is 1.658 Volts OK though? Because that what I needed before to run 7-8-7-20, 1600Mhz. Default voltage is 1.65.

 

So would I see any difference from 1600Mhz 7-8-7-20 vs 1600Mhz 9-9-9-24?

 

 

 

[MercuryDoun]

I simply set it on AUTO. I had looked around about what people thought QPI should be set at and some people did mention keeping it lower. But I assumed AUTO would...well, take care of it; foolish me. Thanks, I'll give it a shot.

 

 

 

OK. Is 1.658 Volts OK though? Because that what I needed before to run 7-8-7-20, 1600Mhz. Default voltage is 1.65.

 

So would I see any difference from 1600Mhz 7-8-7-20 vs 1600Mhz 9-9-9-24?

 

 

 

 

1.658 is fine lol.

 

Personally, i could tell the differences between timings/speed on my RAM doing some testing on speed/timings plus the impact of speed/timings in benchmarks. The biggest thing i noticed is load times were reduced with faster speeds. The effect was quite profound when comparing comparatively low speeds to high speeds. 1050mhz and 9-9-9-27 vs 1730 6-7-6-20 = nearly double load times for 3dmark11, 3dmark vantage, as well as a few games such as Crysis, and SC2. As far as actual gameplay though? I didnt notice is FPS issues with the slower ram, but load times between levels or whatever were much longer with slower speeds.

 

Will you notice the difference only in timings?? Probably not, but faster is faster! So personally id go for it! haha. If you really want to test the difference run a bunch of benchmark programs at 9-9-9-24 and look at the CPU scores as these generally are impacted by memory. 3dMark11 is significantly effected by memory speeds, As well as y-cruncher.

 

 

 

Auto on the QPI usually just means default multiplier, You really ought to see what you can do to get to below 8000MT/s.

 

Also, why arent you using a higher Multiplier?? You can use upto a 24x multiplier with your i3-550, you should be using either 23x or 24x. In my experience with both the 1366 socket and 1156 socket. They both prefer odd numbered multipliers, agian my experience with i7-920's, 930's, 950's, as well as an i3-540 and Xeon X3440(basically a i7-860).

 

With the i3 i mentioned above, i used 23x175 for 4025mhz on the CPU and 1750mhz on the memory, it was 1600mhz memory, but had good overclocking headroom. Looks like you have some decent memory(with heat spreaders so thats good) and according to your sig have gotten up to 1700mhz before, I dont see why you cant push it alittle higher. An overclock like this will keep your QPI down to manageable levels but CPU and memory speed up. Id highly recommend this route. To get the higher clocked memory id start with loose timings such as 9-11-9-27 and something in the 1400mhz area, and get your CPU overclock stable. After your CPU is stable at 4.0Ghz(or higher if you want lol) id push the ram up to 1750mhz, and then work your timings down to whatever you can. I dont see why you cant do 1750mhz at 9-9-9-24 atleast, maybe lower! Just dont forget with faster memory you will likely need more QPI(Northbridge or whatever its called on your MoBo)

 

I suppose you could try a slightly more conservative CPU overclock with 24x160 still nail the 1600mhz on your memory and be over 3.8Ghz still on the CPU.

[momosmokey]

1.658 is fine lol.

 

Personally, i could tell the differences between timings/speed on my RAM doing some testing on speed/timings plus the impact of speed/timings in benchmarks. The biggest thing i noticed is load times were reduced with faster speeds. The effect was quite profound when comparing comparatively low speeds to high speeds. 1050mhz and 9-9-9-27 vs 1730 6-7-6-20 = nearly double load times for 3dmark11, 3dmark vantage, as well as a few games such as Crysis, and SC2. As far as actual gameplay though? I didnt notice is FPS issues with the slower ram, but load times between levels or whatever were much longer with slower speeds.

 

Will you notice the difference only in timings?? Probably not, but faster is faster! So personally id go for it! haha. If you really want to test the difference run a bunch of benchmark programs at 9-9-9-24 and look at the CPU scores as these generally are impacted by memory. 3dMark11 is significantly effected by memory speeds, As well as y-cruncher.

 

 

 

Auto on the QPI usually just means default multiplier, You really ought to see what you can do to get to below 8000MT/s.

 

Also, why arent you using a higher Multiplier?? You can use upto a 24x multiplier with your i3-550, you should be using either 23x or 24x. In my experience with both the 1366 socket and 1156 socket. They both prefer odd numbered multipliers, agian my experience with i7-920's, 930's, 950's, as well as an i3-540 and Xeon X3440(basically a i7-860).

 

With the i3 i mentioned above, i used 23x175 for 4025mhz on the CPU and 1750mhz on the memory, it was 1600mhz memory, but had good overclocking headroom. Looks like you have some decent memory(with heat spreaders so thats good) and according to your sig have gotten up to 1700mhz before, I dont see why you cant push it alittle higher. An overclock like this will keep your QPI down to manageable levels but CPU and memory speed up. Id highly recommend this route. To get the higher clocked memory id start with loose timings such as 9-11-9-27 and something in the 1400mhz area, and get your CPU overclock stable. After your CPU is stable at 4.0Ghz(or higher if you want lol) id push the ram up to 1750mhz, and then work your timings down to whatever you can. I dont see why you cant do 1750mhz at 9-9-9-24 atleast, maybe lower! Just dont forget with faster memory you will likely need more QPI(Northbridge or whatever its called on your MoBo)

 

I suppose you could try a slightly more conservative CPU overclock with 24x160 still nail the 1600mhz on your memory and be over 3.8Ghz still on the CPU.

 

I agree, I did notice a difference between 1600Mhz 9-9-9-24 (stock) vs 1600Mhz 7-8-7-20.

 

I gave up on 200Mhz base clock. It just wasn't stable, even at stock settings with a VTT (QPI/DRAM) of 1.360 volts. Instead, I put it down 1Mhz and it became stable. Apparently odd number do OC better. Right now I'm trying to see how low I can go with VTT voltage to keep it stable.

 

The QPI has several ratios - 22, 20, 18, 16, 14, and 12. At 20 it's 7960Mhz, should I go lower? When should I increase/decrease the QPI freq.?

 

I'm using a 21x multiplier because at 199mhz, it's 4Ghz. When I get everything stable - base clock, RAM, and CPU, I'll up the multiplier. with a 199Myhz base clock and a multiplier of 24x, technically I could achieve almost 4.8Ghz. That's why I'm keeping it at 20/21 for now.

 

When I got my RAM at 1700Mhz before, I forgot to save it and it was reset somehow; however, I remembered all my settings and put them in again. But after that, it just wouldn't stay stable. I'll try again though and let you know.

 

It's rock solid at 160Mhz base clock. That's what I was using previously, but I want a 4Ghz+ CPU!

 

Question:

 

• After I find a stable Base Clock with X number of volts and restore the CPU speed and RAM speed/timings, will/should the Base Clock still be stable? By this I mean: when I achieve a solid Base Clock with X volts, it's been stable for the past 24Hrs (lets say). I input the CPU speed, 199*21 = 4.1/4.2Ghz. I input the RAM speed, 4*199 = 1592 @ 9-9-9-24. The computers unstable, this should not be the Base Clocks fault right? I Should blame either the CPU or RAM, right? Or could the Base Clock stability change after imputing faster speed even after a 24Hr stability pass?

[medbor]

I agree, I did notice a difference between 1600Mhz 9-9-9-24 (stock) vs 1600Mhz 7-8-7-20.

 

I gave up on 200Mhz base clock. It just wasn't stable, even at stock settings with a VTT (QPI/DRAM) of 1.360 volts. Instead, I put it down 1Mhz and it became stable. Apparently odd number do OC better. Right now I'm trying to see how low I can go with VTT voltage to keep it stable.

 

The QPI has several ratios - 22, 20, 18, 16, 14, and 12. At 20 it's 7960Mhz, should I go lower? When should I increase/decrease the QPI freq.?

 

I'm using a 21x multiplier because at 199mhz, it's 4Ghz. When I get everything stable - base clock, RAM, and CPU, I'll up the multiplier. with a 199Myhz base clock and a multiplier of 24x, technically I could achieve almost 4.8Ghz. That's why I'm keeping it at 20/21 for now.

 

When I got my RAM at 1700Mhz before, I forgot to save it and it was reset somehow; however, I remembered all my settings and put them in again. But after that, it just wouldn't stay stable. I'll try again though and let you know.

 

It's rock solid at 160Mhz base clock. That's what I was using previously, but I want a 4Ghz+ CPU!

 

Question:

 

• After I find a stable Base Clock with X number of volts and restore the CPU speed and RAM speed/timings, will/should the Base Clock still be stable? By this I mean: when I achieve a solid Base Clock with X volts, it's been stable for the past 24Hrs (lets say). I input the CPU speed, 199*21 = 4.1/4.2Ghz. I input the RAM speed, 4*199 = 1592 @ 9-9-9-24. The computers unstable, this should not be the Base Clocks fault right? I Should blame either the CPU or RAM, right? Or could the Base Clock stability change after imputing faster speed even after a 24Hr stability pass?

In that case i would blame the CPU (given that base clock by itself is stable and ram is within spec)

 

1. My way of overclocking is to find the maximum frequency of the ram (just to know that later on) (max memory mulitplier, low cpu multi, relatively low base clock) (RAMmax)

2. find max base clock (clock down ram as much as possible and lowest cpu multi) (BASEmax)

3. max out the multi, and slowly increase base clock (keeping ram really slow) to find the maximum CPU frequency (CPUmax)

4. then when cpu is at its max, see what multi you will need versus the max stable base clock (CPUmax/BASEmax gives rough multi, round upwards) (BASEcurr)

5. see if it is stable

6. then find out the ram speed (multi) you can get with that base clock (RAMmax/BASEcurr, round down) (RAMmulti)

7. hope that your board have that multi, otherwise see if another CPU multi can give another ram multi (so that the ram is close to it's full speed)

8. test for stability

9. lower ram timings if possible

10. test for stability

 

This methodology removes most of the variables, so that you can focus on one thing at the time (assuming sufficient multis everywhere)

 

EDIT:

Adding my rigs data:

At cpu multi 6 and ram multi 4 i found my max ram at 980MHz

lowerwed the ram multi to 2.5 and found my base max at 338 (eww!! doesn't even post over 340 )

maxed the cpu multi and found it stable at that base clock (3.05GHz...)

then i found my ram multis are only 2.5/3/4 versus the base clock, and even 3 would be too high (3*338=1014MHz)

so i had to use the lowest ram multi, and thus land on:

CPU 3.05GHz

RAM 845MHz

FSB 338MHz

(i don't remember my memory timings, but i could lower them here quite a bit)

 

My other alternative (to get higher ram speed) would be to lower the base clock to 980/3=~326

and thus land on:

CPU 2.9GHz

RAM 979MHz

FSB 326MHZ

(here only one of the memory timings could be lowered by one small step)

 

But i found in most cases that the first alternative gets better results.

Edited April 13, 2011 by medbor

[MercuryDoun]

 

Question:

 

• After I find a stable Base Clock with X number of volts and restore the CPU speed and RAM speed/timings, will/should the Base Clock still be stable? By this I mean: when I achieve a solid Base Clock with X volts, it's been stable for the past 24Hrs (lets say). I input the CPU speed, 199*21 = 4.1/4.2Ghz. I input the RAM speed, 4*199 = 1592 @ 9-9-9-24. The computers unstable, this should not be the Base Clocks fault right? I Should blame either the CPU or RAM, right? Or could the Base Clock stability change after imputing faster speed even after a 24Hr stability pass?

 

bclk is only part of the equation. So getting a bclk stable then changing multi's changes the end result so getting a certain bclk stable does not guarantee anything. Changes are with a very slight bclk increase 1-2 no voltage chance will be required, but eventually even 1mhz on the bclk is enough to require addition voltage, multiplier changes are most likely to need voltage changes, not always, like if you have alot of head room on the current voltage settings. BTW base clock IS the CPU, heh, so yes, its stability is something that changes after increasing speed.

 

You should use a much lower QPI multiplier to get to down to more reasonable levels, stock is 4800MT/s, so running what you running is QUITE the overclock on it. Normally i go with anything in between 6000-7000MT/s as this is a significant overclock still but doesnt require anything major to get working smoothly, Id change it to the 18 or 16 mutli. But if you wanna keep struggling to make it stable then by all means have at it. Unstable QPI can lead to unstable RAM though, it could possibly why your using issues returning your ram to speeds that are previously known to work. If nothing else, more QPI voltage is quite useful when reaching past the 1600mhz area on memory.

 

Why dont you try 167x24(4008Mhz)? Use slow ram speeds anything less than 1600 would be fine, and get the CPU 100% stable first. Then try to get your ram stable at 1670mhz 9-9-9-24, start with 9-11-9-27 timings. Then slowly work it down to whatever timings you desire. Addition QPI voltage may be required, but DRAM voltage should be perfectly fine at 1.65X.

[momosmokey]

bclk is only part of the equation. So getting a bclk stable then changing multi's changes the end result so getting a certain bclk stable does not guarantee anything. Changes are with a very slight bclk increase 1-2 no voltage chance will be required, but eventually even 1mhz on the bclk is enough to require addition voltage, multiplier changes are most likely to need voltage changes, not always, like if you have alot of head room on the current voltage settings. BTW base clock IS the CPU, heh, so yes, its stability is something that changes after increasing speed.

 

You should use a much lower QPI multiplier to get to down to more reasonable levels, stock is 4800MT/s, so running what you running is QUITE the overclock on it. Normally i go with anything in between 6000-7000MT/s as this is a significant overclock still but doesnt require anything major to get working smoothly, Id change it to the 18 or 16 mutli. But if you wanna keep struggling to make it stable then by all means have at it. Unstable QPI can lead to unstable RAM though, it could possibly why your using issues returning your ram to speeds that are previously known to work. If nothing else, more QPI voltage is quite useful when reaching past the 1600mhz area on memory.

 

Why dont you try 167x24(4008Mhz)? Use slow ram speeds anything less than 1600 would be fine, and get the CPU 100% stable first. Then try to get your ram stable at 1670mhz 9-9-9-24, start with 9-11-9-27 timings. Then slowly work it down to whatever timings you desire. Addition QPI voltage may be required, but DRAM voltage should be perfectly fine at 1.65X.

 

I'll give it a shot and report back. Thanks for the help.

 

 

 

[momosmokey]

In that case i would blame the CPU (given that base clock by itself is stable and ram is within spec)

 

1. My way of overclocking is to find the maximum frequency of the ram (just to know that later on) (max memory mulitplier, low cpu multi, relatively low base clock) (RAMmax)

2. find max base clock (clock down ram as much as possible and lowest cpu multi) (BASEmax)

3. max out the multi, and slowly increase base clock (keeping ram really slow) to find the maximum CPU frequency (CPUmax)

4. then when cpu is at its max, see what multi you will need versus the max stable base clock (CPUmax/BASEmax gives rough multi, round upwards) (BASEcurr)

5. see if it is stable

6. then find out the ram speed (multi) you can get with that base clock (RAMmax/BASEcurr, round down) (RAMmulti)

7. hope that your board have that multi, otherwise see if another CPU multi can give another ram multi (so that the ram is close to it's full speed)

8. test for stability

9. lower ram timings if possible

10. test for stability

 

This methodology removes most of the variables, so that you can focus on one thing at the time (assuming sufficient multis everywhere)

 

EDIT:

Adding my rigs data:

At cpu multi 6 and ram multi 4 i found my max ram at 980MHz

lowerwed the ram multi to 2.5 and found my base max at 338 (eww!! doesn't even post over 340 )

maxed the cpu multi and found it stable at that base clock (3.05GHz...)

then i found my ram multis are only 2.5/3/4 versus the base clock, and even 3 would be too high (3*338=1014MHz)

so i had to use the lowest ram multi, and thus land on:

CPU 3.05GHz

RAM 845MHz

FSB 338MHz

(i don't remember my memory timings, but i could lower them here quite a bit)

 

My other alternative (to get higher ram speed) would be to lower the base clock to 980/3=~326

and thus land on:

CPU 2.9GHz

RAM 979MHz

FSB 326MHZ

(here only one of the memory timings could be lowered by one small step)

 

But i found in most cases that the first alternative gets better results.

 

I'll follow your advice as well.

 

Thanks, all you guys are giving VERY helpful info!

Edited April 14, 2011 by xxmastermushxx

[momosmokey]

OK, here's an update. First off, the advice that I was giving by both medbor and Mercury Duon was what got me started in the right direction. So without further ado, my awesome overclock!

 

This is right after a successful run of IBT, 50 passes, maxiumum. The voltage for the VCore when under load is actually 1.296V.

 

and my RAM.

 

I tried 1750Mhz RAM, but that limited me to 9-8-9-24 with a 175 BCLK and a CPU max of 4200Mhz. While that's nothing to sneeze at, I like my 4500Mhz.

 

So, I now have one question. When will my CPU die?

As you can tell by Realtemp, the max it got to was 64c...I love my CM 212+...Most people said the max voltage is 1.4-1.45, and 1.3-1.35 is safe. My motherboard didn't recommended a voltage over 1.3V. Is that just to cover their buns in case some idiot overvolts the CPU (POP) and then blames the motherboard manufacture?

 

Thanks.

[Locutus]

EDIT: Hah, late advice is late.

 

Nice OC you have there. 64C is a bit high (or at least it would be for me, it really depends on ambient temperature.) Either way, as long as your voltage isn't too far out of line you shouldn't really see much of a decrease in the lifetime of your parts.

Edited April 20, 2011 by Locutus

[momosmokey]

EDIT: Hah, late advice is late.

 

Nice OC you have there. 64C is a bit high (or at least it would be for me, it really depends on ambient temperature.) Either way, as long as your voltage isn't too far out of line you shouldn't really see much of a decrease in the lifetime of your parts.

 

OK, cool. I played a little Crysis (on my IGPU ) and the hottest the CPU got was 40c. Which in my eyes is great! I mean thats...40/5 = 8 * 9 = 72 + 32 = 104 F. That's like a hot tub.

Edited April 20, 2011 by xxmastermushxx

[MercuryDoun]

Nice OC! I wouldnt have expected 4.5Ghz to be only 64c!! My understanding of those little 32nm clarkdales is your completely fine at 64c!! I dont think you have to worry about anything until your in the mid/high 70's 24/7. Which your not! If you got bored and wanted to push the CPU extra hard to try and get 5.0Ghz for a benchmark run or bragging rights. You should be fine to go into the mid-high 70's or even 80-something for short periods of time, but this of course is not totally safe. But alot of people do it, and do it often!(Me lol)

 

I wish my i7 ran at 64c when at 4.5Ghz... When my 930 is at 4.5Ghz it idles at like 54c and under lighter loads such as wPrime, it gets to like 80c in just a few seconds. I ran Prime95 on it for about 10 minutes and it got up to 92c so i stopped it. This is with a heatsink that puts the Hyper 212+ to shame too! You got yourself a NICE chip i think! Dual-core or not, your gonna be enjoying some good FPS in lots of games!

 

Again NICE job on that OC man!

[momosmokey]

Nice OC! I wouldnt have expected 4.5Ghz to be only 64c!! My understanding of those little 32nm clarkdales is your completely fine at 64c!! I dont think you have to worry about anything until your in the mid/high 70's 24/7. Which your not! If you got bored and wanted to push the CPU extra hard to try and get 5.0Ghz for a benchmark run or bragging rights. You should be fine to go into the mid-high 70's or even 80-something for short periods of time, but this of course is not totally safe. But alot of people do it, and do it often!(Me lol)

 

I wish my i7 ran at 64c when at 4.5Ghz... When my 930 is at 4.5Ghz it idles at like 54c and under lighter loads such as wPrime, it gets to like 80c in just a few seconds. I ran Prime95 on it for about 10 minutes and it got up to 92c so i stopped it. This is with a heatsink that puts the Hyper 212+ to shame too! You got yourself a NICE chip i think! Dual-core or not, your gonna be enjoying some good FPS in lots of games!

 

Again NICE job on that OC man!

 

Thanks for the complements. I was pretty amazed I got to 4.5Ghz as well and with such a low voltage. From what I saw, when researching i3 overclocking and the like, people had to use 60 or so more millivolts than what I used to the achieve stability. While I'm rocking a max of 1.328V idle and 1.296V load, their putting 1.386V and more into their I3's to achieve a stable 4.5Ghz OC. Guess I got a lucky chip.

 

I'd love to try and push this thing further. But because I'm just 16 and don't hold a job that could replace this chip if it decides to *POP*, I'm going to have to hold myself back. For now. I'd probably need a liquid cooling system or a Noctua NH-D14 to achieve a 5Ghz OC anyway, at least effectively. When upping the voltage from 4.4Ghz to 4.5Ghz, I had to increase the voltage 2-3 times more than when I went from 4.2Ghz to 4.4Ghz. So going to 5Ghz would mean I'd probably need 1.4-1.45 volts - 70-80C. I'll think about it, but for now, I'm happy.

 

54C idle for an i7 930! Wow, that's like the average tempt I max out at. I usually idle at around 25c with this 4.5Ghz OC, which is room tempt. Man, I'm loving and respecting this chip more and more.

 

Your advice helped a lot and I probably wouldn't have gotten a 4.5Ghz OC without it. I had no idea the QPI frequency was contributing to my stability issues. So here's a cheers emoticon for your help.

Edited April 20, 2011 by xxmastermushxx