My i5 760 Overclocking adventure!?

[Dward]

Haven't messed with mine since November but I think our mobos allow for +0.00625v increments which is great (but ridiculous). I believe I increased 2 steps at a time...so +0.0125v increments.

Once I got to 3.8ghz, my PC was able to pass LinX 100 passes and 3hrs OCCT High Data Set using +0.075v to dynamic vcore but Core 3 failed on Prime 95 around 9hrs. So I bumped just 1 increment at that point for a fine tune to +0.08125v and reran Prime95 and passed 14 hrs on all cores no problem.

As you can see, all tests (LinX, OCCT, and Prime95) are useful in their own way at exploiting your overclock's weaknesses if any.

Ok much appreciated! I'm excited!

I'm up to 200x15. OCCT passed 13 hours w/out errors, will do a Prime95 test over night. Then on to x16

I'm spewing, I'm going away on holidays for 8 days in two days time. I will be itching to get back and continue to push the multiplier lol

 

[Dward]

200x16 has passed OCCT, LinX & Prime without any CPU voltage increase! Will leave it at that until I get back from my holiday.

 

[brencat]

200x16 has passed OCCT, LinX & Prime without any CPU voltage increase! Will leave it at that until I get back from my holiday.

Glad everything is going well for you. I'll bet you'll do 200 x 17 without a vcore increase either. It's really 18x and above that they start needing more.

Enjoy your vacation...thanks for checking in.

 

[Dward]

I should be packing right about now, but instead I'm increasing the multiplier lol.

I've upped it to 200x17. I tried some LinPack only to get BSOD. So I increased the Dynamic Voltage by two steps, so +0.0125V. It then passed 20 passes of LinPack without any errors with a max temp on core 1 of 51C. Will try OCCT now.

UPDATE: Ran 3DMark11, it locked up big time half way through. Maybe I'll leave this till I get back home!!

 

[snerism]

I should be packing right about now, but instead I'm increasing the multiplier lol.

I've upped it to 200x17. I tried some LinPack only to get BSOD. So I increased the Dynamic Voltage by two steps, so +0.0125V. It then passed 20 passes of LinPack without any errors with a max temp on core 1 of 51C. Will try OCCT now.

UPDATE: Ran 3DMark11, it locked up big time half way through. Maybe I'll leave this till I get back home!!

have you tried Dynamic Vcore = "Auto" and QPI/VTT Voltage = "1.250v" ?

coz i've recently discovered if you leave auto QPI/VTT Voltage = "Auto" CPU temp will drastically rise.

 

[Dward]

Snerism: I don't want to set any voltages to AUTO as I find this makes them too high than what they need to be. I have manually set my QPU/VTT Voltage to 1.250v already.

Ok on to some more results...

200x18 was stable with a +0.060v Dynamic Voltage increase. This meant a load Core Voltage of 1.072v and core temps of 58, 56, 56 & 54. (during stress testing)

200x19 needed a few more volts. To get stable I needed to increase the Dynamic Voltage to +0.0875v. This meant a load core voltage of 1.136v and core temps of 63, 61, 60 & 60.

Two things,

1) Very warm ambient temp of 28 in my room today. What do you think of these voltages and temps?

2) I am leaving it at 3.8Ghz at the moment. I will continue to run more stress tests over the next few days at this speed...before trying for 4.0Ghz!!

Cheers

 

[brencat]

Snerism: I don't want to set any voltages to AUTO as I find this makes them too high than what they need to be. I have manually set my QPU/VTT Voltage to 1.250v already.

Ok on to some more results...

200x18 was stable with a +0.060v Dynamic Voltage increase. This meant a load Core Voltage of 1.072v and core temps of 58, 56, 56 & 54. (during stress testing)

200x19 needed a few more volts. To get stable I needed to increase the Dynamic Voltage to +0.0875v. This meant a load core voltage of 1.136v and core temps of 63, 61, 60 & 60.

Two things,

1) Very warm ambient temp of 28 in my room today. What do you think of these voltages and temps?

2) I am leaving it at 3.8Ghz at the moment. I will continue to run more stress tests over the next few days at this speed...before trying for 4.0Ghz!!

Cheers

Great job. Hope you had a nice vacation. Your chip appears very similar to mine in terms of voltage needed for that speed while running only modestly warmer.

Excellent work :thumbsup:

 

[Dward]

Thanks Brencat.

UPDATE: computer crashed while running Heaven Benchmark. I've upped the Dynamic Voltage to +0.0100v. Heaven does not crash now

Temps have gone up to 65, 63, 62 & 61 now with the voltage increase during LinX. However CPU-Z is reading the same Core Voltage as before... 1.136v. Is that normal?

I have double checked in the bios that the Dynamic Voltage is still +0.100v and it is.

Cheers

Here is a 20 pass run with LinX:

 

[Dward]

Dam just had Prime fail on me:

I have bumped the Dynamic Voltage up two notches to +0.1125v.

...on to more testing!!

 

[brencat]

Ya...see that. Core #3 failed round off checking ~ 4 hrs mark. Temps under Prime look good. LinX temps will always be higher. You're doing great. Thanks for checking in.

B

 

[coffeejunkee]

I like this thread, someone who is kinda new to overclocking and actually uses his brain and follows good advice. This gives me new hope for humanity (well, almost).

But your i5 760's seem quite a lot better than my i5 750. I need 1.312 cpu-z vcore to succesfully finish 20 runs Linx at 200x19, qpi/vtt 1.25, eist/c1e/c-states enabled (1.21 qpi/vtt if c-states disabled). Temps still good, 70 degrees Linx. 4GHz needs even more, 1.39V and is impossible without LLC. Energy breakpoint was around 3.6GHz, anything above that required exponentially higher power. Running at 166x20 myself, my chip can do that on 'normal' vcore and vtt.

I've noticed the same in cpu-z, apparently it's not precise enough to notice the 0.0625V increase. Nothing to worry about.

 

[Dward]

Thanks Bren and coffeejunkee! I'm on my dads computer at the moment, mine is running Prime again for a LONG time, so far 8 hours have been run with out any errors

Is 1.152 on the vcore low for this cpu for the speed I'm running it at? (3.8Ghz)

Hmmm just looking at your sig Bren I see vcore is even lower! What temps do you get with Prime/LinX?

Cheers

 

[brencat]

Is 1.152 on the vcore low for this cpu for the speed I'm running it at? (3.8Ghz)

Hmmm just looking at your sig Bren I see vcore is even lower! What temps do you get with Prime/LinX?

Cheers

The image I linked in Post #2 of this thread has my temps. Prime & OCCT peak temps were about the same... 60-58-56-57. For LinX, add 7c to each of those numbers so basically inline with what you are seeing.

Gaming by comparison (like after about 3 straight hours of BFBC2) only gets CPU temps to 55-54-53-53c.

I think you are right there in the sweet spot. I fully expect you to pass 12 hrs Prime95 with no round-off checking errors using your current dynamic vcore offset.

You are welcome to try for 4ghz. I think it will surprise you just how much extra vcore is needed. For me as I've mentioned before, even with a very good 92mm cooler I'm using, I need to leave extra margin for summer temps in the room where my PC is so I am staying at 3.8ghz. PC runs great!

 

[yottabit]

I know it's terrible practice and I'm a bad influence, but I had been using my it 750 @ 3.2 Ghz stock volts for the longest time when I got bored and decided to "guess" and chose 1.25 v and 3.8 Ghz (@21x multiplier, whatever that BCLK works out to be).... worked fine in one shot and it has proved to be stable. IMO if you are lazy doing it this way is better than using those "auto overclocking" utilities. I looked at what my ASUS wanted to do for it's 3.8 Ghz overclock- 1.4 vcore!! With a Hyper 212 it hits 58 deg C under Prime95 torture test, no biggie

 

[Dward]

Cheers guys.

Yeah Bren, I'm not keen on going for 4.0Ghz. On the summer days, the extra voltage/temps would not be worth the .2Ghz increase!

Just passed ~13 hours of Prime pics of full load & then idle:

The only thing I'm concerned about was last night when I was playing BC2, I had a graphics glitch on my minimap. I'm going to load it up again to see if it happens again.

 

[Dward]

Did not see the graphical glitch just then on a couple rounds. Will keep an eye out for it tho. So...

I THINK I'M DONE!!! Just one last (long) stress test to go, OCCT.

Huge thanks to everyone, especially brencat. What a wonderful result, 1.8Ghz during windows @ ~30C. Then a whopping 3.8Ghz @ ~55-60C for gaming!

 

[brencat]

Is that 'graphical glitch' in BFBC2 by any chance a completely blank screen for a sec or two (but sounds are fine, and game continues otherwise) until you move around to a different spot? If so, it happens to me too. I think it's the game. Buggy in places. I have an older GTX 260-216 and I see you have the GTX 460. Could be nvidia drivers but I strongly suspect it's the game.

Anyway, nice work on the overclock.

 

[Dward]

Nah a smaller glitch than that bren. You know how when you are on the flag radius and the small bars fill up blue, the other night a couple times we had the flag and I was on the radius...but some of the bars were not colored blue, they were blank. Never seen that before.

OCCT is still running, no errors!! Think I am all good

 

[brencat]

OCCT is still running, no errors!! Think I am all good

Yep, I think so too... 3 hrs or so on Large Data Set setting should be sufficient for OCCT.

I love a happy ending... take care, B

 

[Dward]

Thanks.

p.s. when are you going to upgrade your GPU?!

 

[brencat]

Thanks.

p.s. when are you going to upgrade your GPU?!

Ha...check my sig again in a few days. Got a 2gb 6950 on the way :biggrin:

 

[Dward]

Ha...check my sig again in a few days. Got a 2gb 6950 on the way :biggrin:

Nooooooooo! Get a GTX 560!

http://www.tweaktown.com/reviews/38...tx_560_ti_golden_sample_video_card/index.html

 

[brencat]

Haven't seen a Gainward card in the U.S. in a long time (maybe I haven't been looking that hard). Regardless, did you also notice the peak power draw of your 'golden sample' GTX 560?.. 395 watts! Compared with only 297 watts for the 2gb 6950 at stock settings. I won't be flashing to 6970 and will probably only do a modest overclock to 880/1325 from 800/1280 base.

I'm NOT buying a new power supply either, which is another reason I opted for the 6950 over the nvidia alternatives.

Edit: Just noticed also that TT used the older Cat 10.10 drivers instead of the 11.1s (article was likely written before their 1/26/11 release) but still those were older than most current regardless. I smell attempted bias.

 

[Dward]

Hmmmm yeah ok. Dam, do you think I would be alright with my HX 650 (http://www.techenclave.com/pc-peripherals/corsair-hx650-146578.html) if I bought a GTX 560.

I would not buy a stock one, I was looking at the Gainward Golden Sample

 

[coffeejunkee]

I know it's terrible practice and I'm a bad influence, but I had been using my it 750 @ 3.2 Ghz stock volts for the longest time when I got bored and decided to "guess" and chose 1.25 v and 3.8 Ghz (@21x multiplier, whatever that BCLK works out to be).... worked fine in one shot and it has proved to be stable. IMO if you are lazy doing it this way is better than using those "auto overclocking" utilities. I looked at what my ASUS wanted to do for it's 3.8 Ghz overclock- 1.4 vcore!! With a Hyper 212 it hits 58 deg C under Prime95 torture test, no biggie

Yes keep your evil ways out of here! No srsly, nothing wrong with what you are doing. If it works for you, fine. As long as you understand what's going on.

My experience with auto vcore is the same, it will give way more than necessary at higher overclocks but it's fine for lower oc's. Fex. I have another system with an E4300. Overclocked to 3GHz auto vcore gives like 1.45V if not more which is way more than necessary. But at 2.4GHz it just gives a little more than really needed. It's an old system so no dynamic vcore and only way to manually set vcore is using fixed value with higher idle temps as result. So I use auto for the moderate oc with eist enabled. But I always wonder when people say: "my cpu can overclock to 4GHz on stock voltages!" if they're actually thinking 'auto' is stock.

As for automatic overclocking features on mobo's, I don't understand the attraction. The people who use these are apparently afraid to mess around in the bios themselves, so they let the mobo mess around in much worse ways. Laziness is not an excuse, it takes about half an hour to read an oc guide and get started. Hey, even I can do it.