Intel Core i5 & i7 Sandy Bridge Overclocking and Feedback

[Chicken Patty]

Just use the voltage your RAM is specified for, I use 1.65v for mine since my Hyper X is supposed to run at 1.65v.

I thought you couldn't run higher than like 1.55v for these SB's...

 

[sneekypeet]

I don't know about "can't", but I don't know that I am not damaging something doing so. So far no issues with multiple kits at 1.65V.

 

[cadaveca]

I thought you couldn't run higher than like 1.55v for these SB's...

Many memory OEMs sell 1.65v SKT1155 ram. Should be fine. Going over that is where it gets tricky.

 

[N-Gen]

Gonna clear my cmos and start over. Most of my BSODs I found out just a couple of days ago to be my SSD, the new firmware has it sorted though, so off to a new start!

 

[Chicken Patty]

Gonna clear my cmos and start over. Most of my BSODs I found out just a couple of days ago to be my SSD, the new firmware has it sorted though, so off to a new start!

I spent a year like that. Finally threw it in my closet until like two weeks ago. New firmware did the trick. . I am in love with my PC all over again.


Thanks for all the replies on the DRAM voltage questions guys.

 

[N-Gen]

Mine's about a month old now, was driving me nuts.

Anyway, had some VTT BSODs, bumped that up a bit, let's see how it goes. Funny how IBT and Prime don't crash it as fast as WCG does.

 

[Chicken Patty]

Yeah, I went nuts for about a year. It was horrible. But it's running sweet now.

...and yeah, I let WCG do my stress testing.

 

[N-Gen]

It passed a 5 pass IBT test, I put on Prime for 10, and the put on WCG @60% usage and it took it less than 5 minutes. Then I let WCG on during the night and I wake up to an error code linking to the VTT. I'm running it as we speak and it seems to be fine now.

Has anyone used the Turbo Additional Voltage option instead of offset? Because, correct me if I'm mistaken, but if I offset say, +0.050, that means the range of the lower and higher voltage goes up by 0.050, right? I tried adding stability by setting offset at -0.005 so that Auto doesn't mess up, still thing it would go higher, and I gave the Turbo Additional Voltage +0.35v.

 

[LAN_deRf_HA]

Depends on CPU..maybe board too..I cannot get lower than 1.725v.

At any speed or just overclocking? The last two 2500k systems I built I put it down to 1.67v and there were no issues, this on asrock pro3s, but they were only set for 4 Ghz.

 

[Darkleoco]

I think I have to conclude I have a crappy 2600K :/ the difference between 4.4 and 4.5 on this chip just seems to be too great, I run at around 70+ degrees on 4.5 Ghz and 1.39v which still wasn't stable but at 1.37v I am running 4.4Ghz at under 60 degrees on air during a Prime95 Blend test. And according to CPU-Z I'm not even running over 1.3v for 4.4 Ghz :/

Any opinions on what is causing the large difference?

 

[N-Gen]

There's the same difference on my 2500k and both my mates 2500k, above 4.4 it seems like it wants to take in too many volts.

 

[Darkleoco]

I'm just going to let it sit at 4.4 Ghz then, if its easier to stabilize and runs that much cooler its not worth the extra 100 Mhz when I can clock my 6950's higher than 6970's for gaming :/

 

[N-Gen]

I left it at 4.4 myself, ambient temps in my room aren't really pretty and stepping from 4.4 to 4.5 was adding +10C.

 

[Darkleoco]

Damn it had my computer freeze after 63 minutes on Prime95 Blend :/ gonna try clocking the memory back down to 1333 because if I remember right 1333 to 1600 is what started this but im not sure ugh

Pretty sure my RAM just needs tweaking to run at 1600 with my OC because I just set it to 1333 and when Prime95 did an almost instant BSOD it was 0x101 which means I need more vcore so I am back at 1.39v trying 4.4 to see if I can have it stable with my memory at 1333 and then work from their, though I do feel like my chip requires rather insane voltages :/

 

[Chicken Patty]

At any speed or just overclocking? The last two 2500k systems I built I put it down to 1.67v and there were no issues, this on asrock pro3s, but they were only set for 4 Ghz.

I tried going down to 1.67v an I didn't' have a BSOD but I kept getting random programs and services quitting unexpectedly, something was definitely up. So I reverted back to 1.77v for now, I'll lower it little by little from here and see.

Damn it had my computer freeze after 63 minutes on Prime95 Blend :/ gonna try clocking the memory back down to 1333 because if I remember right 1333 to 1600 is what started this but im not sure ugh

Pretty sure my RAM just needs tweaking to run at 1600 with my OC because I just set it to 1333 and when Prime95 did an almost instant BSOD it was 0x101 which means I need more vcore so I am back at 1.39v trying 4.4 to see if I can have it stable with my memory at 1333 and then work from their, though I do feel like my chip requires rather insane voltages :/

Bro, try clocking your CPU down a bit to 4GHz or something you are sure it's stable. Once it is, then clock your memory and work on getting that stable. If you do both at once, how do you know what is causing your issues you know? Just do one thing at a time, in whatever order you wish.

 

[Darkleoco]

Bro, try clocking your CPU down a bit to 4GHz or something you are sure it's stable. Once it is, then clock your memory and work on getting that stable. If you do both at once, how do you know what is causing your issues you know? Just do one thing at a time, in whatever order you wish.

If my current test fails thats what I am going to do, taking back the memory to 1333 seems to have done it for now since before trying the memory at 1600 I had never had a freeze and only had voltage related BSOD currently on my longest stable Prime95 run with 4.4GHz and 1.39v and the memory staying at 1333 so I think I have at least isolated the freezing to the memory needing a little bit of tweaking to run correctly at 1600 with my OC, thanks for the advice and hopefully I get this CPU tamed

 

[LAN_deRf_HA]

I also find a big step from 4.4 to 4.5, so I stuck with the former. For 4.5 I had ramp up just about every setting, not just vcore.

As for doing the memory separate it may not narrow things down as upping the core speed later can make previously stable memory clocks unstable. The memory is very dependent on the cpu for stability on 1155.

 

[Darkleoco]

If the memory does turn out to be an issue I might just keep it clocked at 1333 since I doubt I would notice any performance hit from it.

4 hours+ on Blend and still stable with 4.4Ghz, 1.39v and memory clocked at 1333

 

[cadaveca]

The problem with 1155 clocking is that increases in CPU speed also increase cache speed, which in turn affects system memory bandwdith.


So, memory that works @ 1333 MHz with the CPU @ stock, might not when the CPU speed is increased.

But, it might also be the cache that is failing...and not the core proper itself(Core, L1+L2).


That's why I like to clock everything up, and then use programs to test specific workloads that stress certain parts of the system, and then scrutinize the failures for what to change.

 

[Darkleoco]

Since I think I am likely to be stable now since im 20 minutes away from 5 hours on Blend, can you recommend some programs to test other parts of the system?

 

[cadaveca]

I did before...that's all I ever need, is LinPack, SuperPi, Prime95, AIDA64, and a game or three.

Some people like OCCT...I'm not a fan myself, but you can use it too. I prefer the separate tools rahter than the all-in-one interface OCCT offers.

 

[Chicken Patty]

The problem with 1155 clocking is that increases in CPU speed also increase cache speed, which in turn affects system memory bandwdith.


So, memory that works @ 1333 MHz with the CPU @ stock, might not when the CPU speed is increased.

But, it might also be the cache that is failing...and not the core proper itself(Core, L1+L2).


That's why I like to clock everything up, and then use programs to test specific workloads that stress certain parts of the system, and then scrutinize the failures for what to change.

This is the beauty of such communities like TPU. Helping each other out. So now that I read this and know this, I notice my last statement although may be true for other platforms, may not hold true with S1155. Good to know, definitely helpful.

 

[Darkleoco]

7 hours stable on Prime95 , just out of curiosity would I even be likely to notice the difference between 1333 and 1600 on my RAM?

 

[cadaveca]

Daily, not really. Bandwidth gains are not that much, not 50%, and latency differences are where the real difference plays in, but because SB makes such efficient use of the bandwidth available, these gains are barely perceptable except in specific workloads and benchmarks.

If a particular app is system memory heavy, then yes, you can notice a difference, but for web browsing, and gaming, not so much.

 

[Darkleoco]

Daily, not really. Bandwidth gains are not that much, not 50%, and latency differences are where the real difference plays in, but because SB makes such efficient use of the bandwidth available, these gains are barely perceptable except in specific workloads and benchmarks.

If a particular app is system memory heavy, then yes, you can notice a difference, but for web browsing, and gaming, not so much.

Glad to know the memory isn't going to be an issue, I likely just won't bother trying to clock it back to 1600 since I am finally stable, I will probably benefit more from moving on to OC'ing my GPU's now.