The Sandy Bridge OC Thread
- Thread starter yashchitale
- Start date
- Status
devvfata1ity
Contributor
@yash. mate if you are getting rounding off errors in prime that means your system is not 100% stable. Its ok for normal use but you would want to be 100% stable under prime to completely rule out the possibility of BSOD while in normal use. Just like nemesis said lapping the heatsink contact surface is not worth for the casual overclocker, its hard work & top of that you need different grade sandpapers which is difficult to obtain. I must admit though that lapping can actually shave off 4-6 C from the load temps but seeing that you are already getting good temps i would suggest you to stay away from lapping.
what you can do instead is try to improve the application of the TIM and spread it as evenly as possible (i know its difficult but try as best as you can). The moment you get not more than 2C delta between the temperatures of different cores, you know that the TIM has spread evenly.
To become stable under prime, plz increase the vcore a bit and check. for 4.3 Ghz 1.2-1.22 v will get your system absolutely stable under prime. Also enable round of checking option and run the blend test only because all the other tests like small FFT or large FFT will run inside that huge l3 cache and wont be able to stress the SB processors fully as well as wont give you the indication of your overall system stability. Make sure in the test settings you use 1GB less than the total RAM in your system.
As a cross validation use the stability test tool of AIDA64 v1.6 as it uses the latest AVX instructions to take the SB processors to its limits. Your AIDA64 temps should be within +-1 C of your prime temps.
All the best OCing
what you can do instead is try to improve the application of the TIM and spread it as evenly as possible (i know its difficult but try as best as you can). The moment you get not more than 2C delta between the temperatures of different cores, you know that the TIM has spread evenly.
To become stable under prime, plz increase the vcore a bit and check. for 4.3 Ghz 1.2-1.22 v will get your system absolutely stable under prime. Also enable round of checking option and run the blend test only because all the other tests like small FFT or large FFT will run inside that huge l3 cache and wont be able to stress the SB processors fully as well as wont give you the indication of your overall system stability. Make sure in the test settings you use 1GB less than the total RAM in your system.
As a cross validation use the stability test tool of AIDA64 v1.6 as it uses the latest AVX instructions to take the SB processors to its limits. Your AIDA64 temps should be within +-1 C of your prime temps.
All the best OCing
yashchitale
Discoverer
Will do that.. Thanks a lot buddy.. Try and make it to any of the Pune meets.. I'll be there for both..
devvfata1ity
Contributor
stormblast
Forerunner
^Nice esp. the lower core voltage although I think Prime95 might give rounding errors. IRL apps shouldn't be a problem I suppose. What clocks do you prefer for 24x7 usage?
On my measly Transcend 2GB DDR 1333MHz RAM, I was able to safely clock 4.5G at 1.24V. Tested Prime95 stable for an hour without issues.
On my measly Transcend 2GB DDR 1333MHz RAM, I was able to safely clock 4.5G at 1.24V. Tested Prime95 stable for an hour without issues.
yashchitale
Discoverer
@stormblast: The voltage seems high for a 2600k at 4.6ghz.. try reducing it..
@Gannu: Awesome stuff.. good vcore.. you could maybe try going lower.. what thermal readings you getting?
@Gannu: Awesome stuff..
good vcore.. you could maybe try going lower.. what thermal readings you getting?
Actually I am a bit flummoxed with the temperatures. All the 4 cores idle at 42~44'C and do not go beyond 60'C when loaded with Prime95 and monitored using RealTemp 3.60. This is with the TRU120E and a Panaflo fan at medium speeds.
What utility do you guys use for monitoring the temperatures?
What utility do you guys use for monitoring the temperatures?
devvfata1ity
Contributor
I am using both real temp & core temp to monitor temps. Those load temps are awesome. btw are you running in an ac room.
Gannu said:^Nice esp. the lower core voltage although I think Prime95 might give rounding errors. IRL apps shouldn't be a problem I suppose. What clocks do you prefer for 24x7 usage?
On my measly Transcend 2GB DDR 1333MHz RAM, I was able to safely clock 4.5G at 1.24V. Tested Prime95 stable for an hour without issues.
Yeah, the 5.2GHz OC is not prime stable. Probably need a bit more vcore. My CPU was able to boot and run Super Pi @ 5GHz with 1.27V, but for prime stability I needed 1.34~1.35V. I was also able to do 5.1GHz @ 1.34V with almost everything other than Prime benchable, but stuck to 5GHz for my 3DMark benching. For 24/7 use I am still sticking to 4.5GHz @ 1.2V.
Regarding your OC, I think you can try for a lower vcore or a higher multi. Also your RAM doesn't matter as no BCLK overclocking is involved. Even with Ripjaws, you will be able to do the same OC's only
Gannu said:
Use RealTemp 3.67 (has official support for Sandy Bridge) or HWMonitor. Those temps sound right enough. The idle temps seem a wee bit on the higher side though.
devesh38
Galvanizer
hello guys, I am running my 2500k proccy on p8p67 pro board at stock settings on stock cooler. Now I would like to buy a cpu cooler for overclocking. Initially was looking for Noctua coolers but due to unavailabilty at local shops and shortage of budget, I postponed the buying of the cpu cooler.
Now I am in position to buy one. But on seeing the above rigs configs I didnt find any noctua coolers in them. So I am wondering whether i should go with noctua or others (like thermallight ones).
Please guide me with the cooler selection.
budget = 4k (can extend upto 5k)
Now I am in position to buy one. But on seeing the above rigs configs I didnt find any noctua coolers in them. So I am wondering whether i should go with noctua or others (like thermallight ones).
Please guide me with the cooler selection.
budget = 4k (can extend upto 5k)
yashchitale
Discoverer
If you have that kind of budget why dont you invest in water cooling?
Theitdepot - ECO CoolIT Systems ECO-R120 Advanced Liquid Cooling For CPU (ALC)
There are listings for this on ebay too.. Cannot give you the links since i'm in office..
Theitdepot - ECO CoolIT Systems ECO-R120 Advanced Liquid Cooling For CPU (ALC)
There are listings for this on ebay too.. Cannot give you the links since i'm in office..
devvfata1ity said:
It is an ac room but the AC wasn't switched on while running those tests.
I think those temps reported maybe incorrect. Will download Real Temp 3.67 as Lord has suggested and report back.
Will use 3.67 and post back.Lord Nemesis said:Regarding your OC, I think you can try for a lower vcore or a higher multi. Also your RAM doesn't matter as no BCLK overclocking is involved. Even with Ripjaws, you will be able to do the same OC's only
Use RealTemp 3.67 (has official support for Sandy Bridge) or HWMonitor. Those temps sound right enough. The idle temps seem a wee bit on the higher side though.
Guess what, yesterday this weird incident happened to me: I happened to use Prime 95 for a quick test, closed it and started Crysis 2. Only to realise later that Prime 95 was running in the background and there were few stutters in the game which made me ponder why this stuttering now when there were absolutely none for the entire first half of the game. Quit Crysis 2, checked the Task Manager and found Prime 95 was still running in the background. Apparently closing the application by clicking on 'X' doesn't quit it.
And yeah it was running fine with one rounding error reported after about an hour and a half. Talk about 'Maximum stress'. :lol:
stormblast
Forerunner
yashchitale said:@stormblast: The voltage seems high for a 2600k at 4.6ghz.. try reducing it..
@Gannu: Awesome stuff..
1.3-1.35vcore is a lot for 2600k?
isnt that fine for around 4.6ghz if temps are under control
On the Noctua Front, U12P SE2 and D14 are the standard fare in your budget, U12P SE2 costs 3.7k and comes bundled with 2 Noctua fans. D14 costs 4.8k and also comes with fans bundled. There is no doubt that these are good coolers. The TR Venomous X on the other hand does not have any fans bundled (in the default model) and costs 3.6k.
So why did I go with TR Venomous X?
1. Performance wise, the TR VX is in the same league as the Noctua D14 (which is a gigantic cooler), but size and weight wise it is similar to the U12P SE2. I was looking for the best possible performance with the least possible bulk.
2. The TR VX has a wee bit more clearance for RAM heat spreaders. U12 SE2 has 40mm clearance while Ripjaws X measures a bit shy of 41mm. While this doesn't mean that the cooler is incompatible, you will still need to adjust the fan a bit because of the height. Not much of a biggie though. The TR VX has 43mm clearance .
3. I knew from the start that I would end up slapping a high power 38mm Delta or Scythe Ultra Kaze fan to what ever cooler I get.
Most people go for Noctua coolers because of the value added by the Bundled fans (which btw have a reputation to be excellent fans). The Noctua fans alone if you buy separately will cost you 1.8~2k for a pair. If you are fine with the quiet cooling that they offer, by all means go for one of the Noctua's.
So why did I go with TR Venomous X?
1. Performance wise, the TR VX is in the same league as the Noctua D14 (which is a gigantic cooler), but size and weight wise it is similar to the U12P SE2. I was looking for the best possible performance with the least possible bulk.
2. The TR VX has a wee bit more clearance for RAM heat spreaders. U12 SE2 has 40mm clearance while Ripjaws X measures a bit shy of 41mm. While this doesn't mean that the cooler is incompatible, you will still need to adjust the fan a bit because of the height. Not much of a biggie though. The TR VX has 43mm clearance .
3. I knew from the start that I would end up slapping a high power 38mm Delta or Scythe Ultra Kaze fan to what ever cooler I get.
Most people go for Noctua coolers because of the value added by the Bundled fans (which btw have a reputation to be excellent fans). The Noctua fans alone if you buy separately will cost you 1.8~2k for a pair. If you are fine with the quiet cooling that they offer, by all means go for one of the Noctua's.
Gannu, gamed with Prime95 stress threads on. Kewl -- and the system withstood it all.
Best is to use RealTEMP which integrates the Prime95 EXE within itself. Most reliable for temperatures. Also have CoreTEMP to cross reference a few runs. If temperatures are matching then use any one.
Also I like using IBT. It burns the skin off -- and catches the errors within 20 minutes. Less waiting and staring at the screen OR less gaming+Prime95 in the background. he he.
OCCT is a nice killer too, but again the wait is long. Like 60 minutes.
Best is to use RealTEMP which integrates the Prime95 EXE within itself. Most reliable for temperatures. Also have CoreTEMP to cross reference a few runs. If temperatures are matching then use any one.
Also I like using IBT. It burns the skin off -- and catches the errors within 20 minutes. Less waiting and staring at the screen OR less gaming+Prime95 in the background. he he.
OCCT is a nice killer too, but again the wait is long. Like 60 minutes.
- Status