SB Overclocking Thread

[Axon]

Coming from amd for the first time in years
To intel, I'm still confused about the temp readings.

My cores max out around 68 Celsius according to core temp and hwmonitor.

To be honest I Like how amd is setup better for overclocking in there boards bios'

Well they just completely changed everything around. However, I do hope AMD sticks to the classic style of overclocking...that will win them some enthusiast fans, provided BD produces decent numbers.

 

[samboy]

Setup:-

MB: Asus P8P67 MB
Processor: 2600k
Cooler: Noctua NH-D14
Ram: 16GB G-Skills DDR3 1600

I've set the multiplier to 45x and the CPU Voltage to Offset and a short 20 minute run of Prime 95 is stable (I'll do an overnight run later) with the following results:-

Voltages by CPU ID are at the limit of what I would like (1.34 Max Voltage)
And temperatures are under 60C

I haven't tried manually setting the voltage........ I like the offset concept where it will fall back to 1v @ 1600MHZ when not being pushed and not sure if manually fixing the voltage would change this.

Assuming that this is stable overnight, does this look safe for 24/7 use?
I'm assuming a bit of manual tweaking could get some more performance? Any suggestions?

Thanks

 

[Hogan773]

Setup:-

MB: Asus P8P67 MB
Processor: 2600k
Cooler: Noctua NH-D14
Ram: 16GB G-Skills DDR3 1600

I've set the multiplier to 45x and the CPU Voltage to Offset and a short 20 minute run of Prime 95 is stable (I'll do an overnight run later) with the following results:-

Voltages by CPU ID are at the limit of what I would like (1.34 Max Voltage)
And temperatures are under 60C

I haven't tried manually setting the voltage........ I like the offset concept where it will fall back to 1v @ 1600MHZ when not being pushed and not sure if manually fixing the voltage would change this.

Assuming that this is stable overnight, does this look safe for 24/7 use?
I'm assuming a bit of manual tweaking could get some more performance? Any suggestions?

Thanks

I'm gonna post also in an ASRock mobo thread, but I want to make sure I know how to do the following OC (which sounds like what you're talking about):

1) type in new multiplier 44 or 45 or whatever - this I can do

2) set voltage so that it doesn't go HIGHER than 1.30-ish, but still I want it to be able to drop down as much as it wants to.

3) likewise the multiplier - I want it to speed up to a max of [44] but also throttle back like it wants to when not stressed.

Basically I want to leave all the Turbo stuff running, and just expand the top-end multiplier while putting a CEILING on voltage.

So what BIOS settings does this entail changing?

Thanks

 

[Dadofamunky]

Well, Intel's C1E and EIST settings are typically On by default, and those are what enable the SpeedStep capability of throttling back to 1660 MHz when idle. This has been the case for several generations of Intel chips. From what I've been reading on this thread, setting the multiplier on SB effectively sets the speed at which the Turbo speed will run. Turning off the C states and EIST I believe will fix the steady-state speed to be your multiplier settings without the drop down to 1.6G.

Then again, I will be FINALLY putting my assembled mobo into its chassis this evening (CPU, RAM, HSF) and connecting everything up, so hopefully I can put all this into practice FINALLY.

 

[Bryf50]

Having an odd issue with my system. I can run linpack and prime overnight with no stability issues, but at completely random times my computer will just restart. No blue screen or anything. Its happened when i was browsing the web and even when i wasn't at the computer. I'm not sure what to make of this, I tried dropping my oc to 4.2ghz and it still had the issue despite stress testing perfect. My only idea is that its EIST that's causing the issue and that my chip is stable at full load but isn't at idle. How bad would it be to run this thing at full voltage and clock speed 24/7?

 

[Markfw]

Having an odd issue with my system. I can run linpack and prime overnight with no stability issues, but at completely random times my computer will just restart. No blue screen or anything. Its happened when i was browsing the web and even when i wasn't at the computer. I'm not sure what to make of this, I tried dropping my oc to 4.2ghz and it still had the issue despite stress testing perfect. My only idea is that its EIST that's causing the issue and that my chip is stable at full load but isn't at idle. How bad would it be to run this thing at full voltage and clock speed 24/7?

I would say FIRST turn off EIST and c-states, and see if you are correct. I run mine 24/7, and the only bad thing is the electric bill.

 

[Arkainium]

For anyone that's overclocking using voltage offset with EIST and C1E enabled, I recommend performing the following test: open Prime95; set CPU affinity to a single core (CPU 0 w/o HT, CPU 0 & CPU 1 w/ HT); now run Prime95 Small FFTs with enough threads to get 100% load on the core.

My testing has shown that the voltage required to be stable with all cores fully loaded is insufficient for maintaining stability when only one core is actively loaded. The problem with stressing one core is that you don't get the same voltage response that you would get when stressing all the cores, which is what causes the instability. Disabling voltage throttling appears to be the only way around this problem.

 

[Bryf50]

For anyone that's overclocking using voltage offset with EIST and C1E enabled, I recommend performing the following test: open Prime95; set CPU affinity to a single core (CPU 0 w/o HT, CPU 0 & CPU 1 w/ HT); now run Prime95 Small FFTs with enough threads to get 100% load on the core.

My testing has shown that the voltage required to be stable with all cores fully loaded is insufficient for maintaining stability when only one core is actively loaded. The problem with stressing one core is that you don't get the same voltage response that you would get when stressing all the cores, which is what causes the instability. Disabling voltage throttling appears to be the only way around this problem.

Yea since i disabled C1E and EIST my OC is now 100% stable. Its a shame because i felt more comfortable letting it downclock and undervolt when i wasn't using it. I guess that best thing i can do now is re-enable C1E and just use a fixed voltage.

One question regarding LLC. Is there any disadvantage to using it? What i mean is like what is the reason why the motherboard wouldn't default with something like that on.

 

[Arkainium]

One question regarding LLC. Is there any disadvantage to using it? What i mean is like what is the reason why the motherboard wouldn't default with something like that on.

Read the section entitled "Intel Processor Power Delivery Guidelines" (two pages) from the following article: http://www.anandtech.com/show/2404/5

 

[Dadofamunky]

Wellp. got my SB system up and running. BIG disappointment so far...

Yep. FINALLY got my new rig built. Fired up perfectly the first time. UEFI BIOS looks OK. However, the system behaves exactly like it has an i5 2500 in it, not a 2500K. It WILL NOT allow a multiplier over 33. The only way to run this board as it now stands is to accept the turbo default, which is 39, and the board automatically raises the BCLK to 103.

The BIOS appears to be quite old... it is version "0402" which isn't even on the Asus Web site for BIOS downloads. Kinda ticks me off. I've downloaded the BIOS 1053 for this board and will install it when I am next in physical proximity to the new machine (it is in my commuter room near work, which is 120 miles away from me).

So do you think the BIOS update will fix this problem of not being able to overclock AT ALL? That is the situation I'm faced with at this point. The system behaves exactly as though I got an i5 2400 or i5 2500 non-K chip.

Has anyone else seen this?

When I discovered I simply could not set the multiplier (I believe it is called CPU Ratio in UEFI) over 33, that was quite a shock. The odds are it's one of these following issues:

1) I got a mislabeled processor from Micro Center (labeled 2500K when it's actually not)
2) Got a motherboard with no support for K-series chips...

I'm seriously hoping a BIOS update fixes it.

Is there something obvious I'm missing? Is there another crucial setting I need to tweak before I can set the multi beyond the default of 33? Thanks!

Two quick benchies - from Fritz Chess Benchmark, both machines in my sig:

E8600@4.0 GHz - 11.41
2500K@ 4 GHz, 2 cores active - 13!
2500K@4 GHz, all 4 cores - 25.7

The delta for two core operation between my two systems is nowhere near what I expected it to be. Frankly, if I'd known this would be the case, I'm not sure I would have bothered to do the new build.

Intel Burn Test gives me 57 Gigaflops. Also underwhelming.

 

[Arkainium]

When I discovered I simply could not set the multiplier (I believe it is called CPU Ratio in UEFI) over 33, that was quite a shock. The odds are it's one of these following issues:

You need to change the turbo multiplier. I was also confused at first because there are at least three different places in the UEFI where you can set the multiplier, but the turbo multiplier seems to be the only one that actually has any effect.

 

[Bryf50]

Two quick benchies - from Fritz Chess Benchmark, both machines in my sig:

E8600@4.0 GHz - 11.41
2500K@ 4 GHz, 2 cores active - 13!
2500K@4 GHz, all 4 cores - 25.7

The delta for two core operation between my two systems is nowhere near what I expected it to be. Frankly, if I'd known this would be the case, I'm not sure I would have bothered to do the new build.

Intel Burn Test gives me 57 Gigaflops. Also underwhelming.

http://www.anandtech.com/bench/Product/54?vs=288
It gets blown away in every benchmark. Using arbitrary synthetic benchmarks is a poor way to judge performance.

If you want to see massive FLOPS download service pack 1 and the new version of Linpack. With AVX you can get over 100 GFLOPS.

 

[Dadofamunky]

You need to change the turbo multiplier. I was also confused at first because there are at least three different places in the UEFI where you can set the multiplier, but the turbo multiplier seems to be the only one that actually has any effect.

Arkanium, thanks a lot. Others also have pointed this out. I missed that one. Duuuh. I ***Think*** I may have tried that one as well and had no luck.

http://www.anandtech.com/bench/Product/54?vs=288
It gets blown away in every benchmark. Using arbitrary synthetic benchmarks is a poor way to judge performance.

If you want to see massive FLOPS download service pack 1 and the new version of Linpack. With AVX you can get over 100 GFLOPS.

I agree, but what I did find revealing was using the same two utilities across two separate platforms running at what amounts to the exact same clock speed, but completely different platforms and Intel CPU/DDR memory generations.

I once was the author of a technical book about scientific benchmarking and one of the conclusions I came to was: not to optimize the benchmark software to get the results you want on a given system, because you aren't getting a true picture of how the system performs against its competition or previous generations. At the very least, run the very same package on all concerned platforms with no modifications in situ.

(That's why any test that is optimized for Intel and not on AMD is always going to skew its results, or why a game that is optimized for NVidia and not ATI will yield better frame performance. I find these results to be controversial at best and deceptive at worst and rarely ever to be trusted.)

So I did that, and am planning on testing on both of the systems in my sig as they now stand in parallel, and also OCing the SB system to its best practical level, and running the tests on that for the final set. What I also liked was the ability to use the chess benchmark on only TWO cores on the SB system, which provided at least a superficially very interesting result. (i.e. the SB cores not really being that much faster!)

Also, the chess benchmark scales almost perfectly to four cores, so I actually tend to trust it more than I would a 3DMark. These are only the very first two things I've had time to run, and there are many more programs that will be used.

 

[nOOky]

I would say FIRST turn off EIST and c-states, and see if you are correct. I run mine 24/7, and the only bad thing is the electric bill.

No kidding, looking at your sig makes me wonder why you do it unless you're getting paid for all that effort!

My 2600K runs at 4.4, I simply changed the multiplier to 44 and saved and left it there, no voltage changes or nuthin'

 

[Dangerer]

No kidding, looking at your sig makes me wonder why you do it unless you're getting paid for all that effort!

My 2600K runs at 4.4, I simply changed the multiplier to 44 and saved and left it there, no voltage changes or nuthin'

Is that stable? Don't these chips run under 1.0v stock?

 

[Castiel]

No kidding, looking at your sig makes me wonder why you do it unless you're getting paid for all that effort!

My 2600K runs at 4.4, I simply changed the multiplier to 44 and saved and left it there, no voltage changes or nuthin'

Voltage is probably on auto

 

[smakme7757]

Here is my 24/7 Overclock I've spent quite some time on this and it's rock solid!:

2600K: 4.5Ghz (HT ON)
Corsair Vengence Ram (Stock 1600Mhz): 1866Mhz 9-9-9-24 1T
QPI/VTT: 1.080v
Vcore: 1.275v
LLC: Level 1
PLL: 1.70v
VDimm: 1.58v

Idle: 24-28c
Load: 47-52c

 

[Dangerer]

^ Nice temps, whats your cooler? Ambient temps?

 

[smakme7757]

^ Nice temps, whats your cooler? Ambient temps?

GPU and CPU are watercooled with the following components.

Radiator: EK Coolstream 360
CPU Block: XSPC Rasa
GPU Block: EK-FC580 Nickel + Acetal
Pump: Swiftek MCP655
Hosing: 7/16″ Primochill White
Barbs: 1/2″ Barbs

Ambient is 21.2c

I'm very happy with the temps and the overclock. It really does seem like the perfect balance between performance and heat/voltage.

 

[sxr7171]

I was going to say that I wonder why SB is a "tock" and not just another "tick". I don't believe IPC changed much over the last gen. But when I see people's cool and by extension quiet O/Cs I feel jealous.

 

[Dadofamunky]

Successfully hit 4.4 GHz tonight with the system in my sig. Starting an overnight Prime run, did 10 IBT runs successfully (64 Gflops). Not a complete test but all I had time for. I think there's some headroom left. Key was setting a few of the ancillary voltages to non-Auto (CPU PLL, etc... I forget which, I'll post again with more info) and setting the PLL to 1.9 V. CPU is at 1.28V. BIOS is updated to 1053, and the CPU Ratio and Turbo settings are now synchronized properly. Temps are 28-30 idle, 60-65 load. I may have to do some work on my cooler by and by.

Setting EIST and C1/C3/C6 to On or Off seems to have no effect on whether or not the OCs worked. So I turned them off.

Got a lot of advice and really appreciated the help.

 

[Viper GTS]

How are you guys getting 60+ GFLOPS with IBT? I'm running in the ~54 range with my 2600K @ 4.4 GHz. I am running my RAM loose right now at 1600 (16 GB of Vengeance), will that bring it up that much?

On the subject of cooling I ended up picking up a 120x120x12 mm Scythe Slipstream fan (the 2000 RPM version), the pull only config to fit the 212 on my UD5 was not really working too well. Push pull with the 12 mm fan dropped temperatures to a level I'm far more comfortable with. I see some 120x120x20 Yate Loons those might be thin enough to work as well.

Viper GTS

 

[Dadofamunky]

Update: I won't go so far as to declare it stable yet, but I have passed 30 runs of Intel Burn Test standard at 4.5 GHz. Temps peak out at 70C.

Started an overnight Prime test (it passed an overnight test at 4.4). After that I will probably run LinX but at this point nothing I do in the real world will ever come close to the tests I've run so far.

63.5 GFlops
8.360s in SuperPi 1M (ancient single-core benchmark, but useful within its limits)
1.31V CPU voltage in BIOS
1.625V for the RAM
1.9V CPU PLL
Turbo On, EIST and C-states Off

Jeez. 4.5 GHz for a $170 processor on air! THANK YOU, Intel. Wonder what Ivy Bridge will be like.

Sorry, I'm too lazy to do screen caps.

 

[Castiel]

What kind of PLL is everyone running? At 4.6 i'm running 1.4v. What the hell does PLL even do?

 

[Dadofamunky]

How are you guys getting 60+ GFLOPS with IBT? I'm running in the ~54 range with my 2600K @ 4.4 GHz. I am running my RAM loose right now at 1600 (16 GB of Vengeance), will that bring it up that much?

On the subject of cooling I ended up picking up a 120x120x12 mm Scythe Slipstream fan (the 2000 RPM version), the pull only config to fit the 212 on my UD5 was not really working too well. Push pull with the 12 mm fan dropped temperatures to a level I'm far more comfortable with. I see some 120x120x20 Yate Loons those might be thin enough to work as well.

Viper GTS

Just tighten the timings. Mine's running its factory 8-8-8-24. It is possible, mind you, that having 16GB of RAM imposes some kind of performance penalty on the memory controller. Try running it with 8 as well and see what results you get. You should be getting, in fact, more than I am (63.5G) since you have the hyperthreading.

My next step might be to get a fan controller to get my 120mm push-pulls to move air faster. Both my fans are 2000 rpms but they're not running that fast - not even close. However, if there's a noise penalty I might not bother.

What kind of PLL is everyone running? At 4.6 i'm running 1.4v. What the hell does PLL even do?

Dang, you must have some kind of golden chip. I think way back early in this thread some people mentioned that you needed 1.9V on the PLL to stabilize overclocks on the chip. PLL is an electronic underpinning for the frequency generator in the chip and a very widely used electrical construct. Of course, this says it better than I ever could - http://en.wikipedia.org/wiki/Phase-locked_loop.