Has anyone overclocked a Sandy Bridge?

From my understanding, you can't use Throttlestop to set a higher multiplier than what the system booted at. Since notebook bios almost never has multiplier setting options, this leaves us with setting the integrated clock generator via setfsb or some other program. Does any software support the Sandy Bridge clock generator? And if so, what overclocks have people achieved? Desktop users haven't been able to get much past 4% by changing the base clock frequency.

I want to know because I was thinking of getting an Ivy bridge laptop once they come out, and it's supposed to just be a die shrink of Sandy Bridge. To get anything significantly better than what I have performance wise, I'd have to overclock it.

I really hope Intel hasn't killed laptop overclocking I love playing around with this stuff.

 

I dont know how ivy bridge would be regarding that, but currently only extreme models (or K models in the desktop) have the ability to overclock. And they are good at it. People at the aw forums have been achieving @4ghz.

And sincerely I find myself hard to believe that your c2d at that clock can be equal in performance to the x58 chipset, which sb quads are.

 

I think only extreme CPUs can be OC'd, and I haven't seen any K models for laptops

I dunno about SB quads but my OC'd E8335 (T9800) runs circles around a stock 720QM on some benchmarks, on most others it's at least competitive. Yes, you can argue it's not the newest architecture but come on, it's a C2D and it's natively overclockable. Probably beats out midrange i5 SB CPUs, too. IMO Core 2 was a solid line of CPUs with some of the best power consumption up til Sandy Bridge.

 

If he is talking about anything lightly threaded it isn't unbelievable. C2D was a fantastic architecture. At 3.67GHz it probably performs on par with the most powerful mobile SB in single and dual-threaded applications.

 

Do you like manual transmission and carburetors too? From what I read they haven't killed it, rather, they just rendered it unnecessary.

And efficiency? Remember, this isn't just a comparison of raw power but one of size and efficiency as well.

 

uh, yea. poor analogy.

 

If I'm not mistaken, the i7-2920xm has unlocked clock multiplier, but it's up to the OEM to enable it in bios.

 

OP said he wants to get something significantly better performance-wise.

 

I agree, manual transmissions are awesome. Carburetor analogy is probably accurate though.

 

i7-2620M runs at 3.4 Ghz single core turbo and 3.2 Ghz Dual core turbo so I find that unlikely.

 

Look at the single-threaded benchmarks for the T9800 (2.92GHz) vs the i7-2630qm/i5-2410m (2.9GHz single) if you don't believe me. Now extrapolate those numbers to 3.67GHz which OP says his C2D is running at.

 

AnandTech - Bench - CPU

Sandy Bridge at 3.4 Ghz turbo (same as i7-2620M) gets 5423 from Cinebench R10 single-threaded. C2D at 3.33 Ghz gets 4128. Add a tenth more of that for the best case clock scaling scenario to get to 3.67 ghz (412) and you get 5423 vs. 4500 in favor of the SB.

In Cinebench single threaded, SB is 30% faster than C2D at the same clock.

 

so, in 4 years the speed has increased 30%... impressive

 

You aren't taking into account that the score for the E8600 in the link was obtained using Cinebench 32-bit while the SB score uses the 64-bit version. Cinebench scores increase pretty dramatically from 32 to 64-bit

 

Even if they're both single threaded?

 

From every benchmark I have seen, yes. Just look at the scores for the i7-2620m on Notebookcheck

http://www.notebookcheck.net/Intel-Core-i7-2620M-Notebook-Processor.40108.0.html

 

Thanks for the link - interesting.

I can see why you claim that it could be comparable - but my gut feeling is that the C2D wouldn't scale the same as the SNB parts when testing with the x64 bit benchmark?

 

This discussion seems to have drifted from what I originally had in mind. A high end SB will outperform my C2D in 2 or less threads, but not by enough for me to consider upgrading. (I got 4036 in cinebench R10 32bit, and 4536 in 64bit) The lack of major improvement in low threadcount performance is what it making me on the fence about getting a new laptop.

What I was going for was that since SB has a build in clock generator, and desktops have shown that this clock generator is adjustable in 1mhz increments, then all SBs should be overclockable via the 100mhz core clock, just laptops will have to do it through software if possible while desktops have the BIOS option. Have people tried to set the SB clock generator registers through the SM bus like setfsb, clockgen, and other programs do? It seems that if it was possible that someone should have figured it out by now, and Ivy bridge will likely have the same overclocking characteristics as sandy bridge, just clocked higher.

 

Where does it say on AT that their E8600 scores are using 32-bit Cinebench? Especially when Khenglish says that his score is exactly what I predicted extrapolating from Anand's numbers: 4128*1.1=4540 and Khenglish got 4536.
Therefore the AT scores for the E8600 are using 64-bit Cinebench and you are wrong. SB is 30% faster per clock than C2D in Cinebench. Some other tasks show much better improvements going from C2D to SB.

 

AnandTech - Bench - CPU

Hover over the notes for the E8600. I don't know why Khenglish only gets 4000 on 32-bit at 3.67GHz. Judging by scores for other Penryn CPU's, he should be getting about 4300-4500 on 32-bit.

@tiller - I agree but I think the improvement should be something.

 

Here is a Sandy Bridge review with the sample clocked at 4.5GHz (it's easier to look at overclocked reviews because then you know the clock speed exactly rather than trying to figure out what Turbo is doing). In single threaded performance, the 2600K gets 1.81 (this is 64-bit CB11.5). I ran the same test on my T9300 (Core2Duo clocked at 2.5GHz, 6MB cache) and I get 0.78. Scaling it up by a factor of 45/25 works out to 1.404 so SB is faster by 29% clock for clock.

Note, however, that this does not take into account the fact that SB has hyperthreading whereas Core 2 did not. If you take my C2D's multi-threaded score of 1.44 and multiply by the ratio of cores and the 45/25, you get 5.184 whereas SB gets 8.80. That's an improvement of 70%.

In other words, Core 2 was good, but not that good. Even overclocked to 3.67GHz, it should lose to something like the 2720QM by around 10-20% in single-threaded tasks. In multithreaded tasks, it will be utterly crushed.

 

Regardless, OP's point was that only the highest mobile SB processors are going to beat his 3.67GHz C2D tasks with 1-2 threads, and the difference won't even be much. Probably won't even make any difference in most games.

 

Bottom line, SB kills C2D in multi-threaded workloads and is about 30% faster single. C2D was a great architechture for the time, but it is outclassed now. However, its not like its not still useable. Also, the 2920XM is unlocked, and can be OC'd through ThrottleStop.

 

Has anyone verified that the mobile extreme CPUs are unlocked and overclockable in windows? It would be rather unpleasant to fork out that much cash then find out that they aren't.

 

Check the M18x forums. People there have been OC'ing them like crazy, trying to break the 4GHz you can get from Dell.

 

Yes, ThrottleStop is the popular program to do Windows-based OCing.

 

I have been trying to figure out how to do this myself, really interested in overclocking my 720QM in my M15x.

 

For standard CPUs from Nehalem you can use SetFSB, but you'll need the paid version I believe... Might want to check the AW forum section to see if the M15x can use it.

1500 posts, wow... haha.

 

Don't know if the SB's are overclockable via software like SetFSB but if you get a decent laptop that's not a budget build from a good manufacturer then it's definitely overclockable via the BCLK in the Bios. Look in my sig for proof, I've been able to go as high as about 107.0 BCLK stable with no additional cooling. If anyone wants to see benches on the SB I'd be more than happy to throw some out later tonight, I guarantee at the 2720/2820 level or higher it will beat anything out other than the 940xm(possibly) and the top end desktop CPU's. If there was access to the multiplier it would blow everything else out of the water. If you have a multi threaded app/game 4cores/4threads pumping at 3.32ghz each is an awesome thing