would I notice the difference?

[Fistandantilis]

would I notice any real benefit in overclocking my Q6600 from 2.4 to 2.8? 400mhz does not sound all that much faster.
I do a bit of gaming and surfing/office stuff, thanks

 

[BonzaiDuck]

First, the answer to your specific question depends on whether you intend to use the stock multiplier of 9, or drop the multiplier to 8 and then run up the FSB to 1440 and memory to 720 in a 1:1 CPU : DDR ratio.

Those were conservative settings in my log of various over-clocking milestones, either way, but I DID notice a difference with 8 x 360 or 2.88 Ghz.

However, with the stock multiplier, I did not even bother to try a CPU speed of 2.8 ghz, because it seemed at least necessary to get the memory running to 667 with FSB of 1334 and CPU at 3.0 Ghz. Just about everyone I talked to here about their Q6600 over-clocking chose that as a very reliable and still conservative over-clock, or 25% above the stock speed.

I definitely could tell the difference between 2.4, even running the memory with the "Native 800" Mhz setting of the motherboard at the stock speed, and 3.0.

I'm using the B3 stepping, which runs a tad warmer and is less elastic for top-end, stable over-clocks than the G0 stepping. I've found a very stable setting at 3.2 Ghz with the B3, but many people are reporting G0 over-clocks to 3.4 with temperatures that are slightly lower than mine, and VCORE settings that are within the Intel retail/warranty "maximum" for the B3. Without firsthand experience, or anyone to tell me about it, I can only assume that the retail-box "maximum" voltage is the same for the G0, but people running the G0 stepping seem to be reaching these same or higher over-clock levels with lower VCORE settings than my B3.

 

[Fistandantilis]

so if I change my multiplier from 9 to 8 and raise my FSB to 350 would I get better performance than if I went with 9x311 to get the same 2.8?
what about heat, I have heard so many different things about safe temps on the Q6600 GO processors that I dont know what to believe.

Thanks for the help

also, I am right now running my RAM at a 3:5 setting, I have the PC-2 8500 OCZ, what settings would you guys recommend?

 

[zorrt]

I'll say no. But if you can do it without temps increasing then you might as well for the sake of saying your cpu is now 400mhz faster.

 

[BonzaiDuck]

There are various opinions about this.

My personal opinion regarding CPU-to-RAM dividers other than 1:1 is that they introduce wait-states into the transfer of data between the two components.

Some here -- including (I think his handle is "Graysky") the OP of the sticky on over-clocking C2D and C2Q processors -- argue that 1:1 gives the best performance, and that any improvements in other ratio choices only show in synthetic benchmarks, and not real-world performance.

A lot of noobies to over-clocking, and people too timid to get their feet very wet with it, shy away from tweaking the memory timings (or latencies). There is a very good article at Tom's Hardware published between 2003 and 2004 about tweaking the latencies, and it explains the contribution of each timing to the performance of any, single memory operation. For instance, the lower the CAS (column-access strobe), the faster the memory operation. But also, the RAS precharge delay, as I understand it, affects the overall length in clock-cycles of the memory operation. Obviously, various active processes or programs on your computer will execute thousands of those operations in the blink of an eye, but the tighter (smaller or lower) these latencies are, they will affect proportionately the time it takes to perform completely any given task involving memory.

So tighter latencies mean faster speed without stressing the system with higher frequency -- clock-cycles per second in megahertz. Instead, you have reduced the number of clock-cycles at any given megahertz setting to perform an operation in memory.

In the past, some memory manufacturers who produced high-performance (high megahertz) memory created products that were not downwardly flexible in the latency settings. For instance, this was my experience with OCZ Gold EL DDR-500 and OCZ Platinum XTC EL DDR-500 modules: you could not run these modules at DDR-400 and expect to set the latencies at lower than the stock DDR-500 settings. Other modules of both DDR and DDR2 vintage proved to be very elastic. For instance, I could get OCZ Platinum DDR-400 modules with Aeneon chips on them, and the stock latencies at 400 Mhz of 2,3,2,5 were very stable up to 454 Mhz without pushing the VDIMM voltage to the warranty limit. And DDR2 modules with Micron D9 parts have proven to be very "downwardly" elastic in the high-performance flavors -- for instance, my Crucial Ballistix (Micron) DDR2-1000 modules have SPD settings of 5,5,5,15 for the rated 1000 Mhz speed, but I can run them at 3,3,3,6 at DDR2-667, and 3,4,4,8 at DDR2-712.

I get some very good synthetic benchmarks this way with Everest Ultimate, and I don't have to stress the motherboard more to break even with higher Mhz settings.

 

[Fistandantilis]

Originally posted by: zorrt
I'll say no. But if you can do it without temps increasing then you might as well for the sake of saying your cpu is now 400mhz faster.

ya, thats about where I am now, I am wondering though, what temps are considered "fine", under load the highest I have ever seen it is 61-62c, generally they idle at around 42-26c.
would 70c on the processor for any given time be a bad thing?

Thanks for the replies fellas

 

[BonzaiDuck]

. . . . so, to be thorough in what I was trying to say, after trying various multiplier settings lower than stock (you can go as low as 6 on the Q6600 processor), I've convinced myself that it's better to OC to 3.2 Ghz at a multiplier of 9, with CPU_FSB or external frequency of 356 Mhz (DDR2-712), than to set the multiplier to 8 (or lower) and move closer to 400 Mhz (DDR2-800).

 

[BonzaiDuck]

The G0 will show lower temperatures than the B3. I'm stress-testing my B3 today because I decided to lower the 1.2VHT voltage by 0.05V, and I want to "re-certify." At 68F room ambient, my core temperatures (Q6600 @ 3.2 Ghz) are 60, 59,55,56 (Celsius). If I stress-tested at a room-ambient of 75F, they would be approximately 64, 63, 59, 60. This was born out with stress-testing this summer (in So-Cal), but it stands to reason that room-ambient will change load values degree for degree.

As I understand it, these processors are made to throttle back when TCase temperature reaches approximately 62C. TCase is usually somewhere between 9 and 15C lower than the core (TJunction) temperatures. So these core temperatures are very, very acceptable, and you would find things noticeably cooler with the G0 stepping.

 

[Fistandantilis]

Originally posted by: BonzaiDuck
. . . . so, to be thorough in what I was trying to say, after trying various multiplier settings lower than stock (you can go as low as 6 on the Q6600 processor), I've convinced myself that it's better to OC to 3.2 Ghz at a multiplier of 9, with CPU_FSB or external frequency of 356 Mhz (DDR2-712), than to set the multiplier to 8 (or lower) and move closer to 400 Mhz (DDR2-800).

OK I understand, so what about my RAM, it is set at 3:5, I should prolly change that huh?

 

[BonzaiDuck]

Well, that's what I'd do. I don't know the particulars of your BIOS with the eVGA 680i board, but I have the Striker Extreme (ASUS) 680i. If there is a default "linkage" between FSB and memory, disable it. Then manually set the FSB to some level above the stock 1066 setting, and the DDR2 speed to half that.

I had already experimented with an E6600 processor on my system before swapping in a Q6600, so I didn't hesitate to start OC'ing at 1,334 FSB and 667 DDR2. I also set the voltage a tad over what was necessary at that time, and had to drop it in increments before I could get the system to fail -- so I could bump it up just one notch to re-establish stability.