***The Official Q6600 Overclocking Thread ***

[phatboy]

Will dabble with ibt when I get a chance . Going to get an akasa Apache fan for the true on payday as the current zalman is a bit loud

 

[setter]

I got the grit paper at work mate, (painter + decorator). Used a few drops of water and a piece of glass as a base, 10 passes back and forth, 10 side to side, checking the cpu/cooler base with a brand new stanley blade to see how level it was. Took about 3 hours for the cooler, 2 for the cpu, im quite used to sanding things unfortunately.

Edit. Theres no real need to go as high as 2000 grade, all it does is add a bit of shine, flat surfaces are the main priority.

 

[WingZero30]

I got the grit paper at work mate, (painter + decorator). Used a few drops of water and a piece of glass as a base, 10 passes back and forth, 10 side to side, checking the cpu/cooler base with a brand new stanley blade to see how level it was. Took about 3 hours for the cooler, 2 for the cpu, im quite used to sanding things unfortunately.

Nice

Will you be able to send me some sandpapers for free!!?

My kitchen slab is very smooth. Will that do if I want to lap?

 

[setter]

I would if we still used sandpaper, boss only buys sponge sanding pads now, (tight git). Id go with glass as its a lot smoother than even higly polished granite kitchen worktops.

 

[j.col]

Assuming he doesn't use hyperthreading his i7@4Ghz will give maximum value of 64GFlops.

IBT is a double-precision (64 bit) program so it is always 4 Flops with mutli-add operations.

So for i7@ 4Ghz (Quad core)

4 cores x 4 Flops x 4Ghz = 64Gflops

49/64 = 76.56% which is still high enough.

For any dual core you just replace 4 cores with 2 cores and for single core it is 1 core.


E5200@3Ghz

2 cores x 4flops x 3Ghz = 24GFlops

Those values you quoted I believe are the real world predictions by Perspolis in his case. So it doesn't apply to everyone as each case is unique.

Your friend should be able to get over 50GFlops by choosing custom and tweaking memory or closing some background programs. As xsisitor, another fellow ocuker stated, it depends on the algorithm aswell and how well it is optimised for cpu processing. The algorithm in IBT or LinX is the LinPack algorithm/program

The greater an algorithm is optimised the greater output from cpu will be in basic terms. Hence in real world we can never achieve the theoretical maximum but can only try to get as close as possible.

I tend to hit 51 ono gflops on my i7 at 4ghz, same at 4.2, which is proving to be a bit trickier to get LinX stable.

thanks guys for the response

 

[WingZero30]

thanks guys for the response

Here is the Intel GFlops sheet for CPUs at stock speed. If you use the same formula you will end up with same GFlops values as stated in the sheet

http://www.intel.com/support/processors/sb/cs-023143.htm#3




If the floating point operations involved something like subtraction-division instructions, then you may find that FPU (Floating Point Unit) of CPU may well take more cycles to execute those operations compared to multi-add. It seems multi-add instructions are more popular at gauging a CPU's Floating Point operations performance as they are simpler to process and hence can be executed quicker.

 

[DHB07]

Hi guys, I've just pasted the following from my post on the main board, in hope I'd be ble to get a bit more specific help here. Sorry it's so long...!

Hello all, as you can see, it's my first post here so go easy on me! I've been an ocassional reader of the forum for a couple of years now, since building up my machine at the end of 08. At the time I think it was a decent build- Q6600, 4GB ram (800mhz), GA p35 DSL and a GT8800 GPU.

It's been pretty bloody solid... Until recently! For the first time it couldn't cope with something I threw at it - blue ray rip (1080p), 40GB worth of it! So, I decided that I'd prefer to do as much as possible to improve the performance without spending (too) much.
Having read about overclocking in the past, and monitoring my temps I knew I needed to get rid of the stock cooler so ive got an Arctic freezer 7 Pro and a couple of extra case fans.

So, the OC. Currently I'm running at 3.18GHz - FSB 353 x 9. It feels fairly stable here, it stands up to OCCT CPU test for as long as I've run it - 3-4 hours, and prime blend for just as long. However, if I knock the FSB up just 1 click it errors or BSODs within the hour on OCCT, and any higher it bluescreens very quickly. The above is with 1.44 Vcore before OCCT starts, and drops to 1.38-1.4 when it startS testing. (anyone explain this? Resistance in the cores as temp increases?)

Throughout all, the core temps don't go beyond 66C.

So, what I'm asking is where would you guys take it from here.?I've heard that's 1.5vcore is OK, but also that 1.45 is tops, and some people wouldn't go that far on a Q6600.
I'm thinking I might back it down to 3.15GHz, and then start dropping the Vcore until just stable.
But I'm brand new to this all, and would love your advice!

Secondly, one of the first things I had problems with was my RAM speed. The "memory multiplier" was set to auto, and as a result, my RAM was running up past 1000 MHz. Now I've got it on a multiplier of 2, which is a 1:1 divider, or so I've read. It's running at 706 now, so below it's stock. 2.4 is the next option, which would run it well above stock. What's safe, and what's not? And can someone explain in layman's terms how RAM interacts with the CPU?

Sorry for the endless post, but I'm sure there's a lot of guys on here who will look forward to helping me out.

Thanks in advance guys, Ben

P.S. Forgot to mention, it's a G0 revision.

 

[j.col]

I just replied to your other thread.
one thing though, 1.4v is very high for a G0 revision Q6600 @ 3.18ghz
Most G0's can get to 2.9-3.0ghz on stock voltage.
Have a look at winzero30's pics near the top of this page.
He has his G0 @ 3.5ghz with only 1.36v.

what brand/speed is your ram?
better yet, take a screen shot with cpu-z open with the main page and the memory and spd pages open (just open cpu-z 3 times and show the different pages)
then upload the pic to a free pic hosting site like imageshack or flickr.
then reply to this and show us the pic

 

[DHB07]

Here we go:

My RAM is just value stuff, came from "another BIG online retailer", but i remember reviews at the time saying it had been 'clocked to 1000.
I'm looking forward to this... someone explaining to me what all these timings etc mean!

Cheers, Ben

 

[Marine-RX179]

My RAM is just value stuff, came from "another BIG online retailer", but i remember reviews at the time saying it had been 'clocked to 1000.
I'm looking forward to this... someone explaining to me what all these timings etc mean!

Don't think it matters as long as your ram is not exceeding the rated 800MHz.

For my Q6600 overclocked to 3.6GHz, I just use 400x9, so the memory is 800MHz. I also need to use a vcore of 1.5v to get my overclock stable in prime, but that's only because my chip has a weak 4th core which require more volts to stablise.

What CPU cooler you got by the way? Cause highly overclocked Q6600 does generate lots of waste heat.

 

[DHB07]

Ok, so it seems (to me) that something a little bit dodgy has happened...

I pressed on to 414 x 8. ran OCCT and BSOD almost immediately. Fair enough...
Knocked it back to 404. Ran OCCT and BSOD after 5 minutes.

So, I went back in to BIOS to turn the clock down a bit more, came out, opened CPUZ and it told me I was running at 266 x 8. (stock FSB)

Went back in to BIOS - BIOS is readiong 400FSB and Host Frequency Setting is clicked to "enabled". Fired her back up and the same thing again.

So, for some reason it seems that regardless of what I'm tapping in to BIOS, it won't ACTUALLY run the FSB I'm setting it to, despite allowing me to alter it. Any ideas?

As you can imagine, I'm a little worried now :-/

 

[DHB07]

More as a helper to anyone else, I've manage to fix the problem:

It seems that I'd managed to corrupt the Main BIOS, and that it was booting off the Backup BIOS. However, and this is just guess work, Main BIOS still had the final say as to what FSB I was running. That, or the Backup BIOS I believe I was in wouldn't then let me alter the FSB.

Updating BIOS using QFlash did nothing atall, altghough, atleast now I have the latest BIOS version.

I shorted out the CMOS and, just to be sure, remopved the battery from the MoBo for a minute or so. Now, CPUZ reads the same FSB that I'm tapping in to BIOS. Pretty exciting (read "scary") stuff for a fella who's new to all this!

 

[Scougar]

My board: Asus X48 ROG Rampage Formula with Q6600 VID 1.212v

Previous stable overclock 3.4Ghz. (377 * 9) 1.40V in BIOS, 1.36v idle.
New stable overclock 3.6Ghz. (400 * 9) 1.42 in BIOS, 1.408V with LLC enabled.

Only difference really is changing the AI Clock Twister to 'Lighter'. I've even reverted to using Load Line Calibration to brings temps down, as the voltage was dropping too much even at idle.. now rock solid between idle and load, and vDrop is closer to what is set in the BIOS. Seems to be prime stable unlike my previous attempts at 3.6, even though it's actually hotter ambient temp and hotter temps on the core than I previously had. BIOS tweaks ***

I need to reduce the voltage a bit as temps still too high, and I think my cooler paste and seating could be better.

It's a real shame I'll be selling this board in favour of an MSI P35 NEO2 FR.

This weekend I will try out the IC Diamond paste as well before swapping boards, as I need to give them some results

 

[Scougar]

Since it appears stable even at high temps, I lowered the VCore to 1.40v in the BIOS which gives me 1.384v in windows even at load (Load Line Calibration), and increased the overclock to 3.7Ghz (411*9). This reduced temps and seems stable with prime 95. Still too hot for my liking though under prime 95 so will try remounting the cooler etc.

EDIT: The only thing I have really changed is the AI Clock Twister to Lighter. Apart from that I have changed my fan configuration slightly but the stability even at higher temps (Hi 80's previously) leads to me think this hasn't affected anything.

 

[C64]

Have you tried intel burn test scougar ? I'd say it might need more like 1.45 volts to be stable at 3.6 - 3.7ghz even so summer isn't really the time to be pushing these I leave mine at 3.6 402x9 1.46 volts llc enabled speed step enabled I think as long as cpu heavy games are 60c or less load temp I think that's a good region to aim for i wouldn't be too worried about prime and IBT temps as long as it's stable and gaming load is 60 or less.

 

[Scougar]

C64: Game temps are much lower, at least 10 degrees cooler, so I'm not really worried about the temps at all as I'm not a 'bencher' so it will rarely see the prime 95 temperatures. I haven't tried intel burn test in a while, but can do.

The only caveat is that I have swapped my motherboard out for a P35 board that doesn't have any LLC function, so I will likely have to run the voltage higher.

With LLC on the X48 board, it never crashed with 1.408V @ 3.6Ghz, I was quite suprised. That AI Clock Twister really changed the game for me, it went from being unstable at 3.6Ghz at over 1.48v's to being stable @ 1.408v using LLC (LLC made no difference previously).

One thing to note is that I have always had speed step turned off... I should really enable it!

 

[jennidc]

Which p35 board do you have - as I have an ASUS Deluxe and managed to download a modded BIOS from extreme systems forums which had extra Overclocking options which got me to 3.6 stably (I can run 3.8 with 1.5v but thats just too hot). Day to day I tend to run at 3.2 400x8 @ 1.35 volts its rock solid stable and so much cooler.

 

[Scougar]

MSI P35 NEO2 FR

Apparently there is a pencil mod to help with vDroop on this board, but I've also seen a comment saying it was proven "not to work" which I thought was odd as it does seem to close the vDroop gap by all accounts. No idea if a modded BIOS is available. I have to say that the MSI BIOS seems very limited compared to my X48 rampage formula board.

Previously day in day out I was running 3.4Ghz before I tweaked the bios and got higher @ 3.6. I haven't tested 3.6Ghz on this P35 fully yet as not had time (I'm currently working two full-time jobs this week as a favour, 3.5 hours sleep a day so far lol).

 

[danza]

I've decided to try to OC my Q6600 some more. I bought this thing about 2 years ago, and straight away had it running at 3GHz (333 x 9). It will run at that all day, every day on stock volts (about 1.25 according to CPU-Z), even though the VID is 1.325. I've tried upping it by just another couple of hundred MHz, but not had a lot of luck so far despite enabling LLC and upping the volts to 1.325. However, I'm starting to think it's worth trying to eke out a little more from this thing and trying higher voltages.

Currently seems quite stable at 350 x 9 on 1.33 ish volts, but time will tell. Temps are hitting 66 on one of the cores, is this a little high (bearing in mind that Prime95 will get higher temps than every day gaming and normal use)?

I'm only using an Arctic Cooling Freezer 7, which I think is a bit pap. Maybe i should consider investing in something a bit bigger/more efficient...

 

[Scougar]

I swapped from an Artic Cooler 7 as it was fine at a mild overclock, but couldn't really handle a decent quads overclocked, it got wayyy too noisy just trying to keep it cool. (To be fair, I think some of that was bad case airflow at the time!)

Make sure when you overclock to up the voltage on the northbridge slightly else you will really struggle to get the quad stable at higher frequencies