Quick heads-up from the industry: I took part yesterday in a presentation of new motherboards by a global motherboard brand at their Taipei HQ. The new MBs run the E6750, Intel's first (of many) Dual Core chips with 1333 MHz at the front end and it was amazing!! The processor runs stable OC'ed to almost 4 GHz. Launch will be in July, prices for other DC models will drop around 25% around July 20-31.

Just curious, to OC the E6750 to about 4Ghz, does it need very very good CPU air cooling??Alkit

It appears that something significant happened with this next batch of CPUs. The 1333 front side buss spec is really just a higher-is-better sales ploy. O/C-ers have been running 1333+ FSBs since the C2D's day-1 release. 1600-1800 mhz FSB (actual FSB would be 400 to 450mhz, remember?) are entirely ordinary in the O/C-ing community. 2000 mhz FSB (500 mhz) is completely available with the new P35 chip set. The 1333 FSB standard is a non-issue for O/C-ers. What might be actualy happening is that All the CPUs (in this price drop release) have been sprinkled with some sort of magic dust. I expect Intel's new "G0 Stepping" CPU mod is this magic dust. The CPU that are still factory pre-set for a 1066 FSB have this mod too. For instance, the Q6600 will probably run within 100-200mhz of the 6750s. A (potential) 100% increase in Horsepower costs < 5% in RPM. I'll take two! If you are going to O/C, you are going to rock right past any 1333 FSB speed anyway. Careful not to be drawn into this irrelevant 1333 FSB feature. It's not a performance factor, only an indicator of Intel's confidence in this CPU's ability to GO FaSt.

Don't forget, when you're messing around with these high FSB overclocks, you are also relaxing not only the timings on the RAM but the timings on the chipset as well. Thereby affecting the memory bandwidth.I've been running the standard DDR2-1066 with a small bus overclock and at the moment, it seems to have a bit better performance than dropping to DDR2-800 and uping the FSB by a great extent. Not that I have convinced myself that this is best, but chipset timing do affect peformance (memory bandwidth) as well.One thing is for sure, overclocking my C2D is a lot more work than OC'ing all my prvious AMD's over the past few years. Too many variables in the equasion(sp?).

Without doing anything, DDR2-1066 memory will run at 533 mhz on 1066 mhz -QDR-, 266 mhz -actual- FSB system. (BTW, you're right. It's enough to make one dizzy!), so . . . I assume you have ram rated at DDR2-1066 and used the memory multiplier to increase the memory's default FSB's double data rate (DDR, or 266 X 2 = 533mhz) to a quad data rate increase (4 X 266 = 1066). This allowed your FSB to stay at the standard 266mhz and your memory to run at 1066mhz. But then what? You changed the memory multiplier from the default 2X (DDR) to 3X so the memory would run at 800mhz. The memory was rated to run at 1066, so you then turned up the FSB to 355mhz to drive the memory to its rated 1066 mhz (3X355=1066) right? However here, the CPU is also being driven faster by the faster FSB, so a comparison would not be possible. Also, chipset timings do loosen as the FSB is increased. They are called "Straps." But the performance effect is totally negligible. Generally, strap sequencing causes problems when the chipset doesn't auto-shift to a looser timing soon enough. When this happens, you have a FSB "hole." That's a range of FSB speeds where the system won't operate at all. Taking the FSB even higher will eventually cause the the chip-set to finally shift to the next (looser) strap and allow the overclocking adventure to continue. This still sounds like it might be a helpful FSB test. Could you clarify?

Yes, and you're making me more dizzy, LOL...OK, DDR2-1066 on my Corsair PC8500 4x1GB sticks is defualt with 2.66GHz E6700 (4x266) quad-pumped. Right now I'm running 2.8GHz with the DDR2-1066 strapping for an efective FSB of ~ 1120Mhz. I can't exactly explain why exactly but at the moment, this seems more stable an quicker than when I attmped the 400Mhz/DDR2-800/1600 (effective) I found that black-hole to be somewhere b/w 360 & 400CPU. Won't even post. Memtest currently shows about the same bandwidth so I chauked the current performance up to tighter timings of the strap. Does that make sense?Here is the guide I used to get started: http://www.thetechrepository.com/showthread.php?t=41And this is the one that I keep refering to and the crazy chipset strapping: http://www.thetechrepository.com/showthread.php?p=32#post32This is one of my favorite chipset graphics from that same post:http://forums.avsim.net/user_files/175261.jpg

Yea, I've seen this chart too. The way they explain this is enough to baffle a rocket scientist. I'm a bit dense, and that's probably my only salvation.Manually setting the strap only manually sets the chipset's timing. It exerts manual control over a process that could be happening automatically as the FSB speed is increased. No problem there. We like to play. The main takeaway is that manually setting the strap does not effect the clock speed (or "RPM") of anything. However, it could effect the data rate's "bandwidth" (the rate that data is transferred at any given component clock speed). With a 1066 strap manually set, the FSB (and everything else that is driven by the FSB, -- such as the memory and the CPU clock speeds -- ), are still running at the clock speeds that are set in the Bios. The only difference is that the Front Side Buss (the transfer device) will be operating with looser timings and therefore less efficiently. This may result in lower "bandwidth" (data transfer speeds) at AnY given component RPM (i.e., CPU or memory clock). This speed difference is entirely irrelevant to any actual real world application and would be only of interest to the maddest of scientists, like me! This is why you are seeing more stable ops with the 1066 strap set. The Northbridge is running with easier timings and the system will tend to be more stable. Sounds like something else is going on, because the system should be absolutely stable in 'Auto" as long as you stay away from those pot holes! I agree with your Memtest conclusion. As we were trying to show, the faster FSB made no difference in memory performance. The better performance occurred because of tighter Northbridge (FSB) timings. The slower FSB clock speed had plenty of headroom to handle the increased bandwidth the memory could deliver from those tighter timings. And so it will go. Laptops have been running C2Ds on 166mhz (QDR 666) FSBs until just recently. The Santa Rosa line finally upped the FSB to 200 mhz (QRD 800). I've had that on my 865 chip for 4 years now. But even that ancient (?) 166 mhz FSB would still have been plenty of "racetrack capacity." There's not even a real need for the 200mhz FSB, yet. Intel just decided it was cheaper to use the 965 chip instead. . . . and now they tell us we gotta have a 333mhz (1333 QDR) FSB? Geeze!

I think I just found the board for my next custom build. :-jumpyStable at 4GHz? wow...

>I think I just found the board for my next custom build.>:-jumpy>>Stable at 4GHz? wow... Sounds good on paper, but don't look for it to work magic with FSX!

I love listening to you nerds explain this stuff:)

First system is out here in Taiwan, under USD 900: ASUS P5K-VM with the G33 chipset and Q6600 2.4G(Source in Chinese: http://shopping.pchome.com.tw/?mod=item&fu...AAAV16&ROWNO=2)This is the one I saw running at 4.0GHz last week. Now we are talking. ==Marten:Weber==P-D920 oc3.64MHz 2GB RAM WinXP ATI X1900+1024MB 2xSATA250 2xVIEWSONIC 19"Need AI? Check www.world-of-ai.com!

>I love listening to you nerds explain this stuff:-walksmile Takes one to know one!