Taking the C2D plunge.

Going with the E6700 (I'm getting an awesome price on this.)
EVGA 680i A1 revision motherboard (the revision that fixes the RAM problems with the AR)
2 x 1GB XP2-9200 Mushkin.

I have a great power supply already: PCP&C 1KW. I also have watercooling.

This is my first experience overclocking a C2D or a DDR2 based system. My past overclocking experience has been mixed. Good skills (at least with AMD and DDR1,) unlucky at parts. If I buy that hot stepping, I'll get the outlier Basically I take the average overclock I see with a similar set of parts, and drop it by 300MHZ and that's my target. If you say 3.6GHZ, I'll plan on 3.3 fully stable.

Overclocking with DDR2 doesn't seem as straight forward as DDR1 was. Will I be able to hit 3GHZ or higher with a 1:1 ratio on my ram or will I have to use dividers? I've never had luck using dividers with DDR1 and AMD. Sure I got a higher overclock but less performance overall. How is 1T with DDR2?

Basically I consider myself a beginner again with overclocking this new technology so any tips or direction would be considered useful. The stickys seems to be antiquated.

what ram problem? the only reason the revision changed was because the a1/t1 oc quad cores better. not all ram will do 1t, better off just leaving it at 2t.

i was stick with the 1:1 ratio to find the max of the cpu. with a 10x multi its not going to take much. i betting you would be able to get 333/334fsb or 3.33ghz on that cpu with out much trouble. after you find the max of the cpu you then can use the ram ratios to get your ram to run at rated speeds.

The A1 revision did address RAM issues over the AR, perhaps that's why the QC overclocks better.

Anyway it's tangental to the point. No need to get sidetracked.

I'm assuming you mean find the RAM max first then use the dividers to hit my GHZ peak. It sounds like you're saying that I will have to use dividers to hit my CPU max, is that correct? Just trying to clarify so bear with me. Regarding 1T I've heard it isn't the difference maker it was with DDR1, is that correct?

Thanks in advance.

well me being in the NV user group NV stated there was never a problem with ram and the first board. i still have the first run board or AR as they are called. what im getting at though is what is the ram issue you say this board has, cause i have none. there is the problem with people expecting their ram to run find and dandy with 2.4volts + 24/7 with out a problem. you cannot safely run 2.4volts thru ram and expect it to be fine and dandy. there is a post over on EVGA forums about ram issues being with high ram voltage.

about the ocing your using a ratio for fsb to ram either way you look at it. 1:1 ratio just keeps the ram running at the same fsb as the cpu. more over the NV's do it a bit differently then intel boardls. say you tell it 1:1 ratio right, and you actual fsb is 266, now instead of the ram running at ddr2-533 it will be running at the quad pumped speed of the fsb. being that 266x4=1066 the nv 1:1 ratio puts the ram at ddr2-1066. you may have to use the sync mode to get it to run up to 800 if i recall correctly. i didnt take much notes on the board since i started using mine about a month ago. what im saying is this if you had a intel board you would be able to have the ram tied to the fsb being used like 200/266..so on. this way the cpu maxes out first before your ram. this way you can find out the limits of the cpu then go back and get the ram to run at its stock speed.

What you're saying regarding the 2.4V makes sense and is consistent with what I've just researched. In fact, I hear OCZ isn't recommending above 2.2V on the 680i.

I'm glad you keep your predictions conservative. I've seen people advertise 3.6 on air and I'm skeptical of the true stability of the system. I go for completely stable overclocks with multiple runs of multiple benchmarks and real world applications. I've seen 12 hour prime stable systems fail after 2 hours on the subsequent run or in a game.

So, keep the ram ratio at 1:1 and presumably unsynced and I should be able to find my CPU max on my EVGA 680i A1. BTW: I have SLI capable memory. Does that impact overclocking at all?

Regarding 1T I'm hearing that it's not a big difference maker in bandwidth like it was with DDR1. Is that true?

My parts come in Tuesday and I feel like a kid again

Depending on what stepping your E6700 is, I'm guessing you'll get around 3.4 - 3.6 stable. Mine does 3.7ghz (8 hours orthos stable) on air.

well i dunno how to explian the 1t vs 2t on ddr2. while some have been able to get it to run fine some have had problems with it. i really dont mess with that setting, really when i saw benches it was that big of a difference. my predications are more RW the just out of thin air since i look over a lot of posts.

while i can get 3.6ghz on my E6400, mine is the first run with B2 stepping not the L2 stepping. on my cpu though i can run 3.33ghz @1.375 volts with out issues. i have to use something like 1.425 or a tad higher for 3.4ghz stable. i havent done much testing at 3.4ghz as i was looking to break a few of my personal best benches in superpi 1m. i will be going back for 24/7 stable rig, i like to see my temp right around 66-67 max so it if starts peaking around those i tend to back it down.

the one thing with ocing for sure with c2d's is get the highest fsb possible to the cpu and highest ram speed. it makes a big difference, see here why i say that.
http://img.photobucket.com/albums/v733/Evilsizer/ocforums/superpi 1m runs/34g1140g.jpg
http://img.photobucket.com/albums/v733/Evilsizer/ocforums/superpi 1m runs/1m33g11p352.jpg
look how close the second one is the first, as well that is p965 vs p35, e4300 vs e6400.
also look at this one and compare it to the second i posted
http://img.photobucket.com/albums/v733/Evilsizer/ocforums/superpi 1m runs/33g11g.jpg

soon as i get windows reinstall in my ip35 im redoing some benches since it was using the 680i install for that run, the second one i listed.

Got the parts and here are the results.

First of all I had to use a spacer to give my koolance waterblock enough clearance over the capacitors I don't like that one bit.

Right now I've hit 10x360, 3.6GHZ with 1.425 volts (measures 1.36V according to the MOBO sensors.) Temps peak at 50C under load. I'm calling this the CPU max because I needed a really good volt hike to climb from 3.4 so I'm at the point of diminishing returns returns. Maybe I COULD get 100mhz or even 200 more but I'm calling it good for now. It passed two consecutive Orthos passes for 8hours each and 3dmark looping for eight hours. I'm gaming just fine so I'm satisified with its stability.

I set the BIOS to unlinked, running 1440. I set my ram to 720 in the EVGA BIOS and a 1:1 ratio. AFAIK, the RAM is not being quad pumped with this setting and reads 720 total (360X2) SLI Enable for the RAM is disabled.

So now on to maximize my RAM performance. Should I just keep the settings as is and lower my timings or try it at stock quad pumped rate (1143 for my XP2-9200 Mushkin) and work up from there. Right now the BIOS settings for linked-unlinked, SLI Memory, synced, unsynced, etc are really confusing.

I'm also seeing a lot of settings for overclocking the PCI-E buses. Are any of those beneficial?

just to make sure thing run in spec set all the pci-e speeds to 101. if you know what your ram can do stay unlinked and just type in the fsb the ram will do. ie 500x2=ddr2-1000, as such, yea loosen the timings to meet stock and set the stock voltage as well.

Thanks a bunch.

Maybe I'll try to push it to 3.7 or 3.8 with a different waterblock that won't require me to use a spacer. I'm a little paranoid of the spacer.

Some have told me to run it at 800 (400X2) and tighten the timings as supposedly quad pumped DDR2 really doesn't benefit much because of the higher latencies required.

mongoose470 said:

Thanks a bunch.

Maybe I'll try to push it to 3.7 or 3.8 with a different waterblock that won't require me to use a spacer. I'm a little paranoid of the spacer.

Some have told me to run it at 800 (400X2) and tighten the timings as supposedly quad pumped DDR2 really doesn't benefit much because of the higher latencies required.

Click to expand...

cant quad pump ddr2, as ddr2 mean dual data rate. now they board it self when you set to 1:1 the ram is set to match the cpu's quad pumped bus. you will notice that on intel boards this is not the case, when you set 1:1 on a intel board it will match the FSB you are selecting. IE 400mhz = 400mhz, not 400mhz = 800mhz.

Well I upped the RAM to stock speed (1143) and it is crashing. Funny when the CPU is at stock, I can run those speeds on the RAM. I'm just not getting this, I guess.

At RAM=720 it's perfectly stable but memopry bandwidth is low.

Here are my pertinent BIOS settings at stock, which crash.

SLI Ready memory = Disabled
FSB Mem Clock Mode = Unlinked
FSB(QDR) = 1440
Mem(DDR = 1152
No memory ratio is there of course for unlinked.

Voltages are:

CPU Core............................ - 1.425v
CPU FSB............................. - 1.4v
Memory.............................. - 2.30v
nForce SPP.......................... - 1.40v
nForce MCP.......................... - 1.50v
HT nForce SPP <-> MCP............... - 1.20v

Memory Timing Setting
Memopry Timing Setting.............. - Expert
tCL (CAS Latency)................... - 5
tRCD................................ - 5
tRP................................. - 4
tRAS................................ - 12
Command Per Clock (CMD)............. - 2T

tRRD................................ - Auto (5)
tRC................................. - Auto (33)
tWR................................. - Auto (6)
tWTR................................ - Auto (11)
tREF................................ - Auto (7.8uS)

System Clocks
CPU Multiplier...................... - 10x
PCIe x16_1 (Mhz).................... - 100MHz
PCIe x16_3 (Mhz).................... - 100MHz
PCIe x16_2 (Mhz).................... - 100MHz
SPP <-> MCP Ref Clock MHz........... - 200
nForce SPP --> nForce MCP........... - 5x
nForce SPP <-- nForce MCP........... - 5x

Thanks...AGAIN

change the spp's to 4x, give that ago, also some have said with 680i that you dont need to use alot of volts. give the ram ago with 2.2v and use memtest86.

Nothing worked. Memtest showed the memory okay. It crashes immediately in Orthos.

Lowering the CPU back down to 2.66ghz allows me to run unlinked DDR 1150 (which is what the memory is rated for.) If I go to 3.6ghz, I can run orthos without crashing ONLY at unlinked QDR 1440 DDR720, or linked and synced, same thing. Won't even run @ DDR960

I'm thinking the overclock is too high. Perhaps running the memory at full bandwidth is stressing the CPU more than when I was clocking down the ram to find the CPU max.

when you test in orthos are you doing blend or just stressing the cpu with gromacs core?

cause if you stressing the cpu then you need to up the voltage to it.

I'm doing the blend test like I always did when using Prime. It will crash only on large FFTs which indicates a RAM possibility, or who knows what. Small FFTs loop forever. I fired off a post to Mushkin and hopefully I get a reply.

BTW: At those settings I was pulling about 10,268 reads in Everest. At DDR 1150, is that about par? Just curious.

Now I'm at DDR 800 speed. I dropped the timings down to 3-3-3-8 and am getting the same bandwidth. Fully stable. Still, I paid the bucks and I've heard of people pulling 12k in everest with these sticks.

BTW: I have the latest BIOS for my motherboard (p28)

Further strangeness. Let's say I set my QDR to 1440 and Mem to DDR 1150. It will actually report as 1131. That does NOT happen at lower or Stock QDR.

It may all be irrelevant but who knows.