Overlocking E8400 with DDR 1066

[DanielCD]

Alright guys, so this would be my first time overclocking a computer, so I'm a bit confused as to which settings I should use. I want to get to somewhere around 3.6 (or the highest without changing voltages). I've seen that for most people it's just a matter of changing the FSB to 400 and memory ratio to 1:1 since they use DDR2 800, but I'm using 1066 so I'm not sure if I should bump up the FSB to match the memory at 533mhz and just lower the multiplier, or downclock the memory to DDR2 800. But if I downclocked the memory I would also have to change the voltages for it (something I don't want to mess with)

So:

CPU Multiplier: 7
FSB: 533mhz
Memory: 533mhz
1:1
3.73ghz

or

CPU Multi: 9
FSB: 400
Memory: 400 (downclocked)
1:1
3.60ghz

Which would be better? I want to make sure I'm getting everything out of what I paid for so I'm kinda leaning towards the first so I can take advantage of my memory's ability to go up to 533mhz. Am I understanding how it works? Or is there a flaw in my logic?

E8400
2x2gb Mushkin DDR2 1066 5-5-5-15
Gigabyte UD3R

 

[ausch30]

my first suggestion would be to download and run CPUZ and you should get a screen like this

What you'll notice is several different settings for your memory and one of those settings will be for DDR2 800.

You could use the settings from CPUZ for your memory and increase the FSB to 1600 and you shouldn't need a voltage increase. If you wanted to play around with things you could increase the memory voltage and decrease the timings.

BTW, the memory you have is actually DDR2 800, just as the memory shown above, which is guaranteed by Mushkin to overclock to DDR2 1066. The memory above was sold as Mushkin XP2 1066 5-5-5-15 2.05v.

 

[jsc]

Option 1 is theoretically better, primarily because of the higher FSB. However, I have never seen a practical difference in performance.

I also think that you will not be able to reach around 3.6 GHz without increasing CPU voltage and going to an aftermarket HSF.

 

[DanielCD]

FSB to 1600? What do you mean by that?

Already have an after market cooler (not a premium one, but better than the stock cooler). I've heard of people getting 3.8 without changing the voltages, but they probably got a good chip - for the most part I've also heard that 3.6 is the max without changing it. (but maybe I could hit 3.7)

The only thing I don't like about the first is that speed step won't do much as to decreasing the frequency/power usage at idle (3.1ghz at idle and 3.7 at load)

 

[ausch30]

I meant 400FSB

 

[DanielCD]

I tried the first one out, started up but it failed prime95 after around 10 minutes. But for some reason it was locked at a 6 multiplier @ 3.1ghz instead of 3.7 how it should have been. So I went back into my bios and loaded the "optimized defaults" - now when I start it up it only goes to 2.0ghz (6*333mhz) - even under load. I know speedstep lowers it to 6x, but only under idle. Anyone know what could be wrong? If I can't get it working I'll try out lowering my RAM instead

 

[Ancient_1]

If you haven't changed the voltage of your memory that is probably why it failed p95 so quickly. I am actually surprised that it would even boot at that speed without an increase in the memory voltage.

I have no idea on why you are stuck with the 6x multi but would suggest using the stock 9x with ram @ 1:1 and try tighter timings after you get your cpu at a stable speed that you are happy with.

I have two bios profiles saved that I run. The 1st is 430fsb with cpu voltage @ auto plus 3% and mem @ 1.95v with 4-4-4-10 timings. The 2nd is 467fsb with a cpu voltage of 1.38 and disable vdroop which gives me a load of about 1.39 and no drop when idle and mem @ 2.05 with 4-4-4-12 timings.

For the 1st few months I normally ran the 2nd settings but lately I have been running the the 1st since it is much easier on the components and I want this to last a few years as a back-up and file server (in about a month I am going to build an i7).

I believe that you should end up 3.6~3.8 with auto or small increase on your vcore for 24/7 use. It is all trial and error trying to find the best settings for 24/7 use taking into account cpu speed/heat/voltage and what works best for you, others may make suggestions but each setup has its own sweet spot.

Here are a couple shots of how mine runs with the settings.

Idle

Load

 

[Mondoman]

Also remember that on many MBs, "auto" does not mean "no CPU voltage increase"; it just "automatically" boosts the CPU voltage proportional to the FSB speed increase.

 

[Ancient_1]

Yes that is why I said auto instead of stock.

On auto mine sets these values with my 8400 and I haven't seen them increase with increased fsb speeds.
NB 1.3v
CPU 1.2v
VTT 1.1
MEM 1.8

 

[DanielCD]

i was trying with 1.225v/1.23 I think. What's really bugging me is the multiplier.. it's locked at 6 so it's really limiting me. I think that's why it even booted up, even though it was at 533mhz, the multiplier was at 6x for a total of only ~3.1ghz. another thing, when i use 533mhz for the FSB the CPU cooler is noticeably louder

 

[DanielCD]

- 3.6ghz (9 * 400) - vcore auto - wouldn't boot up (think I had the wrong ratio)
- 3.7ghz (7*533) with vcore on auto (cpuz reporting 1.4v..), started up but failed P95 instantly
- 3.4ghz (8.5*400) vcore at 1.25, testing right now, seems stable. I chose 2.66D in the gigabyte bios for the memory multiplier, and cpuz reports that i have a 3:4 fsb:dram ratio. But now it says 5-7-7-24 in cpuz for CAS latency, tRCD, tRP, tRAS where before it was 5-5-5-15. Did I do anything wrong?

 

[Ancient_1]

But now it says 5-7-7-24 in cpuz for CAS latency, tRCD, tRP, tRAS where before it was 5-5-5-15. Did I do anything wrong?

I would guess that is a setting in the spd of the ram for 533/1066 speeds which it looks like it is running now. You should be able to change the settings to what the ram is rated at as long as you set the voltage that is stated for those speeds and timings.

Also at the 533 fsb speeds you may need to add some more NB voltage to get stable at that speed even if the ram and cpu are stable.

 

[DanielCD]

I set it to 5-5-5-15 and 2.060v (it says 2.050 for 533mhz in the timing table) but it still comes up as 5-7-7-24 in CPUZ. Also, how would you tell if the NB is causing instabilities at the current voltage?

 

[Ancient_1]

Also, how would you tell if the NB is causing instabilities at the current voltage?

All you can do is raise it 1 or 2 notches and see if that solves it. Reasons why overclocks fail is usually trial and error and trying to take out as many variables as possible as you go. That is why I would run the mem at a 1:1 ratio and the cpu multi at default until you find the cpu's top speed so you are taking out the ram and high fsb speeds as likely causes of failure.

That is what I do while I am testing for the top cpu speed and use occt to test untill it fails and then either raise the voltage to regain stability or lower the cpu speed to get stable.

it says 2.050 for 533mhz in the timing table

I would guess that is under an EPP profile which your motherboard wont read since you have an intel chipset (the EPP are read on nvidia chipsets).

To see your actual ram timings and speeds here is a link to memset a nice mem util.

 

[DanielCD]

Just did 3dmark06 with the CPU at 3.6 and got an 8k (1500 cpu score). Last time I did it when I bought the system I got 14k . I think I'm just gonna put it back at stock for now and do some more research

 

[DanielCD]

Decided to put it back at stock and I'm now getting a score of 6500--if any of you guys know what I could have messed up the info would be greatly appreciated

http://www.tomshardware.com/forum/255618-30-drastic-3dmark06-performance-plummets-bios-ud3r

 

[ausch30]

Decided to put it back at stock and I'm now getting a score of 6500--if any of you guys know what I could have messed up the info would be greatly appreciated

My first question is did you use the same video drivers for all the tests? Also you didn't list the other components of your system so I don't know if you have a Nvidia video card. There are some drivers that offload certain duties, such as physics, from the CPU to the GPU so just by changing drivers can have a huge effect on 3DMark scores.

Here's an example I did a while ago with my system. The only difference was an Nvidia Beta physics driver in the second test.