e8400 4.212ghz stable, next step ?

[Hardup]

So I have finally come to an OC I am happy with and willing to settle with =) The voltages as well as the temps all seem to be fine, as well as stay in check. It is currently 18 hours prime stable, and honestly, that is long enough for me.

Temps under load on Core1 dont pass 57c and on Core2 61c.

After achieveing a stable OC, what would you recommend I do with OCing my memory? This has to be the most unfamiliar area of OCing for me, as im new to it all honestly. This is my first built rig, and first time OCing.

Right now everything in the Bios for mem settings is at Auto, im not really sure where to go with timings, freq, and or any of that.

Current stable 18 hour prime OC.

No mattter what I do I can not get 4.3+ stable but just for the sake of it here is a 3Dmark score at 4.3ghz.

Critiques and comments welcome. Should I be content with this and just stop now ? =)

 

[Shiggity]

Very very nice

If you want the cpu to last longer (2+ years) and don't really need the extra cpu power I'd see how low of a voltage you can get with a 4GHz OC. You could also try to find your max 'benching oc'. By that I mean how high your CPU will go just to run 3d06 or another benchie.

There is a lot of talk going around about 'cpu degredation' with the new 45nm cpus. 1.47V might be a tad too high for 24/7 use if you want the cpu to last a long time. However this is a hot debated topic because there really is no true concrete proof of how long they will last at certain voltages because they haven't been out that long.

But that OC is pretty freakin' sweet for your first OC, good job

 

[Hardup]

Thank you Shiggity!

Would you have any insight on where to go with the memory clocks, and freq's ? That is my next step or so I think.

Just show me the light =)

 

[Albuquerque]

Welcome BTW I'll echo Shiggy's comments and say that's a great OC you have there, so good job...

Tell us what memory hardware you currently have and how much you want to abuse it! 6-6-6-18 seems awfully loose for only 900Mhz. And yeah, I see the 1T command rate, but the 1T really is just icing on the cake -- at this speed, you'd likely get far better performance by running 5-5-5-12 with a 2T command rate.

The obvious remark is simply to continue squeezing it until you can't get any more. Generally speaking, you want that last memory timing number to be the sum of the first two plus two more... So 4-4-4-10, 5-5-5-12, 6-6-6-14, et al. Also, you may not be able to get them ALL down to say 4-4-4-10 timings, but maybe a 5-4-4-11 would work.

Of course, extra voltage and cooling will help too.

Finally, keep in mind that timings are relative to speed. A pair of 800mhz sticks at 4-4-4-10 will actually be slightly slower than a set of 1066mhz sticks at 5-5-5-12... You can figure out the cycle latency by taking (1000 / <speed of ram>) * <clock wait>

So...
DDR2 800 4-4-4-10: (1000/800) * 4 = 5ns latency
DDR2 1066 5-5-5-12: (1000/1066) * 5 = 4.7ns latency
DDR2 900 4-4-4-10: (1000/900) * 4 = 4.4ns latency

Get the idea now? Use this as a guide to help you determine the best speed strategy for your ram's speed and timings ability.

 

[theELVISCERATOR]

what kind of scores do you get in everest for memory?

just wondering..

 

[Hardup]

Albuquerque,
Im messing with some stuff now, thx for the info.

theELVISCERATOR,
How do I go about getting my "scores" in everest for my memory? Let me know and I will get it done for you.

 

[Joeflyer]

Welcome BTW I'll echo Shiggy's comments and say that's a great OC you have there, so good job...

Tell us what memory hardware you currently have and how much you want to abuse it! 6-6-6-18 seems awfully loose for only 900Mhz. And yeah, I see the 1T command rate, but the 1T really is just icing on the cake -- at this speed, you'd likely get far better performance by running 5-5-5-12 with a 2T command rate.

The obvious remark is simply to continue squeezing it until you can't get any more. Generally speaking, you want that last memory timing number to be the sum of the first two plus two more... So 4-4-4-10, 5-5-5-12, 6-6-6-14, et al. Also, you may not be able to get them ALL down to say 4-4-4-10 timings, but maybe a 5-4-4-11 would work.


Finally, keep in mind that timings are relative to speed. A pair of 800mhz sticks at 4-4-4-10 will actually be slightly slower than a set of 1066mhz sticks at 5-5-5-12... You can figure out the cycle latency by taking (1000 / <speed of ram>) * <clock wait>

So...
DDR2 800 4-4-4-10: (1000/800) * 4 = 5ns latency
DDR2 1066 5-5-5-12: (1000/1066) * 5 = 4.7ns latency
DDR2 900 4-4-4-10: (1000/900) * 4 = 4.4ns latency

Get the idea now? Use this as a guide to help you determine the best speed strategy for your ram's speed and timings ability.

This is interesting, as my Corsair RAM(as viewed in sig) is stock at 5-5-5-18 22 2T(800Mhz) So you are saying that with my overclock at 3.4Ghz currently would be better suited with RAM timing to 5-5-5-12? Just curious.

 

[renegade44]

What you running that E8400 on?

Batch number? Should be on the box if retail

 

[blazed]

Different memory works better with different timings, some 4-4-4-12 some 4-4-4-15 etc. As a general rule the last number only needs to equal the total of the first 3, any lower and you can lose performance.

 

[lordkosc]

I'd agree with Shiggity , 4Ghz is a mighty fine overclock, anything more and the performance / heat / lifespan issue comes up, and its not worth it.

 

[Albuquerque]

Different memory works better with different timings, some 4-4-4-12 some 4-4-4-15 etc. As a general rule the last number only needs to equal the total of the first 3, any lower and you can lose performance.

Nope, it's tRCD + tRP + 2, as mentioned here: http://www.anandtech.com/showdoc.aspx?i=3184&p=10

You may find it easiest to set all primary memory timings (CL-tRCD-tRP) to the same value when first testing (i.e. 4-4-4, 5-5-5, etc.), and as a general rule of thumb, cycle time (tRAS) should be set no lower than tRCD + tRP + 2 when using DDR2 - for DDR3 try to keep this value between 15 and 18 clocks inclusive.

If you need more, use more, but that's where you want to start if you can. That linked Anand article has several good pages on memory subsystem tweaking; it's certainly worth the read.

 

[theELVISCERATOR]

Albuquerque,
Im messing with some stuff now, thx for the info.

theELVISCERATOR,
How do I go about getting my "scores" in everest for my memory? Let me know and I will get it done for you.

down at the bottom there is a benchmark tab...do mem read and writes..

 

[Hardup]

theELVISCERATOR,
Here is the Memory Read:

Here is the Memory Write:

I dropped the multi to x8 and up'd the FSB to 500 for 4.0ghz. It is 14 hours prime stable with much lower voltage than the x9 at FSB 468 for 4.212ghz stable.

Not sure which I plan to stick to but if you would like I can run the read and write at 4.212ghz as now I am curious to see how it compares x8 with the lower FSB.

Anyway, let me know if the screens look good or bad ? Im clueless. Heh.

 

[theELVISCERATOR]

just wondering what DDR3 is bringing to the table....hmm...

some of my own scores...

 

[Hardup]

Here are some new shots, better scores, need to mess with clocks a bit

Read:

Write:

 

[MooMasster716]

What GPU are you running?

 

[Hardup]

2x HD 2900 XT's in Crossfire

 

[GTengineer]

A 1.47V vcore may be pushing it with 45nm CPU's. Intel states absolute max voltage for 45nm around 1.45V, which is equivalent to 1.55V for the 65nm parts (Q6600 for example is 1.55V max voltage). Now I know some people run 1.55V with 65nm CPU as long as your temps are in check (I run 1.53V) but ask yourself if you had a 65nm CPU would you be pushing it beyond 1.55V? I am not saying it is going to fail on you but it is a gray area where no one can really say....

45nm (duals)
Page 17, Table 3
http://download.intel.com/design/processor/datashts/31873201.pdf

(sorry cant find the link for the 65nm right now)

 

[agentbad]

Add a flux capacitor?

 

[MadMan007]

A 1.47V vcore may be pushing it with 45nm CPU's. Intel states absolute max voltage for 45nm around 1.45V, which is equivalent to 1.55V for the 65nm parts (Q6600 for example is 1.55V max voltage). Now I know some people run 1.55V with 65nm CPU as long as your temps are in check (I run 1.53V) but ask yourself if you had a 65nm CPU would you be pushing it beyond 1.55V? I am not saying it is going to fail on you but it is a gray area where no one can really say....

45nm (duals)
Page 17, Table 3
http://download.intel.com/design/processor/datashts/31873201.pdf

(sorry cant find the link for the 65nm right now)

I was going to say something about the voltage too. You are risking the chip by running it that high 24/7. It might not die fast like the xtreem voltage people have had happen but I'd be willing to bet it will have problems within it's useful lifetime.