• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

RAM possibly Bad?

Z

Zayla74

Weaksauce
Joined
Apr 26, 2008
Messages
64
Having a odd issue .. trying to keep stable 3.6 Ghz.

I have ran MEMtest 86 @ rated speeds for my ram and passed 3 passes. Maybe not enough .. which was 2.1V 800Mhz @ 4-4-4-12. I also tested it using a 333/400 strap which puts it at the 960mhz rang 5:6 ratio (5-5-5-15) and passed there to.

BUT ....
When I go to stress test...

With 1.36 NB and 1.18 VTT 1.3Vcore
I seem to always fail Linpack at these settings/timings ususlly within the first 5-7 tests.
Never even make it to the Prime testing.
(went as far as 1.3V to CPU). Both ratios 333/667 or 333/400 turn up the same.

Eventually I turned down the Ram. to 2.0V and 5-5-5-15 timings. @ 333/667
Now heres whats odd ...

I was able to reduce the Vcore. 1.28 (1.26 under load)
Also .. turned the VTT back to default 1.14 AND NB I lowered to 1.32
ending up with 3.6Ghz core speed and 800Mhz RAM speed @ 5-5-5-15.
Now when I test:
25 passes of Linpack ok
Prime 95 went for 11Hrs 58min. and failed on 1 core. Im thinking this could possibly be NB Voltage .. might be wrong.
At any count, considering 25 passes of Linpack was OK, I tend to think Vcore Ok.
Or am I completely wrong here?


So now my question is, is it possible my RAM is bad?
I cant see why I can not keep it stable under Linpack if i set to advertised timings 4-4-4-12 and 2.1.
But at 5-5-5-18 it will run linpack all day at 3.6Ghz and 1:1 ratio

Any advice where to look from here? Thanks.


Question 2:
Is it adviseable to turn EIST and C1E on after I find a stable OK? Or is this a bad idea.
I have seen mixed reviews on it. I do understand the reason turning it off while finding a stable OC .. but just curious what ,if any , disadvantages may come from it.
 
Question 2 first, it is the easy one.
I advise to turn it on to reduce stress on the machine when under light loads. It should not affect your OC at all ( should being the operative word). In general its a good thing and works great. As soon as you start to run something that needs the power/speed in a millisecond the computer ramps up to full speed. You will not even know its been enabled it works so well and it will save you a couple of bucks of the electric bill too. I ran my machine full bore for a year and then started using it. I never noticed any difference when using the machine and now I leave it enabled all the time. At least after you get your OC issue sorted out, give it a try. If it does not work for you it only takes a minute to reboot and disable. I recommend it.

No I don't think your memory is bad. I think its the northbridge but it is hard to say. That board is very complex and it is almost impossible to point to one setting and say "do this". All I can do is tell you what I would do in that situation and you may have done it already but here goes:


Reduce the CPU multiplier and give the CPU the max voltage you have tried that is stable. This will under-clock (compared to your normal OCing settings) the CPU and take it out of the picture. Now start a careful methodical testing plan of the Northbridge and memory subsystem by increasing the FSB until the memory starts showing errors on memtest86+. Increase NB voltage a notch and retest. Rinse and repeat until you cannot get any more or reach a NB voltage that scares you. Find the FSB limit at those memory timings. Now play with Vtt a bit and see if you can get any more. (probably not). Now loosen the timings reset the NB voltage and do the same thing. You now have a good idea of what you memory system can do at performance and loose timings and what NB voltages are needed to obtain those goals. Perform this testing like a scientist and take notes. Done correctly this will kill an entire afternoon or more but the information you gain is invaluable. Use small FSB increments once you start to get into the range where you know the memory will be stressed. For normal operation you typically have to back off a few MHz from whatever you find as your max FSB.

You have a zillion settings on that board. it is a tweakers dream but I strongly suggest you leave as many as possible on auto until you do the basic testing. Then you can go back and add to your base of knowledge what adjusting one setting at a time, does to your abilty to OC.

I will also comment that that DFI is extremely sophisticated with all the adjustments and you might want to find a forum specifically for DFI boards with DFI expects hanging around.all the time. But dont go away forever! LOL. I would look for reviews on that board that might contain hints as to what settings are condusive to OCing and which should be left alone. It is a beast and my best advice is to keep it as simple as possible to start and change only one thing at a time and take notes. You have hours and hours of tweaking fun ahead (well I think its fun - but I am absolutely insane ).

Best of luck. And hopefully a DFI expert will wander by.



Oh and keep in mind it is a SYSTEM, while the memory will test good at some frequency using memtest, its a whole different ballgame when you start doing real world stuff and the cpu is processing complex commands, video is being pumped through the NB as well as memory data etc. etc. etc. So finding the upper limit of a cpu or memory in standalone testing does not mean it will do that when you have all the sub-systems working at the same time.

1)
 
Thank you for the advice.
I tried dropping the multi to x8 and raising the nb using the 4-4-4-12 timings but didnt work for me.

I went as far as 1.46 ... but stopped there as I personally felt it was a bit much seeing as the FSB is only 400. Just seems like such a huge leap just to tighten the timings :(.
I even tried with the Vtt @ 1.2 instead of 1.14 but still didnt help .. :( failed Linpack within first 3 tests. :( I am really running out of ideas where to look.
Starting to think it has to be something in the mem section of bios im missing beyond the stand 4-4-4-12 setting.

There is a setting i noticed in my Bios for MCH ref and 1 &2 are at 00. and a 3rd at 80 with tighter timings. But no idea what its for ... of course theres several settings im not sure about.


On a good note .. I did manage to get 3.8 Ghz with ease. 2.8Vcore and 3.6 NB. RAM set at 800mhz 5-5-5-15 tho. Stable through 30 Burn test passes.
With RAM at this setting its also rock solid stable at 3.6 and a 1.24 Vcore.
Another reason I found it odd it would need more than 1.46 on the NB just to drop timings a bit. Somethings not right :( and I have no idea where to look now.
Though the diff. between 5-5-5-15 and 4-4-4-12 may not even bbe noticeable from what ive read.

The issue seems to be def. be isolated around RAM or NB somehow... im guessing theres a setting that needs to be adjusted somewhere in bios im missing .. but ... I am fairly new to OCing Intels and although i love my DFI, its is NOT a newb friendly board :(
 
You must log in or register to reply here.
Back
Top