DDR3 1333 stable, 1366 fail?

kidscruff

kidscruff

kidscruff

Associate
Joined
23 Feb 2003
Posts
372
Hi, I'll spit this into two parts. A quick question, and then a longer query if you don't mind reading a bit:

First, my memory (OCZ3P1333LV4GK) is rated at 1333 7-7-7-20 1.65V. Set these stock timings up (actually 1.64V in the BIOS), which actaully proved unstable (OCCT failed after 50 minutes). Noticed the motherboard AUTO base clock was 136.7, so I locked this to manual 133. This seems to have stabilised the RAM, all other settings are the same. As a test, I dropped the memory voltage to 1.60volts, which with the AUTO base clock instantly errors in Prime95 every time. Locked the base clock back to 133 and it survived for almost an hour before erroring.

Is this normal that only a few extra MHz can make that much difference? I presume a small increase in voltage may stabilise the ram at 1367MHz, but I don't want to push outside the recommended limits.

As a note, memtest86+ v4.0 passes every time, even at timings/voltage settings that I know will fail Prime95/OCCT instantly.

ok, and now for the longer query:

I've been testing stability and memory timings without the video drivers installed (using Win7 64bit default drivers), as I've previously been getting a lot of graphics card VPU recovers and random crashing at idle (strangely, not when gaming). This may have been a symptom of my system being unstable, which is why I'm getting it all working before I move on to diagnosing the video card.

I originally had a Corsair HX450W PSU, but since that's pushing it for the Core i5 and 5850, I upgraded to a Corsair HX650W PSU. That's the point when I started getting ramdom freezing, from the very first boot! (i.e. nothing responds, except a hard reset). I hadn't had a single freeze with the 450W, just the VPU recover messages. I have no idea why my system seems more tempremental with the new PSU, perhaps it's just subtle voltage differences.

Before the PSU upgrade, I had been running my memory on auto bios settings (1066MHz, 7-7-7-19 1.5V), and whilst the timings seem fine, I think the voltage was just too low. Perhaps that made it worse over time?

Another possibility is the CPU socket having bent pins. Seeing as I spent such a long time getting a neat and tidy build, I'm reluctant to start pulling out all the components to get at the CPU. And I need some more thermal compound if I do so! I really don't know if my problems could be related to this?
 
I can't remember where I got to with this memory, but it doesn't like higher speeds, even with slacker timings.

I'm running mine at 1140 (6*190), 6-6-6-18.
 
i have been getting this freezing you mention, latest one was yesterday, the memory i have is ocz brand, the psu is corsiar hx650, i am OC on my cpu to 3.6 but im sure thats perfectly stable as it will run prime95 for over twelve hours, also noticed its fine if the system is gaming , it generally does it at idle for me, mainly browsing internet, i never experienced freezing on my older system with same psu, so im leaning more towards the memory, just very hard to pinpoint as its not repeatable
 
hyperst, I've read what appear to be your posts on the OCZ forums, as well as the replies you've made to several of my posts here at OcUK.

It does seem there's something strange going on, either with the P55/i5 combination, or perhaps even OCZ memory. Like I said, a few extra MHz seems to cause system freezing. The memory could be extremely fussy with bios settings, but I find it hard to believe that the memory (at stock speeds) is operating at it's absolute limit! Or less than stock in your case.

But then I've also had most of my freezes/crashes at idle (NB: video crashing only occurs with drivers installed), which kind of points toward the CPU power-save modes?

I may disassemble my computer to check the CPU socket, but I don't think I'll find anything. Which will be a pain... but I'm at a position where I might actually try some different ram, even though memtest gives zero errors.
 
well i headed over to ocz forum aswell;), there are quite a few users there complaing about the freezing thats happening, trouble is there the response from the mods/people actually giving out the advice seem to know very little about the p55 chipset and its voltage limits/its memory dividers, they seem to randomly say throw 1.7 v at the ram , when its quite clear intel states 1.65v max or damage to cpu could occur.

if you do try different ram please update with the results ,as im thinking of swapping this ocz out
 
Last edited:
Today I ran LinX for about 5 runs, using ALL memory. Ran fine, so I closed the program. Then went to click on Chrome browser and freeeeze!

So updated the BIOS (again) to latest, which coincidently drops the Auto base clock back to 133. Memory timings back to standard, vdimm 1.64V. Froze during web browsing!

Upped VTT/QPI (uncore) voltage to 1.15, in keeping with the 0.5v vdimm-qpi rule. Currently running ok. But problems are so intermittent... LinX and the others may draw out the error eventually, but I'm amazed it can pass multiple LinX runs, or an hour of OCCT, then just freeze when nothing's happening.

[edit:] I also noticed the OCZ guys always suggesting very high voltages. I got shot down by suggesting a vtt/qpi voltage of 1.35V was too high for Core i5, but constantly recommended by OCZ.
 
Last edited:
yeah i have just learned the maximum you should put through the VTT is 1.21v mines at 1.31 since my last freeze and so far it has not froze on me, trouble is it seems my motherboard see's this voltage as ok i.e it's in the bios under a white setting and not a red, i know wonder if msi have over looked this about the vtt voltage
 
Mine does also. Perhaps 1.35 is ok, perhaps it's not been changed from the i7 specs.

You're running Vista 64bit right? I have Win7 64bit, and just noticed the "High Performance" power setting stops the CPU multiplier dropping to x9 at idle. Will see if that helps things at all. Ran 20 loops of LinX at size 20000, no problems.
 
yeah vista 64, one thing is my multi has never dropped below x20, even when loading defaults in bios, so what do you have you vtt v set at
 
The freezes are definitely associated with the ram?

Thank you for this, I'm going to throw more voltage at my ram, qpi & loosen the timings to see if it runs stable. Cheers
 
Perhaps not "definitely" the ram. CPU core voltage may help run a system more stable, along with many factors! At the moment, memory timings and voltages seem to be the most influential, but it's possible it might be something entirely different!

Currently VTT at 1.15V, but I'm running everything at stock. The x9 multiplier is the Intel Speedstep kicking in. In XP I don't think this was supported, so only got the x21 to x24 turbo running single core. Pretty sure all this disables if you change the base clock.
 
Well, it refuses to boot at cas9 so I guess I should stay at 8. More qpi voltage appears to have made no difference whatsoever, same with more ram voltage. Weird issue, the system is passing ibt but crashing in use.

Trying silly vcore now to see if that resolves it, currently running at 1.425V.
 
Surely CAS9 is looser/slower than 8? I really don't know enough about overclocking to help you with the voltages. I presume everything is fine at stock speeds right?
 
one thing i did note aswell when it happens is all my internet explorer tabs i have to sign back into, usually when starting up im allready signed into ocuk, ebay, and such stuff, when it freezes and i reset everything is gone, like i cleared my cookies
 
Yep, CAS9 really should have worked. Odd that it didn't. I'm sure I'm missing something fairly trivial here.

Stock speeds are fine, 4ghz is very stable, 4.2 I think is stable. Currently 15 minutes into 4.4 and it hasn't frozen yet, but I won't feel confident for another hour or so. Xtreme systems thinks any vcore less than 1.5 is fine, but I'm not sure they're as keen on 24/7 use as I am. At least it's cold.

Regarding your first post, a few mhz can make all the difference. CAS 7 is pretty tight, when taking the frequency over 1333 it's well worth changing to cas 8 manually, and probably increasing the voltage too. Your ram will be fine whatever voltage you feed it, cpu might be upset over 1.7V. The intel limit of 1.65V isn't the whole story, as long as ram voltage is within 0.45V of the QPI/VTT voltage you'll be fine. So if your qpi is 1.2, the default, don't set the ram voltage over 1.65. If your qpi is at 1.4, ram is good up to 1.8 or so, and I wish you luck keeping it cool.

Ram these days seems to use xmp, which changes loads of settings. Increases bsck, changes qpi voltage etc. I don't like this much, preferring to set it manually. You may find you have more luck disabling xmp, setting voltage to 1.5 to 1.6, timings to cas7 and increasing qpi slightly if needed.

Don't worry about the cpu socket. The issues with foxconn sockets just don't matter until you're running well over 4ghz. Keep playing with the settings until you get it stable, then start pushing bsck upwards. The guys at ocz know what they're talking about, though understandably won't care that much if your cpu dies. The guys at intel don't really care if your ram runs at spec or not. Hence ocz voting for more voltage and intel voting for slower ram.

Enjoy :)

edit: Your copy of windows might be knackered by this point too, I'm confident that my system freezing isn't windows failing since it freezes under linux and in the bios too. Otherwise I'd be reinstalling windows at default settings.
 
You must log in or register to reply here.
Back
Top Bottom