Corsair CM3X1024-1600C7DHX v3.1 & v4.1

just to be clear, its the northbridge voltage i will be increasing. ram voltage i have always kept at the 1.80v mark, by using the bios setting of ddr1.5v +0.30 to get 1.80v, everest confirms that memory runs at about the 1.810mark ish.

so i will first try 7-7-7-20 with northbridge volts at +0.050 and then i will try 9-9-9-24 at same +0.050 northbridge voltage if the 7-7-7-20 settings fail.

yellowbeard you ever had success at running 4 of these ddr3 sticks in a p35 mobo at 1600mhz before?
 
i guess its going to be a case of testing and seeing if it works out for me. yellowbeard would you say memtest is the best way to find out if themmemory configuaration is stable? are there any other tools that can get the testing done faster/ any more stressful memory tester out there that can work the memory harder to see if it errors out sooner?
 
Memtest is not really a stress test at all. But, when OCing I use it as a boot test when going for big OCs. If a high OC won't boot and run Memtest for 1 pass, I don't even try the OS.

Try Prime95 or Hyper PI.
 
ahh, for me if i want to stress test the cpu i run intelburntest, that uses up a fair amount of memory as well and its the only app that gets the cpu up to the highest temps iv ever seen.

iv heard of other programs that test the memory, such as docmemory, or goldmemory, but never used them i think goldmemory you have to pay for.

as for superpi it crashes on me when i use vista, seems to works fine though under xp when i used to dual boot.

i will try your suggestion out tonight, which is to stick the ram to the 9-9-9-24 timings and leave memtest running overnight.
 
Cyber-Mav said:
criky, and here is me planning on doing a +0.050v bump on the northbridge and running memtest overnight.

gordon what are your results like with 4 sticks of ram in memtest? do you get as many errors as i did in an 8 hour run? i may just bump up northbridge volts to +0.10 tonight and then let memtest run to see how it is.
Click to expand...

I find memtest to be woefully inadequate for testing speed/overclocking stability. I've had memory pass days of memtest without any errors, and the system be stable under heavy gaming load, and even folding with multiple clients for days, only to have OCCT error out in under 10 seconds every time (I kid you not). memtest is, IMNSHO, completely useless in testing OC stability. I only consider a machine stable If it passes 24 hours of OCCT. I've also found that most DDR3 memory will fail this level of rigorous testing at stock settings. My Patriot modules, for example, require to be run at +0.1V (1.90V set instead of 1.80V as per the spec, plus the mobo's over-volting by about 0.10V anyway, so 2.00V (+0.20V) in total) to be stable at stock settings. With 4, it's downright impossible (well, I didn't want to push the DIMMs past 2.30V (2.40V measured), but since that's approaching magic-smoke-escape-threshold voltages, it's irrelevant.
 
gordon what you have said is also very true, memtest passed fine for me for 13 hours at 7-7-7-20 using 2gig of ram but intelburntest failed few iterations, i had to drop back to 8-8-8-20 for full stability. which makes me think that 9-9-9-24 at 1600 may pull it off yet.

best thing is to use a combination of apps to test the system out. but i find memtest to be adequte enough to prevent me booting into windows without nuking the operating system. wont dare boot into windows directly after altering memory configuration without running basic memory tests.
 
This differs from module to module, but relaxing the timings TOO far actually makes the memory stop working completely. Yes, that's right - relaxing timings can lead to instability. For example my DDR3 won't even POST at CAS 10 at 1600, but it's stable enough to boot at CAS 8.

You could use something like a Windows LiveCD if you're really worried about trashing the disk contents through cache corruption. Have a look at BartPE
 
im currently running memtest at 9-9-9-24 timings as suggested by yellowbeard, if memtest passes then i will do further testing in windows to make sure its stable.
 
You're likely heading for a disappointment and wasting your time. As I said before, I've had memory easily pass 24 hours of memtest and boot Windows just fine, only to fail in OCCT in under 8 seconds every single time. memtest failing indicates a major problem (e.g. massively unstable OC or genuinely faulty memory). memtest passing, however, proves nothing.
 
which is exactly why im not just stopping at memtest testing. once done i can boot into windows and run stability testing to see if the memory controller is making any errors or not. but first step is to do memtest to isolate the memory only and see if its making any errors. cpu is deffo stable since it passes 12 hours of intel burntest with ram running at 7-7-7-20 on a divider at 1280mhz, or with 2gig of ram running at 1600mhz 8-8-8-20 is fine too.

if 4gig works at even 9-9-9-24 1600mhz id be happy with that. if not i will stick to 7-7-7-20 on the divider and run at 1280mhz, either way its faster than having 2gigs running at full speed :)
 
looks like a no go with 1600mhz at 9-9-9-24, memtest froze up at 10 hours 21 minutes. thought i would give this goldmemory program a try, very interesting results so far, memtest just locked up at 10 hours and 21minutes into testing, goldmemory picked up 2 errors 6 minutes into testing and 1 more error 8 minutes into testing, so thats 3 errors in picked up in under 10 minutes of testing using goldmemory.

memtest is a waste of time will be using goldmemory from now on since its testing methods seem to be more harsher on the memory and are picking up errors in minutes where memtest is taking hours.

back to 1280mhz now with 7-7-7-20 timings.
 
a question for you yellowbeard, do you think the corsair ram i have can do tighter timings at 1280mhz? or is 7-7-7-20 the lowest it will do?
 
Cyber-Mav said:
memtest is a waste of time
Click to expand...

Much as I hate to say this, but - told you so. ;)

Re: timings - can you get a dump of memory SPD (I think CPUZ offers a way to do this)? This will tell you the speeds/settings it's supposed to run at. In my experience, settings up to 1333 are achievable at stock timings even with quad DIMMs.
 
the spd table seems a bit dodgey to me:

xms3dhxspd.PNG


im running at 640mhz ram speed (1280ddr) and 7-7-7-20 timings, was wondering if they could go lower on timings at the speed i run at, but according to the spd table i should be going far more slack on the timings.
 
Last edited:
Indeed, that's right. Your timings are already "overclocking" the memory.
The XMP profiles are not guaranteed (or even expected) to work with 4 DIMMs. They are overclock settings, and for up to 2 DIMMs only. With 4 DIMMs, you should be running 9-10-10-25@1333MHz. Also note that settings up to 1333 (667 DDR) are at 1T, and XMP is at 2T.

One thing I feel I should point out - this is nothing specific to the Corsair memory. It is no worse than any other memory (sadly, it would appear it's no better, either). All DDR3 memory seems to be suffering from the exact same issues. XMP settings/timings are only expected to work with up to 2 DIMMs. And FWIW, yes, I also found out about this the hard (and expensive) way.
 
although to note im still running the ram at 1.8v which is probably why the 777 timings work. im wondering if timings such as 6-6-6-18 would work at 1333mhz with this ram.

my ocz reaperx did 6-5-5-18 i believe at 1333mhz but then again, the reaperx was specced for that job.

another question:

i have the system setup as 400x8 = 3.2ghz with ram on 3.20 multi to get 1280mhz,
would it be better to run at 356x9 = 3204mhz with the ram running with the bus speed at 712mhz (1424mhz) assuming it even runs at this speed.
 
Last edited:
It depends on how optimistic/lucky you feel. There's only one way to find out for sure. Personally, I nowdays to consider it a great success and/or luck if memory works even at stock settings. The only exception is the full fat ECC server memory, but then again, that still only comes in DDR2 form, and the fastest modules are 800MHz rather than the latest 1066 standard.
 
You must log in or register to reply here.
Back
Top Bottom