Discussion in 'Processors and motherboards Intel' started by chinobino, Nov 12, 2015.
You probably meant DDR3L.
Just another small update, just passed the 18 month mark for my main system with the i7 6700k and 10 months for the i3 6320.
Both are still going fine with 4 x DDR3 DIMMS @ 1.65 V.
Been running DDR3-2400CL9 at 1.65v for over a year on my test bench doing overnight stress tests and it hasn't degraded or had any overclock stability issues thus far. The click-bait sites that said it would kill your CPU did were just bunk. It's the same morons that said Intel was magically going to make me stop overclocking my non-K Skylake
I'm revisiting this thread because of the new 4600MHz DDR4 that use 1.5V and I wondered if it is safe with Skylake.
Looks like it is easily ok by your experience Chinobino.
I am using Corsair Vengeance 3000C15 @ 1.35V which I have clocked up to 3600MHz at 1.4V but decided to back down to 3333MHz at 1.33V to be safe.
I tried to get 3733MHz at 1.42V but it was no go.
They are B-die or D-die Samsung chips rated to around 2V, sweet.
After realising 1.5V and perhaps a bit higher are no problem for the CPU I want to see how far these can be pushed.
Fingers crossed nothing dies because I cant cover the repair costs atm.
But this looks a no brainer so its got to be tried
Hi Mufflore, well I have 2 Skylake CPU's still going strong on 1.65 V - so yeah 1.5 V should be fine.
I just got my arse handed to me after breaking the golden rule of not using my main OS to test overclocks.
First boot attempt at 3600MHz 1.4V the screen went black before the OS loading starts and there it stayed.
Removing the overclock didnt help, nor did slowing down the ram below stock.
At first I thought I had harmed the ram, it took 20 mins of troubleshooting before I tried my backup OS drive which booted fine, phew.
I booted my Windows install disk and attempted a repair of my OS (on SSD). It couldnt repair the version of Windows I am using (Win7 vs Win7 SP1).
So I wrote an SP1 disk, booted that and it couldnt fix the boot problems, it wanted to send info about it to MS.
I tried the full slew of boot recovery methods from CMD and none worked.
Doh! I wondered if the SSD had been harmed.
Time for a nuke.
Booted my backup OS drive, backed up up the complete OS from the SSD.
Installed Win7 SP1 on the SSD to the point where it had written the boot sector.
Booted to my backup OS, recovered the full OS backup to my SSD again.
Bingo, I'm back.
Thats 2 hours I wont get back.
I'll have another go when my heads screwed on
Just read through the whole thread. Interesting stuff! I would love to be able to do tests like this, but I don't have extra money to spend on computer parts for testing. Staying tuned...
@Mufflore, did you test your RAM speed incrementally? i.e. 3200 MHz, 3400 MHz, 3600 MHz
I would run memtest86 through all tests (except maybe Test 10 Bit fade test because it takes so long) to see if errors are occurring.
You would want to make sure the PC is stable at each speed increment by running it for at least a few days doing all the things you normally do in a day.
From my experience memory issues tend to pop up under specific conditions, for example browsing websites that use flashplayer or games that use the memory a certain way.
Also if you will likely need to adjust the RAM timings manually (which you probably have already).
If you check over at HWBot there is a lot of DDR4 info such as this thread, [GUIDE] Skylake Memory Timings on Asus Motherboards !
I have corrupted my OS with a bad RAM overclock and the culprit was the tertiary timings, I had to increase WRWR_dr + WRWR_dd to 8 for 4 DIMMS.
Previously stable settings were used as a starting guide to see if it would cut out some leg work.
I realised it might not work because there have been numerous bios updates since.
Timings were set a bit looser to start which got me into the CMOS ok but then no further.
The dim part was forgetting to test with a spare drive/OS because beer, after all, I had already unplugged the other drives.
Then a bit of bad luck completely screwing up the boot sector so well it wouldnt recover - on the very first boot attempt. Gotta laugh.
The first time in over 10 years I managed that, it normally just screws the OS a little.
Thats a good guide thanks, I've used it a few times.
I'll have another go when theres time and I havent found the beer first
The ram didnt suffer at all, it performs exactly as before, there is that...
Hehe, sounds good
I've been using ddr3 1.65V with haswell for more than 3 years know. It should die when reading first rumors. Nothing happend. Everything works fine. I game a lot.
I had a good go at clocking my ram again with odd results.
Ram is Corsair Vengeance 4x4GB 3000C15 1.35V D-Die
I got it up to 3866MHz C19 at 1.52V and it was rock stable for a few hours hard testing with many successful reboots as I tried a few things (this becomes significant).
Tried for 4000MHz which failed. I didnt try too hard because I dont want to push the voltage further.
But then 3866MHz gave problems POSTing with exactly the same settings that were stable before.
It would occasionally POST and once in Windows it was rock stable but it was a lesson in frustration getting it to POST.
I tried for a couple of hours to find what went wrong and occasionally had 2 to 4 successful boots in succession which sent me down wild paths of troubleshooting.
But it was ultimately futile, they were just down to luck.
I conceded defeat for now.
Back at 3733MHz something has changed because I can now use lower voltage than my previously stable 3600MHz (from last years testing).
I was running 3600MHz C19 at 1.4V.
Now I'm using 3733MHz C18 at 1.37V.
I may push the voltage higher and reduce CAS etc.
But I'm happy for now.
I think the reason for the initial crash at 3600MHz was DMI voltage.
This was reset to Auto while running at 3333MHz but it needs to be higher to reach 3600MHz+.
I have DMI set to 1.2V now.
Hi deadinhollywood, thanks for your input.
Haswell is 22nm (like Ivybridge before it) and as you say has proven to be fine with 1.65V VDIMM, even Broadwell which is 14nm (same as Skylake) can be used with 1.65V VDIMM.
@Mufflore - what is your VCCIO and VCCSA set to?
Both at 1.2375V
Ok, that sounds about right, I am using 1.165V VCCIO and 1.21V VCCSA for anything above 2666 MHz (DDR3).
I set it manually in the yellow zone, default was actually higher in the red lol.
I havent verified it for a long time though, they might have fixed that in a bios update.
When I first got my mobo XMP used to set my ram at 1.5V.
They sorted that one.
I'm happy with the 25% overclock, some things are quicker/snappier.
I didnt have fps issues before this so doubt I'll notice while gaming.
Best CAS without going over 1.45V is
3733MHz C17 @ 1.43V
In my own experience....i`d say 8 weeks, and mine was on stock clocks on an Asus Maximus VIII Hero with corsair pro 2400hz
Both systems are still going strong (even after the PSU shorted on the 2nd PC).
Only a couple of days before my 6700K's Intel Performance Protection Plan runs out.
I may attempt to install a 9900K once the price has settled down, there are already reports of successfully running one on a Z170 board.
Skylake is supposedly more susceptible to failure from high memory voltage than preceding processors. I don't believe it though. As long as the processor is properly cooled, it should last several years just fine. Supposedly Sandy couldn't take 1.65V DDR3 either, according to early reports.