What is my point of failure?

[AManWithPlan]

Dear LTT,

 

You can find a detailed description of my rig in my profile.

 

Essentially I have a 5930K processor that I overclock to 4,000MHz using a 40x multiplier. I set the voltage to be "Adaptive" within the Bios of my Asus X99-Pro board, so that the processor doesn't always work on overvoltage, but only when it needs the juice. It has been completely stable in this setup.

 

I have 64GB of Corsair Dominator memory rated for 3000MHz currently running at 2600MHz 14C-15-15-32-1T. My strategy has always been to buy a higher rated memory, and then overclock its timings at a lower than rated clockrate. XMP on the 4 DIMMS are 15-17-17-36-2T at 3000MHz and 1.35 volts. I run them at a slower than rated clock speed, but at 1T and a better than rated timings at about 1.29V.

 

I post successfully, but after enough time passes (roughly hour or so), I lose stability and completely hard crash the system. All monitors display the last image they displayed, and system completely unresponsive. When I JEDEC clock the RAM at 2133MHz 15-15-15-32-2T, I have complete system stability with no such crashes.

 

Interestingly, following a crash, sometimes all of my memory is no longer detected on system reboot. It will only detect 32GB, ignoring the 4th DIMM, or sometimes misreading completely the specs of DIMM 2 and 3 as 8GB even though they have 16GB each. I have seen 32GB and 40GB detected before in various crashes. Reseting does not help, only a reset of the timings and clock speed of the RAM along with a reset will detect all 64GB.

 

The question is, what is my point of failure?
1) The 5930k is only rated for 2133MHz Quadchannel RAM, is it unable to keep up with the faster speeds?

2) The ASUS X99-Pro is rated up to 3,000 MHz, but perhaps it can't handle both the Quadchannel RAM overlock as well as the processor? Perhaps a voltage increase?

3) Is the RAM incapable of running at any rate beyond its rated timings, no matter how underclocked the clock rate is?

4) Is it a combination of one or more of these issues?

Thoughts?

 

Thanks,

Amanwithplan

[Jurrunio]

1. Yea, that's why it's called memory overclocking. It's enabled by the chipset.

 

2. The problem here is the RAM. Give it 1.35, these things draw little power. You could push 1.5V and they will still last long.

 

3. Not really, minimum timings depends on frequency as well. Again, stability above factory rated speeds are never guaranteed.

 

 

[AManWithPlan]

1 minute ago, Jurrunio said:

2. The problem here is the RAM. Give it 1.35, these things draw little power. You could push 1.5V and they will still last long.

I have given it 1.35V before, it will still crash. The instability is elsewhere.

There should be no problem overclocking timings on RAM rated for 3000MHz operation when I am running at a 13% slower clock speed than rated.

[TheGlenlivet]

You can test the ram for errors too.  I just RMA'd some Gskills and got back the new DIMMS last night.  Ran my workload all night and it was stable, where it crashed before.

I used memtest86 to verify which of my DIMMS were bad.

[Radium_Angel]

2 minutes ago, AManWithPlan said:

I have given it 1.35V before, it will still crash. The instability is elsewhere.

There should be no problem overclocking timings on RAM rated for 3000MHz operation when I am running at a 13% slower clock speed than rated.

Return everything to default. No overclocking.

See what happens then

[AManWithPlan]

1 minute ago, TheGlenlivet said:

You can test the ram for errors too.  I just RMA'd some Gskills and got back the new DIMMS last night.  Ran my workload all night and it was stable, where it crashed before.

I used memtest86 to verify which of my DIMMS were bad.

I have run memtest86 on the modules, it resulted in no errors.

[AManWithPlan]

2 minutes ago, Radium_Angel said:

Return everything to default. No overclocking.

See what happens then

I have done this with the RAM, still overclocking processor. I have completely stability at JEDEC.

[TheGlenlivet]

1 minute ago, AManWithPlan said:

I have run memtest86 on the modules, it resulted in no errors.

All tests x8?  Just making sure...

[AManWithPlan]

1 minute ago, TheGlenlivet said:

All tests x8?  Just making sure...

No, I didn't bother doing 8 passes. Did 3.

[TheGlenlivet]

Just now, AManWithPlan said:

No, I didn't bother doing 8 passes. Did 3.

AH.  Maybe try again.  I had all 4 dimms fail on pass 7 of test 4 then the specific dimm fail again on one at a time testing on pass 4 of test 4.  It was always test 4 of 13.  Yes, this takes some time...

So far, the system is stable with the replacement RAM.

[Jurrunio]

13 minutes ago, AManWithPlan said:

I have given it 1.35V before, it will still crash. The instability is elsewhere.

There should be no problem overclocking timings on RAM rated for 3000MHz operation when I am running at a 13% slower clock speed than rated.

The shorter the timings, the higher the transfer rate and more likely to be unstable. The frequency might be lower, but it is doing more work.

[AManWithPlan]

1 hour ago, Jurrunio said:

The shorter the timings, the higher the transfer rate and more likely to be unstable. The frequency might be lower, but it is doing more work.

I sometimes wish there was a website where timings/clock speed combinations are tested on various RAM and the stability results are posted for all to see.

[Jurrunio]

4 hours ago, AManWithPlan said:

I sometimes wish there was a website where timings/clock speed combinations are tested on various RAM and the stability results are posted for all to see.

I wish there's a site that tells me whether my new settings give higher transfer rate or just the same as it did before tweaking.

[AManWithPlan]

As an update to this topic, my computer suffered a similar hard crash with stock RAM speeds. I have since underclocked the processor and have no had a crash since, I am strongly suspecting that it did not get enough voltage on the CPU under load.

Still investigating.