i7 6700 + DDR3 1.5v possibly causing BSODs?

[Droidbot]

My friend has a system with a 6700, 1060, and a AsRock B150 Pro4/D3 mobo. It only supports DDR3, and he has a kit of Corsair Vengeance DDR3 full voltage. 

He is getting BSODs and crashes when he resumes from sleep, and I suspect it is a memory issue. 

He's been running this system for a year and the crashes have only started in the last 2-3 months. Has he fucked his IMC up bad? 

 

obligatory:

@MageTank

[SammoFS]

4 minutes ago, Droidbot said:

My friend has a system with a 6700, 1060, and a AsRock B150 Pro4/D3 mobo. It only supports DDR3, and he has a kit of Corsair Vengeance DDR3 full voltage. 

He is getting BSODs and crashes when he resumes from sleep, and I suspect it is a memory issue. 

He's been running this system for a year and the crashes have only started in the last 2-3 months. Has he fucked his IMC up bad? 

 

obligatory:

@MageTank

what le fuck

I have a diff system but, After i resume from overnight sleep I crash too. (after about 5 mins of being on)

[drkknghtavngr]

woah! I also have 1.5 ddr3 ram on my skylake chip, tho ive been only been using it for 2 months... looks like my destiny is set

[Droidbot]

Just now, drkknghtavngr said:

woah! I also have 1.5 ddr3 ram on my skylake chip, tho ive been only been using it for 2 months... looks like my destiny is set

try to get it to 1.35v, even if you go down to 1066 speeds its fine

[Energycore]

It was said before that the Skylake IMC doesn't like voltages over 1.35, whether for DDR3 or DDR4. That's why the DDR3 motherboards required "low power" 1.35V DDR3

 

Your options are downclock the memory / looser timings so it runs stable at 1.35, or getting some new LPDDR3

[Energycore]

6 minutes ago, Droidbot said:

He is getting BSODs and crashes when he resumes from sleep, and I suspect it is a memory issue. 

This BSOD might also be because the PSU isn't Haswell-ready. In that case disable sleep or the C6 and C7 power states so the PSU doesn't have to handle them.

[drkknghtavngr]

1 minute ago, Droidbot said:

try to get it to 1.35v, even if you go down to 1066 speeds its fine

o rly? undercloocking it will work? its has 1333 stock but i ocd it 1600 to gain a bit more fps with my integrated graphics...

[Guest]

Memtest? 

[Droidbot]

Just now, Energycore said:

This BSOD might also be because the PSU isn't Haswell-ready. In that case disable sleep or the C6 and C7 power states so the PSU doesn't have to handle them.

That's quite possible, it IS a shit unit. He has some 80+ White 'WIdeTech' unit that I wouldn't trust as far as I can throw it.

Just now, deXxterlab97 said:

Memtest? 

Passed without issues. I already asked him to run it, ran it overnight on his system.

[MageTank]

2 hours ago, Droidbot said:

My friend has a system with a 6700, 1060, and a AsRock B150 Pro4/D3 mobo. It only supports DDR3, and he has a kit of Corsair Vengeance DDR3 full voltage. 

He is getting BSODs and crashes when he resumes from sleep, and I suspect it is a memory issue. 

He's been running this system for a year and the crashes have only started in the last 2-3 months. Has he fucked his IMC up bad? 

 

obligatory:

@MageTank

It's unlikely to be the IMC not liking 1.5v. The IMC doesn't care at all about what vDIMM you use. vDIMM is supplied from the board, to the ram itself, completely bypassing the IMC. The only voltages that goes from the IMC, through the CPU, to your ram, is VCCIO/VCCSA. Start there, and let me know what his voltages are for both values.

 

Also, what does "DDR3 full voltage" mean? 1.65v? I have my DDR3 box behind me running a 6600T on an ASUS Z170-P D3 at 2133 C9 (1.65v) just fine. In fact, I've even overclocked the ram manually as tight as it can go, and it's been running non-stop without a single crash. 

2 hours ago, Energycore said:

It was said before that the Skylake IMC doesn't like voltages over 1.35, whether for DDR3 or DDR4. That's why the DDR3 motherboards required "low power" 1.35V DDR3

 

Your options are downclock the memory / looser timings so it runs stable at 1.35, or getting some new LPDDR3

Intel completely screwed up the definition of VDDQ in their own whitesheets in regards to voltages. They also say max DDR4 voltage is 1.26v, and yet we have people running 1.35 (or 1.41 like myself) 24/7. Max DDR3 voltage is 1.4175v according to the very same whitesheet, and yet I run a 1.65v XMP (with manual tweaks) on the box behind me. As long as you have VCCIO/VCCSA under control, you won't kill an IMC at all. The issue with these DDR3 XMP's, is that if left on auto, their VCCIO/SA values are absurdly high for 24/7 use. Mine, when loading a 1.65v profile, had 1.25v VCCIO and 1.35v VCCSA. To put this into perspective, I only needed 1.1v VCCIO, and 1.14v VCCSA to be perfectly stable. There are 4 tertiary timings that can drastically reduce the required VCCIO/VCCSA voltages, and they don't impact performance that much at all. These are tRDWR_SG, tRDWR_DG, tRDWR_DD, and tRDWR_DR. _DD only matters if you have 2DPC, and _DR only matters if your sticks are dual rank. Otherwise, these specific timings are the most stable, when you match them to your CL. If your CAS Latency is say, 10, then set all 4 of these timings to 10. If they are tighter than CAS (which, by default, they are) they put unnecessary stress on the IMC, for relatively no performance gain at all. When they match the CL, you can lower VCCIO/VCCSA voltages or tighten other tertiary timings for basically no cost in performance. 

[Droidbot]

5 minutes ago, MageTank said:

Also, what does "DDR3 full voltage" mean?

1.5v. Never dabbled in memory OC or anything around it, so I'll have fun learning these terms

[MageTank]

4 minutes ago, Droidbot said:

1.5v. Never dabbled in memory OC or anything around it, so I'll have fun learning these terms

Can you link me the exact kit he is using?

[MageTank]

2 minutes ago, Strike105X said:

Actually from what i heard this DDR3 boards come with a chip that is supposed to take the load of the imc, so voltages shouldn't really be a problem.

This is simply not true. There is no "chip" that will take the load off the IMC. The power delivery from vDIMM, comes from the board itself, to the memory. It does not pass through the IMC at all. This is why vDIMM cannot kill an IMC, because the two are completely removed from each other. The only memory voltages that can kill an IMC, is VCCIO (the voltages that govern the lanes into and out of the IMC) and VCCSA (your PCIe/memory subdomain voltage). There is no voodoo magic involved, and it's pretty black and white. If vDIMM could kill IMC's, @Lays and I would have been whining a really long time ago, lol. 

[Droidbot]

4 minutes ago, MageTank said:

Can you link me the exact kit he is using?

He's away from his PC at the moment, but all I know from him is that it's a CL15 Corsair Vengeance kit clocked at 1600Mhz

[MageTank]

Just now, Droidbot said:

He's away from his PC at the moment, but all I know from him is that it's a CL15 Corsair Vengeance kit clocked at 1600Mhz

What? A DDR3 1600 kit clocked at CL15 and 1.5v? This can't be real, lol. 

[Droidbot]

Just now, MageTank said:

What? A DDR3 1600 kit clocked at CL15 and 1.5v? This can't be real, lol. 

I may have read wrong, shitty photos sent over MMS.

Yep, I read wrong. 

1600Mhz 10-10-10-27 1.5v