RVIE & 7960X: Shutdown during RealBench... Due to OCP?

[gridironcpj]

I was stress testing my 7960X's OC of 4.6GHz at 1.185V with my memory at its XMP speed and primaries (manually tuned in) and my system shut down after 20-30 minutes of RealBench, then rebooted on its own. I thought it was perhaps the vcore not being high enough, so I increased it to 1.195, but the same issue occurred. I then tried 1.205, but again, the same issue occurred. The hard shutdown made me think it must actually be related to OCP. I randomly came across this review, which describes my situation perfectly:


https://youtu.be/gz9HBVh57T8?t=15m7s

So with that said, is this just an issue the RVIE has? If so, then that means 4.6GHz is my ceiling... which is very sad since this CPU still has a lot left in it since it isn't even delidded and I'm still hitting under 90C max in RealBench with the above settings. Any thoughts or possible solutions for this?

[Marko911]

Hate to burst your bubble,but forget 4,6Ghz even at 1,20v with that CPU...you should be aiming more closer to 1,30v,which I wouldn't recommend. Try 4.5 if the same happens at 1,20V,try 4,4Ghz or lower....These HCC CPUs aren't capable at that high freq rate.At one point,even if CPU is delided thermal transfer rate will be limiting factor,among other things.
This have nothing to do with RVIE,it's just CPU limitations...For example,I can OC my at ,4,7Ghz at 1,21V with temps at 85C max core.But i9 7900x have 10 cores,yours have 16 cores.You can't expect to have the same OC as lower core count CPUs.

[gridironcpj]

Marko911 wrote:
Hate to burst your bubble,but forget 4,6Ghz even at 1,20v with that CPU...you should be aiming more closer to 1,30v,which I wouldn't recommend. Try 4.5 if the same happens at 1,20V,try 4,4Ghz or lower....These HCC CPUs aren't capable at that high freq rate.At one point,even if CPU is delided thermal transfer rate will be limiting factor,among other things.

I passed 2 hours of RealBench at 4.5GHz 1.15V. SiliconLottery was even selling binned 7980XEs that were hitting 4.5GHz under 1.15V (they are sold out and no longer listed). 4.6GHz was stable for 2 hours in RealBench at 1.205V after I set my memory and mesh to default. The memory (running at XMP, but manually tuned in along with the appropriate voltages) and mesh must put it over the edge to trip OCP.

[LiveOrDie]

gridironcpj wrote:
I passed 2 hours of RealBench at 4.5GHz 1.15V. SiliconLottery was even selling binned 7980XEs that were hitting 4.5GHz under 1.15V (they are sold out and no longer listed). 4.6GHz was stable for 2 hours in RealBench at 1.205V after I set my memory and mesh to default. The memory (running at XMP, but manually tuned in along with the appropriate voltages) and mesh must put it over the edge to trip OCP.

If thats the case thats called a sweet spot, most CPUs will do well to a point then need a large amount to do higher which seem to be your case, Random shutdowns can mean low vcore voltage try increasing it.

[Marko911]

At one point,higher freq would demand a lot higher voltages,so benefit of OC is diminished.(like if you get stable 4,5Ghz at 1,15v and 4,6Ghz isn't stable for example at 1,20V than you have spotted your limit). -P.S because you have vaguely described what have you changed in your BIOS,I can only speak about standard OC .(like no mesh settings changed,no manual RAM settings or others non standard settings in OC)-in regard to your first post.
Also,I I don't believe in Silicone Lottery or Der8auer marketing strategy(at least the second one sells his CPUs OCed with max 1,4V which in my opinion is ridiculous),but 7980XE 4,5Ghz at 1,15V smells very fishy.

[gridironcpj]

Marko911 wrote:
At one point,higher freq would demand a lot higher voltages,so benefit of OC is diminished.(like if you get stable 4,5Ghz at 1,15v and 4,6Ghz isn't stable for example at 1,20V than you have spotted your limit). -P.S because you have vaguely described what have you changed in your BIOS,I can only speak about standard OC .(like no mesh settings changed,no manual RAM settings or others non standard settings in OC)-in regard to your first post.
Also,I I don't believe in Silicone Lottery or Der8auer marketing strategy(at least the second one sells his CPUs OCed with max 1,4V which in my opinion is ridiculous),but 7980XE at 1,15V smells very fishy.

There are users with 7980XEs who have managed 4.5 under 1.15V and 4.6 under 1.2V. I'm sure there are some users here who can share. I also have a 7920X, which is capable of 4.7GHz at 1.25V without a delid. The issue I'm getting at here is tripping OCP, as I'm not getting a BSOD like I would with my 7920X when hitting instability while stress testing. The main difference between these CPUs is their power draw. The video I linked describes my situation perfectly. I'm just curious if these CPUs are too power-hungry for these boards beyond a certain point, with that certain point being much lower than I thought. If so, then the power draw is the bottleneck to maintaining these overclocks in stress tests rather than temps being too high (not even hitting 90C) or voltages being too low.

[Marko911]

Yeah,I think we would see those owners bragging with those CPUs... 😉 But,there is literately no evidence of these CPUs with those values.
The 7920x can hit 4,7 GHz at 1,25V even without delidding because die surface is greater than 7900x so heat transfer is much much better,but also less cores than 7980XE allow to OC with lower voltage values too.They don't scale gradually 7900x,7920,7940,7960 and 7980XE ....
Hm,I see what you're getting at...You think that asus should've installed two 8 pin EPS cables on RVIE? But again,I doubt your voltage values could cause OCP tripping... it seams way too low..

[CSN7]

I doubt this is OCCP related, but you can raise CPU current capabilities in the bios under Digi+ Control to get this out of the equation.

As your system crashes aren't solved by adding vCore I suspect your system hangs and restards are related to Mesh OC or insufficient Input Voltage. Instabilites in Mesh OC can often not be discovered in somewhat short stress tests but are reveiled in prolonged tests instead. Also keep in mind, for Mesh you need to use manual voltage override instead of the adaptive mode which doesn't seem to work properly. Still with Mesh you are going to hit a wall at some point, where the next step takes exorbitant more voltage to be stable.

As a general guideline you can use 1.15v Mesh and 1.125v IO voltage for 3000-3200 MHz Mesh on big skylakeX without issues. 4.6-4.7GHz should be obtainable with 1.2 vCore and 1.98v Input voltage (LLC disabled). Preferably use adaptive mode and an offset usually start by -20mv and find your sweetspot.

Keep in mind realbench uses AVX2 instructions. Therefor it is mandatory to use a decent offset, or else you need significantly more vCore. On my 16-core I use an offset of -5.

Current is roughly 375W during a run.

Important to remember is to find stability for non-AVX (games and the usual software) and AVX2/3 offsets are a must on the big chips.

[Silent_Scone]

Set current capability to 200-240% in the UEFI and see if the shutdowns persists.