LLC (Load Line Calibration) not working... Bios fix required ASAP

[bloot]

It's not a rumor anymore, it's a real issue http://www.overclock.net/t/1638955/z...#post_26394902

And it affects other boards too, Asrock just released a beta bios fixing it https://www.asrock.com/MB/Intel/Z370...index.asp#BIOS

Cheers

[elmor]

https://www.io-tech.fi/uutinen/paivi...0-emolevyilla/

Use google translate. That will give some of the you the answer to this hoo-haa. Unfortunately, they aren't measuring from the right place, either. To get a more accurate reading, you'd use one of the MLCC caps behind the socket. You want to get as close as you can to the socket pins so that you can see how much resistive loss you get through the power plane. Review sites simply aren't equipped to do this stuff properly. The non-technical guys profess to know things they don't understand, and those that try to be technical mess it up, too.

The requested VID can be affected by the IA/AC and DC load line settings. Has been like that for a while now. I'll see if I can get a board here and check what's going on.

IO-tech got told by Asus to measure this way, which was a mistake. It will not yield accurate results. It would be best to measure at the MLCCs at the back of the socket, or even better using the on-die sense pins which are a bit trickier to find. Basically the closer to the CPU the better.

As a comparison on latest test BIOS with 1.360V manual mode. 4.7G loaded with Prime95 29.1 SmallFFTs. LLC levels are not fully tuned yet so can't reference which one yet. Probably around 4-5. CPU-Z has been relatively accurate and at most ~15mV off.

CPU-Z 1.264V
Measuring at IO-Tech indicated location 1.345V
Measuring at socket MLCC 1.253V

[Raja@ASUS]

Hi Raja,

Does this mean you guys agree there seems like there is a FW bug? My comment here is just to point out several other potential customers are monitoring this issue...

I'm not stomping around angrily, I also bought 1 of these boards and although I am waiting for the 8700k, and if there is no formal statement from you guys or acknowledge/debunk in the next 1-2 weeks, I will likely return (within the 30 days) unfortunately. It's got nothing to do with branding; I bought the 8700k for playing with Overclocking (else I would have bought the 8700), and being limited by a mobo FW bug seems disappointing.

It's easy to say "Oh Asus will not care about the 1 unhappy customer," but the Overclock3d video now has 25,000+ views in just 3-4 days... That's real potential revenue loss for you guys, so I'm glad you're taking it seriously... As a customer I am obviously rooting for you guys to get it solved.



...Also as a fellow senior and highly experienced Hardware Electrical Engineer (PCB, mixed signal both Analog and Digital), I tend to disagree that risk of this problem is low. I'll be blunt; I know little about Overclocking; I don't design for MOBO's or Intel processors. You guys know more about LLC than I'll ever know (hence why I'm relying on your expertise)... That said I've done several PCB's with FPGA/Arm CPU's, Analog, RF, ADC, DAC, and dozens of 1k+ component count PCBA's. #1 rule is get your power supply/rails correct. Don't screw that up... Over-Voltage transients can kill any chip depending on the bandwidth, overshoot, and duration (on-chip ESD diodes really are not that good)... And I'm assuming the #'s people are posting are wrong from the SW issues like you said, but having that much ground bounce just an inch away seems pretty sketchy to me... Add more PWR/GND planes, layers are cheap, and plenty of vias ...

Afterall we are only talking about a little cpu processor (not the 10kW 100A+ RF Generator I'm working with now

R&D has found the issue. Awaiting updates.

[Raja@ASUS]

IO-tech got told by Asus to measure this way, which was a mistake. It will not yield accurate results. It would be best to measure at the MLCCs at the back of the socket, or even better using the on-die sense pins which are a bit trickier to find. Basically the closer to the CPU the better.

As a comparison on latest test BIOS with 1.360V manual mode. 4.7G loaded with Prime95 29.1 SmallFFTs. LLC levels are not fully tuned yet so can't reference which one yet. Probably around 4-5. CPU-Z has been relatively accurate and at most ~15mV off.

CPU-Z 1.264V
Measuring at IO-Tech indicated location 1.345V
Measuring at socket MLCC 1.253V

Thanks, Jon. Someone needs to learn where to measure.

[Tech Hog]

Thanks, Jon. Someone needs to learn where to measure.

You mean whoever at Asus told them to measure there? :P

[Raja@ASUS]

You mean whoever at Asus told them to measure there? :P

More so the person that told them, but also for everyone that measures. If you need to ask...

[Tech Hog]

More so the person that told them, but also for everyone that measures. If you need to ask...

True. I think I'll order a new DMM just in case. You never know after all.

Fingers crossed for a fix this week!

[Raja@ASUS]

True. I think I'll order a new DMM just in case. You never know after all.

Fingers crossed for a fix this week!

Good choice. Expecting updates this week.

[Tech Hog]

Good choice. Expecting updates this week.

Is this only affecting Strix boards btw? I've heard people saying that they think Prime-A and even M10H have the issue as well.

[Raja@ASUS]

Is this only affecting Strix boards btw? I've heard people saying that they think Prime-A and even M10H have the issue as well.

The Prime, yes. Elmor would know about the Hero. Either way, all boards that require an UEFI update will get one.