03-17-2017 05:21 AM - last edited on 03-05-2024 02:04 AM by ROGBot
08-15-2017 03:19 PM
gupsterg wrote:
Had C6H since launch. Used 4x differing CPUs and 3 differing RAM kits on same board. Used pretty much every UEFI released. What I found with my R7 1700 was UEFI after 1403 seemed to require some retweaking for same OC. My R7 1800X also as well was unstable in UEFI 9920 and 1501 with same setup as 1403, again retweaking sorted it.
These kinda of variations can occur, especially on a developing platform as Ryzen IMO. I had been on Intel since 2007 prior to Ryzen, even on that you can have some variability. Even when a BIOS/UEFI hasn't changed sometimes the 'HW' just needs a retweak as it can't hold the OC at previous settings.
Not saying your stability testing was not valid, but IMO 1-2 programs pass does not qualify an OC as stable for me. I have experienced on my Ryzen setups cases where I have passed lengthy testing in IBT AVX, custom x264 and Y-Cruncher, only to find HCI Memtest has found low count errors (~<10) in ~6hrs of testings, where as the other programs were passing.
A setting that has aided gaining stability at 3333MHz RAM on later UEFIs when using my R7 1800X has been lowering VDDP. I had noted in past screenies that VDDP [Auto] was ~900mV and in UEFI's after 9943 it is ~975mV, using 900mV and doing some other minor tweaks nailed 3.9GHz with 3333MHz Fast and 3466MHz The Stilt preset for me.
08-15-2017 04:24 PM
antony209494 wrote:
Thanks for your help. I am now on 1.4 and have tried every intermediate value between 1.375 and 1.40. The same thing happens in all of them: When I run a stress test or cinebench the voltage dips! It dips as low as 1.35 and it crashes in a matter of seconds as 1.35 is understandably not enough to hold a 4.00 Ghz OC. Which has me thinking: Should i tinker with load line calibration to prevent the VCORE drops (that's what it does right?). I noticed two settings: CPU load-line calibration and CPU current capability. The first one is in levels 1-5 and the second in percentages 100-140%. Should I also use the current capability option alongside the load-line calibration? And what should I set them to?
08-16-2017 06:06 AM
Pandemix wrote:
You just need LLC level 4 to stop the majority of the drops. This will result in the thing running at .2V higher than what you have it set at during full load, it will still dip, but when it dips it will dip to what you have it set at in the bios, instead of undervolting like it does now. You might actually be able to drop your voltage .1 to compensate for this, but I probably wouldn't recommend it. I don't think you need to mess with current, I think this is more of an extreme overclocking tool.
I have read that Ryzen CPU's and CPU's in general have a break in period where you might be able to achieve higher overclocks for a few weeks, but once the chip settles in a bit you will need more voltage. I've had this board since day 1 pretty much and have been through just about all the bios revisions. When I first got the thing it ran stable at 4GHz at 1.37v LLC 3, bios 1201 it ran at 4ghz at 1.38v LLC4, now i need 1.46v LLC5 to hit 4ghz. So there may be some truth to it. I actually bounced back to Bios 1001 to test this theory, I had 4.25ghz on there at one point when it was first released, but it didn't help. I could still only get 3.925 GHz without cranking voltage.
Honestly your probably just better off dropping your overclock in .25 ghz increments until it's stable if you are already pushing 1.4V The difference between fully stable 3.925ghz (stable) and 3.950ghz (half ass stable) for me is like .5v. You have to find your sweet spot with Ryzen, which is typically anywhere between 3.8 to 4Ghz with a non X chip.
Also when you test with cinebench, don't just run it once, run it like 5 times in a row. Then run it a few times every time you start up your computer for like a week. You can think you have your overclock good and a few days later it will crash. Eventually you will find your sweetspot, at least until the next bios 😃
EDIT: I tried dropping the VDDP, this did nothing for me, but Ryzen is a tricky beast and you have to master your specific CPU, which can be tough.
08-16-2017 08:27 AM
08-16-2017 09:43 AM
antony209494 wrote:
I ended up using 1.4V with LLC level 2. This keeps voltages between 1.373 and 1.395 which is just great and it is now perfectly stable after 1 hour of prime 95, realbench and multiple cinebench runs. But it still seems weird to me that bios updates destabilize OCs that much. Thankfully temps are still fine (only slightly increased - I am still well within the 60-70 mark with quite high ambients (30-32 C) cause it's summer). Thank you all for your help and lets hope future bios updates don't mess up things further in that regard!
08-17-2017 02:15 AM
08-17-2017 06:42 AM
Pandemix wrote:
I think you can probably get the same results @ 1.38v, maybe even 1.37v with LLC 4, basically with LLC2 1.4v you are just overvolting the chip when it isn't needed, then it drops down to the 4ghz stable voltage range at max load. Basically you are throwing an extra .3 to .5v at your processor all the time under light loads (normal everyday usage), when it's almost guaranteed not to be needed. So it's a bit bass ackwards, but if it works for yah, more power to you =P
08-18-2017 06:42 AM
08-18-2017 09:20 AM
08-18-2017 11:13 AM
sadaharu wrote:
Did anything changed over beta 1501
or
the official 1501 is as same as beta 1501 ?
Thanks.
Btw, with the new official 1501 ; AURA is working now.
3900 mhz at 1.375v and docp standart works like a charm.
cpu temps goes a little bit higher. like 3-4 C.
Main problem is does anyone know how to fix it ?
I am Turkish. My system language is in Turkish. But RyZen power plan is in russian ?