I didn't study the benches last night, just did them and posted. Will do a graph of data, also be interesting to compare MemTweakIt timings between straps. We know from Elmor's guide lower straps have some improved timings, hoping to do a 3.8GHz / 3200MHz profile with lower strap but upped BCLK to see how that improves over 3200MHz strap + 100MHz BCLK.
Originally Posted by hotstocks Warning: Spoiler! (Click to show)
This platform is goofy as hell. I will try and report back over the weekend when I get max stable ram and cpu overclocks, but I started 2 weeks ago with a 1700 on 1002 bios and never changed the bios. I also have 8X2 G.Skill Tridentz 3600 C16 B-die. I just got an 1800x and this is looking bad and weird.
1700 Memory ran completely fine at 104,108,109,110, 117 BCLKs up to 3430mhz ddr 16-16-16-16-36 perfectly stable. But REFUSED to run at 3200mhz no matter what I tried!
1800x Memory will ONLY run at 3200mhz. ANY BCLK overclocking at all makes memory drop down to like 2126mhz and Cas 15, WTH? I changed NOTHING BUT THE CPU. The cpu is what is dictating what your memory will run at. I can't believe this, but if true I will have to decide between a 3.9ghz 1700 with 3430 memory or a 1800x with 3200 memory but hopefully 4 or 4.1 ghz, will take some testing. This is all incredible odd and I might flash to bios 81 to see if my memory will go any higher than 3200 like it did before.
Or is there something I am missing, like after I swapped the cpus I have to set something in the bios or do something to retrain the memory? Usually all I had to do was his reset 3 or 4 times on a faild 0d and then it would just retrain and go back to the 3430mhz I was running at 100% stable.
P.S. If it matters, the 1700 is Malaysia and the new 1800X is China
Do you have IHS stamp info? cheers
Originally Posted by elmor Warning: Spoiler! (Click to show)
Your DRAM fails to train from cold boot and causes AMD CBS settings to reset, including P-states. BIOS is not aware of failed DRAM training and has all other settings still intact, meaning you get default voltage + your specified offset. Not sure what we can do about that at this point. Future updates might help with the cold boot issue and maybe a reliable way to detect DRAM training failures. It's also possible to have a fixed P-states. An alternative at this point would be an OS application which applies your P-state settings (FID/VID).
+rep for above and CPU socket thermistor location, will add that and the mobo temp sensor location in OP of my thread
Thank you for your time
, I agree RAM training was most probable cause of borked boot.1st setup
As an update, last night I CMOSCLR > set up base profile which has no OC/manual voltages > base with manual voltages (except CPU default UEFI) with 3200MHz 14-14-14-14-34
= borked boot again
So did CMOSCLR again, loaded base profile, then applied only manual voltages (except CPU default UEFI)
, then 3200MHz 14-14-14-14-34, then applied 3.8GHz Pstate 0 / Offset +162mV.
Now I may be barking up wrong tree, but may help someone.
I reckon in 1st instance of setup above, underlined process, due to VBOOT being set with other voltages = powerdown/up to apply sent 3200MHz training iffy. In 2nd setup I did manual voltages on their own so when went for 3200MHz C14 there was no powerdown/up of mobo.
Originally Posted by finalheaven Warning: Spoiler! (Click to show)
Raise your SOC volts. It fixed my cold boot up issues. I never had trouble rebooting, but cold boots were a problem until I raised my SOC.
Yeah, I reckon I will go 0.968V or 0.975V. I saw when I used 3.8GHz / lower strap + 134MHz BCLK = 3200MHz C14 1T I needed 0.975V SOC to get into OS, not yet stability tested. So 0.975V using 3200MHz strap + 100MHz should have increased "stability" vs 0.962V I use currently.
+rep for share chap Edited by gupsterg - 4/14/17 at 2:04am