Crosshair VII Hero Essential Info Thread
Printable View
-
09-17-2018, 02:38 PMRageTooL
i hope its bigger now. if not
Manufacturer
G.Skill
Series
Trident Z RGB
Part Number
F4-3466C16-16GTZR
Serial Number
00000000h
JEDEC DIMM Label
16GB 2Rx8 PC4-2133-UB1-10
Architecture
DDR4 SDRAM UDIMM
Speed Grade
DDR4-2133
Capacity
16 GB (16 components)
Organization
2048M x64 (2 ranks)
Attachment 75830 -
09-17-2018, 03:20 PMgupsterg
Cheers :) .
What I said applies. The memory presets are for single rank/sided dimms. You would need to invest time setting up your own tweaks to timings or use 1usmus calculator. Even if you use calculator you may still need to tweak things.
Ryzen IMC on gen 2 is defo weaker IMO for higher density RAM. For example on gen 1/C6H I could hit 34xxMHz on 2x8GB, add 2 dimms more and need minimal tweaks to stabilise. So far I struggle to gain 3333MHz rock solid on gen 2/C7H :o . Same gen 2 CPU can do 3466MHz on 2x8GB and depending upon what you call stable could be classed as 3533MHz stable.
What RAM frequency are you able to hit? -
09-17-2018, 03:43 PMRageTooL
Never clocked memory before so i haven't tried anything cause im not too familiar with memory tweaking. Thats why i asked about the XMP Profiles :(
-
09-17-2018, 04:01 PMgupsterg
On the Extreme Tweaker page is setting Ai Overclock Tuner, pick D.O.C.P option this will use XMP profile settings :) . If it has set option Memory Frequency to [DDR4-3466MHz] I'd scale that back to [DDR4-3200MHz], see if you can attain that first.
Within Extreme Tweaker > External Digi+ Power Control is DRAM VBoot Voltage make sure this matches DRAM Voltage on Extreme Tweaker. Selecting Ai Overclock Tuner as D.O.C.P should have made both those voltage options 1.35V, if not set them as that initially IMO.
I would opt for manually set SOC Voltage (feeds the IMC juice). As stated to akarabamba a few posts back when you change CPU SOC Voltage [Manual mode] you will see a sub menu VDDSOC Voltage Override where you enter the voltage you'd like. I'd opt for something like 1.05V. -
09-17-2018, 05:18 PMRageTooL
-
09-17-2018, 05:43 PMRageTooL
Even scaling back to 3000mhz , it wont boot
-
09-22-2018, 02:16 PMgupsterg
IMO you'll need to set manually ProcODT and or RTT settings.
-
09-22-2018, 05:16 PMRageTooL
Got it to run at rated speeds with only 1 stick. Tried everything with both sticks but sadly it wont run :(
Attachment 75912 -
09-23-2018, 11:41 AMgupsterg
Increased RAM density defo needs manual tweaks IMO, may that be 2x16GB or 4x8GB. I have spent fare few hours to get this far :o .
Attachment 75924
Attachment 75923
There is no shortcut I can give, I can only share what I do. So how did I get to above as being a repeatable stable pass?
The experience with the CPU on 2x8GB Single Sided/Rank kit showed it doesn't need greater than 1.35V VDIMM for 3466MHz C15 1T even with some of the primary, secondary and tertiary timings tightened further than the preset by The Stilt in UEFI. SOC voltage again I do not need the increased [Auto] value, so manually set less than or equal to 1.025V is more than ample for 3466MHz 2x8GB.
So my target was to use SOC 1.025V and VDIMM 1.35V for 3333MHz when on 4x8GB. I loaded The Stilt 3466MHz timings, which are C15, which I thought would be loose for 3333MHz, I had HCI pass stability testing at times, then other times crash when stopping testing and or error during testing. So only option was to setup my own timings :o .
First step I just set voltages as stated and had RAM timings on all [Auto] but command rate 1 and gear down / power down enable mode disabled.
Attachment 75925
Now these benched like a slug, but did pass stability testing consistency. After several hours of further tweaking/testing I got to the first screen shared top of this post and thought :) . So I further tightened 4 timings.
Attachment 75926
BOOM 1 error at ~97% of ~120% :( . So rerun previous setup and all good again second screenie shared top of this post.
I have added in OP section RAM Timings Info today, I may expand on it but wanted to keep it as simple as possible.
All I can say is be systematic.
Change as few things at a time so you can see effect and if it is right thing to do.
As stated above, I started with SOC/VDIMM set as experience with CPU/RAM had shown me should work for targetted setup. I tried to reduce some timings and at the point I got some errors I decided I try tweaking VTTDDR Voltage from [Auto] (found in Tweakers Paradise). I changed this to 0.687V, it improved stability for my setup. VTTDDR should be set to 1/2 of VDIMM, but setting it notches below or above that "rule" could help. Stability was still problematic in repeat testing. I then bumped CAD Bus from 24 20 24 24 to 30 24 30 30 and finally 40 30 40 40 seemed to be right. As shown above further reduction in timings caused an issue, so now I'm back on an earlier setup. Now either I will figure out which of the 4 timings caused the issue before moving to tightening other timings. Currently I do not believe I need to tweak ProcODT, RTT and CAD settings, as system posts well and seems pretty stable for length in HCI. I also think currently I do not need to increase voltages, but I may have to, to gain some improved timings. Regardless it does seem to bench well.
Attachment 75927
Yes I am on C6H at present, I wanted to test how 4x8GB is on it with ASUS T-Topology and then transfer same CPU/RAM to C7H and see if I can attain or improve on RAM timings/MHz. -
09-24-2018, 07:04 AMgupsterg
In my previous post, second screenie from top, I left that "run" of HCI go to ~550% (end shown below).
Attachment 75939
Next I decided to revisit dropping some of the timings I previously did and had an issue. As I wanted to lower tRAS I felt I must lower tRTP, as I'm sticking to VDIMM 1.35V I did not think I could improve tCL, but perhaps tRCD could be in next round of testing. I avoided lowering tWR, as felt this may have been the reason for failure before (reference previous post, fourth screenie).
tRTP 12 > 10
tRAS 40 > 38
tRC 54 > 52
(Again as you meddle, it will get etched into you, how when you drop x, the effect on y ;) )
Attachment 75940
So had a pass of ~240% :) , next I thought I'd like to know if the SOC of 1.025V is excessive or not for 4x8GB @ 3333MHz C14 1T; with 1.012V = fail :eek: .
Attachment 75941
Above failure occurred at 140% and as 1 error not highly unstable IMO. Now I went back to SOC 1.025V.
Attachment 75942
We have a pass again :) . It could be said this was a waste of time to check SOC 1.025V again, but this next run was to cover "POST to POST training variance" which was not occurring IMO :) . Let's now go over time covered within screenies of this post.
~550% run is ~5hrs
~240% run is ~2.25hrs
~270% run where fail occurred at ~140% was ~1.25hrs in of total run of ~2.5hrs
~250% run is ~2.25hrs
Total of good 12hrs testing :eek: . To wrap up the day's testing, I wanted to get an extra length test under the profile's belt prior to tightening it more.
Attachment 75943
Will carry on tweaking this setup :) , here is settings of above pass, link (CPU OC setup found at "Custom Pstate0"/line 267).