| Topic Review (Newest First) |
| 01-20-2020 08:29 AM | |
| myionzolo |
Hi I have the asus rog zenith extreme and this ram F4-3200C14Q-32GTZRX, (TOT 64GB) if I make a memtest86 one ram at a time, all the ram are good I have no errors. While if I use all the RAM in total 64GB I have errors! I have the latest firmware version especially 2001. Maybe I should try 1003 and see if I still have errors? has anyone had the same problem and solved it somehow? I would be glad if anyone could help me. Thank you |
| 01-19-2020 03:42 AM | |
| neil_tohno |
Quote:
Originally Posted by dejanh
Thanks for the detailed feedback. I took a quick look at the results you posted and that does seem a bit discouraging. I would have expected 2nd generation of chips to be more, not less forgiving of higher speed memory. I have plenty of issues as-is getting 4x16GB dual-rank running at rated 3200MHz CL14. I do need the PCIe lanes since I use a lot of PCIe storage so that generally rules out mainstream platforms like AM4 (this plus multi-GPU setups is the same reason why I stuck with HEDT platforms for a long time). It sounds to me like your advice would really be to save some pennies and stick with the 1950X for the time being and when I decide to upgrade move up to newer generation of TR. I am not keen on blowing $3K on a new TR3 setup at this stage since I do not have a lot of confidence in early boards for the TR platform (the disaster that was 1st gen has me sufficiently weary).
On a positive note, I came across some useful information from one of the Gigabyte forums of somebody who was running the same dual-rank memory as I am and had plenty of trouble getting 3200MHz CL14 rated going on more recent UEFI, and his brute force solution was actually just rooted in bumping SOC voltage to about 1.12V and driving the actual DRAM with about 1.45V each. Doing the same today in my setup allowed me to get a (for now apparently) stable 4x16GB 3133MHz CL14 with the CPU also overclocked to 4GHz. Next I am going to try for 3200MHz. *fingers crossed* Attachment 320030 About the new 2001 BIOS, ZE + 2950X Gskill [Trident Z RGB] F4-3200C16S-16GTZR [DDR4-3200 64GB (16GBx4) ] 3200MHZ 16-18-18-38-56 1.36v, and ProcODT 60Ω, CAD_BUS 24 24 24 24 You see, I just bumping SOC voltage to 1.11250V and adjust the DRAM voltage with 1.36V (don't need 1.45V that high) , and use this setting, the DRAM stability issues is solved: Quote:
DFE Read Training [Enable]
FFE Write Training [Disable] PMU Pattern Bits Control [Manual] PMU Pattern Bits [5] |
| 01-15-2020 01:26 AM | |
| gupsterg |
@dejanh
NP  , I hope you attain what you'd like. The training issue may not be voltage related. For example I could make POST intermittently fail fully so board reverted to fail safe defaults by using incorrect ProcODT, etc. So you may find a setting, may mask the incorrect setup of another.There is no doubt in my mind Zen2 is way more forgiving with RAM than Zen/Zen+. So defo don't blow more on X399 and if you can't do without the PCI-E lanes then best is to save up and go TR3+TRX40. It's not just the RAM side of things that have improved. TR2 exposes core performance ordering to Ryzen Master, but OS can not take advantage of it. Where as Zen2 has CPPC (Collaborative Processor Performance Control), so on TR3 OS will be aware and can take advantage of best cores. |
| 01-14-2020 09:19 PM | |
| dejanh | @gupsterg Thanks for the feedback. Let me play around a bit with those settings and see what comes of it. Seeing as I get immediate failure to train off of a cold boot, I am going to try adjusting Vboot to start and see what impact it has, then go from there. |
| 01-12-2020 02:38 AM | |
| gupsterg |
@dejanh
My experience with dual sided/rank dimms is pretty much none. If what you say was happening to me on single sided/rank dimms, then it's the usual things to tinker with and see if it solves it. SOC/VDIMM & VBOOT/VTTDDR/ProcODT/CAD Bus Drive Strengths Don't over look playing with Phy menu in AMD CBS > UMC Common Options > DDR4 Common Options, near the end is Phy. When I moved a 1950X from ZE to ZEA I found the same CPU & RAM would not do 3400MHz without adjusting Phy menu. I'd been running 3400MHz on ZE since about Aug 18 when it became available in UEFI 1402. For the single sided/rank 4x8GB kit I had in use, I needed to set on ZEA:- DFE Read Training [Enable] FFE Write Training [Disable] PMU Pattern Bits Control [Manual] PMU Pattern Bits [6] Link to post by Elmor on Phy menu. |
| 01-11-2020 05:55 PM | |
| dejanh | @gupsterg - Quick question for you...it appears that I can successfully clock my memory to 3133MHz by boosting the SOC and AB/CD DRAM voltage and pass 400%-500% of Memtest HCI and 8 hours of Prime95 Blend test without issues. 1/3 times however when the system is starting from cold boot, the memory training fails and the system goes to the F1 screen after failing to POST. I then literally enter the UEFI, hit F10, the system reboots and posts without issues. Any clue what I may be missing here? This entire issue with the memory, including not being able to run 3200MHz is all training related. I just have no idea what to do about it. |
| 01-11-2020 12:11 AM | |
| dejanh |
Quote:
Originally Posted by gupsterg
Just before Christmas I got 2x 2920X, they weren't that great if I really thought about it, see here.
TR2 is Zen+ and compared with my Zen+ AM4 CPUs the IMCs were not great on ones I had. I had 3x 2700X, one reached max 3466MHz on 2x8GB, another 3533MHz and last one I had was 3600MHz. 4x8GB I'd lose significant MHz, ~3333MHz which would be like say 4x16GB on TR2. The Zen+ AM4 CPUs were used on a daisy chain topology. Also ZE/ZEA use T topology, AFAIK AMD FW is more tuned towards daisy chain topology, AFAIK AMD don't allow vendors to tweak FW like Intel, see this post. If you don't need the PCI-E lanes of X399 and don't wanna go TR3+TRX40, I'd say your better off going AM4. A R9 3900X does very well against a 2950X and has better RAM MHz/density support, due to the IO die you don't have latency penalty like in UMA mode on TR1/2. You also have option of 3950X, plus AM4 may see Zen3 and AM4 as mainstream gains more UEFI updates than the halo line up. On a positive note, I came across some useful information from one of the Gigabyte forums of somebody who was running the same dual-rank memory as I am and had plenty of trouble getting 3200MHz CL14 rated going on more recent UEFI, and his brute force solution was actually just rooted in bumping SOC voltage to about 1.12V and driving the actual DRAM with about 1.45V each. Doing the same today in my setup allowed me to get a (for now apparently) stable 4x16GB 3133MHz CL14 with the CPU also overclocked to 4GHz. Next I am going to try for 3200MHz. *fingers crossed* |
| 01-10-2020 02:47 PM | |
| gupsterg |
Just before Christmas I got 2x 2920X, they weren't that great if I really thought about it, see here. TR2 is Zen+ and compared with my Zen+ AM4 CPUs the IMCs were not great on ones I had. I had 3x 2700X, one reached max 3466MHz on 2x8GB, another 3533MHz and last one I had was 3600MHz. 4x8GB I'd lose significant MHz, ~3333MHz which would be like say 4x16GB on TR2. The Zen+ AM4 CPUs were used on a daisy chain topology. Also ZE/ZEA use T topology, AFAIK AMD FW is more tuned towards daisy chain topology, AFAIK AMD don't allow vendors to tweak FW like Intel, see this post. If you don't need the PCI-E lanes of X399 and don't wanna go TR3+TRX40, I'd say your better off going AM4. A R9 3900X does very well against a 2950X and has better RAM MHz/density support, due to the IO die you don't have latency penalty like in UMA mode on TR1/2. You also have option of 3950X, plus AM4 may see Zen3 and AM4 as mainstream gains more UEFI updates than the halo line up. |
| 01-10-2020 11:55 AM | |
| dejanh |
Quote:
Originally Posted by neil_tohno
At last, I found the PMU Training have some bugs,
 when I disable PMU Training, everything seems OK. Also, any thoughts from anyone on whether upgrading to the 2950X instead of the 1950X would allow me to run 3200MHz CL14? Finally, any idea on the increased CPU overclocking room using the 2950X and the ZEA board? |
| 11-02-2019 04:48 AM | |
| neil_tohno |
Quote:
Originally Posted by neil_tohno
Then I pressed Clear CMOS Button, but it doesn't work anymore. I have flashed 1701-BIOS three times, until I use Afuefix64.efi, again, the problem is solved. https://www.overclock.net/forum/26403470-post1.html At last, I found the PMU Training have some bugs, when I disable PMU Training, everything seems OK. |
| This thread has more than 10 replies. Click here to review the whole thread. | |
| Posting Rules | |