chew*

It can boot above 4300 and that "wall" is around 4200 on DR 2x16gb but not full manual. I mentioned this bug awhile back. Touch no voltages set primaries loose and touch nothing else (18-18-18 4400 20-20-20 4600 ) even 32gb can boot 4600. after 4400 vccsa gets excessive ( 1.6 vccsa) though so fair warning. Just manually setting voltages alone and setting them to exactly what board is auto setting is enough to hang up the boot sequence but I have not pin pointed the problem.

Here's a tad more info on what it does. It existed on z390 but not everyone knew much about it.

RTL IOL are impacted by alot of things. Density. Cas latency and speed. Sometimes you can pull them in tighter if a board defaults loose. Sometimes you can't. typically speaking if 4 dimms 58,59,60,61 for example would be normal for 4 dimms. 58,59 for 2 dimms of course all this can change based on CL and speed so the simple way to drive RTL lower is lower CL. IOL at least on this board is rather loose with DR dimms.

100% in agreement but at least we are not taking steps backwards. Newer bios are improving operation.

It's 32g and that profile is conservative and still won't load everytime unless I set them to 0 not 2x8. I explored beyond and went balls out. this was impossible with out them set to 0. This also loads every time now. I'm fairly certain it would be way easier on 2x8

you guys are probably 1 version behind me then. Screenshot shows defaults. It's right under dll bandwidth. setting that to 0 alone does not buy me much. setting these all to 0 got rid of my random retrains something fails OC. Not only that but without it say I set trefI to max my rtls would go wacky even though set manual when I set trefI max. now they don't with 0,0,0,0. In fact I can pull a lot of timings in now and boot fine. I should probably do a video. My settings are prime stable hci stable tm5 anta stable occt stable aida stable but if I loaded defaults then loaded profile fail OC. just changing those now when I load profile it works every time.

Here's a guide although it mostly follows the rules which some can be broken. https://github.com/integralfx/MemTestHelper/blob/master/DDR4 OC Guide.md

Have you tried setting DLL bandwidth 0 and all the dllbwEN settings to 0? That seemed to solve my random load profile and sometimes train and sometimes not issues. I don't have a post code reader on unify itx but i have debug lights without those set it goes 1-2-3 repeats 3 times fails oc. With set to 0 it just goes 1-2-3 boot. Not to mention those settings are magical as i can run way tighter timings yet i can't measure any losses anywhere. I've played with Rtt's also but maybe I just have not found the right combo yet. So far its not buying much with DR on my particular setup.

Just exploring options in the new bios and what the board can and can not do. 2x16 DR 36ns flat @ 4200 c14.

I don't think 5 gig is an artificial issue. My chip can bench 5.2 ( multithread like CB ) and I've run xtu @ 5.1 on sub par cooling. If anything I gained 100mhz benchable over my buggy/broken from day 1 ASUS z490 Itx -013 serial # board. I'm sure if I swapped the cooler in the prototype i'm working on to my liquid freezer 2 280mm I could get over 5 gig stable.

you would be wrong. If i was not running memtests right now to screenshot i'd go screenshot bios right now. Unify I memory timing page bottom of page power options PPD enabled 2 or 3 settings down from that I have PPD 0 set. These are not even try hard memory timings and I'm sub 40ns because I have PPD 0 set in bios. And yes it absolutely on it's own does. My latency would be 45-46ns if i change nothing but that. I've already stated this in this thread if you read. Also the Asus being at 37ns @ 4400 c17 is not impressing me. My Asrock gets 37ns @ just c15 3600. Looks like PPD to me in screenshot 3 but maybe I forgot how to spell. Added as of 1.2w. screenshot 4 shows what was added with 1.2z bios.

I would go with the latest. Your latency issues is just the PPD setting bottom of mem timing page. Set PPD enabled in power options then set the PPD to 0.

define stable. I probably run more stability then most. I could pass anything anyone asked before at 2k yet not load the profile 100% and get a failed Oc reset. Now I don't get a failed Oc reset loading a 2k profile I'm let see at 23 hour mark in prime 95 which 99% of users won't even run.... That would denote more stable. Oh before I forget. At least on my bios they added the DLLbwEN options theres 4 of them.

Artificial manual tuning DR dimm wall at 2000 is gone. No clue how far it can go yet. Seems that training may be improved I'll test more my 2000 profile would load from defaults sometimes and sometimes I had to baby step to 1900 then 2000 but seems to handle shotgun approach better now. Not 100% sure but looks like have lost 1ns latency.

As of 1.2w yes. txp on its own makes no difference in the unify I latency. I tested just PPD 0 and that alone solves the latency issue roughly lowering it 4-5ns. I still think its maybe 1-2 ns off but then I'm comparing to a very mature z390 board I have pulling sub 40ns at only 16-16-16 3600 with similarly conservative sub timings like the above screenshots. Considering we are only 3 1/2 bios in on z490 I feel like this is rather reasonable as the platforms certainly not mature yet. As far as stability goes. The above settings still work except trefi. It passed hci 2000% task and tm5 anta extreme but failed prime 95 ( hard crash reboot ) custom full mem use. Set trefi to auto retested and now passing prime like its literally about 1hr and 15min away from a full 24 hour pass now. Could just be the added strain to IMC+32gb+PPD 0 and most likely will not effect 2x8gb users. This is at 2000 ( 4000 effective ) 1900 ( 3800 ) might be able to handle it. Keep in mind I'm using a NEO ( AMD ) 3600 c16 bin of b die as well. *EDIT* Added screenshots. What I passed in HCI/TM5 versus prime custom full memory usage which is why I still trust prime to this day. Also tossed in the z390 3600 latency expectations that I'm comparing to my z490 latency expectations.

1.2w bios definitely helps with the poor latency results. PPD 0 txp 5 gave a little boost to bandwidth not much. I tried it auto ( 12 for my memory config ) and 5 it made no impact to latency. The 1.2w bios or the above PPD 0 setting got the latency down roughly 4-5ns even running rather relaxed timings. Back to running memory tests to verify it sacrificed no stability.

Thought I might add that I was wrong. The board does not have a wall at 2000 with 2x16gb bdie. It has a bug artificially limiting it if you set vccio vccsa or vddr manually. 2 examples and be warned around 4500 the auto vccio/vccsa gets excessive. Excuse the crappy timings but since you can't set vddr vccio or vccsa manually or it won't post I only set the 3 primaries to just test what speeds it could possibly boot. Limited to 1.35 vddr I did not have much choice You will only want to set desired memory speed. 3 primaries. turn lucky mode on if you would like to test this yourself. Do not touch anything else. 4600 was max I could get into windows. 4633 tried to bork my OS but I got past bios. YRMV.

I'm sure they may have a problem and with single rank diims one could expect 1T CR to scale above 3600 but my expectations are realistic and I would not expect to be able to pull 3600 1T CR or even higher with a 2x16b DR configuration so just speaking on behalf of 2X16gb the only abnormality I'm experiencing is lower than expected latency performance. Clocks/timings are about what I expected for 2x16gb bdie. I'm not expecting $500 apex results out of a board that retails between $240-270 USD Just better than average memory oc's which when compared to say the Asus z490 itx it's doing better than average vs the ASUS with 2x16gb. Just the latency is annoying as is the lack of vtt volts/sensors for an allegedly tuned to OC memory board. Well that and the useless VRM fan that stole potential USB port space and poor passively cooled performance requiring you to at least get some airflow over m2 SSD/Southbridge at the very least and lastly the power reset location. Literally other than that the board works great. It's recovery from failed overclocks is nothing short of amazing ( I've yet to need clr cmos for anything ) I think the only other thing I found somewhat annoying was the inability to "unlink" fan control. Like say one fans quiet even at max as its a low rpm. I want to run that full speed I can't unless all run full speed. It applies to all or none unless you set custom profile for each.

Yah 2x8 can pull in a lot easier. IO-L with 2x16gb 8 is the lowest that will post. For reference sake my asrock z390 9900k with 4x8gb bdie and only 3600 c15 can pull 37ns. It should not be that hard or need that much power to get latency down. 3800 is even worse in the 45ns range. 2000 gets me into mid 41ns

Here I did a quick write up on the board. http://www.xtremesystems.org/forums/showthread.php?296970-An-unbiased-and-unsponsored-MSI-Z490I-Meg-Unify-Motherboard-Review

Ryzens limitations has always been the chiplet. I've told people this many many times but they don't listen. it's why latency is poor. It why we have struggled to clock ram although they kinda but not really fixed that on 3000 lets just say they unlinked it but the results were not what they were "hoping" for. They needed the chiplet gone but...I'm not sure how it will impact other results and I am also not sure how they will deal with multiple die product like 3900/3950 without it.

The latency issue at least for DR dimms lies in RTL IO-L I believe. At 3600/3800 you should be able to run 6/6/6/6 or 7/7/7/7 but currently board is freaking out if set lower than 8 and having a mind of it's own and loosening them up to 9/10 if you attempt to do so. This is my findings on B die 2x16gb. your results may differ on diff IC and density. 1T is not happy on DR but usually not a realistic goal DR. Included a screen to show the state of latency currently. 32gb not 16gb........ EDIT FOR A BUG REPORT ( Opening dragonball z or whatever app even if you close it completely and then open AIDA and run cache/memory = a result of 0/0/0/0 across the board )

be a good 2 secs faster if you dump the 3800x and use 3900x

c9 might not be as fast as c10. It wasn't on zen1 at least according to the data I collected. at 1333 c9 is easily doable yet I did not use it I always follow the philosophy I got from tony @ OCZ. Use what IS fast not what you THINK is fast. That's why we tested multiple boards under multiple speeds etc etc etc. I must have tested at least 10+ motherboards on zen 1. Even cpu's impacted the results, tested a lot of those 2. Example I ran multiple tests last week with 3800x over 3900x. The closest I could get was within 1 second of 3900x on good runs and more often than not it was 2 secs slower. I think the largest problem with 32m pi and zen 2 will be the inability to use very low dividers and bumping back up via ref clock to get the boards to "train" at tighter inaccessible "settings" when you go cold to make up for the lack of speed. You can try this on intel and pay attention to RTL in AMD land I think it was MRL on older arch. Not sure if they call it that anymore on AMD but I know as of now and zen1 its not accessible in bios or software but is via lower dividers and ref clock. On zen+ it may not work because its not achieved natively but on a chip separating PCI but I never tested it so who knows. I also know on zen1 when we set timings they had no effect at higher speeds but when we used a lower divider + ref clock to "trick" the cpu they actually worked. I can guarantee you 1 thing for certain. The OS is hardly optimized. I'm not even using max mem. The "spectacle" others put on about the OS was a bunch of dogs barking up the wrong tree. I found it mildly amusing for a little while but the amusement wore off quickly as the drama escalated and I decided to go back to hibernating as that's what's best for me to avoid drama. The only thing people really did not try was a giga or asrock to beat us and we never used the asus to try to beat anyone ( I had it, it was slow cpc ). The asrock and giga used a ref memory trace layout. the asus did not....the asus clocked higher and easier. quite obvious what was going on. they slacked board to "gain" visual speed. This is why infras fastest run is with very high refclock and high mem speed. He got the MRL or whatever they call it now down and bypassed all the slacking crap asus did to achieve higher speeds than other boards could do. Also as to why we didn't bother with ref at 3333-3400 was simple. our boards were not very fond of over 110 and you needed 1866/2133/2400 dividers to get the real gains. 2933 gained nothing and 2666 was not all that fast either. Once again use what IS fast not what you THINK is fast. These cpus have a lot of features that can be used towards your advantage. I believe error correction is 1 so using maxmem to gains stability because your actually unstable....I wonder how that works with the cpu's error correction feature

yah im not sure but based on my run and loud_silences 5.2 untuned run and comparing that to my tuned 5.0 run I personally feel that zen 2 when I run some 7 will be faster. its a different kind of fast but faster none the less I think. I think if you cold bug imc hard though and you had to run both zen 1 and zen 2 @ 1333 10-9-9 etc zen 1 would win however but I don't think that will ever be the case where you need to go that low on zen 2 cold. If you were able to maintain your above runs mem speeds cold @ 5.0 there is no doubt in my mind that it would be faster than my 6:53 5.0 @ 1333 ddr. its looking like a 6:50-6:40 range run to me if you ran @ 4999.