The AMD Ryzen All In One Thread /Overclocking/Memory Speeds & Timings/Tweaking/Cooling Part 2

This is a little weird ...
Tweaked 3200 MHz gives avg 72 FPS in Assassin's Creed Odyssey while tweaked 3333MHz gives avg 67 FPS in 1440p.
Shouldn't it be the other way around ?
Testing continues....

 

You need to find Sweet Spot IMO
LLT 1T -> Maby some Timings are too High? and thus you having some bootleneck/ or chock point in IF.

AIDA is good but Raw Gaming is always better to check things Up

 

Have you tried the Stilt preset timings ? - https://www.overclock.net/photopost/data/1603652/f/f7/f749b56e_The_Stilt_Presets.jpeg , there good settings to try , give them a go to see if you can get the performance up. Also have you tried more voltage at vccio 1.2~1.5v for high ram clocks. Airflow over the rams also helps with stability. I hope this helps.

Kind Regards: Chispy

 

I'm using the Stilt FAST memory presets. I have no stability issues which is why it's a bit odd.

 

Hi my friend , this is weird indeed :/ , maybe as OnnA suggested try to loosen some subtimings and check again if it helps.

 

Starting to feel like there is a line one crosses. When the line is passed, performance goes down. Whole thing reminds me when I setup my Prime-X370. The agesa at that time had the line at right around 3500; performance just tanked at that speed. Also, voltage after a certain point is an enemy for performance. You might be stable but the voltage needed to get there could be the reason. Just my 2 cents.

 

Yes, that could be why. With both 3200 Mhz and 3333Mhz memory is running with 1.4v.
But I'm wondering if running infinity fabric with 3333 Mhz makes xfr2 clock the CPU lower than with 3200 MHz, which could explain the difference.
I will try both again with a fixed multiplier.

 

I will try with loose / auto sub timings to see if it makes a difference.

 

So I just tried running 3200 MHz and 3333 MHz with normal / safe timings.
Meaning 14-14-14-14-34-48 for 3200 MHz - the Flare X default timings.
3333 MHz was using 14-14-14-14-30-52 everything else on auto.
The difference in Assassin's Creed Odyssey is 1 frame !!!





I'm beginning to suspect XFR2 and I think a fixed multiplier will tell a different story.
But so far 3200 MHz with XFR2 is just as fast or even faster with tweaked timings than 3333 MHz.

 

,
I would assume the reason is that it has to modify the process priorities/affinity, which has to be done with a program (similar to what process lasso can do), they could bundle it with the chipset drivers, but it still has to be a separate background process, the reason this exists is due to windows 10's lackluster numa support for threadripper, these problems dont manifest on linux nearly as often, and continues to be improved.

 

G.SKILL Packs 32GB per DDR4 Module

G.SKILL is announcing their new Trident Z RGB “DC” DDR4 modules. The DC stands for double capacity, since these RAM sticks pack 32GB of DDR4 in one PCB.
This allows G.SKILL to offer dual-channel kits with 32GB x 2 configuration. Obviously, this is only made possible with custom PCBs that are much larger than the ones on the regular Trident Z RGB module. In fact, it is almost twice as high since DDR4 ICs have not exactly been miniaturized.

This technology has been closely developed with ASUS, so naturally these modules will be fully compatible on their boards primarily.
This includes the new ROG Z390 MAXIMUS XI APEX, ROG Maximus XI Gene, and ROG STRIX Z390-I GAMING.
The last two are particularly special since they only come with two DDR4 DIMM slots.
Now users can max it out to 64GB despite having limited DIMM slot availability.

G.SKILL offers these new “DC” Trident Z RGB variants in either 3000MHz or 3200MHz versions. Both having CL14 timings at 1.35V.
Although the 3200MHz version also has a secondary 14-15-15-35 variant as well.

128GB for everyone (Yup we can cache whole game thanks to this)

 

IMO Assassin's Creed Odyssey should be exclusive in DX12 like The Division is (I know it has DX11.1)
HW is here but API is not

 

Funny mine acts different, I have CPU all on default XFR, voltage ect. & core speed at 4.348 listed on AIDA. I have same ram as you and with 3200 and G Skill's default timings and 1.35v my Latency is 67.5 but with Stilts fast setup 3200 & 1.40v my Latency is 65.3 and default voltage on SOC is running at 1.13 - 1.14. CPU clock read at 4348 on both runs for Aida I noticed yours are different for each run according to Aidia. I'm curious what is your SOC at? I'm just now getting into messing with this it has been running great but I'm not trying for extreme settings either but seems from most charts that 3200 is sweet spot anyhow but tighter timings never hurt...lol. I haven't tried 3333 yet just been running stilt fast for few days to see how it ran so far so good.

 

SOC is on auto. Will check what it say in BIOS.
When I was running 2 x 16 GB B-die I used SOC with 0.914v. But ever since I got the Flare-X kit it's been on auto.

Edit
Just checked in BIOS and SOC is 1.13v

 

I'm using HWiNFO64 to get readings from just FYI. Don't know what your using. Ya the auto setting on SOC seems to be good, sure I would like to see lower but heat isn't a problem on my side and running great and no hiccups so I figured why mess with something that is working. I don't think I'll mess with cpu at all just tweak on memory my games are fine at stock settings for cpu so just no need to. But the faster ram does make tasks snappy, windows open, prog's opening ect.

 

DRAM Calculator for Ryzen™ 1.4.0
-> https://www.techpowerup.com/download/ryzen-dram-calculator/

 

Just a heads up. A new BIOS 1101 was just released for the CH 7 Wi-Fi with AGESA 1.0.0.6 !
https://www.mediafire.com/folder/jkoxu4r003qu7/Release

It's coming ...

 

I'm assuming your going to check it out, let me know how it is if you don't mind. Anything noticable anyway that makes a difference I don't have Wi-Fi but doesn't matter there. Just curious I'll again assume your going to check memory again, drop some numbers if you don't mind. Appreciate thanks bud ...

 

Hey guys, just testing a new (possibly now stable) Memory OC of 2933 Mhz.

Oh, also, this is with a Ryzen 5 1600 (currently at 3.6Ghz, which is an over-clock that I actually did a few days ago and isolated; made sure it was stable and it still is as of now).

The Memory is the Corsair CMK16GX4M2B3000C15 (2x8GB). They are a variant with Hynix memory on them (they do exist with Samsung memory but they've been quickly superseded by Hynix in more recent versions; the Samsung variants of that specific memory model is very tough to find brand new by now). On the sticker, it is shown that they are rated to run up to 2933Mhz, and 1.35v. Previously, I had them running at "only" 2666 Mhz, because anything I tried to push them more would result in BIOS Post failures, even when I loosened the Timings dramatically (like 18-24-24-50, things like that; it never helped regardless). The only frequency above that mark I managed to get to the Windows startup screen with was only 2700Mhz (can't recall the actual number but it's around there somewhere). However, as soon as I got to the Desktop it all failed (freezes, weird behaviors from apps, blue screens, etc).

Recently, however, I decided to give a try to increasing the SOC voltage. I read around about it a bit and realized soon enough that the default SOC voltage on my Motherboard (the ASUS Prime B350-Plus) was lower than most other "stock" SOC voltages I've seen being reported around (on other Motherboards, of different brands mostly). The BIOS was updated since the last time I checked, so I thought why not give another go with the Memory (the BIOS update says that compatibility was improved, I assumed it talked about making the Mobo more compatible with more types of Memory modules around).

The default SOC voltage was sitting at around 0.850 I believe (just an estimation, I can't check as I type this). The 'safe' over-clock on that voltage is usually reported to be up to around 1.1v (beyond that it becomes too much of a risk, apparently). So I went for a 'middle point' OC on it and gave it +5 'ticks' of incremental increases in the BIOS (Offset) setting. So right now the SOC voltage is set to about 0.925v. Considering that I've seen reports around that some Motherboards have a default SOC voltage even higher than that (at around 0.950) I thought that my setting wasn't too risky (then again, 1.1v is usually where that voltage can go to without too much problems). After that, I bumped the Memory voltage itself from the previous 1.35v (which I was using for my 2666 Mhz setting) to 1.40v. I restarted and... no errors. It booted to the Desktop, no errors. I haven't done any testing yet, not even started a single game to check. I don't know if it's stable, but it's definitely the first time I even see my Memory running at the rated spec I bought them for to start with, so there's that (it's truly a relief to be honest, feels like I wasted no money anymore).

So, with all this said, even though I don't know yet if my Memory OC is stable, I'm simply curious concerning the Timings. The thing is that I only moved the base BIOS settings in AI Tweaker, and most of it is done 'automatically', which includes the Timings. The "manual" stuff I did was only in two steps: 1) Choosing the preset of 2933 Mhz in a drop down list for Memory speeds, and 2) Going with a SOC voltage setting of "Offset" (to get manual OC going) and bumped the small incrementals five times on my + key on my keyboard and that was it (the increments are very low every time, and doing it five times brought me to a voltage of about 0.925v, from a default of about 0.850v).

BUT... the Timings themselves are ALL set to "Auto", which leads to the picture below. I'm wondering if those Timings seem to be 'within the norm' for the speed on my platform. Or should I really try to 'tighten' them a bit? The only thing I read about Timings that was simple enough to understand is that as long as the "CL" Timing (CAS Lantency" sits to an acceptable speed it should be 'within the norm' (say... if "CL" stays around 14 to no more than 17, it should be fine). Then again I'm not sure if that's really accurate or not. I do have 16 on CL, but the rest... I don't know, seems 'ok'? I've seen some user reviews around (for my specific memory) and some of them do report having better (lower) Timings for the same speed, but some others report higher Timings (again, on the same 2933 Mhz speed). But none of them report things like "faster" or "slower" performance or anything dramatic like that. Maybe if it does impact performance it's too negligible anyway.



So I'd appreciate some comments / advice about this if I do have to 'worry' about anything here.

Thanks.

EDIT:

So I read a bit more about the Timings and checked some more reviews and user over-clock reports. The Automatic Timings were indeed too high outside of the CAS Latency (that one is ok at 16, it's the other ones that were too low). So I simply decided to go with the actual specs of my Memory according to Corsair themselves. Which gives Timings of 15-17-17-35-52 (1T) at 2933 Mhz (also shown as such in the BIOS itself as recommended timings at that speed).

I also decided to lower the SOC voltage since I think it might be a little too high and wouldn't be necessary to let it be higher than it has to, to keep stability going (I'll check the SOC voltage separately over time until I get instability, then I'll tick it up again one or two increments and leave it there). For now I went from my initial 2 ticks to just one. However, curiously enough, coming back to the BIOS from the Desktop and checking on the SOC voltage actually reported it being considerably higher than what it was shown to be before I booted to the Desktop (it was originally shown in the BIOS to be at 0.950v after just a normal BIOS restart, but ever since I'm back at the BIOS from Desktop activity it's now showing as 1.118v... strange).

Now, my Timings (still for 2933) are set at 16-17-17-35-52 (1T). So far it seems stable, I've only done a minimal amount of testing though (not much beyond 3DMark and 30 mins of MemTest64, I'll try a full over-night MemTest and another over-night Prime95 stress test for the coming two days or so).

The only "problem" I'm running into right now is this:

In my BIOS, when I go in my Timings section to manually adjust them, I can type any number I want and it properly appears (and stays) as such. However, one of the Timings - for some reason - is STAYING at a specific number after the BIOS reboot and doesn't move from there, regardless of what number I type. The Timing in question is none other than the CAS Latency Timing. It is "stuck" on 16. Indeed, I can actually left-click on its line to type and enter 15, or 14 if I want, and if I do (and I did) enter 15 the actual number does stay and shows as being 15. But, once I restart and go to the Desktop and check CPUz, AIDA64 and other Memory / Benchmark / System Info software around they ALL report my Timings as being 16-17-17-35-52, instead. It might actually be a genuine 'bug' in the current BIOS version of my Motherboard, and only ASUS would be able to fix it.

So I'm not going to fight it, I'll leave it at 16 CAS, and the rest is set to Corsair specifications.

The one 3DMark test I completed (regular Time Spy) shows a bit of an increase in the score (without touching anything else beyond just going from 2666 to 2933). It's not much but now my total score is 7178. I've seen better scores with the exact same system specs being reported though, it's confusing me a bit but whatever, at this point I think I'll be happy that I finally got my Memory running at the specified rated speeds from Corsair (which is why I bought them to start with, didn't want to go below 2933 but I had been 'stuck' at 2666 for months now).