The AMD Ryzen All In One Tread /Overclocking/Memory Speeds & Timings/Tweaking/Cooling

So far I love my 1500x. Got my AMD 4 bracket in for my 212 Evo. Gonna use it for a bit till I get a new AIO. I sold my h100 with my 3930k.
Ram is Corsair DDR4 3000. XMP for it wont work. Best I can get ram at is 2666 @ CL 14 1t. CPU is stock right now with 3.6 boost. I wanna push for 4.0 but I dont know if its possible even with a 212 EVO with single 1200 RPM fan. CPU is around 43c in BF1 with 80% usuage avg.
I am seriously surprised in a great way with Ryzen. Its my first main AMD rig since 2008. I was worried going from a 12 thread CPU to a 8 thread CPU, but damn does this 1500x stock actually feel smoother then my 3930k did that was @ 4.1. Forza Horizion 3 is nearly 100% butter at all times. No stutters, nothing. And most of the system is the same except CPU, Ram, and Mobo.

 

Came across this post when trying to help someone to get their ram functioning and thought it would be helpful to post here as well. The bottom of the post makes a very interest statement.

https://www.overclockers.com/forums...-with-higher-RAM-frequencies/page4?styleid=36

 

Version 090220 17/09/18

PRIME X370-PRO BIOS 0902
Update AGESA 1.0.0.6B

6.1 MBytes

https://www.asus.com/uk/Motherboards/PRIME-X370-PRO/HelpDesk_Download/

 

Flashed it a couple hours ago. For me there is no change compared to 0812. Ram is still limited to 2933 MHz. ( 2 x 16 GB Trident Z F4-3200C14D-32GTZ)
I can get it to post with 3200 MHz if I set ProcODT_SM to 80 Ohms, but it reboots as soon as The windows logon screen shows up. If I set ProcODT to 96 Ohms system won't post with 3200 MHz.
But I haven't really played around with it. Waiting for AGESA 1.0.0.7 since 2933 MHz (14-14-14-14-34 CR-1) is rock solid.

 

Last night I decided to test those parameters on the URL that Closeau posted, and the gaming performance went up a notch and the latency tests went down from 84ns to 81ns, not bad. I disabled BanksGroupSwap and BanksGroupSwapAlt. I couldn't disable GearDown as then it would make my memory unstable. The solution was either GearDown enabled or disabling it with CR 2T but that would increase the latency as well.

 

I've installed this also and have everything setup the same way as the previous bios, no problems but also no improvements noticed. All 4 channels still running at 3200Mhz stable, will test more over the weekend with beers and see if i can get the mem running closer to their 3466 rating. Its doubtful but worth a try

 

Hi All,

Just looking for a little bit of tweaking advice for my 1800X overclock settings - I may be at the limit, but thought it worth asking if there's additional fiddling I could do!

(This is using taichi + 3600Mhz CL16 B-Dye single-rank ram)

Currently - I am stable at the following settings:

CPU @ 4025 - 1.392 vcore reading (set to 1.4 in Bios, LLC 3)
SOC @ 1.12
DDR @ 3333 - timings below:


I'm guessing the below is due to the infinity fabric speed affecting stability -

I can get 4.1 on my 1800x with very little additional vcore - however, that's with the ram scaled back significantly.

I can also run the ram @ 3466 CL 16 - but that's not with anything about 3.9 on the CPU or so.

Am I right in thinking this is due to the infinity fabric, and thus at my limit of them combined?

 

I don't think it is related to your Infinity Fabric but more like the IMC aka Integrated Memory Controller. Some of them due to silicon variations, can withstand better timings and latencies than others, along with the motherboard/BIOS compatibility factor.

on;

 

I started on an Asrock x370 Gaming board and could only do 2933, and to be fully stable 1.37 V and 1.10 V SOC. Now can do 3333 with tighter timings, standard voltage on RAM, and 1.05 V SOC. Could even go a bit lower on that! 3466 works but doesn't feel as smooth. Only change was the motherboard. I think the Samsung B-die RAM helps.

 

I've just installed this bios. When re-entering the memory settings I decided to give 3466 another go. I don't know whether it will still be a bit iffy, it seems to be fine though! I'm guessing I might be pushing it though. It seems to be better though than earlier bioses, I managed that 3466 with geardown disabled, 1.36 V on RAM and 1.075 V on SOC. I'm pretty sure I couldn't have Gear Down disabled before at 3466. Those voltages aren't tweaked, I could possibly go slightly lower, but I don't want to push it. From memory I needed slightly more voltage previously to get 3466 working. I don't think 3600 will work just yet (if at all), so I'm not going to try it unless there is a reason to - maybe the next major AGESA. So far it seems stable. My timings are 16-16-16-36. The memory is DDR4-3733 17-17-17-36, so I don't think I really need to try an push the memory timings any further since it will yield little benefit, and any instability I would not know whether it were the timings or the RAM speed. I'm currently using 1T timing as well, I guess if there are signs of instability later I'll drop this to 2T. It's probably the highest speed where 1T would be stable anyway, I don't even know if Kaby Lake/Coffee Lake could do DDR4-3733 at 1T (at least without extra voltages or whatnot).

I'll report back if it proves not to be stable. I don't overclock the CPU, because I do a lot of encoding etc and it's approaching summer, I prefer having the CPU on the cool side for longevity. So keep that in mind with the following score. CPU-z benchmark has the single thread score for me now at 425, which is 25 higher than the 1800X (I have 1700X) at default. I guess the memory speed even helps pure CPU benchmark due to the way the memory speed is linked.

 

For the last 4.5 hours I was encoding two things in parallel (maximise CPU use), complex scripts and h265, web browsing, and just started watching something. In stability tests it looked stable, but just had a GSOD. I guess that means I was pretty close, but I knew some of my settings were pushing it like the 1T timing. I've dropped to 2T, geardown on auto, tRFC to 520 (from 500, 480 was stable at 3333 on previous bios), and see how it goes this time . If it happens again I might drop back to 3333, but at least I can say that this bios is better with faster RAM speeds. I really hope it is stable with 2T set, I think people can agree I was probably hopeful having 1T at 3466 as stable being a little too hopeful. I could probably do it with 1.2 V SOC and 1.4 V RAM voltage, but I would consider that very much not worth it.

Reason why you couldn't go higher initialy is the CMOS probably wasn't flushed properly. Where is the RAM training data held? Did you let Windows start when you couldn't initially go past 2133? Once the DMI pool data is updated which only occurs at the end of the time where you can enter setup, you might have been able to use the bios normally. I would almost certainly say it was an issue with trained data and info not being flushed, rather than a bad flash. The bios flash does checksum tests on each block of data (or at least it should), so it shouldn't be related to a bad flash which would more likely make your computer unusable rather than not being able to go faster than 2133.

EDIT: I forgot that enabling gear down overrides your manual CR setting, and that Geardown should be just as stable as 2T without quite so much penalty. Will see how the system goes. I ran memtest64 (a Windows based RAM tester) for the last hour, no errors. I guess if it GSOD's now it just means it's exceptionally close. I really do think (hope!) it was me pushing 1T with geardown disabled that lead to the last GSOD.

The recommended procedure (which I haven't done... but should!) is:
Restore bios defaults, reboot, flash, restore bios defaults again, and start the tweaking. If you can't tweak like you experienced, you let the DMI pool information update by not entering setup, and then re-enter the bios once rebooted.

 

That's a bit worrying, maybe MFlash (MSI bios flasher) doesn't do checksum comparisons? That would be a big no-no when it comes to something critical like a bios flash.

 

As far as timings, you may want to try

Spoiler

Am happy with the results and staying here till 3500 or 3600 is stable

Spoiler

have since raised the cpu multiplier to 38.75 and all is still stable. Corona went down 1 sec to 2:00. Those bus and Rtt settings is what allowed the timings to run this tight. Kit being used is F4-3600C16D-16GTZ. They will not boot at CL14 at 3466. Try adjusting the bus and Rtt settings to eliminate that GSOD. On a side note, the cold boot issue on this board boiled down to an optical drive had connected. Replaced it with one on the qvl and it has vanished for good. Good luck.

 

The MSI beta bioses are numbered x.xx, and release bioses x.x. The extra 'x' for the beta bios is the test version of the x.x bios, meaning 1.96 is older than 1.9, and why 1.91 was out ages ago. It's all explained on the beta bios page .

@Clouseau , nice results there! I see you're running 1.395V on the memory, does any lower work? I personally would prefer 1.37 V or less, I guess it possibly made your 1T timing work. What's the performance difference between geardown disabled and enabled at 3466 for you, including system 'feel'? Next time I reboot I may fine tune some settings, although it's probably not worth it unless a timing or two is way out since new bioses and AGESA will likely improve on things.

 

The voltages are needed for stability, any less and errors are thrown. Tried with geardown enabled. Could not get the ram stable with it enabled. Tried by changing up the bus settings and some settings in the tweakers paradise section...would not boot into windows with different bus settings (40-20-40-40 it stays). Ended up to where the bios got corrupted and had to reflash it. But have to say it was interesting that the closest to stable I got was with 1.37v being the bios setting; only ten errors with HCI Memtest to 400%. System felt snappier with it enabled but just could not get it stable without tweaking the timings; something was not willing to do. Maybe will try more relaxed timings with it enabled at a later date; but not now.

 

Gear Down is meant to be more stable than 1T whilst retaining most of the benefit, but 1T should be faster. As such, it's interesting that it's not stable. When you consider 1.395V is just shy of a 33 percent voltage increase, such voltages probably haven't been tested long term on these modules. Also long term high SOC voltage and the effect of memory voltage on the memory controller etc hasn't been tested.

 

Well for the heck of it, finally got 3500 stable. Would agree with you on running this speed. My SOC voltage was a -.01875 offset. Making 3500 stable required upping CLD0_VDDP to 900 from 875. 3500 was not trainable at 875, nor was it trainable at 925. 900 moved the hole away perfectly. SOC went up to a -.00625 offset. VDDP had to be lowed to .870v. Going back to 3466. So am not worried about the SOC voltage; works out to just over 1v under load.


EDIT: well according to my stupidity, I tried getting geardown stable while still on 1T; go figure.

 

I need to find my Hole in CLD0_VDDP now im on 888
I know i can get 3200-3400 on this RAM but it will require some time to get there

Happy with my 2940 CL14 1T

 

Same thing for me. I came from an Asrock X370 Gaming K4, where the best I could get would range from 2933 CL18 to 3066 CL14, depending on the bios.
Now with a different board, 3200 C14 works perfectly fine, even with 32GB installed. I haven't tried any higher speeds yet.

I'm still trying to figure why some settings report much higher latencies in Aida64.