Asrock X370 Taichi Overclocking Thread - Page 737 - Overclock.net - An Overclocking Community
Forum Jump: 

Asrock X370 Taichi Overclocking Thread

Reply
 
Thread Tools
post #7361 of 7406 (permalink) Old 03-25-2020, 05:17 PM
Stock is too casual~
 
Veii's Avatar
 
Join Date: Aug 2018
Location: Austria / Bulgaria
Posts: 700
Rep: 37 (Unique: 27)
Quote: Originally Posted by polkfan View Post
OMG HAHA these are the default timings and settings for 3800mhz for the LOVE OF JESUS people please never ever keep it at these settings i gurentee these are worse then stock 2133/15 cas timings lets do the math here

3800/28=135.7
2133/15=142.2

Oh and it turns on geardownmode haha *** so its way worse then 2133mhz/15

Also as one can tell auto settings for all the memory and timings and training everything default. Now i'll bump that down to 3600 and 3200 and see

Just for laughs i'll re-run Sisoftware and also test memory latency at my optimized settings the test in ryzen calc completes in 107 seconds.

Link for the Sisoftware results for the horrid default 3800/1900FCLK

https://docs.google.com/document/d/1...it?usp=sharing
Spoiler!
I mean tRFC 666 takes a long time to complete
This "defaults" did indeed load your procODT of 34.4Ω, at least it look to behave better than procODT of 60Ω

At the end we need to find out, why it doesn't let you use optimal low voltages, to give you a bit of OC headroom, but defaults to over 1v VDDP and VDDG
Quote: Originally Posted by polkfan View Post
As for the ProcODT Settings 28.2 is the lowest setting i can pick at 3800/1900 nothing lower is available as one can see it boots fine and still runs at 1.1V VDDP+VDDG and 1.2V SOC
Spoiler!
Yea, that's great hopefully it's stable - but voltages are a bit too high
Maybe cross compare if anything changed between low procODT and your Record Settings

We need to get this voltages down , glad procODT 28ohm boots
Even tho it's not flat 28ohm
You don't have any option for 30ohm ?
Btw, Changing procODT through ryzen master including CAD_BUS, can indeed cause HW damage
There is a bug where it will run 120ohm procODT and 120ohm CAD_BUS Drive Strengh (first one)
Be cautious to anyone who reads this and want's to experiment with it

SiSandra Inter-Core Bandwith 14nm Ryzen global #1 | Twitter
Personal Records
RAM > Viper Steel @ 3466C14 (64.2ns) & Alt | Hynix-MFR 3333CL14 (72.5ns) & Alt

GPU > Gigabyte 970 G1 @ 1595|1900 (Hynix) | RX 580 Red Devil GS @ 1620|2200 (Micron)
CPU > TR 1950X [email protected]1.2v & [email protected]1.365v[105Mhz FSB](3.23Beta)
#Main: 1700X [email protected]1.365v1.73Ghz IMC CL14 (64.2ns)」+ Dark Rock 3 | X370 Taichi 1.0.0.3 ABBA by The Stilt | Viper Steel [PVS416G400C9K]| Gigabyte 970 G1 | Samsung 850 Evo
Veii is offline  
Sponsored Links
Advertisement
 
post #7362 of 7406 (permalink) Old 03-25-2020, 05:22 PM
Noctua Fanboy
 
polkfan's Avatar
 
Join Date: Apr 2015
Posts: 1,041
Rep: 16 (Unique: 13)
Ah i never once used RM to change anything but PBO settings sometimes i always just use it to check the settings as it seems to always report the correct settings.

I was at least impressed that 3800/1900FCLK even booted with auto settings for sure a accomplishment on Asrock haha. I do believe i seen 30Ω.


Quite a lot has been improved to say the least and to be honest even the latency on memory seems improved on crappier setups and good setups so that's a great thing for all.

I kind of wish i had a 2000/1000 CPU to play around with and see if any improvements made it to them. EDC bug is still their and indeed with the FIT bug i can get 100mhz or so more MT speed but to be brutally honest ST turbo works almost as good as the FIT bug.

Have to be honest i wasn't expecting anything new for the Ryzen 3000 parts i really wasn't with Zen 3 coming so i'm guessing its to get ready for Zen 3 and like always the previous Zen family's get to see improvements too.

3700X PBO (IF, 1900mhz)16GB G.SKILL TridentZ Series Samsung B-die 3800mhz 14-17-10-32-42
MSI GTX 1080 Gaming X
EVGA SuperNOVA 750 G3
Fractal Design R6(All fans replaced with 5 Noctua A14 fans)
H150i(Fans replaced with Noctua NF-A12)

Noctua fanboy detected! Fans all run at 500RPM

Last edited by polkfan; 03-25-2020 at 05:27 PM.
polkfan is offline  
post #7363 of 7406 (permalink) Old 03-25-2020, 05:25 PM
Stock is too casual~
 
Veii's Avatar
 
Join Date: Aug 2018
Location: Austria / Bulgaria
Posts: 700
Rep: 37 (Unique: 27)
Quote: Originally Posted by polkfan View Post
3600/1800
3200/1600

auto settings down below it keeps the same voltage as the 3800/1900 settings for VDDP+VDDG but 3200 drops the SOC down to 1.1V
Spoiler!
mm, so it only behaves & predicts fine voltages, if memory is under 2800MT/s haha
Either your old memory training or CAD_BUS of flat 24-24-24-24 does mess it up, instead of 24-20-20-24 or 30/40-20-20-24
At the end , memory training result is worse with 24-24-24-24 set and still uses too high voltages for this low frequencies
Guess Thracks (Robert Hallock) & The Team still need to finetune a bit better the PMU
Good reports

SiSandra Inter-Core Bandwith 14nm Ryzen global #1 | Twitter
Personal Records
RAM > Viper Steel @ 3466C14 (64.2ns) & Alt | Hynix-MFR 3333CL14 (72.5ns) & Alt

GPU > Gigabyte 970 G1 @ 1595|1900 (Hynix) | RX 580 Red Devil GS @ 1620|2200 (Micron)
CPU > TR 1950X [email protected]1.2v & [email protected]1.365v[105Mhz FSB](3.23Beta)
#Main: 1700X [email protected]1.365v1.73Ghz IMC CL14 (64.2ns)」+ Dark Rock 3 | X370 Taichi 1.0.0.3 ABBA by The Stilt | Viper Steel [PVS416G400C9K]| Gigabyte 970 G1 | Samsung 850 Evo
Veii is offline  
Sponsored Links
Advertisement
 
post #7364 of 7406 (permalink) Old 03-25-2020, 05:39 PM
Stock is too casual~
 
Veii's Avatar
 
Join Date: Aug 2018
Location: Austria / Bulgaria
Posts: 700
Rep: 37 (Unique: 27)
Quote: Originally Posted by polkfan View Post
Ah i never once used RM to change anything but PBO settings sometimes i always just use it to check the settings as it seems to always report the correct settings.

I was at least impressed that 3800/1900FCLK even booted with auto settings for sure a accomplishment on Asrock haha. I do believe i seen 30Ω.


Quite a lot has been improved to say the least and to be honest even the latency on memory seems improved on crappier setups and good setups so that's a great thing for all.

I kind of wish i had a 2000/1000 CPU to play around with and see if any improvements made it to them. EDC bug is still their and indeed with the FIT bug i can get 100mhz or so more MT speed but to be brutally honest ST turbo works almost as good as the FIT bug.

Have to be honest i wasn't expecting anything new for the Ryzen 3000 parts i really wasn't with Zen 3 coming so i'm guessing its to get ready for Zen 3 and like always the previous Zen family's get to see improvements too.
Yea 1st gen was fun, usually i would say - a finetuned 2700X performs exceptionally well
While we had this limits up
Now what most of it is gone, and the difference between coupled and decoupled mode is only 12ns inter-CCX
We should be able to move up and up without having much penalty

I was expecting AMD to play their hidden card if Intel would launch something dangerous , as there where 30GB/s InterCore bandwith lost for nothing
Which relates to a huge bump in IPC
I struggled a lot to get 4-5GB/s out of the 1700X, only worked if everything was running perfectly
But alone that lead 50-60cb in R15 and some nice boost after lowering L3 Latency with performance Boost and Performance Bias settings
(applying the same inter-core latency fixes that 2nd gen was shipping with)
It was fun, wish i could play a bit more with it - it had a bit of headroom left
CL12 3467MT/s would've been fun, the 14-12-14-12 where stable, but still a bit offsync according to SiSandra
Well, whatever - i had a lot of fun with it
You where forced to keep signal integrity clean, else you could forget anything over 3200MT/s

At least since 2nd gen , stuff didn't change that much , rules still apply
I strongly think, that we should be able to to reach that 2000FLCK wall after all artificial limits are gone
Question would be, how clean does it need to be

Down from 34.4 to 28ohm is already a big step in the right direction
Let's play a bit with voltages
950mV VDDG, 900 VDDP does work well for 8x3733MT/s dual-quad channel according to Yuri
According to IMC design, it should feel very well under 1.05v vSOC
Yes, let's continue to play on that part - we already are making wondeful progress
Hitting 1950 stable FCLK is the next step now

Actually you do pass Y-Cruncher first 3 tests with 1:1 1900 ?
These ones:
Spoiler!

LinX is still too unreliable, but at least Y-Cruncher is consistent so far

SiSandra Inter-Core Bandwith 14nm Ryzen global #1 | Twitter
Personal Records
RAM > Viper Steel @ 3466C14 (64.2ns) & Alt | Hynix-MFR 3333CL14 (72.5ns) & Alt

GPU > Gigabyte 970 G1 @ 1595|1900 (Hynix) | RX 580 Red Devil GS @ 1620|2200 (Micron)
CPU > TR 1950X [email protected]1.2v & [email protected]1.365v[105Mhz FSB](3.23Beta)
#Main: 1700X [email protected]1.365v1.73Ghz IMC CL14 (64.2ns)」+ Dark Rock 3 | X370 Taichi 1.0.0.3 ABBA by The Stilt | Viper Steel [PVS416G400C9K]| Gigabyte 970 G1 | Samsung 850 Evo
Veii is offline  
post #7365 of 7406 (permalink) Old 03-25-2020, 05:45 PM
Noctua Fanboy
 
polkfan's Avatar
 
Join Date: Apr 2015
Posts: 1,041
Rep: 16 (Unique: 13)
I can pass Y-Cruncher in 20 passes the thing is i wonder if it will in a month or so when my 3900X comes in? From what i have been reading 1900FCLK is much harder to come by with dual CCD's for some reason.

3700X PBO (IF, 1900mhz)16GB G.SKILL TridentZ Series Samsung B-die 3800mhz 14-17-10-32-42
MSI GTX 1080 Gaming X
EVGA SuperNOVA 750 G3
Fractal Design R6(All fans replaced with 5 Noctua A14 fans)
H150i(Fans replaced with Noctua NF-A12)

Noctua fanboy detected! Fans all run at 500RPM
polkfan is offline  
post #7366 of 7406 (permalink) Old 03-25-2020, 05:58 PM
Stock is too casual~
 
Veii's Avatar
 
Join Date: Aug 2018
Location: Austria / Bulgaria
Posts: 700
Rep: 37 (Unique: 27)
Quote: Originally Posted by polkfan View Post
I can pass Y-Cruncher in 20 passes the thing is i wonder if it will in a month or so when my 3900X comes in? From what i have been reading 1900FCLK is much harder to come by with dual CCD's for some reason.
Yes i mean right now, with 28ohm and the new Inter-CCX & Infinity fabric changes
With your great settings - the first 3 tests, one round is enough to show "HW error"

I don't think it's thaat hard
I mean if we look at how many CCDs Threadripper has
VDDG nor VDDP is an issue - can say bios, user error and procODT user error
Shouldn't be that hard - IMO 3467 on 1st gen was called to be "hard"
It's not hard, 3600 is hard
Just needs low procODT and low voltages ~ no one cares still about SiSandra and thinks Signal Integrity is an issue
People push CCX over the limits and wonder why there is no room left for FCLK OC
The IMC still hasn't moved away from the SoC - it still depends on chiplets OC and signal integrity will suffer from that

2nd gen can hit 1900FCLK without thaat much struggle, and here it should be even a better scenario for it with far less procODT required
People should struggle with 1920-1950, not 1900

SiSandra Inter-Core Bandwith 14nm Ryzen global #1 | Twitter
Personal Records
RAM > Viper Steel @ 3466C14 (64.2ns) & Alt | Hynix-MFR 3333CL14 (72.5ns) & Alt

GPU > Gigabyte 970 G1 @ 1595|1900 (Hynix) | RX 580 Red Devil GS @ 1620|2200 (Micron)
CPU > TR 1950X [email protected]1.2v & [email protected]1.365v[105Mhz FSB](3.23Beta)
#Main: 1700X [email protected]1.365v1.73Ghz IMC CL14 (64.2ns)」+ Dark Rock 3 | X370 Taichi 1.0.0.3 ABBA by The Stilt | Viper Steel [PVS416G400C9K]| Gigabyte 970 G1 | Samsung 850 Evo
Veii is offline  
post #7367 of 7406 (permalink) Old 03-25-2020, 06:13 PM
New to Overclock.net
 
thomasck's Avatar
 
Join Date: Dec 2017
Posts: 462
Rep: 8 (Unique: 8)
I've tried 6.20A for a short while, and I've noticed is;
Cannot use same timings as 6.20, no boot at all, so I came back to 6.20 as I needed de pc,
VDDP stock is 0.928. and does not change when fclk 1:1 is over or equal to 3600. However master ryzen reports accordingly to bios when changed, hwinfo keeps showing 0.928
Both vddg are shown correctly.
soc is correctly shown in hwinfo accordingly to bios, and ryzen master always show 1.2V.
cpu all auto pbo/xfr off gives 1.36v to the 3900x while running linpackx or cinebench. and boosts fine, at least numerically, then I applied my normal offset (so 1.28+- on load), I get same scores but no boosts. score will be better if you can sustain less 70C. Whatever.

PCDEBAIANO
(7 items)
CPU
Ryzen 3900X
Motherboard
Taichi X370
GPU
AMD Radeon VII
RAM
HyperX Predator 4000
Cooling
3x360mm
Case
Lian Li o11d
Monitor
HANNspree HG324
▲ hide details ▲
thomasck is offline  
post #7368 of 7406 (permalink) Old 03-25-2020, 06:40 PM
Stock is too casual~
 
Veii's Avatar
 
Join Date: Aug 2018
Location: Austria / Bulgaria
Posts: 700
Rep: 37 (Unique: 27)
Quote: Originally Posted by thomasck View Post
I've tried 6.20A for a short while, and I've noticed is;
Cannot use same timings as 6.20, no boot at all, so I came back to 6.20 as I needed de pc,
VDDP stock is 0.928. and does not change when fclk 1:1 is over or equal to 3600. However master ryzen reports accordingly to bios when changed, hwinfo keeps showing 0.928
Both vddg are shown correctly.
soc is correctly shown in hwinfo accordingly to bios, and ryzen master always show 1.2V.
cpu all auto pbo/xfr off gives 1.36v to the 3900x while running linpackx or cinebench. and boosts fine, at least numerically, then I applied my normal offset (so 1.28+- on load), I get same scores but no boosts. score will be better if you can sustain less 70C. Whatever.
Thank you for confirmation too
What where your old memory settings ?
What do you mean by "same scores no boosts" - no boost in score or ?
Which offset did you apply ?
By applying personal offset, you wipe the pre'applied auto offset making only SMU and PMU difference speak for themself
PMU does speak for themself, because you see a difference with your memory settings
SMU, hmm it's hard to say - boosting behavior will change if you aren't clock stretching or overvolting from the start
PSP firmware stays updated, even if you downgrade (because it's inside the chip)
~ tho speaking of downgrade, did it allow you just to downgrade by itself or did you use flashtool

An updated PSP firmware inside the ryzen, could make issues with low bios versions
It's at least good to hear that it still works will 6.20 after updating - although if PSPTool works anymore, unsure

@polkfan & people who update, please double-read this message
Quote: Originally Posted by Veii View Post
To everyone who hasn't updated:
Please doublecheck 4 things:
- visit https://github.com/irusanov/SMUDebugTool/ and help infraredbg a bit by submitting SMU reports before this big update and after with corresponding new SMU Version
It's important to get a readout before update, and after update
Also important for the future, to check if we got an SMU update (if the version changed) or it's still masked as the old one
Because it clearly was updated
If SMU got updated, throttle temp should've shifted too (IF they changed them)
Although you can override it from AMD CBS , it should potentially make a difference
* well not only that, but you'll contribute to the PSP tool and infrared's zenstates tool, which both go in the direction of more control on the ryzen (PB & PE, including PBO)=more performance, as also PCIe 4.0 unlock for X370/X470 after PSPTool is able to decrypt AMDs secret-key, which gives it full control to the chip's internal PSP firmware
(we did have control, but needed still the registers where how it crosstalks with SMU ~ which again is zenstate's project
^ soo please support that)

SiSandra Inter-Core Bandwith 14nm Ryzen global #1 | Twitter
Personal Records
RAM > Viper Steel @ 3466C14 (64.2ns) & Alt | Hynix-MFR 3333CL14 (72.5ns) & Alt

GPU > Gigabyte 970 G1 @ 1595|1900 (Hynix) | RX 580 Red Devil GS @ 1620|2200 (Micron)
CPU > TR 1950X [email protected]1.2v & [email protected]1.365v[105Mhz FSB](3.23Beta)
#Main: 1700X [email protected]1.365v1.73Ghz IMC CL14 (64.2ns)」+ Dark Rock 3 | X370 Taichi 1.0.0.3 ABBA by The Stilt | Viper Steel [PVS416G400C9K]| Gigabyte 970 G1 | Samsung 850 Evo
Veii is offline  
post #7369 of 7406 (permalink) Old 03-25-2020, 07:10 PM
Noctua Fanboy
 
polkfan's Avatar
 
Join Date: Apr 2015
Posts: 1,041
Rep: 16 (Unique: 13)
Quote: Originally Posted by thomasck View Post
I've tried 6.20A for a short while, and I've noticed is;
Cannot use same timings as 6.20, no boot at all, so I came back to 6.20 as I needed de pc,
VDDP stock is 0.928. and does not change when fclk 1:1 is over or equal to 3600. However master ryzen reports accordingly to bios when changed, hwinfo keeps showing 0.928
Both vddg are shown correctly.
soc is correctly shown in hwinfo accordingly to bios, and ryzen master always show 1.2V.
cpu all auto pbo/xfr off gives 1.36v to the 3900x while running linpackx or cinebench. and boosts fine, at least numerically, then I applied my normal offset (so 1.28+- on load), I get same scores but no boosts. score will be better if you can sustain less 70C. Whatever.
This is so weird how can it be different for other CPU's? Maybe Veii is right that the board saves profiles or training's or something even after a update?

I mean beleive me the OC profile worked perfectly for me and all my memory timings stuck that would drive me nuts to if i had to redo them haha.

3700X PBO (IF, 1900mhz)16GB G.SKILL TridentZ Series Samsung B-die 3800mhz 14-17-10-32-42
MSI GTX 1080 Gaming X
EVGA SuperNOVA 750 G3
Fractal Design R6(All fans replaced with 5 Noctua A14 fans)
H150i(Fans replaced with Noctua NF-A12)

Noctua fanboy detected! Fans all run at 500RPM

Last edited by polkfan; 03-25-2020 at 07:35 PM.
polkfan is offline  
post #7370 of 7406 (permalink) Old 03-25-2020, 09:22 PM
New to Overclock.net
 
thomasck's Avatar
 
Join Date: Dec 2017
Posts: 462
Rep: 8 (Unique: 8)
@Veii

Memory settings are the one we did together, a time ago, I still using it,

Spoiler!



What I mean by same scores is the following. If I leave cpu completely auto, pbo/xfr disableb in both menus, I get around 1.36V under load. CPU will boost to 4500~4575 frequently while browsing, youtube and stuff, and here and there will boost to 4600 too, but not that often. When benching I'll score around 7200 in cb20 and around 3190 in cb15, that's due to the cpu breaching 70C, and sustaining 70+> it throttles down the clock to around 4075~4100 and stays like that.
When I apply a more aggressive offset (bios comes with offset enabled and on AUTO) of -0.1 straight, it gives me ~1.296V under load, I do score the same, or even more some times, cause the cpu won't breach 70C quickly, keeping around 4100-4200 for longer, hence higher scores sometimes (depends on how hot the loop already is). But I don't see any clocks reaching 4500mhz (only 4450ish) whatsoever while youtubing etc. So clock speed is down to maintain the cpu below 70C, which is not possible in my case, I already cooling it and the gpu with 3x360. Would I benefit from pushing forward the vcore hence the fit is "1.36"? No, I won't, thermals will not allow, unless I keep my window opened. I got close to 1.35V before testing this and thermal it's a wall. I'll yes, see more boosting but when under load it will breach 70C and clocks will be drastically reduced, and performance will same, or slightly worse comparing with -0.1 offset.

And yes, it allowed to just downgrade using the tool in the bios, like always, no change about that. actually I've never used a flashing tool.

PCDEBAIANO
(7 items)
CPU
Ryzen 3900X
Motherboard
Taichi X370
GPU
AMD Radeon VII
RAM
HyperX Predator 4000
Cooling
3x360mm
Case
Lian Li o11d
Monitor
HANNspree HG324
▲ hide details ▲

Last edited by thomasck; 03-25-2020 at 09:47 PM.
thomasck is offline  
Reply

Quick Reply
Message:
Options

Register Now

In order to be able to post messages on the Overclock.net - An Overclocking Community forums, you must first register.
Please enter your desired user name, your email address and other required details in the form below.
User Name:
If you do not want to register, fill this field only and the name will be used as user name for your post.
Password
Please enter a password for your user account. Note that passwords are case-sensitive.
Password:
Confirm Password:
Email Address
Please enter a valid email address for yourself.
Email Address:

Log-in



Currently Active Users Viewing This Thread: 5 (1 members and 4 guests)
hesee
Thread Tools
Show Printable Version Show Printable Version
Email this Page Email this Page


Forum Jump: 

Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off