Home Artists Posts Import Register
Gumroad has gone full retard. While we watch it sink, check out the archive password sharing.

Downloads

Content

In the SETTINGS menu there is a new HYBRID OC operation mode - Frequency lock mode. It is designed specifically for systems with a huge amount of garbage background applications, which significantly affect the performance.

The basic principle of its work is similar to the boost mode from Intel processors - the maximum frequency for all cores. Since the Windows Task Scheduler is still unable to intelligently distribute the load between cores and constantly juggles them, this mode is the only way out. We avoid the process of constantly resetting the frequency to a base state and ramping up the frequency in the next step. The processor is immediately ready for all the surprises.

Since this is a test build there is also one drawback, some users will have to manually adjust the left table CO. This is a temporary difficulty.

As an example, I show my own performance results of the new mode (CPU-z and 3D Mark).

____________________________

Changelog:

* Frequency lock mode - new HYBRID OC MODE (better performance for "clogged" systems).
* Improved stability of diagnostics for processors whose cores do not have CO reserve.
* LibreHardwareMonitorLib updated (improved compatibility of the GAME profile with the new drivers).
* Overboost. Tolerances changed, stability improved in ultralight tasks.
* Improved stability for AVX2 profile.
* Improved HYBRID OC wake-up function after sleep (however, I still cannot guarantee 100% protection against rebooting).

____________________________

No new diagnostics required!
____________________________

UPDATE:

For NVIDIA users who have encountered an issue with the GAME profile not working correctly, I am posting a FIX (FIX GAME profile for NVIDIA . zip). Unzip the contents into your HYDRA folder, confirm the file swapping.
____________________________

Special thanks to @eXp.fr,  @profitkaa, @Riwwelorsch, @Ann Martiné, @ACE, @arcanazen, @DayDreamer, @Gadfly, @rizlah , @&&Devin, @Bloax , @RdWing and @wanderinarcheologist  for help with development and testing.

Files

Comments

Anonymous

Awesome

Anonymous

I just copied the file into my old Hydra folder, is it okay to do so?

1usmus

I suggest that in the new version you load profiles (LOAD BACKUP)

Anonymous

It appears that Game Mode is broken, at least on my system, in this release. Game mode is not triggered presumably because Hydra is unable to read the GPU information per this log message: Unable to get information about GPU load. I should not that I am running Windows 11 with the latest Nvidia drivers. Windows 11 very well could be the culprit. But in combination with the new LibreHardwareMonitorLib and without rolling back to Windows 10, it is difficult to tell. 1.0A did not have this problem.

Anonymous

Thanks, Yuri!

Anonymous

I think the wrong PDF was uploaded, it is still a 10A pdf.

1usmus

Because 1.0B - test build and there are no visual differences except for one new test function

Anonymous

Thanks for the work! should I keep the AVX1/2 PPTs the same or can I increase them with a 5950x, dark hero, full mono block with 3x 360 of rad space? Also aren't there PPT/EDC/TDC settings that should be changed in the bios?

Anonymous

Also a nice feature would be a debug view of the log in real time with like maybe 5 verbosity levels?

1usmus

I've been thinking about this for quite some time. The main problem is the speed of logging information. Because there can be more than 100 actions per second.

1usmus

My proposal for the safety system - PPT 220 EDC 220 TDC 160 For ideal volt search function - PPT 165

Anonymous

I seem to be struggling to get good clock speeds (5950x) with this version compared the 1.0A - all cores come up as unstable and downclocking about 300mhz on each core?

Anonymous

Hi Yuri, thanks for the great work made here! Is the compare button currently working? I would love to see the comparison between before and after, like you did with CTR.

Anonymous

I might wait on the next version 1.0a has been good for me. How long before out of Alpha/Beta? Thanks for all the work!

1usmus

The current stage of the project is the beta version. The main bugs are removed, but needs polishing. I plan to release 2 versions this month. One in the middle of the month and one at the end of the month.

1usmus

Tell me which mode you used or if you want I can take a look at your system through TeamView

Anonymous

Apart from the ''Unable to get information about GPU load'' message, it seems to be working fine for my 5950x. Temps are normal again compared to 1.0A

1usmus

Attached to this post is a fix in a zip archive, unzip the archive with the replacement files.

Anonymous

04:14:33: Test 1 started! 04:14:33: 1T-8T EDC limit reached! 168.7A 4:14:34: Cinebench stopped! Can't open Cinebench. But after I close Hydra 1.0B, I can open Cinebench.

Anonymous

***HYDRA 1.0B PRO by 1usmus*** 10/02/2021 16:29:32 Microsoft Windows NT 6.2.9200.0 AMD Ryzen 5 5600X 6-Core Processor ASUS ROG STRIX B550-E GAMING BIOS ver. 2423 SMU ver. 56.53.00 TABLE ver. 3672325 DRAM speed 3200 MHz Restoring the diagnostic process. Please do not take any actions. Unable to get information about GPU load. Unfortunately, the test data has been lost! Performs a retest of the CO.. Phoenix deactivated!

Anonymous

***HYDRA 1.0B PRO by 1usmus*** 10/02/2021 17:25:02 Microsoft Windows NT 6.2.9200.0 AMD Ryzen 5 5600X 6-Core Processor ASUS ROG STRIX B550-E GAMING BIOS ver. 2423 SMU ver. 56.53.00 TABLE ver. 3672325 DRAM speed 3200 MHz Information about the last failure Unstable frequency, APIC ID: 4 If the failure occurred in a game: Decrease by 30 the values for (GAME CO offset, SETTINGS tab). If the failure occurred in a multi-threaded load (CPU usage > 70%): In the CO table #2 reduce the CO for CORE#3 by 10. If the failure occurred in a low-threaded load (CPU usage < 70%): In the CO table #1 reduce the CO for CORE#3 by 20. If the above information is difficult for you - press the MINUS button for both CO tables. Unable to get information about GPU load. Still fails to resume after crash.

Anonymous

Hey @Yuri! At first this is seems to be the first release which is game stable (New World in my case) - thanks for that! My default setup is fully Curve Optimizer verified, means i tested ever single core over weeks with SSE, AVX and AVX2 load - 24/7 stable i can tell. When using the balanced power plan my 5600x drops down to around 0.960V and 3600MhZ - with Hydra i`m at 3500MhZ and 1.100V? Thats not much i know, but i would love to understand. TIA!

Anonymous

Good afternoon! My parameter for 5900x: ALL(GAME) VID-1350, CCD1-3700, CCD2-3700. What am I doing wrong?

Anonymous

So, out of curiosity, which setting does most users use ( Undervolt, Normal, OC, XOC ) and why? Which provides the best performance?

Anonymous

In summer i prefer undervolt, in winter i use XOC. Gaming performance difference is negliglible.

Anonymous

it's a real catch 22 - I'd like to see 5gz in afterburner, and when I'm used to it I may revert back to undervolt from OC (5950x) - however, keeping the CPU cool means the gpu does not get as much fresh air from case fans - but one thing I know - I think I love tweak voltages and fan curves more than gaming right now

Anonymous

How can I bypass the voltage for single core? The max I can use is 1475; anything higher than that I cant activate profile. Thanks.

Anonymous

Seems to work great for me. Slight bump in single and threaded tests. Have you ever looked into the dual CCD chips with one CCD disabled? Chances are small they will hit the same speed as the active CCD but it would be nice to have a dynamic scaler activating these cores under heavy threaded workloads. Probably far fetched but it tickles me to think about it.

Anonymous

Hi Yuri , Whats the max watt I can put in Diagnostic settings for Ryzen 9 5900x ? avx1 PTT-Avx2 PTT

Anonymous

Hi Yuri, I've upgraded to Windows 11 and Hydra is giving me an error "Unable to get info about GPU load" - Is this a knows issue with Win11 or is there something I can fix for this?

Anonymous

Saw your comment on another post in here for replacing DLLs :) thanks

Anonymous

Is it normal, that I get Golden Sample 4.12 Eff. on CTR 2.1, but Bronze Sample 3.94 Eff. on Hydra for my 5800X? Also, which is currently more efficient/performant?

Anonymous

i will crashing even with normal oc profile, apic id 24 on window 11, is it the window 11 cause it or what? cause as i remember before is okay with no problem at all. 5900x dark hero & diy water cooling system.

Anonymous

I hit the same "EDC limit reached!" problem. 5900x with EDC at 220, running Windows 11.

1usmus

CTR has an overestimation of cores. Also Hydra uses a different range for evaluating cores, so the energy efficiency coefficient will be different

Anonymous

Sorry is 26, later I go try again. This not amd common issue ? I search web seem many people is like that

Anonymous

I previously used CTR2.1RC5, recently switch over and now CPPC information does not load onto CTR or Hydra, I tried deleting the files, resetting CPPC setting under bios without luck. HWInfo is still picking up CPPC info. A very old CTR2.0 still shows the CPPC number on CTR, any suggestions? https://i.imgur.com/hMTiX8u.png

Anonymous

Excellent Software. For the sake of clarity, in the instructions "Manual CPU LLC" is mentioned with recommended settings. Do you recommend setting both CPU LLC and SOC LLC to these?

Anonymous

What should I set my Max Voltage and temps for the AIO settings part? Should I adjust these to match the other PPT and Temps in the settings? Or are the default settings good and I just have to pick what cooling I have? Or should I just leave it disabled? I have a 360 rad NZXT Kraken AIO, Thanks

1usmus

You have turned on Preferred Cores and CPPC in bios? P-state you don't use? Also, some custom versions of Windows are not able to display the information correctly

1usmus

Hey! For your system I suggest using the Normal preset, the other default settings are already optimal. Maybe Max PPT 175 should be used.

1usmus

https://www.amd.com/en/support/kb/faq/pa-400 I think this is part of the problem and it is worth waiting for a solution from AMD

Anonymous

Hi, first of all thanks for everything, I have used CTR before and it worked perfectly. Now I have a problem with this program. Whenever it is time to perform the OC (CORE) test, the pc crashes. After restarting, the program does not start automatically, so when starting it manually, the following message appears. What can it be? Thank you. https://ibb.co/k5PmqYw

Anonymous

So, the main issue I have been having is a hard computer shutdown coming back to an idle system. The computer is not going into a sleep state. This is the typical scenario. I play a game, Black Desert Online, that you can minimize to the tray. After leaving the computer idle for a fairly significant amount of time I will return to the computer, bring the game back up, and within 30 seconds, the computer will shutdown. The game profile is enabled. With the game in the background, the game profile is occasionally triggered. I do not have this problem simply putting the game into the tray and bringing it back, even 5 minutes later. I tried a new test today. I disabled all profiles other than the game profile. I was doing so for some other reasons ( which I will discuss in another post later) I was testing as well. Even in this case, the crash just occurred. Generally things seem otherwise stable. This is a very consistent failure. Note, in this case, Hydra does not log any particular issue with a core, though likely that is simply because it fails before it can. Any thoughts on this? I can certainly disable Hydra in this situation to avoid the problem, but I hoped the above information may be helpful, particularly that it occurred with only the game profile enabled.

1usmus

Send me the diagnostic logs, please. 1usmus#0529 - Discord usmus_amd - Telegram

Anonymous

So select AiO, Leave the temps 75 degrees for both, move the max ppt up to 165w on both avx1 & avx2, and run?

Anonymous

Interesting update to this. I've found the near root cause of the problem. In a recent Windows 10 update, Modern Standby was enabled by default for any hardware that supports it. This is the new S0 sleep state. Despite the fact that the sleep configuration is set to never sleep, it seems that my system was still entering the S0 state. That may be typical. Unfortunately, this S0 state is obviously unstable on this system. I imagine due to voltage fluctuations of some sort at resume time. I suspect there is something that could be tweaked, but I'll leave that for another day. You can read more about the SO state as it is rather intesting since it allows processes to continue running in a very low power state. Interestingly, I had noticed there was something odd about power configurations on my system, which started me down this path. The only profile that was available on my system was Balanced and nothing I tried seemed to be able to enable other profiles. It turns out that only having a Balanced profile available is normal when Modern Standby is turned on, as confusing as that may be. For reference, since it isn't obvious how to disable this, the following registry configuration will disable it: reg add HKLM\System\CurrentControlSet\Control\Power /v PlatformAoAcOverride /t REG_DWORD /d 0 I suspect Hydra did exacerbate this situation as there are already some known issues with sleep states, but it was not the cause certainly as it could occur, albeit less frequently, even without Hydra running. I hope others find this information helpful.

Anonymous

It might be a stupid question, but do I have to run the diagnostics for every preset separately? Prior thanks for clarification ;)

Anonymous

no such thing as a stupid question....yes is the short answer. I've done exactly this for undervolt and oc and saved the profiles as backup's