Home Artists Posts Import Register

Downloads

Content

CHANGELOG:


MONITORING

* Replaced "LOAD TYPE" with the numeric value "EDC T". This value determines the throttle level of the EDC. The higher the value, the more severe the load on the system. Helps you adjust AVX1, AVX2 and FMA3 thresholds (SETTINGS tab) more efficiently. For experienced users only.
* Added tooltips for some monitoring data.
* For all monitoring values, the weighted average values are now displayed.
* Monitoring speed for the game profile has increased by a factor of 2. Affects system stability.
* LibreHardwareMonitorLib.dll is not used temporarily.

HYBRID OC
* Automatic disabling of HYBRID OC when user starts to edit the main table.
* Improved performance and stabtility of multithreaded profiles.

HYBRID OC (CUSTOM PROFILE)
* Added a second profile for the second list of applications.  
* Added the ability to sort and filter processes.  
* Custom black list for processes.
   
HYBRID OC (GAME PROFILE)
* New triggers : DEDICATED VRAM and GPU UTILIZATION. The evaluation is now performed for each running process. This allows you to perfectly determine the presence of a running game and improve the stability of the system.  
* Improved compatibility with new OS builds and video drivers.
* Determining the state of the application (whether the game is maximized or not). This determines the activation of the game profile. Significantly affects the stability.

OVERRIDING BEST CORES (in limited mode, error correction is in progress)
* A number of improvements and fixes, 6 cores instead of 4 are available to manage.  
* This function will now only provide each process with free cores (if available). If there are no cores available the Windows thread scheduler will be used.
* Capable of filtering system processes that cannot be accessed.   
* The "speed of applying" the new core list is 200ms.
* The core number input check has been improved.

DIAGNOSTIC
* Additional temporary files so as not to lose important data during the process.  
* Asynchronous recording  temporary files to improve performance.
* Added the ability to configure AVX mode for diagnostics.  
* Automatic filling of Overriding best cores.
* Outputs the intermediate energy efficiency value into the log.
* Improved CO search for UEFI.

OTHER CHANGES
* Added possibility to enter fractional numbers for AVX1, AVX2 and FMA3 thresholds.
* Status bar removed due to unfriendliness to the user. Added a pop-up notification if the user entered values in the tables incorrectly.  
* Added RE-CREATE PROFILE button for extra convenience.
* Added additional threads for monitoring, redesigned the synchronization of the threads. CPU usage is reduced.
* Added information about processor stepping.
* Modified buttons with voltage presets.
* Other important fixes and improvements.

A reminder that version 1.1C will be released at the end of the month (January 31st)!

Thank you for your patience!

Comments

Tom Parkinson

Thanks as always for your dedication and efforts. Looking forward to the future of Hydra!

Anonymous

You asked on the 7th December for user requests, I added a few items. Are any of them included in this change "Other important fixes and improvements." ? Could you perhaps expand more on this last change seeing that you consider them important. I suspect you were under pressure to release on Monday, thanks

Anonymous

Diagnostics sets invalid values (clock instead of core number) for "best cores" at the moment and cant be activated with those values: Hydra sets "Best core#1: 4925 Best core#2: 4950" and so on. EDIT: After a system restart all best cores are "67"...

1usmus

Thank you for saying that. If you manually edit these values, are they valid after rebooting?

Anonymous

Still crashing when CO values is calculated. When disgnose mode is running - - - >>CO core#2 "abnormal high delta" and Crash. Then it will do this step forever and reboot windows like zilion times. I have to stop the disgnose. This has been the case since 1.0E...5950x, msi x570 unify, 3800Mhz ram, Windows 11. Latest bios and updates. Hydra Basic settings.

1usmus

I remember your request to save the position of the application window as well as the seamless log. This will be implemented. Also a pause mechanism. Resetting the counters. Please explain the value of this function. Reset EDC/TDC/PPT values. These values do not affect the performance of the processor and exist only for the stop signal in HYDRA. You do not want to reset all the setiings, only the EDC/TDC/PPT?

1usmus

Please send a personal message with the logs. Also enable "Safe mode" (SETTINGS tab). This will help you avoid this problem.

Anonymous

hey juri, unfortunately I can only agree with that. With one difference, with the first two versions I did not have this behavior. >>CO core#2 "abnormal high delta" and Crash. @r5900x on Dark Hero, Bios3801, 3800Mhz ram, Win11, Hydra Basic settings.

Anonymous

Thanks for your hard work and dedication! I cant restore my settings Backup from v 1.1A to 1.1B " HYDRA INFO --------------------------- STATUS : unable to load data! --------------------------- EDIT: Reason 1.1B hast two Custom Profiles, 1.1A only one. Adding P9 Profile values in the Backup.ini and its working now

Anonymous

"* Added possibility to enter fractional numbers for AVX, AVX2 and FMA3 thresholds." Okay, please let me know if I've got the right idea here then... When I run (on my 5800X) a 16-thread AVX Prime95 workload and EDC T % shows (at the point that it "steadies out") 26.5% while running this load but something like 26.3% while running the same Prime95 workload, but with SSE instructions only, then I'd set my AVX threshold to "26.5" in Hydra's "Settings" menu... Correct?

Anonymous

I did try the " safe mode" settings, but still crash and the computer reboots and the hydra procedure starts again, but can't get passed the "abnormal high delta", what ever it means....

1usmus

Thank you for this question. In most cases even with "SSE" instructions you can do very hard work and warm up the processor, as you can with "AVX".

1usmus

I have reviewed your logs and have already prepared a fix. In a future update and after each unsuccessful attempt, the starting frequency will be lowered to avoid an endless cycle. The update will be available on the 31st.

Anonymous

Не могу пройти тест диагностики, на 10м ядре в определенный момент температурный скачек 95 градусов и система уходит в перезагрузку, в логе пишет температура 95.1 тест остановлен. Не пойму как быть, вроде не плохие параметры подобраны на 10 ядер, как их применить или как все таки пройти тест кто подскажет ? процессор 5900х

Anonymous

Hello, Yuri! Thank you for your work! May be you can add to HYDRA "Profile folders" - support for different folders, from which HYDRA reads all ini files (settings, profiles, may be logs) and a command line switch for that, something like --profile="some_path". So, we can make different shortcuts to HYDRA with command line switches, pointing to different "profile folders". By default HYDRA can use program folder itself, like now. This way we can have completely independent profiles AND settings (IDLE_VID, settings for diag et.c.) for different use cases. For example, I use different DIAG and program settings for undervolting and for XOC and get different CO values after DIAGNOSTIC. P.S. Set "Best cores" manually, enabled it, started AIDA64 - PC hanged. Rebooted, checked settings one more time, started "Cyberpunk 2077" - got veery sloow game work. As I could see on the second display with opened HYDRA window - only one core (C03) had boost frequency (it's even not from the "best cores" list, but has hieghest CO by AMD version) and there were very low frequencies on the other cores.

Anonymous

Hi, thank you for the new release. I was verifying the CO values per core (for UEFI) that the program gives. Now, instead of SAFE/MID/FAST CO values, I get "GUARANTEED CO" (I assumed that they are equivalent to "SAFE CO", right?) and "GUARANTEED+RESERVED CO", whose value is much much higher than the previous FAST CO. For example: C06 GUARANTEED CO: -12 GUARANTEED + RESERVED CO: -30 I understand that -30 is the highest performance possible but I am quite certain that this core cannot post with that value. Is that a normal behaviour? Previously, as FAST CO, it had something around the value -20 or so.

Anonymous

Hello, I got some interesting behaviour on one core. Instead of the base clock (4625 I think) it started CO-testing with 4500 MHz. After cancelling and restarting it began at the last measured frequency of the run before the 4500 MHz at 4675 MHz. More odd stuff was the results table and message at the confirmation run of the each core. "Restoring the diagnostic process. Please do not take any actions. ... CORE| CPPC | BASE | RES | TEMP | CO | UNSAFE C01 | 129 | 4700 | 4700 | 84 | 216 | True C02 | 133 | 4625 | 4675 | 84 | 66 | False ... Phoenix is ready to continue testing! ... 17:06:21: CO (CORE) test started! Diagnostic VID: 1450mV Diagnostic FFT: 84 17:06:24: Step: 14 CORE#2 BASE FREQ: 4675MHz REAL FREQ: 4675MHz 17:06:24: BACKUP#1: saving intermediate values... 17:06:31: Test#1 Attention, CORE#2 has abnormally high DELTA. CORE#2 CO: 54 DELTA: 219 TEMPERATURE: 83°C ... Results CO testing CORE| CPPC | BASE | RES | TEMP | CO | UNSAFE C01 | 129 | 4700 | 4700 | 84 | 216 | True C02 | 133 | 4675 | 4675 | 83 | 54 | True" Watch the base frequency and that it mentions the high DELTA which isn't really there. Greetings Edit: Endless reboot at last core :( Single core test yesterday of this core succesfull, but today with all cores failure. Please give CO Values for successful tested cores during diagnosis and not after. Hoping for next Version.

Anonymous

I cannot get game profile to engage, no matter what settings I choose. RTX 3090 and 5950X. Windows 10.

Anonymous

...try lowering your vram percentage to 25%. However, at 25, game profile might kick in while watching YouTube.

Anonymous

This version no longer has a percentage it is in MB instead I have set it as low as 1 MB and up to 6144 MB under no circumstance does the gaming profile engage.

Anonymous

I've tried this version and it hasn't gone well for me. Running diagnose using manually set voltages seem to get reset to UV profile default upon diag completion. Also, I've noticed micro-stutters in Doom Eternal every few minutes. I've gone back to 1.1A as it seems to perform more consistent for me.

Anonymous

...from memory, I may have disabled the per core CO diagnoses, instead just wanted a recalculation of the base frequencies given an updated voltage

Anonymous

This program gave me virus warning from window defender.. why?

Tom Parkinson

Have you read any of the previous Hydra posts? This has been explained numerous times, and Yuri attaches regular VirusTotal reports to his release notes. It's a false positive due to the way in which Hydra interfaces with the CPU.

Anonymous

이전꺼 포스트 보시면 바이러스 검사 내역이랑 다 있습니다. 오진이에요;

Anonymous

running the diag in safe mode seemed to help my stability issues....

Anonymous

Thanks Mathew - my default CO diag voltage is 1375, and hovering over this field, a tooltip says, that this should match the single thread voltage. As my 1/2T profile was set to 1300mv, I also set my CO diag voltage to 1300. I think this produced very aggressive CO offsets and is the likely cause for the instability that I experienced. Using Hydra 1.1c and keeping CO diag voltage at 1375 and 1/2T at 1300 it seems A LOT more stable. In fact, just thinking about it now, I may had set my multithread CO diag (default 1175) to 1050 - because that's what my 16t profile was set to - extreme undervolting !! So my conclusion is, calculate the CO from the most upper voltage limit as practical, so as to produce the most stable results