Home Artists Posts Import Register

Downloads

Content

This version has some pretty radical changes. I completely removed the old diagnostics and created it from scratch. The main goal was to make it easy to use and to combine all kinds of diagnostics into one, so you can use both HYBRID OC and AMD PBO2 at the same time.  And I took care of all those cases when you don't know what you need to do in BIOS to prepare the system for diagnostics. No requirements, HYDRA automatically configures all the startup settings. The disaster recovery system called "Phoenix" was also created from 0. 

You can also see the search function for the ideal voltages. The desired maximum temperature and power consumption are the only parameters you have to provide.

I would like to note that the structure of diagnostics is modular. You are free to choose only what you think needs to be done.

Control of internal BCLK from Windows was added (thanks to sbski#6362 and Shamino). Thanks to it user has a chance to get a little more frequency for auto-boost.

Dynamic PBO2 is a brand new feature that allows you to change the AMD CO in real time depending on the load (conditions). As I noticed there is a small margin of safety for multithreaded loads. User can specify offset how much to increase CO in multithreaded load relative to static CO values. Note: you cannot use Dynamic PBO2 and HYBRID OC at the same time.

Information about BCLK frequency and percentage core load (according to Windows) for each CCD has been added to the tab. By hovering the mouse pointer over the item of interest you can get help information.

Once again I paid attention to HYBRID OC and in particular to low-threaded performance. Due to imperfection of thread scheduler Ryzen processor family suffers from record amount of context switching (load jumps between cores). Every switch is a loss of performance. Unfortunately HYDRA is not able to approach hardware frequency sampling rate and cores (1ms vs 6ms), that is why I introduce the concept of "waiting cores". For example profile 1T-2T used to be allowed to use only 2 cores. Now to those cores added 2 waiting cores which are not sleeping and have the same frequency as 2 working cores (idling). These 2 additional cores are waiting for Windows to decide to move the task to other cores. This allows for no loss of performance during profile holding or switching.

The concept of soft profile switching has received a number of improvements to avoid excessive frequency loss during profile switching.

COMPARE tab has received a number of enhancements. You can see the percentage difference between the default and tuned state. The normalization of the scales has been changed for better visual comfort.

I'm stuck in testing today, so a full list of changes + a small guide will be added tomorrow (no download restrictions).
___________________________________


VIRUS TOTAL REPORT

___________________________________

The old profiles and configuration files are not compatible!!!

___________________________________

Once again, thank you so much for your support during these difficult times. I wish you and your families would never have to experience what is happening in my country right now.

Files

Comments

Anonymous

Glad to see you're still with us Yuri

Anonymous

Lets gooooooooo!

Anonymous

The temp target for the profile creation is brilliant!!

Anonymous

says file is corrupt EDIT: NVM, it extracts fine but you can't open it with 7zip

Anonymous

On the last core of a 5950X i had a reboot by diagnostic process and following lines in the log: "Restoring the diagnostic process. Please do not take any actions. Attempting to recover data from the general log... Attempting to recover data from the temp file... Unfortunately, the test data has been lost! Phoenix deactivated!" What could have been happened? Can i check something or do to prevent this?

Doug Wilkinson

Wow, Yuri. This is really a fantastic update. 🤩🤩

Anonymous

So after a couple of hours of core-by-core testing, it crashed a few dozen times while testing all cores just to tell me the test data have been lost. This same thing happened with the last version. This software is not working as expected for me. There were also some .NET issues when restarting a couple of times, but closing it and starting it again worked each time. Edit: the all-core testing crashed instantly as it started every time. Less than 1 second after Hydra launching after the reboot it would crash.

Anonymous

There are any Windows 11 specific settings?

1usmus

I will ask you to send a folder of logs, as well as offer in this case to carry out diagnostics in 3-4 stages. Since the system is modular it will be easy to do. An additional question. In your power profile is sleep for drives enabled or not?

Anonymous

This is GOLD the most stable Release . Love it and better performance

Anonymous

Thank you again and awaiting the guide on how to use this version to achieve the best results! When and how to set up parameters for a test? When to change BCLK? What to do with bios? Windows 10-11 specific settings?

Anonymous

Should I be concerned about tha DLDO_VID hitting 1.55V during diag on a core? That seems insanely excessive

Bill Bittel

Same exact thing happened to me. 5950X. Started diagnostic with everything in Bios at defaults except RAM OC to 3200CL14 (stable). This was my first time running Hydra. It took many hours but seemed to get through finding CO values for all 16 cores (98 steps). Then it started "The search a CO for HYBRID OC". The log file shows a bunch of errors at Step #25 of this process. It rebooted and started up again OK. More of the same errors during Step #26, It got up to Step #56 (Core #16), then rebooted again. When it restarted it was now at Step #27. Got to Step #47, rebooted. When it started again it said all test data lost. Breaking this diagnostic into sections sounds like a good idea. Hydra took about 10 hours to do all of this, then lost everything, I just booted up and opened Hydra. It blue screened before I even did anything.

Bill Bittel

Any update on a Guide for this version? I posted in a response further up but short version is after about 12 hours of Diagnostics (5950X), after several reboots I got the "Unfortunately, the test data has been lost!" After this, and a fresh reboot, I got a blue screen soon after opening Hydra even though I did not do anything. After that, it would not even boot. I had to Clear CMOS to get it back. I have the logs and some screen snips of the error messages I was getting.

Anonymous (edited)

Comment edits

2022-12-04 03:43:23 I tried again to run it but ended up dying again, but much sooner this time: ***HYDRA 1.1G PRO by 1usmus*** 06/02/2022 17:00:19 Microsoft Windows NT 6.2.9200.0 NVIDIA GeForce RTX 3090 AMD64 Family 25 Model 33 Stepping 0, AuthenticAMD AMD Ryzen 9 5950X 16-Core Processor ASUS Pro WS X570-ACE BIOS ver. 4101 SMU ver. 56.69.00 TABLE ver. 3672069 DRAM speed 3200 MHz Operating system does not see CPPC core tags! CPPC, CBP, Cool & Quiet and Preferred cores should be ENABLED! Also try another BIOS! Restoring the diagnostic process. Please do not take any actions. Attempting to recover data from the general log... Attempting to recover data from the temp file... Unfortunately, the test data has been lost! Phoenix deactivated! Try another BIOS? Is that a serious suggestion?
2022-06-02 22:05:37 I tried again to run it but ended up dying again, but much sooner this time: ***HYDRA 1.1G PRO by 1usmus*** 06/02/2022 17:00:19 Microsoft Windows NT 6.2.9200.0 NVIDIA GeForce RTX 3090 AMD64 Family 25 Model 33 Stepping 0, AuthenticAMD AMD Ryzen 9 5950X 16-Core Processor ASUS Pro WS X570-ACE BIOS ver. 4101 SMU ver. 56.69.00 TABLE ver. 3672069 DRAM speed 3200 MHz Operating system does not see CPPC core tags! CPPC, CBP, Cool & Quiet and Preferred cores should be ENABLED! Also try another BIOS! Restoring the diagnostic process. Please do not take any actions. Attempting to recover data from the general log... Attempting to recover data from the temp file... Unfortunately, the test data has been lost! Phoenix deactivated! Try another BIOS? Is that a serious suggestion?

I tried again to run it but ended up dying again, but much sooner this time: ***HYDRA 1.1G PRO by 1usmus*** 06/02/2022 17:00:19 Microsoft Windows NT 6.2.9200.0 NVIDIA GeForce RTX 3090 AMD64 Family 25 Model 33 Stepping 0, AuthenticAMD AMD Ryzen 9 5950X 16-Core Processor ASUS Pro WS X570-ACE BIOS ver. 4101 SMU ver. 56.69.00 TABLE ver. 3672069 DRAM speed 3200 MHz Operating system does not see CPPC core tags! CPPC, CBP, Cool & Quiet and Preferred cores should be ENABLED! Also try another BIOS! Restoring the diagnostic process. Please do not take any actions. Attempting to recover data from the general log... Attempting to recover data from the temp file... Unfortunately, the test data has been lost! Phoenix deactivated! Try another BIOS? Is that a serious suggestion?

Anonymous

I thought it may be due to me choosing the "find VID" setting, but even with this not enabled I'm seeing the Diag pushing 1.5V STEP #48 CORE#3 CO: 140 09:33:39: Saving intermediate values... 09:33:42: Saving intermediate values... 09:33:42: Test #3 started! FFT MIN: 84 FFT MAX: 84 FMA3: OFF DURATION: 30 TELEMETRY VOLTAGE: 1.362 TARGET DLDO VID: 1.518 FREQUENCY: 4900 TEMPERATURE: 62 09:34:21: Test #4 started! FFT MIN: 84 FFT MAX: 84 FMA3: OFF DURATION: 30 TELEMETRY VOLTAGE: 1.362 TARGET DLDO VID: 1.526 FREQUENCY: 4900 TEMPERATURE: 63 09:35:00: Test #5 started! FFT MIN: 1344 FFT MAX: 1344 FMA3: OFF DURATION: 30 TELEMETRY VOLTAGE: 1.36 TARGET DLDO VID: 1.47 FREQUENCY: 4850 TEMPERATURE: 62.6 09:35:38: Test #6 started! FFT MIN: 1344 FFT MAX: 1344 FMA3: OFF DURATION: 30 TELEMETRY VOLTAGE: 1.356 TARGET DLDO VID: 1.524 FREQUENCY: 4900 TEMPERATURE: 62.8 09:36:17: Saving intermediate values...

Anonymous

Mine doesn't seem to have lost any data, but instead seems to really not like the first core on my 5950X. On previous versions it put me at around -20 CO or so, but this time it seemed to stabilize at -1 and continued. However, after ~step 122, it decided to continually crash on core 0 every single time, increasing CO by 2 each time. After leaving it on for 24 hours I eventually stopped it and looked at logs, it's currently trying to set the CO of my core 0 to +137 (which is probably why it's crashing lol). I can send over logs if you'd like, just lmk. 5950X w/ a X570 Dark Hero.

Anonymous

Yeah. Running both PBO and HOC just crashes all over the place. That's after the almost 24h the diagnostic took this time (before it was like 6h). Also, don't you just release a patch? Can't seem to find it anywhere now. Well, doesn't matter anyway. It seems the PBO2 settings just deleted themselves after the recent crash. So I guess I have to run a new diagnostic to get them back. Definitely not happening. Lol

Anonymous

I just finished a diagnosis in hybrid or c the system crashed only once and started again by itself. Before I had set the Global C-states to disabled in bios so the result was a bit misleading ... PBO from it's just fine and it's OK, but as far as Hybrid OC + PBO is concerned the test result is worse than the default ... I hope it helps someone and doesn't do things like I did ...

Anonymous

Entering the bios I found the global parameters C 'states re-enabled and then the program set it by itself, at least I think to my surprise, because I believed in a mess ... So I wonder why I got lower results ...

1usmus

I added tooltips everywhere. Please don't be lazy to hover your mouse cursor over the item of interest. The DLDO VID value is a theoretical value, relative to which the processor calculates frequency in real time. Not to be confused with the real volts the processor got. CPU TEL/VID informs user about real voltages and voltages the processor asks for.

1usmus

I am a programmer and tester and someone who writes guides. Due to the fact that I am alone, it is not possible to do everything at the same time. The guide will be later, as promised, in the public domain. About the blue screen. I can't explain anything without a log. I also advise you to read the tooltip before turning on any option. For example turning on the internal BCLK may cause stability problems. Also, especially for cases where something went wrong with the diagnostics in a future patch I am adding the function of forced continuation of the diagnostics.

1usmus (edited)

Comment edits

2022-12-04 03:43:23 Operating system does not see CPPC core tags! CPPC, CBP, Cool & Quiet and Preferred cores should be ENABLED! Also try another BIOS! This message indicates that CPPC tags were not loaded by the operating system during OS startup. At this point, I have no explanation as to why this is the case.
2022-06-04 09:40:11 Operating system does not see CPPC core tags! CPPC, CBP, Cool & Quiet and Preferred cores should be ENABLED! Also try another BIOS! This message indicates that CPPC tags were not loaded by the operating system during OS startup. At this point, I have no explanation as to why this is the case.

Operating system does not see CPPC core tags! CPPC, CBP, Cool & Quiet and Preferred cores should be ENABLED! Also try another BIOS! This message indicates that CPPC tags were not loaded by the operating system during OS startup. At this point, I have no explanation as to why this is the case.

Anonymous

Oh I'm sorry for not understanding your very helpful tooltip that says "Target DLDO voltage of the core (V)" which I then googled what "DLDO voltage" was supposed to mean and found no useful information. In fact, Google searching "TARGET DLDO VID" actually returns more results for sex toys than anything to do with voltages. https://www.google.com/search?q=TARGET+DLDO+VID Throwing Dildos at Target - YouTube https://www.youtube.com › watch 20 Feb 2022 — We threw big dildos at the Target sign to see if they would stick!

Bill Bittel

Thanks Yuri. I do understand that you are doing all of this basically on your own, and under less than optimal circumstances. To be honest, I am amazed that you are working on this at all with the horrible insanity that is happening in your country. I can send the log file but an not sure how to do that on this platform. Otherwise I will just wait for the guide. No hurry, and thanks again for all that you do.

Anonymous

If Phoenix does lose the diagnostic data, you can go into the logs folder in the Hydra folder and read all the logging details before it crashed. On mine, it did have many restarts as it found the optimal CO values, and created the profiles too, up until the Game profile when it crashed and lost all data. Not all lost though since the logs captured everything which I can enter manually into the Hybrid OC page. Edit: Just re-ran profile creation. It was really quick and no problems this time around. All set up now and performance with 1.1G is somehow better than 1.1A which I was on previously, while consuming the same power. Thanks Yuri!

Anonymous

I've been running this one instance of diagnostic for 6 days now, and it's going at the rate of 1 core/day in terms of speed. Is this normal? Assistance mode : Moderate Overclock/Gaming/Mid tier Motherboard+Liquid Cooling My rig Ryzen 9 5950x Asus Rog Strix X570-e Bios 4403 1:1 fclk with ram LLC Level 2 iinm

1usmus

Persuasive, my bad. I will add an additional transcript. Anyway there you can see even 1.68 V, that's not the limit. The better the sample relative to the AMD estimate - the higher the value.

1usmus

I am now preparing a patch, which will record the same 3 logs, but at different intervals. I think this is a chance, otherwise we will have to add cloud storage to the project, but this too may not solve the problem because network devices are able to cache data and data loss is still great. I've been struggling with this problem for quite some time, Windows specifically writes any information, sometimes not immediately, even if you give a command to clear the I/O buffer. I'm sorry this is the way it is, I'm trying to solve it. What drives do you use?

1usmus

The internal BCLK control is available if the user does not have an external BCLK. There are no specific requirements. Note that the maximum value may not be that great 101-102 MHz. The internal BCLK does not have super overclocking capabilities

Anonymous

It shouldn't take that long. Have you seen what the logs say? Is it just slow in testing per core or is it going through many many steps for the curve optimizer? Another thing you can do is to run each of the three diagnostic sections separately.

Anonymous

I have a SATA SSD. I just tried again and it died during Core 16/16. I the software has not yet worked fully; it has had fatal problems somewhere in the process each time.

Anonymous

I always got TDC limit reached, even after i increased it manually in the settings. For example the first result is TDC limit reaceh! 119.1A, after I increased it in the settings to 120, the next test shows TDC limit reached! 125.2A. After that I increased it again in the settings to 130, the next test shows TD Limit reached again with an increased value.

Bill Bittel

How do you run them separately? It took about 12 hours for my 5950X and then I got the message that all data had been lost. Yuri mentioned doing the diagnostics one at a time but did not say how to do this. I have been waiting for him to release an updated guide.

Anonymous

How much longer should I support this project? This is not working for me at all. I cannot complete a full cycle, even ignoring the worst cores. Today, I got through the individual core tests but it crashed a couple times in multicore. After 5 times or so, it says the PPT limit was reached. This keeps happening and I have to babysit my computer most of the day just for it to fail. Then, if I input the saved values, obviously the computer crashes if I run an OCCT test on it, but I can't tell which core is causing it. I can't get a stable OC, I can't finish the PBO diagnostic... this software isn't working. I'm glad others are having success but something is wrong here.

Anonymous

I think it just tells you what TDC it hit, but I think this is to be expected because you raised the TDC so the tune can go higher and suck more power, Set your TDC to between 90A - 150A I just go straight to 150A on a 5900X PPT 215-230W for anything but a 5900X or 5950X PPT 200W EDC 200A

Anonymous

My advice was for Zen3 up to a 5900X for 5950X it is a little different. See this for power limits https://www.reddit.com/r/overclocking/comments/qikcbd/zen_3_pbo_curve_optimiser_tweaking_guide/ The gist of it all is this- 5950X PPT: Between 200 - 300W Depending on cooling ability. (not a lot of things will cool this CPU at that power limit I suggest 215w and go from there up to about 230w stay under 220W if you have not good liquid cooling) EDC: Between 140A - 220A TDC: Same as EDC Thermal Limit: 90c

Anonymous

Happy assist, if you give me the CPU and MOBO and RAM specs. Not too detailed, what you should probably do as a start is confirm the stock system works don't touch the UEFI bios whatsoever just set XMP memory test stability for a 30m and then come back to trying Hydra. There is a workaround for babysitting that do that makes the OS ignore user account login so when the test resets the PC (crashes) the computer reboots and logs itself back into the user account no PW is needed ( you may turn that after you have the diagnostic finished) PPT limit reached is caused by the safety limits being too low for the diagnostic, I have told Yuri this in past I am not sure if he has any thoughts about making the safety dynamic during the diagnostic. I don't like how the safety is a fixed limit during the diagnostic either because when it runs certain tests it will fail and for obvious reasons the tested loads are not realistic it will suck a lot more than what it probably would under everyday use, like for example I may want a 90c limit and a PPT limit but for testing, I don't care what the limit is for PPT to an extent I would like the option to set a BETWEEN value for diagnostic so that I can say I am aiming for 220W but I will accept 300W during diagnostic. It will always fail multi-threaded testing on me due some safety limit reached usually its TDC or temp for me as TDC is 150A and my temp limit is 90c because that is the tjmax of the CPU anyway. The diagnostic process does not throttle when it hits the limit it just stops which I would not mind something done about throttling would be better than stopping because for the long time you have to go and restart the whole thing again because it was like 1c over the safety or 5w over or something. I just don't let the testing do comparisons and stability testing I will do that myself or it will probably fail because of safety limits. I can give you starting values for power limits for any Zen3 chip

Anonymous

He means too Toggle On/Off PBO2 or Hybrid OC in the diagnostic section and if you must you can select how many cores to do during that test, in your case, I would start with half the cores and then do the other half later. This is how https://imgur.com/a/o71brdn