neurotix

Hey @J7SC_Orionthanks so much, I am going to rep you up. I did 104 eclk with +10 all core and now my chip is hitting 5500-5600 in single thread benches. It still has the odd problem where at I see it hitting 5750 or even 5800 at idle in hwinfo64, but under a load (Fire Strike, 3dm11) it doesn't hit those high idle frequencies, it does like 200-300 less. If you have any idea why my chip clocks down during a single thread load, please let me know. Should I try 107 eclk? If so, how much voltage should I add? Additionally, it might not work anymore but there's a specific Ryzen Cinebench tweak: do manual per-CCX overclocking, set it to like 52 first CCD and 48 second CCD with like 1.3v, and run Cinebench in realtime mode. You should get a score that is a few thousand points higher than running it using PBO or CO. Give it a try. EDIT: If I aim for 107 eclk what amount of voltage should I use? 25? Even with it at 104 just now with +10 all cores and trying to bench Fire Strike, it froze up before running the bench and the whole program crashed, Steam wouldn't allow me to reopen it even after killing the Futuremark tasks, I had to reboot but then decided to go back to folding in Linux with eclk on auto and no extra core voltage.

Yeah I kind of remember you switching to just regular PBO because of that. I can't have it crashing on me at idle randomly given I run folding in Linux pretty much 24/7 on the GPU, need the CPU stable for that.

The Gigabyte didn't have it anywhere in it's options, there was nothing like in the screenshot you have. However, I got my ROG Strix X670E Gaming Wifi back from a socket repair and we put it in today and it supports eclk. So I have a question. Would it be reasonable to set eclk to 104 and CO to +5 all cores as a starting point? I'm really a novice with CO overclocking and it's like black magic to me. I have really no clue how to figure out the values or test cores individually, do I run Prime95 or something with CO on and see which cores fail and then give them more voltage? I was also thinking of just a straight CO overclock, no eclk, is setting all cores to -18 then running P95 and see which cores fail and raise the values til they don't? Thanks Orion, I really appreciate your help.

In bigred

I can't even find eclk on my Aorus X670 Elite AX. Under the base clock setting at the very top there's no option for manual and expo, just the bclk speed. If I go to EXPO and turn it on there's a few options but none are or give eclk. Where do I find it in this BIOS?

Thanks so much for this! Question though- do you have to enable EXPO for this to work? How do I still retain my 7600MHz cas36 RAM overclock?

Hey, because of the bclk being tied to PCIE and southbridge, I think I am actually wanting to do eclk overclocking. Can you please point me to a good beginner's guide? Thanks

Got it. Thanks.

I'm a little confused about the list. It has the names of people (mostly Mr.Fox) who presumably did a write up or guide on how to do certain things, yet there's no links to the writeups or content. Shouldn't there be like links to whatever write up with the persons name beside it? How do I see Mr. Fox's tutorials?

Thanks so much buddy. And I don't see an eclk option anywhere in my BIOS. It was base clock at the top of the tweaker menu. Should I be disabling PBO as well, because of the 5750 multiplier thing?

Hello, I am trying to overclock my 9950x on a Gigabyte Aorus X670 Elite AX. I still have the problem where in hwinfo64 I can see my 1st CCD cores hitting 5650 or 5700MHz using PBO+200 at idle, however when put under any load (usually 3dmark11) the cores go to their default boost which is 5450MHz. I decided to set PBO to auto and try overclocking using bclk 104Mhz, I lowered my memory multiplier so it would be 7600 (which I have stable). 104x54.5 is 5668MHz or something. Voltage was auto. I tried this more than once and the system refused to post. I'm a novice at bclk overclocking and never really done it. Do I need to go into AMD Overclocking in the bios and adjust the uclk and memclk or something? Are there secondary voltages for the bclk that need to be raised? I'm stumped. Any help would be great, I know @J7SC_Orion knows how. Thanks in advance.

This is pretty cool Ultra. Definitely makes for some nice photos.

Hey you two, can you help me out here? I have a 9950x now and it runs my memory at 7600MHZ just fine, but I am aiming for 8000MHz. It made it almost all the way through the test and then I got this with no errors afterward: That was the only error. What should I do? Raise voltage or a particular timing? I did have the flag --max_errors 1 but it kept going. Is this something minor that is safe to ignore, or should I not run my system at 8000MHz ram? (Orion, I am still using your timings and running the sticks at 1.435v)

Congrats on the win firedfly. Great to see you in it again. How many 4090s were you running, 3? Otherwise what's your GPU makeup? I'm pretty happy as this is my personal best for a two day foldathon. Last month I did 78m, the month before that 77m, now I have set a personal record of 81m in two days, which I think is pretty great

Thanks for the info. I'll just wait then.

Hey E, I am having trouble with rig creator and cannot edit any of my rigs. There's no pencil icon by them to edit them. I was going to go put prices in. I'm on mobile in Firefox using a lot of addons, they have never affected it in the past. Thanks.

What happened to your GPU?

Sorry for the late reply, but I wonder if the meager gains in Windows will be rectified with Windows patches or chipset driver updates. AMD FineWine™ and all that. Who knows but here's hoping.

I found something interesting: AMD Ryzen 9 9950X & Ryzen 9 9900X Deliver Excellent Linux Performance Review WWW.PHORONIX.COM Last Wednesday was the review embargo for the Ryzen 5 9600X and Ryzen 7 9700X Zen 5 desktop... Check out the last page. In over 400 Linux benchmarks, they're saying the 9950x is 17% faster than the 7950x. Also saying the 9950x is 33% faster than the 14900k. Good news for me as Linux is my main OS.

Try watching "8-Bit Guy" on YouTube. His focus is mainly on the Commodore 64 and similar 80s computers, but he fixes a lot of them and also does "retrobrite" on them (hair bleach and water in a tub and soak a pc case or whatever in that outside in the sun to get rid of yellowing). Try and find his videos including soldering, I think he did extensive soldering on a Macintosh SE as well as a Mac LC III. I pretty much learned to solder just by watching that guy, but just enough to fix old game systems, he doesn't really do SMD soldering. Louis Rossman is another great one, he mostly repairs Macs but he does do SMD soldering, as well as reworking on CPUs or GPUs. For general PC knowledge, and up to date stuff, GN is a very good source though personally after a while how fast Steve talks gets kind of annoying to me. Certainly they are better than Linus or Jay though. I also attended tech school twice but only for one semester each time, then I dropped out. I did get a 4.0 both times but the classes were braindead easy. I have a Cisco certification in PC repair from the last time I went. Coursera is another great option if you've never heard of it, you can finish the courses at your own speed and generally they are supposed to be a month long but you can do them in a week if you focus on them 8 hours a day. They will give you a certificate at the end you can attach to your resume. Here's a link. https://www.coursera.org/learn/network-security Other than that, look into PC repair best practices and things to do in order to diagnose a failed machine if it happens to you. For example, I have fried two PSUs by folding on my 4090 24/7 in Linux. The first was because of very bad practice on my part, fiance pointed it out to me more than once and I ignored her, well I fried that power supply because I didn't listen. More recently, about two months ago I woke up and checked my rig, I had left it folding overnight but it was off. Power button did nothing. I/O shield RGB lighting and RAM lighting was on, but it didn't boot or even turn on and spin fans. We thought maybe a cable came loose and was causing a short and needed reseating. So we went through every cable connection point in the system while trying to power it on, but nothing worked. Then someone in discord recommended hooking up another power supply, just the 24-pin and CPU EPS 8-pins and see if it would boot. We did that and it booted right up, so we figured out it had a bad PSU. No cables were damaged or melted. Learn all the common diagnostic tricks, like booting with 1 stick of memory, etc. Deductive reasoning goes very far here.

I'm going to have to be the detractor here, and say I just bought a 9950x, mainly for all the reasons I stated earlier in the thread in my last post. (Being unhappy with my 7900X3D that hasn't worked properly since last April) P.S THANK YOU @bridgypoo

I have a 7900X3D currently that has some glitch or defect that makes it not boost on the 2nd CCD for games that need frequency, or for enthusiast benchmarking on hwbot.org I *might* still get a 9950x. Reasons why. - 12 core to 16 core. - It would boost up to 5.7GHz for games that like frequency over L3 cache. The only game I really play is Forza Horizon 4 and 5 and I think they benefit from clock speed - It would boost to 5.7GHz for 3dmark which doesn't use 3D V-Cache or benefit from that at all and afaik in my experience none of the 3dmark benches ever made benefit from V-Cache. - I would also be able to run all the hwbot processor benchmarks on it. - If it runs cooler and is more energy efficient I would probably see higher boosts in the applications I mentioned with lower temps. - it might have a better IMC and allow me to OC my ram to 8000+. My current config is only stable with ram at 7600MHz, I can run 8000MHz for benching though. If my 2nd CCD didn't have this boost glitch where in Hwinfo64 on the desktop I see it boosting to 5650MHz but in any benchmark or under any single-thread load the 2nd CCD is stuck around 5150MHz, I wouldn't consider the 9950x.

Hmm, this might cause me to change my mind on getting a 9950x.

In bigred.