[Solarisfps]

Which CPU did you come from ?

If it's showing an increased frequency in the bios for CPU and ram when you adjust bclk or eclk, I'd say the next step would be to try a fresh windows install, sounds like it might be borked.

Its showing in the bios, it just zentimgs never updates the mt/s to the correct value ignore the timings at this moment dram frequency is incorrect in zen timings

 

[NobleNecromance]

Got my ram about as tuned as I think they can go, the only thing holding it back now is that it reaches mid 50s celsius when using OCCT's CPU+RAM+GPU test. I would just point a fan at them, but my air cooler pratically sits right on top of them(probably why they get so hot as well.). Anyone got some suggestions for a decent AiO and maybe ram cooler?

 

[zbug]

Yeah I hear you mate, in the same situation.

I'm actually finding starting from absolute default, just focusing on CPU for now, I'm not bothering with eCLK, just tuning negative CO, getting good results.

I'm finding most tweak settings either do nothing for my performance, and / or introduce instability.

So we are coming back full circle it seems ! But i'm glad someone else is seeing similar behaviors, on a game focus machine/tests with eclk vs pbo -CO.

I was about to try the same settings you posted 2/3 pages ago (with eclk+all core pbo) but I was doubtful it would really show improvements versus just negative co seeing it is a low enough eclk.
eclk might start to get ahead at 105 or more, but up to 104 (with curve shaper, i have not tested eclk and negative all core co except a crazy -30 that was obviously crashing), at least in game/3dmark benchmarks, it has been resulting in lower scores than simple pbo for me as well. But it is nice to have some testing on that front!

 

[Nico67]

Does zentimings update on a bclk overclock because my ram frequency doesn't change in windows

Nah, Zen Timings is not showing bclk adjustment on a numbers of boards with 9800x3d. You can see it in HWInfo and CPU-Z, but I recommend using 2.10 as 2.12 does seem to cause BSOD with bclk set.

 

[Solarisfps]

Oh no i'm reinforcing your point. To many people reaching for the moon imo.

And i'm not afraid to run any stability test that someone says is difficult.

I ignored the SHA tests on my 7800x3d it was absolutely ridiculous thing to pass. FPU Julia was 100% stable on my -co test i just went with that.

Is that a static overclock you have there 5460

 

[Solarisfps]

Nah, Zen Timings is not showing bclk adjustment on a numbers of boards with 9800x3d. You can see it in HWInfo and CPU-Z, but I recommend using 2.10 as 2.12 does seem to cause BSOD with bclk set.

Thanks,I just use the CPU-Z from bench mate which is 2.1, thanks for your insight. Ill ignore getting zen timings to show correct figures it is annoying why it is like that

 

[Xdrqgol]

thanks mate - using eCLK, set cpu core to offset, negative symbol (positive is missing entirely), 0.05, it boosts voltage instead of subtracting.

My gene only display negative offset 😩 - positive is missing as well - bug most likely

 

[Arizor]

Yeah @zbug little to no difference, might as well stick with your curve optimiser for now, seems to get me the same or better results at lower voltages and temps.

 

[Arizor]

My gene only display negative offset 😩 - positive is missing as well - bug most likely

Thanks for confirmation. It's really weird, made weirder that it actually BUMPS UP voltage rather than reduces it too with the negative sign. Nuts.

 

[Xdrqgol]

Thanks for confirmation. It's really weird, made weirder that it actually BUMPS UP voltage rather than reduces it too with the negative sign. Nuts.

No worries! Yes, i am trying to keep it as simple as possible and see where it lands and stick to it for now.

If Apex is a thing - cannot wait to see it released - but if the bios continues to be buggy well …

 

[GRABibus]

I ignored the SHA tests on my 7800x3d

That is what we all should do in fact.

 

[zbug]

Yeah @zbug little to no difference, might as well stick with your curve optimiser for now, seems to get me the same or better results at lower voltages and temps.

Lower voltages and temp that is for sure, when doing 104 eclk at various negative and positive values of curve shaper, i would still see over 10°C+ higher temp compared to pbo, not to mention very high voltage spikes.

That being said, in some tests, there would be "a bit" of differences, but essentially non-gaming. 3D mark cpu and cpu-z would have pretty massive single core results improvements.
Sure, in moderate gaming loads, the boost clock would be at 5.6ghz but these ~200mhz did not seem to matter at all and would fluctuate a lot more (probably due to temperature) compared to staying constant at 5.4 under pbo and pretty much never move for hours.

Still worth playing with but ill resume eclk adventures when i'm done with Indiana Jones or if someone make a bigger breakthrough

 

[Antsu]

Why tempt fate with stability over less than 1% performance gains?

1%, huh? I tried to keep my nose out of this topic but this broke my back. xD

Let me start by saying that I 100% agree on the fact that you should never claim to be stable if you can only play games, but is this a thing really? I feel like all the guys running borderline settings specify that it's "gaming stable" = passing shader compilation in games. This means you are pretty solid while in the game if you are using fixed clocks without limits. Maybe I've missed something, but in general at least I don't see this as an issue, and I'm sure anyone running such settings will gladly admit they aren't 100% stable if challenged on the matter.

Why? Because if the pixels on my screen switch to blue color in the middle of my manchild activities, my life won't end. And who can say they have so much CPU overhead that they can easily bottleneck a 5090? I can't at least, so I'll gladly take all I can get. You can have a separate profile for serious stuff, but I don't even feel the need for that personally because I don't crash while in the game.

1% is the difference between 5700 and 5650, are we sure we are going to lock in that you can run ALL the test suites at -50Mhz from the maximum clocks you can boot up a game? I'd say it's closer to 500 than 50.

 

[AdamK47]

1%, huh? I tried to keep my nose out of this topic but this broke my back. xD

Let me start by saying that I 100% agree on the fact that you should never claim to be stable if you can only play games, but is this a thing really? I feel like all the guys running borderline settings specify that it's "gaming stable" = passing shader compilation in games. This means you are pretty solid while in the game if you are using fixed clocks without limits. Maybe I've missed something, but in general at least I don't see this as an issue, and I'm sure anyone running such settings will gladly admit they aren't 100% stable if challenged on the matter.

Why? Because if the pixels on my screen switch to blue color in the middle of my manchild activities, my life won't end. And who can say they have so much CPU overhead that they can easily bottleneck a 5090? I can't at least, so I'll gladly take all I can get. You can have a separate profile for serious stuff, but I don't even feel the need for that personally because I don't crash while in the game.

1% is the difference between 5700 and 5650, are we sure we are going to lock in that you can run ALL the test suites at -50Mhz from the maximum clocks you can boot up a game? I'd say it's closer to 500 than 50.

What ragged edge are you running where the difference in stability is a massive 500MHz?

People be runnin those clapped out ragged edge dirty clocks round here.

 

[bmagnien]

Just tried 1.27 vcore 5.5ghz static cpu craps out on r23.

You need to stay below like 65c to manage this - gonna be hard at that voltage

 

[Goose3080]

I ignored the SHA tests on my 7800x3d it was absolutely ridiculous thing to pass. FPU Julia was 100% stable on my -co test i just went with that.

Is that a static overclock you have there 5460

I wondering how important SHA3 really is, especially as I dont see people like skatterbench testing with it, the only thing I can find regarding it is something to do with password hashing, at the moment, one core on my 3D CCD on my 7950X3D is at -3 because anything else fails sha3, but I can pass Y-Cruncher and Prime all day long with that core set to -24, all the other cores are at -20 to -30

 

[Solarisfps]

I wondering how important SHA3 really is, especially as I dont see people like skatterbench testing with it, the only thing I can find regarding it is something to do with password hashing, at the moment, one core on my 3D CCD on my 7950X3D is at -3 because anything else fails sha3, but I can pass Y-Cruncher and Prime all day long with that core set to -24, all the other cores are at -20 to -30

I have yet to find a game or anything else that i could possibly do on my pc that would use it.

 

[Solarisfps]

You need to stay below like 65c to manage this - gonna be hard at that voltage

I have it on direct die so i have best case scenario . I will just play about with BCLK and ECLK to find my max stable.

 

[Goose3080]

I have yet to find a game or anything else that i could possibly do on my pc that would use it.

I agree, but I got a funny feeling Windows uses it for security and keep phishers out or something like that.

 

[Wolverine2349]

I am not sure if this is the right place to post this but I thought it is interesting and maybe related none the less at least realized in that I was lead to this test to test FCLK scaling and stability based on Googling for Ryzen 7000 and 9000 FCLK stability (due to its error correction which makes it hard to ensure stable OC with no performance regression due to error correcting). Have a 2nd 9800X3D seemingly stable at 6400 RAM 1:1 at 2133 FCLK best on passing lots of memory and other tests with no errors.

I read that it was a good idea based on Googling to run OCCT VRAM stress test while also running Prime95 Large FFTs at the same time to ensure FCLK and RAM is stable if overclocked and tuned timings as it goes through PCIe bus which stressed FCLK.

So I do no then I notice this in Event Viewer an error when it finished:

The description for Event ID 153 from source nvlddmkm cannot be found. Either the component that raises this event is not installed on your local computer or the installation is corrupted. You can install or repair the component on the local computer.

So I wonder why am I getting that error does that mean my 6400 RAM 1:1 and FCLK at 2133 is not totally stable? So I decide next to not run Prime95 and just start OCCT VRAM test. Strange thing about OCCT VRAM test it puts little load on actual GPU and just fills the GPU RAM which is an RTX 4090.

So I run the OCCT VRAM test for a minute or 2 then stop it check Event Viewer and low and behold same error there and got it to do it almost every time running that test indicating also running P95 Large FFTs along side had nothing to do with it. So ok maybe still an unstable FCLK and/or RAM OC because the settings are still applied despite no P95 also running.

So I turn off all XMP/EXPO/DOCP settings and FCLK reset back to motherboard default 2000 and RAM SPD default 4800 with all PBO/OC settings auto. Then boot into Windows and run just the OCCT VRAM test and boom same exact nvlddmkm error right after stopping the test or when test ends and stops itself after 1 hour run., So nope good thing not an unstable FCLK and RAM overclock.

Does this mean GeForce RTX 4090 video card going bad or just a benign or weird driver issue with OCCT VRAM test? Never had any issues with this RTX 4090 gaming so hopefully just an app issue and OCCT VRAM produces no errors itself.

Cause it had nothing to do with any FCLK or RAM overclock/XMP/EXPO.