[PJVol]

Is there a way to read this in software?

Yeah, in Bios it may be under Core Performace Boost, but if not, if you have a way to read MSR C0010015, 25-bit is CpbDis. 1 means CPB disabled.

 

[BHS1975]

I can under-volt the vcore with about 0.065v before performance drops

 

[Luggage]

I can under-volt the vcore with about 0.065v before performance drops

YC, p95 small?

 

[BHS1975]

I can run 1866 fclk at 1v soc but no post at 1900 and whea city beyond that. My 5600x wouldn't go below 1.06 SOC without whea errors.

 

[BHS1975]

YC, p95 small?

R23 OCCT stable as well.

 

[tcclaviger]

It's very temp sensitive, I've found. Counterintuitively the colder it stays the less neg offset works, because it boosts higher for longer.

Once it's holding at max multi steady, any neg offset becomes a hindrance.

For example at 44.5 mutli I score 15200 R23. Add any neg offset and it drops to 14800. At 4550 all core auto voltage is too little and it needs +0.0625 to score 15600. +0.0125 scores 15720.

 

[MrHoof]

It's very temp sensitive, I've found. Counterintuitively the colder it stays the less neg offset works, because it boosts higher for longer.

Once it's holding at max multi steady, any neg offset becomes a hindrance.

For example at 44.5 mutli I score 15200 R23. Add any neg offset and it drops to 14800. At 4550 all core auto voltage is too little and it needs +0.0625 to score 15600. +0.0125 scores 15720.

That was the case already for every Ryzen i had 3600 and 5800x. They always have temp thresholds they try to hold and if they cant they clock down to keep the temps steady.
So if you can cool higher voltages there is no point for a negative offet unless it can do the desired frequency at lower voltages too.
edit: My cooling solution isnt the best and at stock it seems to start clocking down for me at 65°C 4.45GHZ and further at 80°C 4.4GHZ and then further down to 4.375GHZ to keep it 80°C

 

[BHS1975]

It's very temp sensitive, I've found. Counterintuitively the colder it stays the less neg offset works, because it boosts higher for longer.

Once it's holding at max multi steady, any neg offset becomes a hindrance.

For example at 44.5 mutli I score 15200 R23. Add any neg offset and it drops to 14800. At 4550 all core auto voltage is too little and it needs +0.0625 to score 15600. +0.0125 scores 15720.

Do you mean 0.125?

 

[bloot]

I can under-volt the vcore with about 0.065v before performance drops

How do you undervolt? The only way I found is using @PJVol's PBO2 Tuner in Windows but i'd like to make it permanet and not having to set it everytime system reboots. Also setting LLC to the lowest value helps reducing some voltage on load at least.

Please AMD, give us vcore control so we can undervolt it on bios, just block it up to 1.35V if you don't want us to fry the cache, it shouldn't be so difficult, I just want a lower vcore than stock.

 

[Luggage]

MSI to enable Ryzen 7 5800X3D BCLK overclocking for X570 MEG motherboards, up to 7% higher single-core performance - VideoCardz.com

AMD Ryzen 7 5800X3D to gain BCLK overclocking feature on MEG X570 series MSI is set to release a new AGESA firmware update for Ryzen 7 5800X3D processor that will enable BCLK overclocking. Wccftech reports that MSI will be rolling out an update to AGESA 1.2.0.7 in the coming days. This firmware...

videocardz.com

Hmm - the wants just increased a smidge, now will this fsck up my two nmve drives?

 

[yzonker]

Just got my 5800x3D. No issue dropping it in and just resetting everything back to what I had with the 5800x. 3800 fclk was no issue or my tuned setting for the B-die. I'll have to try going higher. My 5800x wasn't happy above 3800.

I'll have to do some research. I tried setting BCLK to 101 and got a boot loop immediately. But I haven't dug in to it as far as what might make it work or not. (Gigabyte X570 Aorus Ultra).

One question though, which parts of the PBO2 tuner tool on the first page of this thread is supposed to work?

Setting CO values seemed to work to a point like I think others have reported. In Cinebench it would lock at 4450mhz on all cores. Score was right at 15000 just running it in my normal OS. Looks like it's hitting 4550mhz running CoreCycler.

It doesn't appear the CO setting survives a reboot though.

 

[BHS1975]

How do you undervolt? The only way I found is using @PJVol's PBO2 Tuner in Windows but i'd like to make it permanet and not having to set it everytime system reboots. Also setting LLC to the lowest value helps reducing some voltage on load at least.

Please AMD, give us vcore control so we can undervolt it on bios, just block it up to 1.35V if you don't want us to fry the cache, it shouldn't be so difficult, I just want a lower vcore than stock.

I used the vcore offset in the bios.

 

[tcclaviger]

Do you mean 0.125?

No 0.0125, 12.5 mv not 125mv.

How do you undervolt? The only way I found is using @PJVol's PBO2 Tuner in Windows but i'd like to make it permanet and not having to set it everytime system reboots. Also setting LLC to the lowest value helps reducing some voltage on load at least.



Please AMD, give us vcore control so we can undervolt it on bios, just block it up to 1.35V if you don't want us to fry the cache, it shouldn't be so difficult, I just want a lower vcore than stock.

Vcore offset in bios works. It's not AMD blocking it it's the vbios.

MSI to enable Ryzen 7 5800X3D BCLK overclocking for X570 MEG motherboards, up to 7% higher single-core performance - VideoCardz.com

AMD Ryzen 7 5800X3D to gain BCLK overclocking feature on MEG X570 series MSI is set to release a new AGESA firmware update for Ryzen 7 5800X3D processor that will enable BCLK overclocking. Wccftech reports that MSI will be rolling out an update to AGESA 1.2.0.7 in the coming days. This firmware...

videocardz.com

Hmm - the wants just increased a smidge, now will this fsck up my two nmve drives?

Maybe. I've not had m.2 corruption due to bclk, ever. Bad memory config yes, bclk no. It is possible though.

That's across WD SN550, Samsung 980 Pro 1tb, and CS3030 1tb and 2tb.

SN 550 tolerated 110 bclk fine too.

 

[bloot]

I used the vcore offset in the bios.

Which agesa version is your bios? I have no vcore offset mode in 1.2.0.6 bioses, and on 1.2.0.5 it's there but the cpu won't boost past 3.4GHz 😞

 

[bloot]

Vcore offset in bios works. It's not AMD blocking it it's the vbios.

So it's MSI who's to blame that I have no vcore offset setting on agesa 1.2.0.6b or c? I thought nobody had it.

 

[ilmazzo]

Yuk! Asrock has a 1.2.0.6 agesa in the latest beta bios for x470 taichi (L4.88)

“Update AMD AM4 AGESA Combo V2 PI 1.2.0.6b”

 

[yzonker]

I did some basic 3Dmark runs just for consistency. No matter what, it scores lower in PR now,

Result

www.3dmark.com

Anybody else see this? Some issue maybe with my mobo (bios)? I flashed it to the latest. F36b ("Update AMD AGESA V2 1.2.0.6 B for Ryzen 7 5800X3D support")

Guess I'll have to put the 5800x back in to verify.

 

[tcclaviger]

Edit: PR, duh me. Not TS lol.

PR is so gpu bound the cpu is only minimally impactful tbh.

Would guess a 200 point swing could just be a temp difference if the GPU was on the same clock profile, or multiple tests were run back to back on a GA102. The score degrades slightly when run over and over, GN, J2C and Kingpin all mentioned observing it happen, and so have I.

So it's MSI who's to blame that I have no vcore offset setting on agesa 1.2.0.6b or c? I thought nobody had it.

Correct. AMD didn't disable it.

 

[yzonker]

Edit: PR, duh me. Not TS lol.

PR is so gpu bound the cpu is only minimally impactful tbh.

Would guess a 200 point swing could just be a temp difference if the GPU was on the same clock profile, or multiple tests were run back to back on a GA102. The score degrades slightly when run over and over, GN, J2C and Kingpin all mentioned observing it happen, and so have I.

So I've gone full circle now and it is something else with the system. Can't get back to that first run with the 5800x either. Grrr...

Result

www.3dmark.com

I wasn't expecting a gain in PR, just not a loss. But it appears to be something else.

It did gain in Firestrike that runs at a much higher framerate. And this is against the "Magic" runs I made originally.

Result

www.3dmark.com

I also tested CP2077 and Tomb Raider. Both showed significant gains, so I would say it's working as expected.

 

[tcclaviger]

Yeah I've fought that PR score regression in the past too. Haven't been able to isolate it in my case beyond, sometimes it just runs poorly sometimes not, wish I could help.