AGESA 1.2.0.8

[Sptz]

Gigabyte boards seem to overvolt and overclock by default. Just about every Gigabyte board I've seen will set the FSB or BKCLK to a few MHz over spec and the same with voltages, but the DRAM voltages on thier B550 boards are concerning. If the sensors are too close to the VRMs, Gigabyte engineers could be compensating for VDROOP. Nonetheless the "Tweaker" section of the BIOS is useless but the AMD Overclocking section in Miscellaneous has a setting for VDDG that can be tweaked as a temporary workaround to bring VDIMM down to 1.35v ...

VDIMM can be set in the "main" / "tweaker" page, the only thing is that it will always be overvolted anyway.
On my board, loading XMP (3600c16) vsoc on auto = 1.18v which is pretty insane imo.

I'm testing ram oc, using 1.5v vdimm and it's actually reporting 1.55v lol pretty cool ey.

So, is this EDC bug only relevant for non 3D chips? I noticed that my cpu has been running substantially hotter, I'm on the latest bios for Gigabyte b550i, i think I was 2 bios down when it was fine, -30 co, cinebench never crossed 70C, now it jumps to 78-79C almost instantly, maybe it's not related anyway.

 

[Mussels]

Gigabyte boards seem to overvolt and overclock by default

Confirmed on three boards now, two x570 and one B550.
Even my x370 seems to be overvolting, so it's not a generation thing, it's their choice in the current firmwares
All three run XMP for 1.35v RAM at 1.41v, two run 1.09v SoC and one is running 1.15v (The bug that makes auto use previously set settings?)

On my board, loading XMP (3600c16) vsoc on auto = 1.18v which is pretty insane imo.

That shouldnt happen, i have the same board too (VR system). Did you check if it's doing the "previously set manual setting doesnt go to auto" bug, or is it genuinely changing with XMP on/off?

 

[lawood]

The auto for my vsoc is also 1.18v if I go over 3600 frequency.
On true auto below 3600 is only 0.900v.
Gigabyte board.

Oh, and with 2000 fclk, the vddp goes straight to 1v on auto.

 

[Mussels]

The auto for my vsoc is also 1.18v if I go over 3600 frequency.
On true auto below 3600 is only 0.900v.
Gigabyte board.

Oh, and with 2000 fclk, the vddp goes straight to 1v on auto.

Interesting, i'm at 3600 and got 1.09v - the same as XMP disabled
It's very weird how the defaults change so much between boards and bios revisions

 

[tabascosauz]

Interesting, i'm at 3600 and got 1.09v - the same as XMP disabled
It's very weird how the defaults change so much between boards and bios revisions

This is not new. The v1.0 B550I Aorus always overvolts VDIMM by up to 0.06V. Even for Gigabyte, which has always had a reputation of overvolting about ~0.03V on pretty much every AM4s, it's probably the most egregious offender.

It's more or less accurate, the board is actually applying somewhere between 0.03-0.06V extra. You can always just dial down VDIMM manually to compensate.

Auto VSOC rules vary by every board, not really overall vendor rules.

 

[lawood]

Interesting, i'm at 3600 and got 1.09v - the same as XMP disabled
It's very weird how the defaults change so much between boards and bios revisions

I was surprised by the very low vsoc below 3600, 0.900v is very low. Also 0.900 on both vddg's, but thats more normal.
But it works.

But when I dial in 3800 the vsoc jumps to 1.18v, I experimented with some values and found that 1.12v works the best for 1900 fclk. 1.18 seemed very excessive.
The vddg's also go to 1v on auto, didn't touch that, seems perfectly fine.

The vddp is 900mv across the board on all frequencies but if I go 2000 flck it jumps straight to 1v, the other voltages values dont change at all. Sadly, that doesnt help at getting rid of whea but its an interesting behavior.

In gigabyte eyes high fclk (+2000) requires high vddp, they dont touch vddg's at all.

 

[Mussels]

I was surprised by the very low vsoc below 3600, 0.900v is very low. Also 0.900 on both vddg's, but thats more normal.
But it works.

But when I dial in 3800 the vsoc jumps to 1.18v, I experimented with some values and found that 1.12v works the best for 1900 fclk. 1.18 seemed very excessive.
The vddg's also go to 1v on auto, didn't touch that, seems perfectly fine.

The vddp is 900mv across the board on all frequencies but if I go 2000 flck it jumps straight to 1v, the other voltages values dont change at all. Sadly, that doesnt help at getting rid of whea but its an interesting behavior.

In gigabyte eyes high fclk (+2000) requires high vddp, they dont touch vddg's at all.

I'm surprised because mines the same with XMP on or off, it doesnt seem to change (2400 to 3600) - it remains the same at 1.09v


The contradictory behaviour is what has me alarmed, as it should be consistent

 

[OneMoar]

could be worse could by my board and default to 1.2v

 

[Mussels]

After the messes on the Intel side (totally ignoring PL1/PL2 settings and defaulting to hundreds of watts higher than TDP) to the points boards overheated with 65W CPUs, and AM5's exploding CPUs and with hindsight AM4's erratic voltages...

We really, really need a lot more attention paid to motherboard voltage accuracy, and default settings. AMD and Intel need to crack down on partners who go too far out of spec.
voltages need to be within reasonable limits, and they have to be honest - no one cares about a 1-2% difference, but when boards are just adding 50mv or more to what it claims to be running, it's an outright danger.

Since intel and AMD are the ones who suffer the reputation loss and sales loss from hardware death, they need to smash these board makers into line and stop them selling trash.

 

[JoeN]

Hi Just in case this is useful to anyone, Gigabyte have released a version f37d BIOS, at least for the X570 Aorus Pro v1. It should let you change the Dynamic Vcore (DVID) with the 5800x3d installed.

I cannot test this now as I am now using the R9 5900X and very happy with it.

 

[izy]

MSI released a new bios for B450
7C02v3H for B450 tomahawk max

- Modify language related issue.
- Add ASPM Control item.

Ill try it out in the next days but i dont think it has any improvements for RAM / CPU OC

 

[Mussels]

Hi Just in case this is useful to anyone, Gigabyte have released a version f37d BIOS, at least for the X570 Aorus Pro v1. It should let you change the Dynamic Vcore (DVID) with the 5800x3d installed.

I cannot test this now as I am now using the R9 5900X and very happy with it.

Oh goody, if it lets you use an offset voltage that's what i've been using on my 5800x3D on asus and it's been a game changer

-30 PBO and -.06125v lets me flatline 4.55GHz at 100-113W (depending which sensor reading you use)

the same would apply to any 7000 x3D's, if the board allows it. It's why i USED to recommend asus, before the current shenanigans came out.

 

[Sptz]

Oh goody, if it lets you use an offset voltage that's what i've been using on my 5800x3D on asus and it's been a game changer

-30 PBO and -.06125v lets me flatline 4.55GHz at 100-113W (depending which sensor reading you use)
View attachment 305518


the same would apply to any 7000 x3D's, if the board allows it. It's why i USED to recommend asus, before the current shenanigans came out.

Yeah, damn. Wish I could do that, can't really find anything for offset on my Gigabyte B550i Aorus. :[

 

[tabascosauz]

Oh goody, if it lets you use an offset voltage that's what i've been using on my 5800x3D on asus and it's been a game changer

-30 PBO and -.06125v lets me flatline 4.55GHz at 100-113W (depending which sensor reading you use)
View attachment 305518


the same would apply to any 7000 x3D's, if the board allows it. It's why i USED to recommend asus, before the current shenanigans came out.

Interesting performance. I see your effective clocks are also 4450, but 14200 is a score more typical of bone stock 5800X3D. I can't even stick to 4450 for the duration of a loop, but am always at 15000, at below 100W stock limits. Not even at -30 I think, like -25 or -28. Strange how these samples vary so wildly.

Tried Vcore offset once, didn't seem to work so great on my sample in terms of performance.

 

[JoeN]

Yeah, damn. Wish I could do that, can't really find anything for offset on my Gigabyte B550i Aorus. :[

Might be worth raising a ticket with Gigabyte support, it did take around 2 months but they notified my within a few days of the BIOS release.

The F17 BIOS release includes U[dates for AGESA V2 1.2.0.8 and optimisations for the 5800x3d so the request might bump the release of the next update?

 

[Sptz]

Might be worth raising a ticket with Gigabyte support, it did take around 2 months but they notified my within a few days of the BIOS release.

The F17 BIOS release includes U[dates for AGESA V2 1.2.0.8 and optimisations for the 5800x3d so the request might bump the release of the next update?

I noticed the F17 bios is an agglomeration of the previous betas. The release notes are exactly the same per beta. I did flash it anyways even though the latest beta supposedly is exactly the same as F17 but they removed them from the website.

What did you raise them a ticket for? And where would one do that?

 

[Cowboystrekk]

Oh goody, if it lets you use an offset voltage that's what i've been using on my 5800x3D on asus and it's been a game changer

-30 PBO and -.06125v lets me flatline 4.55GHz at 100-113W (depending which sensor reading you use)
View attachment 305518


the same would apply to any 7000 x3D's, if the board allows it. It's why i USED to recommend asus, before the current shenanigans came out.

Hmm, a bit low score and high consumption? I get 15150 using -30CO using 93-96W, EDC 100, TDC 80, temp 72-75C with D15 cooler.

 

[Ferrum Master]

Hmm, a bit low score and high consumption? I get 15150 using -30CO using 93-96W, EDC 100, TDC 80, temp 72-75C with D15 cooler.

Polly also ram. With GDM it means ~CL20 at 3866?

 

[JoeN]

I noticed the F17 bios is an agglomeration of the previous betas. The release notes are exactly the same per beta. I did flash it anyways even though the latest beta supposedly is exactly the same as F17 but they removed them from the website.

What did you raise them a ticket for? And where would one do that?

I noticed that higher temps on my 5800x3d following a BIOS update, F37b on my x570 Aorus Pro, I managed to underclock using easytune but could not find the equivalent settings in the BIOS. I'm not entirely sure of the timeline and am not sure of the full story but it looks like the AGESA V2 1.2.0.8 changes also included the removal of some overclock settings for the 5800x3d specifically, due to the voltage limitations of additional cache. Also the implementation of this AGESA update seems to have a few issues that actually result in unnecessary high temps in some CPUs, this might be addressed in later versions of AGESA, 1.2.0.B I think, but we are still waiting for these.

I sent Gigabyte a screenshot of my useable easytune settings, and a few messages later they sent me an update this morning. If you look up the Gigabyte page for your board, click support, then online support you can login and submit a ticket.

 

[Cowboystrekk]

Polly also ram. With GDM it means ~CL20 at 3866?

Ram means very little in CB23. I run mine at 3800cl14 with tight timings, but I got same scire with 3200 xmp.

 

[mclaren85]

It doesn't do much aside from patch some security vulnerabilities and unhide PBO controls for 5800X3D on some boards. Performance and boost behaviour is largely the same as 1207.

Does ageasa really affect system performance? Boot time etc?

 

[tabascosauz]

Ram means very little in CB23. I run mine at 3800cl14 with tight timings, but I got same scire with 3200 xmp.

Maybe background programs, I know some can definitely add up to 1k R23 loss

Otherwise cooling is good, 4450 no clock stretching, should be 15k minimum

I have to wait till Friday but I'll take the 120A BIOS for a spin then

Currently 15000-15200, don't even need to touch limits, it limits itself to 100W at about 75C

Does ageasa really affect system performance? Boot time etc?

AM4 is way past the point of major performance boost or reduction in boot time, but you can just look at AM5 now or AM4 2 years ago to answer your own question

 

[Sptz]

Hmm, a bit low score and high consumption? I get 15150 using -30CO using 93-96W, EDC 100, TDC 80, temp 72-75C with D15 cooler.

Hmm, I'm all core -30CO, Limits: PPT 122 TDC 82 EDC 124
RAM @ 3733C16 (tightened, 145ns RFC)

Depends on ambient temp of course, but with ambient at around 20C I always get above 15k, around 15100-15200 but it does consume more than that, around 110W or so. Don't really get why. No clock stretching, all effective clocks pegged at 4449.9 MHz

I used those power limits as that's what I found from this on reddit:

Could I possibly limit it a bit more without affecting real world performance?

I noticed that higher temps on my 5800x3d following a BIOS update, F37b on my x570 Aorus Pro, I managed to underclock using easytune but could not find the equivalent settings in the BIOS. I'm not entirely sure of the timeline and am not sure of the full story but it looks like the AGESA V2 1.2.0.8 changes also included the removal of some overclock settings for the 5800x3d specifically, due to the voltage limitations of additional cache. Also the implementation of this AGESA update seems to have a few issues that actually result in unnecessary high temps in some CPUs, this might be addressed in later versions of AGESA, 1.2.0.B I think, but we are still waiting for these.

I sent Gigabyte a screenshot of my useable easytune settings, and a few messages later they sent me an update this morning. If you look up the Gigabyte page for your board, click support, then online support you can login and submit a ticket.

Hmm, I did notice that moving from the early F16's to F17 did raise my CPU temps overall. But it could also be coincidental with summer time so it's hard to be precise in placing any sort of blame.

I'd just love to be able to undervolt a tiny bit. Already run my 4090 at max OC on an undervolt, higher performance much less power consumption and temps, with very minimal clock stretching.

 

[Cowboystrekk]

Hi Just in case this is useful to anyone, Gigabyte have released a version f37d BIOS, at least for the X570 Aorus Pro v1. It should let you change the Dynamic Vcore (DVID) with the 5800x3d installed.

I cannot test this now as I am now using the R9 5900X and very happy with it.

I tried lowering DVID. It also lowers SOC for some reason. Down to -48mv seemes to give no performancedrop, -60mv seems to give slight drop.

In CB23 with -30CO:
-No offset 15150p 96W, 74C
-60mv 14950p 85W, 68C
-48mv 15050p 87W, 70C

Quite impressive

 

[mechtech]

Do you use bios values for reported voltages or software in windows or both?