Monday, May 8th 2023

Latest AMD AGESA that Nerfs Ryzen 7000X3D Voltage Control Also Limits Memory Overclocking

by
btarunr
 Discuss (73 Comments)
The latest AMD AGESA 1.0.0.7 AM5 platform microcode that the company recently released to improve stability of machines powered by Ryzen 7000X3D processors, more importantly, prevent them from physical damage due to increased voltage in voltage-assisted overclocking scenarios; reportedly impacts memory overclocking capabilities, too, reports g01d3nm4ng0. The "PROCHOT Control" and "PROCHOT Deassertation Ramp" toggles that were available in the oldest versions of AGESA for AM5, are not available in the latest production AGESA.

The memory compatibility is also affected. AMD recently added support for odd-density DDR5 memory modules, such as 24 GB and 48 GB, which make up 48 GB and 96 GB 2-module (dual-channel, four sub-channel) kits. It is possible to max out 192 GB, but while the older AGESA 1.0.0.6 allowed memory frequencies of up to DDR5-6000 with SoC voltage of 1.3 V, the newer AGESA is only stable up to DDR5-4400 at this density. To be fair, most motherboards advertise maximum memory frequencies of under DDR5-4800 for memory configurations where there are two DIMMs per channel, and both DIMMs are dual-rank (so four dual-rank DIMMs in all, which is the least optimal memory configuration from a memory frequency and latency perspective).
Source: g01d3nm4ng0 (Twitter)

Related News

Add your own comment

73 Comments on Latest AMD AGESA that Nerfs Ryzen 7000X3D Voltage Control Also Limits Memory Overclocking

#51
AusWolf
Minus InfinityZen 4 is the gift that just keeps on giving to Intel, So glad I didnt rush out and update my 3700X for 7800X(3D). Definitely hoping they get their crap together for Zen 5, but openSIL is years away so no guarantee AGESA won't continue to be a crap shoot for the next release. Arrow Lake is looking more and more appealing if those clowns can release on time.
I'm not saying that these issues aren't real, but the news are a bit overblown. You'll find many people here on TPU who have had a problem-free experience with Zen 5.

To be fair, I had more trouble with Zen 4 and B550 in its early days with the USB dropout issues and such.
Posted on Reply
#53
Minus Infinity
AusWolfI'm not saying that these issues aren't real, but the news are a bit overblown. You'll find many people here on TPU who have had a problem-free experience with Zen 5.

To be fair, I had more trouble with Zen 4 and B550 in its early days with the USB dropout issues and such.
I think you mean Zen 4 and Zen 3/B550.

Zen 4 like RDNA3 seem like beta products. I thought Zen was bad enough with all the BS memory issues. I'm sticking to Zen 2 and Zen 3 in my current PC's for a long while. Zen 5 will need to be a near flawless update for me to stick to them. I'm also having plenty of issues with 6800GT that are annoying all all hell.
Posted on Reply
#54
Zubasa
R-T-BMy man you are literally posting in the TPU thread on the exact same news piece.
Yup and as usual it is the internet playing telephone all qouting the same twitter / reddit post.
Often the news sites cites each other as the source. At this point I feel like ChatGPT ain't so bad after all. :roll:
Posted on Reply
#55
AusWolf
Minus InfinityI think you mean Zen 4 and Zen 3/B550.
Yes. I'm at work and a bit tired, sorry. :D
Minus InfinityZen 4 like RDNA3 seem like beta products. I thought Zen was bad enough with all the BS memory issues. I'm sticking to Zen 2 and Zen 3 in my current PC's for a long while. Zen 5 will need to be a near flawless update for me to stick to them. I'm also having plenty of issues with 6800GT that are annoying all all hell.
I think we can count on everything being a beta product these days. Things are getting more and more complex, and development cycles tighter and tighter. Something's got to give.

6800 GT?
Posted on Reply
#57
laxman10100
I must've gotten very lucky recently switching over from Intel CPUs to the 7800X3D with the motherboard that I chose.

Even before this CPU burn-up fiasco, my CPU SoC never went over 1.25V with EXPO enabled rocking 64GB 6000 CL28. Granted, I manually entered most of my settings (memory timings included) based on what I read was safe as I had done a ton of research on which BIOS settings should be modified since I wasn't at all familiar with the AMD platform...

Regardless, I believe that the default CPU SoC was 1.3V, however, my usage of the performance settings of -30mV CPU SoC & PBO 75°C limit kept the CPU voltage rather low. After the BIOS update with AGESA 1.0.0.7, nothing has really changed for me aside from my latency going from 64ns to 65ns. Which does seem a bit odd considering I have always set my BIOS to ignore AGESA so that my manually entered memory timings wouldn't be messed with/loosened.

I wonder if my CPU was at all a little degraded from all of this anyway and I just haven't realized...
Posted on Reply
#58
AusWolf
laxman10100I must've gotten very lucky recently switching over from Intel CPUs to the 7800X3D with the motherboard that I chose.

Even before this CPU burn-up fiasco, my CPU SoC never went over 1.25V with EXPO enabled rocking 64GB 6000 CL28. Granted, I manually entered most of my settings (memory timings included) based on what I read was safe as I had done a ton of research on which BIOS settings should be modified since I wasn't at all familiar with the AMD platform...

Regardless, I believe that the default CPU SoC was 1.3V, however, my usage of the performance settings of -30mV CPU SoC & PBO 75°C limit kept the CPU voltage rather low. After the BIOS update with AGESA 1.0.0.7, nothing has really changed for me aside from my latency going from 64ns to 65ns. Which does seem a bit odd considering I have always set my BIOS to ignore AGESA so that my manually entered memory timings wouldn't be messed with/loosened.

I wonder if my CPU was at all a little degraded from all of this anyway and I just haven't realized...
Like I said earlier (or maybe in another thread), the issue is way overblown, as always. Most people never had any problem with their setups.

Welcome to the forum, by the way! :)
Posted on Reply
#59
beedoo
somebodys_kidMy 2x16 GB GSkill EXPO profile became unstable after my BIOS update (DDR5 6000 CL 30 at 1.35V). Which memory are you all using that keeps the speed and latency but is stable?
This weekend I built a new machine with ASRock Taichi X670E and 2 x G.Skill 16GB EXPO 6000 CL30 at 1.35v, latest BIOS and been totally stable.
Posted on Reply
#60
laxman10100
AusWolfLike I said earlier (or maybe in another thread), the issue is way overblown, as always. Most people never had any problem with their setups.

Welcome to the forum, by the way! :)
Ahh, yes. I believe that it was earlier on in this thread. I do recall reading that message. But also, I agree. From what I have read, it seems that the issue depended on which particular motherboard you had, as specific boards from specific manufacturers would most likely be affected by the issue. Also, thanks for the welcome =D.

Edit - I just scrolled up and it was only a few comments above mine. I suppose that I am occasionally too forgetful, haha.
beedooThis weekend I built a new machine with ASRock Taichi X670E and 2 x G.Skill 16GB EXPO 6000 CL30 at 1.35v, latest BIOS and been totally stable.
Yep! That board should be good - as I have the Carrara version of that board and it was stable before the latest few BIOS updates with my 7800X3D. But perhaps that was due to what I mentioned in my comment above with the undervolting setting that I had selected. *shrugs*. I have also been told by a few users on a different forum that their non-ASRock DDR5 motherboards did not have voltage options for both CPU SoC and Mem_SoC/VDDIO_Mem. So, I imagine that those particular motherboards applied the same voltage to the CPU and the Memory. Of course, I am not sure if that is true at all, but our ASRock boards have separate voltage settings for all three of those things.
Posted on Reply
#61
Slizzo
I haven't had any issues with my 7950X3D system. -30 all core offset, had memory at 6000C30 or 6400C32, either worked just fine for me. I'm using an Intel kit of memory purchased well before the rest of the parts, so haven't been using EXPO profiles.

Actually, haven't ever used the XMP profiles either. I've been using the Memory Try it! profiles that MSI baked into the board to run different speeds just to test etc...
Posted on Reply
#62
Braegnok
Asus dropped AGESA 1007a today, I loaded it this morning, it's working very well on my system.

Posted on Reply
#63
harm9963
So glad I went for a new GPU, not a new platform, my 5950X is servings me well .
Posted on Reply
#64
Taraquin
DavenAMD’s biggest weakness is the memory controller. It needs improvements across the board (pun intended).
Yeah, same shit as Ryzen gen 1 and 2 usually maxing out at 3333-3466. On gen 3 compability was finally very good and most could to 3733/3800.
Posted on Reply
#65
Zubasa
"Our 'Scumbag ASUS' video is up -- not relating to the Ally. We want to note also that ASUS emailed us last week after Part 1 of exploding CPUs - an unprompted email - and asked if they could fly out to our office this week to meet with us about the issues and speak "openly." We told them we'd be down for it but that we'd have to record the conversation. They did say they wanted to speak openly, after all. They haven't replied to us for 5 days. So... ASUS had a chance to correct this. We were holding the video to afford that opportunity. But as soon as we said "sure, but we're filming it because we want a record of what's promised," we get silence. Wanting to comment on something and provide a statement is not only fine, but encouraged; we're always happy to provide that opportunity. See: Newegg interview with the executives. However, we're not going to let it be done without accountability and in the shadows. They could have done this the right way."

Note that this is Auto SOC voltage not memory voltage.
Posted on Reply
#66
b.infinite
MarsM4NI mean that's just some basic common sense. :) If AMD is setting new maximum limits & RAM vendors selling RAM kits that run outside of spec, there will be incompatibility.
Just looked at the listenings over at heise.de. 170 kits listed with AMD EXPO, and their voltage specs:



If you narrow it down to kits that are specified for 6000MHz (AM5 sweet spot) there are left 65 kits. Their specified voltages are:



Which means you have only 5 kits that run within AMD's specs at 6000MHz (from TeamGroup, Apacer & Lexar). 60 kits are certified by RAM vendors to run outside of AMD's specs! Which btw. does not mean they won't run with lower voltages. It's just not guaranteed to run at lower voltages. Each kit is of different qualitiy and RAM vendors just set higher voltages to guarantee compatibility across the board.

I guess the solution would be to use/develop RAM kits that run with lower voltages. It kinda reminds me of the DDR3 times when all vendors had kits running on 1.5+V and G.Skill came around the corner with their "G.Skill ECO Series" which did run on only 1.35V. It also marked the rise of the 1.35V kits.

Those aren't the soc voltages everyone is talking about. Those voltages won't cause this problem or explosion, from what gn reported.
Posted on Reply
#67
Hugh Jass
DavenAMD’s biggest weakness is the memory controller. It needs improvements across the board (pun intended).
I agree it would be nice, but it's already running twice as quickly as Intel's is, so I'm not really sure what you're expecting this early on in the DDR5 game.

Good explanation video
Posted on Reply
#68
Nino88
somebodys_kidMy 2x16 GB GSkill EXPO profile became unstable after my BIOS update (DDR5 6000 CL 30 at 1.35V). Which memory are you all using that keeps the speed and latency but is stable?
I have a 7800x3d CPU, a 7900xtx, on a mobo B650M ASUS TUF PLUS WIFI, and im currently working with the T-Group Create Expert 6400Mhz CL34-44-44-84 memories but my motherboard did not recognize the ram memory, so I boot with frequencies of 4800mhz, and different latencies, what I had to do was set it manually without any EXPO profile, since I just changed to one of those profiles to try to establish my memories as they are predestined by the factory and I started the system and everything, but playing it felt very unstable, playing Fortnite it crashed twice, so I have to manually re-establish a configuration leaving them working at 6000 MHz, with voltages specified by the same memory 1.35V and lower latencies 30-36-36-74.

Once these changes have been made, my device works so far without any crashes, I will continue testing to see how it goes.
Posted on Reply
#69
Taraquin
Nino88I have a 7800x3d CPU, a 7900xtx, on a mobo B650M ASUS TUF PLUS WIFI, and im currently working with the T-Group Create Expert 6400Mhz CL34-44-44-84 memories but my motherboard did not recognize the ram memory, so I boot with frequencies of 4800mhz, and different latencies, what I had to do was set it manually without any EXPO profile, since I just changed to one of those profiles to try to establish my memories as they are predestined by the factory and I started the system and everything, but playing it felt very unstable, playing Fortnite it crashed twice, so I have to manually re-establish a configuration leaving them working at 6000 MHz, with voltages specified by the same memory 1.35V and lower latencies 30-36-36-74.

Once these changes have been made, my device works so far without any crashes, I will continue testing to see how it goes.
You will probably get 6200cl30-37-37 working with 1.4v voltage :) Most important is lowering RFC and raising REFI. M-die can do 450-500 RFC and 65536 REFI, those 2 changes alone counts for 5-10% better performance :)
Posted on Reply
#70
rv8000
TaraquinYou will probably get 6200cl30-37-37 working with 1.4v voltage :) Most important is lowering RFC and raising REFI. M-die can do 450-500 RFC and 65536 REFI, those 2 changes alone counts for 5-10% better performance :)
M-die will bottom out around 160ns for trfc, its best not to recommend a timing value as that depends on the speed they're running at.

(tRFC value x 2000)/effective memory frequency = tRFC in ns

Lowest on m-die that is likely stable would be 496 @DDR5 6200 (exactly 160ns)
Posted on Reply
#71
Taraquin
rv8000M-die will bottom out around 160ns for trfc, its best not to recommend a timing value as that depends on the speed they're running at.

(tRFC value x 2000)/effective memory frequency = tRFC in ns

Lowest on m-die that is likely stable would be 496 @DDR5 6200 (exactly 160ns)
M-dies RFC scales with voltage. I build a setup recently and it did 480@6200 with 1.4v. At 6000 it ran at 470, but had tobraise it to 480 with 1.35v. If you are really lucky with binning 450-460 at 6000@1.5v should be doable :)
Posted on Reply
#72
rv8000
TaraquinM-dies RFC scales with voltage. I build a setup recently and it did 480@6200 with 1.4v. At 6000 it ran at 470, but had tobraise it to 480 with 1.35v. If you are really lucky with binning 450-460 at 6000@1.5v should be doable :)
What are you using and how long are you running said tests to establish stability?

Afaik its pretty common m-die isnt 100% stable below 160ns. Of the three M-die kits I have none go below 160ns at 6400 c30 1.45-1.46 vdd.

Once you’ve increased your refresh window by raising tREFI to max, theres little to gain by over tightening tRFC.
Posted on Reply
#73
Taraquin
rv8000What are you using and how long are you running said tests to establish stability?

Afaik its pretty common m-die isnt 100% stable below 160ns. Of the three M-die kits I have none go below 160ns at 6400 c30 1.45-1.46 vdd.

Once you’ve increased your refresh window by raising tREFI to max, theres little to gain by over tightening tRFC.
20 rounds of TM5 usmus cfg without errors. Build it for a friend, no crashes since May so seems stable :) I agree REFI counts the most, but I got a small boost from lowering RFC after REFI. Guess I was a bit lucky with that M-die kit.
Posted on Reply
Add your own comment
Jan 11th, 2025 12:08 EST change timezone

Latest GPU Drivers

New Forum Posts

Popular Reviews

Controversial News Posts