MPG X870E CARBON WIFI Beta BIOS

Sort by date Sort by votes
Alyred said:
If you're running a 9800x3d it sounds to me like you're pushing it to its very limits by boosting it's bclk and then dropping it's voltage via CO to about the maximum available, and then having stability issues. I agree, this board should have features the tomahawk doesn't for the extra money spent on it. There's still problems to be worked out and MSI is still probably just spinning up from Chinese new year. But you're really pushing RAM and Processor to the limits there.
Click to expand...
Hi,
Even at 103 BCLK, it would pass Cinebench R23 , OCCT, AIDA and Y-Cruncher ( to an extent..), the voltage fluctuating too much would make sometimes my overclock not stable, depending on what numbers it chose to dial in ( using auto on Overclocking AMD for Voltage, overclocking page, setting a fixed voltage for PBO doesn't work well)
The real stability issues are more USB related and that does happens even at stock, I use my computer for simracing in VR, so I have a huge amount of USB connected at all time.
regards,
Victor
 
Upvote 0 Downvote
Alyred said:
GPIO (for Promontory) will always fail because that's the codename for the chipset for AM4 (Promontory). I've read that the AMD Interface driver installs but always shows fail because of an installer log issue, but that apparently fails on pretty much everyone's installation.
Undervolting on your cores won't improve performance specifically, it allows you to raise your scalar and boost clocks for longer while keeping your temperatures lower. You could also increase your base clock but that's more complex and a bit more "dangerous" to stability to keep the base clock running at higher frequency. Ideally you'll want to set your core COs individually as each core will have different tolerances for undervolting. Finding an all-core undervolt that is fully stable means that there are probably some cores that could be undervolted a little more but that takes time to test.
Memory clock that @Anaksagoras mentioned are ideal but not every CPU/RAM can reach those numbers. I also have been looking into FCLK settings and there's apparently a synchronicity when the FCLK is set to an even ratio to the UCLK/MEMCLK, at 3:2 it will lower your latency by a few nanoseconds. I've seen that in tests myself, though it's less than 2ns. Which ain't nothing, but not mind-blowing either.
You can test your bandwidth and latency by downloading AIDA64 as a demo, it'll show you your read speed and latency. You can also use it to stress test your system's combination of CPU, FPU and RAM. Throwing that at it will let you know if you have issues lurking when the CPU starts getting warm or gets saturated with a lot of work. You can have a system that runs on day-to-day stuff fine but gets unstable when working hard. The problems aren't really the crashes in those situations but the "errors" - think it returning a wrong number when doing math stuff, where it'll return the wrong information but won't throw an error.
Click to expand...
So, after resetting the CMOS several times, it is possible to identify that the same BIOS settings give me different performance results. Currently I use individual CO configuration per core and the central issue is to have the same performance with a CO of 0 and CO of -20, for example.
Working with several VRM configurations also does not give me performance gains, but when using LLC 6 it is certain that the system freezes.
Today I'm using FCLK@2000, but going up more than that doesn't make me gain performance.
I've tested several configurations and it's difficult to extract more than 45000 points on the CB23, and I've seen many get 47000 without making much effort. I know that each silicon is different, but I myself have already managed to reach 47000 but with the same settings I used now they don't give me the same score regardless of the values configured in the BIOS.
I've been thinking about not using any overclocking settings in the MSI section and using it only in the AMD overclocking section. I found a post on Reddit where a user also lost performance when using the overclocking section of the motherboard manufacturer.
 
Last edited:
Upvote 0 Downvote
RobsonByte said:
So, after resetting the CMOS several times, it is possible to identify that the same BIOS settings give me different performance results. Currently I use individual CO configuration per core and the central issue is to have the same performance with a CO of 0 and CO of -20, for example.
Working with several VRM configurations also does not give me performance gains, but when using LLC 6 it is certain that the system freezes.
Today I'm using FCLK@2000, but going up more than that doesn't make me gain performance.
I've tested several configurations and it's difficult to extract more than 45000 points on the CB23, and I've seen many get 47000 without making much effort. I know that each silicon is different, but I myself have already managed to reach 47000 but with the same settings I used now they don't give me the same score regardless of the values configured in the BIOS.
Click to expand...
Just my two cents, I tried, like you, to set a negative offset per core, but after spending hours finding the optimal value for each core, I realized that Cinebench R23 and CPU-Z would only score based on the worst core offset I could achieve. So, I ended up using that value for all cores.
From what I understand FCLK works best when it can be synchronized with an exact multiple with ram speed like 2000 on DDR5 6000 kit or DDR5 8000.
(If it's not in sync it would introduce latency, saw that on Reddit, true or not it's about the rule I'm following for getting best results in benches )
 
Upvote 0 Downvote
vic123159d02e5 said:
I never said I had a problem overclocking beyond standard PBO, but the voltage is completely erratic. Sometimes it boots at 1.245V, then on the next reboot it hits 1.28V, and I’ve even seen it push to 1.34V! Even with BIOS A25, it’s almost impossible to boot if I leave a USB stick plugged in.
I’m running a –34 curve with a 103 BCLK and 2100 FCKL; the max boost is 5587 MHz, and my RAM runs at 8240 MHz at CL36. The chip never exceeds 73°C even during OCCT.
It can also achieve an all-core overclock at 5625 MHz with SMT disabled, but then it hits 88°C, so I reverted to the standard boost. This motherboard has no external clock (imagine that for a $500 board), so 103 BCLK is about the maximum achievable,103.5 is possible, but my RAM would need significantly more voltage (around 1.56V) for stability.
I don’t consider using standard PBO +200 mhz boost even with a negative curve as overclocking; most $200 boards are capable of similar performance. If I’d be happy with only a 200 MHz boost on PBO, I would have bought one of those cheap boards.
Click to expand...
Aha so you went way over specs and complaining ..im with -20 ,6000 ram at 1.35 soc 1.1 and i can boot wit and without usb no stuters no bs on win 10 ..you try to push memory over 10k and pbp at -99 and then complajn..
 
Upvote 0 Downvote
vic123159d02e5 said:
Just my two cents, I tried, like you, to set a negative offset per core, but after spending hours finding the optimal value for each core, I realized that Cinebench R23 and CPU-Z would only score based on the worst core offset I could achieve. So, I ended up using that value for all cores.
From what I understand FCLK works best when it can be synchronized with an exact multiple with ram speed like 2000 on DDR5 6000 kit or DDR5 8000.
(If it's not in sync it would introduce latency, saw that on Reddit, true or not it's about the rule I'm following for getting best results in benches )
Click to expand...
Right, having the FCLK in an even resonance with the UCLK and MEMCLK is best for latency, otherwise the processor on FCLK speed is waiting for the next RAM refresh cycle each time it need to put or retrieve something from memory. Resonance numbers are at 3:2 since FCLK can't reach 3000+ speeds.
2000 for DDR6000 (MEMCLK/UCLK:3000) or DDR8000 (MEMCLK:4000 UCLK:2000)
2067 for DDR6200 (MEMCLK/UCLK:3100) or DDR8400 (MEMCLK:4200 UCLK:2100)
2133 for DDR6400 (MEMCLK/UCLK:3200)
etc.
 
Upvote 0 Downvote
zonachristiany12280298 said:
Aha so you went way over specs and complaining ..im with -20 ,6000 ram at 1.35 soc 1.1 and i can boot wit and without usb no stuters no bs on win 10 ..you try to push memory over 10k and pbp at -99 and then complajn..
Click to expand...
Jesus, do you even read posts before answering? The USB problems happen at stock speed. The only thing I hate about overclocking with this board is the voltage being different on each boot. Read the entire thread before rejoicing as if you just discovered the Earth isn't flat... FYI : I have DDR5 rated@8000 , running it at 8240 is not that much of an overclock.
 
Last edited:
Upvote 0 Downvote
Alyred said:
Thanks for the hint. I did just check and it does seem to be maintaining 4th gen between reboots. I do automatically set the M.2, chipset, and PCIe slots as fixed at PCIe v4 in the BIOS, I hate letting the board determine it automatically and have some weird response set it down to 1.1 or something. :D So at least on the carbon wifi it's not experiencing that issue. If the Tomahawk is, that's a fairly significant issue that a lot of folks might not notice. I wonder if your BIOS upgrade went wonky or something? Any other folks that you've spotted reporting this? I was unable to find a Tomahawk x870 motherboard beta BIOS page...
In my (and another commenter's case a few pages back, I forget which right now) the speeds are definitely PCIe v4 level, but random write IOPS is much lower than it should be on the first M.2 slot:
View attachment 199084
This was from just after I upgraded to A25. The random write on A21 is around 1.2 million IOPS with the same hardware.
Strangely, the M2_3 slot has better performance in Random Write (which is what you really want over Sequential), but a sequential read that is slightly lower than it should be, even when factoring in the overhead from going through the chipset:
View attachment 199085
I expect to see a slight difference, like the difference between the random read on the two drives, but everything is fairly significantly lower on that D drive, except the random read IOPS which is about the delta it should be (around 3%).
Click to expand...
Pretty widespread it seems. I have seen mention of this on the Carbon and -P as well.

https://www.reddit.com/r/MSI_Gaming/comments/1iexixz
Such an odd bug. Pretty nice board otherwise.
 
Upvote 0 Downvote
vic123159d02e5 said:
Hi,
Even at 103 BCLK, it would pass Cinebench R23 , OCCT, AIDA and Y-Cruncher ( to an extent..), the voltage fluctuating too much would make sometimes my overclock not stable, depending on what numbers it chose to dial in ( using auto on Overclocking AMD for Voltage, overclocking page, setting a fixed voltage for PBO doesn't work well)
The real stability issues are more USB related and that does happens even at stock, I use my computer for simracing in VR, so I have a huge amount of USB connected at all time.
regards,
Victor
Click to expand...
I am normally an Intel overclocking enthusiast so I don't know if it will affect AMD in the same way. Maybe worth investigating if you have not already. USB dropout issues and hanging during POST with USB ports populated on Intel platforms are often experienced when system agent and/or memory voltage is set too low, even though all memory tests may pass without error. Have you already tried elevating SOC and VDIMM to see if that reduces or solves the problem? If not, I would suggest giving that a try.

My AM5 platform is the most stable and performs best with fixed core ratios and fixed (override) voltage values. No adaptive behavior also works much better for me on my Intel platforms. More reliable and performant on both platforms for me when dynamic core boosting and voltage behaviors are eliminated.
 

Attachments

  • 2025-02-19_06-25-05.jpg
    2025-02-19_06-25-05.jpg
    1.1 MB · Views: 40
Last edited:
Upvote 0 Downvote
Mr.Fox said:
I am normally an Intel overclocking enthusiast so I don't know if it will affect AMD in the same way. Maybe worth investigating if you have not already. USB dropout issues and hanging during POST with USB ports populated on Intel platforms are often experienced when system agent and/or memory voltage is set too low, even though all memory tests may pass without error. Have you already tried elevating SOC and VDIMM to see if that reduces or solves the problem? If not, I would suggest giving that a try.

My AM5 platform is the most stable and performs best with fixed core ratios and fixed (override) voltage values. No adaptive behavior also works much better for me on my Intel platforms. More reliable and performant on both platforms for me when dynamic core boosting and voltage behaviors are eliminated.
Click to expand...
Ah, looks like you're on a 9950x which has more flexible settings, IMC and a higher base clock as well. I think a lot of the problems we are hitting are due to the 9800x3d chips (probably other 3d vCache ones as well) and interactions with the mainboard BIOS/IMC on that chip, which seems to be a little less able to push really low latencies on the RAM.
 
Upvote 0 Downvote
Mr.Fox said:
I am normally an Intel overclocking enthusiast so I don't know if it will affect AMD in the same way. Maybe worth investigating if you have not already. USB dropout issues and hanging during POST with USB ports populated on Intel platforms are often experienced when system agent and/or memory voltage is set too low, even though all memory tests may pass without error. Have you already tried elevating SOC and VDIMM to see if that reduces or solves the problem? If not, I would suggest giving that a try.

My AM5 platform is the most stable and performs best with fixed core ratios and fixed (override) voltage values. No adaptive behavior also works much better for me on my Intel platforms. More reliable and performant on both platforms for me when dynamic core boosting and voltage behaviors are eliminated.
Click to expand...
Hello, I do only have hanging issues with USB sticks plugged in right now, disconnections have gone since A23 Bios version (If I recall right) . The SOC is already super high at default 1.306 v , so I do not have much Headroom here, I also played quite a lot with Vdimm but without any luck.
 
Upvote 0 Downvote
@Svet I'm waiting for carbon 870 from China (out of stock in my region). But now there are problems with the bios on the tomahawk 870. Tell me, maybe there are beta builds for the tomahawk to try? Cyclic reboots and sometimes the pci e crashes with vanguard 5080 soc
 
Upvote 0 Downvote
pogulyajkodim128f02a9 said:
@Svet I'm waiting for carbon 870 from China (out of stock in my region). But now there are problems with the bios on the tomahawk 870. Tell me, maybe there are beta builds for the tomahawk to try? Cyclic reboots and sometimes the pci e crashes with vanguard 5080 soc
Click to expand...
I'm hoping someone will post a link to them, if so. I've seen enough questions about the Tomahawk x870 and nowhere to point people, and unable to find it by searching in the downloads forum.
 
Upvote 0 Downvote
pogulyajkodim128f02a9 said:
@Svet I'm waiting for carbon 870 from China (out of stock in my region). But now there are problems with the bios on the tomahawk 870. Tell me, maybe there are beta builds for the tomahawk to try? Cyclic reboots and sometimes the pci e crashes with vanguard 5080 soc
Click to expand...

Alyred said:
I'm hoping someone will post a link to them, if so. I've seen enough questions about the Tomahawk x870 and nowhere to point people, and unable to find it by searching in the downloads forum.
Click to expand...

Oks, I've opened a new topic for this board betas here:
 
Upvote 1 Downvote
I appreciate the Betas; I wish it was made more clear on the web-site that these exist. I stumbled on here from another thread.

I'm hopeful the A25 prevents the random blue screens with my NVMe drive not showing up on reboot.
If it does, I know I have at least one stable option available so I don't have to return my MSI board before my Amazon return window closes (March 8).
(And if it doesn't, will try out that A26 and will jump on that new thread.)
 
Upvote 0 Downvote
Svet said:
Updated.
Click to expand...
Thank you for providing beta bios. I have a lot of issues with my second nvme drive (gen5) whatever the slot (M2_3 or M2_4). It seems writing on it trigger the disconnect more quickly but that could be a bias. Yet if I do nothing with it, it does not disconnect (but not very useful is it?).
(a lot less issues with M2_1 since A25 I believe especially if I force gen4 on my T705 crucial (gen5)).
I will install the A26.
Could you provide some kind of changelog or something or did I miss it somewhere? (just to give me an idea if it's worth testing again?)

=> still the random disconnect issue :(, will try with another nvme gen4 drive soon, quite disappointed for now
 
Last edited:
Upvote 0 Downvote
gael.blottier15a102de said:
=> still the random disconnect issue :(, will try with another nvme gen4 drive soon, quite disappointed for now
Click to expand...
I wanted to ask if you've tried the drive itself in the top slot, and on another board to see if it still disconnects. It could be the board has an issue specifically with the crucial T705, but it would be good to eliminate the possibility that that specific drive has problems, and is properly cooled. Are your forcing Gen4 (instead of Auto) in the BIOS for those slots?
 
Upvote 0 Downvote
nopone156702dd said:
FYI, I resolved my problem with the NVME in M2_4 not being detected by raising the Chipset Core Voltage. Perhaps that could help with the disconnects too if you are using the slots driven by the chipset and not the CPU.
Click to expand...
Dang, good idea. Several of the USB ports are through both chipsets, i winder if that might eliminate some of the other USB issues folks are having?
Where specifically is the chipset voltage setting in the BIOS?
Have you noticed any heat issues with the increased voltage? The second chipset already runs fairly hot according to HWInfo, and there's no built-in active cooling.
 
Upvote 0 Downvote
Back
Top