MPG X870E CARBON WIFI Beta BIOS

[Svet]

Spoiler: Older Versions

>>> E7E49AMSI.1A1E <<<

>>> E7E49AMSI.1A1H <<<

>>> E7E49AMSI.1A21 <<<

>>> E7E49AMSI.1A22 <<<

>>> E7E49AMSI.1A24 <<<

>>> E7E49AMSI.1A25 <<<

>>> E7E49AMSI.1A26 <<<

>>> E7E49AMSI.1A28 <<<

>>> E7E49AMSI.1A29 <<<

>>> E7E49AMSI.1A31 <<<

>>> E7E49AMSI.1A32 <<<

>>> E7E49AMSI.1A33 <<<

>>> E7E49AMSI.1A34 <<<

>>> E7E49AMSI.1A41 <<<

>>> E7E49AMSI.1A44 <<<

Notes:

• * Unstable RAM OC
• * Hang 0D when saving BIOS settings or doing M-Flash

>>> E7E49AMSI.1A62 <<<

>>> E7E49AMSI.1A65 <<<

>>> E7E49AMSI.1A66 <<<

 

[sko8155902d1]

First, I noticed your Curve Optimizer is at -38. While the 9800X3D is efficient, -38 is quite aggressive. Cold start issues are often linked to "low-load instability." When the system is cold and idling at the desktop, a high negative CO offset can cause the voltage to drop too low for the CPU's internal PCIe controller to maintain a stable link with the Gen5 GPU. I would suggest testing with CO at Stock or -20 just to rule out CPU instability affecting the PCIe lanes.

I'd have to test it with the 5080. With the 4080 (since December 2024), I run aggressive UV without any problems. I've been running +200/-42 for a very long time, but testing has shown that +200 doesn't really make much difference, especially if you're playing with FPS cap anyway. Just pumps in unnecessary heat and high voltage. That's why stock clocks with high UV.

If the cold start problem doesn't happen, the system runs perfectly. With FCH Spread Spectrum enabled (+ new AMD chipset driver), the problem doesn't seem to occur for now.

Disable "Fast Boot" in Windows:

I tried disabling fast start. It didn't work.

Disable "Power Management Mode" in NVIDIA Control Panel: Set it to "Prefer Maximum Performance" to see if the stuttering stops. Sometimes the aggressive power saving on Gen5 cards causes link drops during idle.

Since I initially suspected PCIe 5.0, I tried many settings that prevent the GPU from changing the PCIe version. Max Performance in NVCP isn't an option for me cuz the power consumption in idle/office load is absurdly high. I use two 1440p monitors (165 Hz & 360 Hz). With standard settings, the 5080 consumes 20-25w avg. With max performance, it consumes 50-60w, just for doing nothing.

I also tested other cables (DP & HDMI), tested lower and identical Hz ratings, ErP on/off, the newest monitor firmware etc., but without success. Now, with FCH Spread Spectrum, the problem no longer seems to occur.

Your "recovery count" going crazy is a clear sign of PCIe Signal Integrity issues.

This behavior occurs only when the cold start bug occurs. Interestingly, a user from the HWL forum with the same issue has exactly the same image errors (from cold start bug sometimes happen) as I do, but with an RTX 5070Ti and an X870 Tomahawk WiFi. Also with 9800XD, but without any UV. Translated link post. That's what I read most often in this situation. A common problem with new MSI boards and X3D CPUs.

If forcing Gen4 didn't work, it might actually be the CPU's SOC voltage or the CO offset being too low to drive the high-speed bus stably.

CPU/IO-related settings such as vSOC, VDDGs, VDDIO, UV, etc. may be a good tip. I will continue to keep an eye on this.

Keep GPUz installed and check it regularly - you might see some 4.0's in there that you might not notice otherwise, but it largely does take care of the problem with FCH Spread Spectrum enabled.

I would say atm everything is normal (without the cold start bug). Under load, it jumps up to x16 5.0, and under idle/light load, it goes down to x16 1.1.

 

[Vendex_AMD]

I'd have to test it with the 5080. With the 4080 (since December 2024), I run aggressive UV without any problems. I've been running +200/-42 for a very long time, but testing has shown that +200 doesn't really make much difference, especially if you're playing with FPS cap anyway. Just pumps in unnecessary heat and high voltage. That's why stock clocks with high UV.

If the cold start problem doesn't happen, the system runs perfectly. With FCH Spread Spectrum enabled (+ new AMD chipset driver), the problem doesn't seem to occur for now.





Since I initially suspected PCIe 5.0, I tried many settings that prevent the GPU from changing the PCIe version. Max Performance in NVCP isn't an option for me cuz the power consumption in idle/office load is absurdly high. I use two 1440p monitors (165 Hz & 360 Hz). With standard settings, the 5080 consumes 20-25w avg. With max performance, it consumes 50-60w, just for doing nothing.

I also tested other cables (DP & HDMI), tested lower and identical Hz ratings, ErP on/off, the newest monitor firmware etc., but without success. Now, with FCH Spread Spectrum, the problem no longer seems to occur.


This behavior occurs only when the cold start bug occurs. Interestingly, a user from the HWL forum with the same issue has exactly the same image errors (from cold start bug sometimes happen) as I do, but with an RTX 5070Ti and an X870 Tomahawk WiFi. Also with 9800XD, but without any UV. Translated link post. That's what I read most often in this situation. A common problem with new MSI boards and X3D CPUs.


CPU/IO-related settings such as vSOC, VDDGs, VDDIO, UV, etc. may be a good tip. I will continue to keep an eye on this.


I would say atm everything is normal (without the cold start bug). Under load, it jumps up to x16 5.0, and under idle/light load, it goes down to x16 1.1.

That’s a very interesting finding about FCH Spread Spectrum. Enabling it usually helps with EMI (electromagnetic interference) by smoothing the clock signal, which suggests the PCIe 5.0 link is indeed on the edge of stability.
However, I wouldn’t dismiss the Curve Optimizer (CO) just because it worked with your 4080. Here is why:

• PCIe 5.0 vs 4.0 Sensitivity: The RTX 4080 uses PCIe 4.0, which is far more forgiving. PCIe 5.0 doubles the bandwidth and is extremely sensitive to voltage fluctuations in the CPU's I/O die. Even if the cores are stable in Cinebench, the internal interconnect (Infinity Fabric/L3 cache) might struggle to maintain the Gen5 link during the low-voltage states of a cold start.
• The "Cold Start" Logic: During a cold boot, the VRMs and the silicon are at ambient temperature. Resistance is lower, and voltage transitions happen differently than when the system is "warm." A -38/-42 offset might be pushing the voltage below the physical floor required for the PCIe controller to initialize a Gen5 link correctly.
• vSOC and VDDIO: Since you mentioned these, have you checked your BIOS-reported CPU SOC Voltage? For 6200MT/s RAM, the Auto settings might be pushing it too high or keeping it too low. Manually setting SOC to 1.20V - 1.25V often stabilizes the PCIe bus on X870E boards.

Regarding the idle power consumption: I totally understand not wanting to use "Max Performance" in NVCP. Instead of that, you could try setting "PCI Express Link State Power Management" to Off in your Windows Power Plan settings. This prevents the link from dropping to 1.1 when idling, which might stop the stuttering without forcing the GPU clocks to stay at max.

It definitely sounds like a BIOS maturity issue (AGESA) combined with very tight voltage margins. If the FCH Spread Spectrum fix holds, that’s great, but if it happens again, I’d strongly suggest backing off the CO to -20 just for a week to see if the "cold start bug" disappears entirely.

 

[Blackbird7]

Instead of that, you could try setting "PCI Express Link State Power Management" to Off in your Windows Power Plan settings. This prevents the link from dropping to 1.1 when idling, which might stop the stuttering without forcing the GPU clocks to stay at max.

I have that off, but still there are drops from 4.0 to 1.1 when idle (I have 4090). But never noticed stuttering. So when GPU-Z shows 4.0@1.1, it's all good. Stuttering happens only when it shows 3.0@3.0 (or less for second number), after cold boot.
Good point with SOC, if I get the problem again, I'll try raising that a bit, now is 1.20V. Few days without problem, so maybe disabling fast startup did help...

 

[Alyred]

Stuttering can happen at a number of different "PCIe" gen speeds. It's trying to step back up to 5 but can't. It's normal for it to drop down to 1.1 or 2 when the card isn't being used. But if you try to do a render test in GPUz, and it does NOT rise back up to 4 or 5, you are experiencing the PCIe locking and should suspend or warm-restart your computer because it will continue stuttering and likely crash.

FCH Spread Spectrum has mostly solved the issue for most of us in this 69 page thread. There's been a lot of discussion on it over the past year.

 

[L E G I T]

Although I own the MPG B850 EDGE TI WIFI, I have been experiencing an issue where the PCIe bus gets locked at version 1.1 since the launch of the RTX 50XX series in January last year. I am running an MSI RTX 5080 with a 9800X3D.

No BIOS update, Nvidia driver update, or chipset driver update has resolved the problem. During system startup, PCIe often gets stuck at version 1.1 or 2.0, which causes noticeable lag, mouse stuttering, and after several seconds usually leads to a system restart. In most cases, after the restart PCIe finally runs at version 5.0, but this does not happen every time. With some BIOS versions, I had to restart the system several times because PCIe remained at version 2.0 or 3.0, far from the expected 5.0.

I noticed that BIOS version 7E62v1A20 dated March 18, 2025 is the most stable so far. With this version, the issue occurs only occasionally instead of daily or every second day, as was the case with other BIOS versions, including the latest official release 7E62v1A46 with AGESA 1.2.0.3g.

I have tried every possible workaround, including changing many BIOS settings such as Memory Context Restore, Global C state Control, Power Supply Idle Control, ASPM Control for CPU PCIe, and ErP Ready, as well as testing all available Nvidia drivers and chipset driver versions. Despite all of this, the issue still persists.

Fortunately, the problem only appears during system startup. Once PCIe successfully initializes at version 5.0, the system works perfectly and remains stable under load. However, the fact that MSI has been unable to fix this issue for over a year is, in my opinion, unacceptable, especially considering that ASUS and Gigabyte released fixed BIOS versions as early as March last year and the issue was completely resolved on their platforms.

On January 12, 2026, MSI released a beta BIOS that introduces AGESA version 1.3.0.0, but since it is still a beta version, I do not intend to install it or risk potentially more serious problems.

I wish good luck to both myself and you, and I would be happy to learn about any tips or options that I may have missed or not heard about.

Edit: I should also add that I have iGPU disabled, and manually setting the PCIe slot to version 5.0 or 4.0 does not help. I am only using the EXPO1 RAM profile, and all other CPU and GPU settings are at stock. I experimented with GPU undervolting and adjusting the PBO for the 9800X3D, but I stopped doing this because I was concerned that the frequent system restarts caused by the PCIe bus getting stuck could damage something.

 

[sko8155902d1]

Since you mentioned these, have you checked your BIOS-reported CPU SOC Voltage? For 6200MT/s RAM, the Auto settings might be pushing it too high or keeping it too low. Manually setting SOC to 1.20V - 1.25V often stabilizes the PCIe bus on X870E boards.

These profiles work for me (vSOC in ZT). All of them have been tested multiple times with TM5 (anta X3D) and some of them additionally with OCCT Memtest. These profiles, as well as the UV, ran mostly without any issues over the past year (old logs, use atm ZT 1889).

Here is why:

However, this doesn't explain why the user with an X870 Tomahawk and 5070Ti having exactly the same problems, as are some users here and on Reddit. The user from the HWL forum has everything set to Auto/Default except for EXPO. Completely new hardware.

That’s a very interesting finding about FCH Spread Spectrum. Enabling it usually helps with EMI (electromagnetic interference) by smoothing the clock signal, which suggests the PCIe 5.0 link is indeed on the edge of stability.

It definitely sounds like a BIOS maturity issue (AGESA) combined with very tight voltage margins. If the FCH Spread Spectrum fix holds, that’s great

It seems I've found the solution, at least one that works reliably for me.

I had wondered why FCH Spread Spectrum seemed to eliminate the cold start bug. This setting affects the (FCH) bus clock. When I compared my old logs (from testing), I noticed that the bus clock is now only at 99.8 MHz. Depending on the setting (PBO/ CO) and BIOS, this value changes between 100 MHz or 99.8 MHz, even though FCH Spread Spectrum and Bus Clock are (always) set to Auto.

To me, it looks like there is a bug in the BIOS (A70)/ default settings. I set the bus clock to 100.000 with FCH Spread Spectrum enabled, but it still stays at 99.8 MHz. Maybe this is intentional behavior, idk. One thing is for sure: unstable/incorrect bus clock affects PCIe and USB, and basically all other communication.

Solution: Manually set the FCH bus clock to 100.000 MHz and FCH Spread Spectrum back to Auto. Then the bus clock stays at correct 100 MHz and the system boots up without any problems every time. Hopefully it will help other users too.

 

[L E G I T]

Solution: Manually set the FCH bus clock to 100.000 MHz and FCH Spread Spectrum back to Auto. Then the bus clock stays at correct 100 MHz and the system boots up without any problems every time. Hopefully it will help other users too.

I also have 99.8 MHz, even under load. I will set it to 100 MHz and see if that helps.
PS. But let me ask for clarification. You're on AGESA version 1.3.0.0 or 1.2.0.3g?

 

[sko8155902d1]

PS. But let me ask for clarification. You're on AGESA version 1.3.0.0 or 1.2.0.3g?

BIOS (A70)

AGESA 1.2.0.3g

 

[marceluss19]

These profiles work for me (vSOC in ZT). All of them have been tested multiple times with TM5 (anta X3D) and some of them additionally with OCCT Memtest. These profiles, as well as the UV, ran mostly without any issues over the past year (old logs, use atm ZT 1889).

View attachment 208906


However, this doesn't explain why the user with an X870 Tomahawk and 5070Ti having exactly the same problems, as are some users here and on Reddit. The user from the HWL forum has everything set to Auto/Default except for EXPO. Completely new hardware.



It seems I've found the solution, at least one that works reliably for me.

I had wondered why FCH Spread Spectrum seemed to eliminate the cold start bug. This setting affects the (FCH) bus clock. When I compared my old logs (from testing), I noticed that the bus clock is now only at 99.8 MHz. Depending on the setting (PBO/ CO) and BIOS, this value changes between 100 MHz or 99.8 MHz, even though FCH Spread Spectrum and Bus Clock are (always) set to Auto.

View attachment 208916

To me, it looks like there is a bug in the BIOS (A70)/ default settings. I set the bus clock to 100.000 with FCH Spread Spectrum enabled, but it still stays at 99.8 MHz. Maybe this is intentional behavior, idk. One thing is for sure: unstable/incorrect bus clock affects PCIe and USB, and basically all other communication.

Solution: Manually set the FCH bus clock to 100.000 MHz and FCH Spread Spectrum back to Auto. Then the bus clock stays at correct 100 MHz and the system boots up without any problems every time. Hopefully it will help other users too.

View attachment 208917

99.8 MHz Isn’t it normal and expected behavior with Spread Spectrum enabled? Setting 100 and enabling auto, you disabled Spread spectrum.

 

[dlomond]

File: E7E49AMSI.1A70
Size: 32,768 KB
AGESA: ComboAm5PI 1.2.0.3f

64 is G that is why we tell everyone to go with 64 bios.

 

[dlomond]

File: E7E49AMSI.1A70
Size: 32,768 KB
AGESA: ComboAm5PI 1.2.0.3f

64 is G that is why we tell everyone to go with 64 bios. Also I think that 64 works better. That is what most of us see anyway. With msi you can't just take that last bios and think it is the latest version MSI goes back and forth. Don't know why they do it like that.

 

[sko8155902d1]

99.8 MHz Isn’t it normal and expected behavior with Spread Spectrum enabled? Setting 100 and enabling auto, you disabled Spread spectrum.

It was a bit unclear to me which setting overwrites the other.
Bus clock auto + FCH Spectrum enabled = 99.8 MHz - OK
Bus clock 100 MHz + FCH Spectrum auto = 100 MHz - OK
Bus clock 100 MHz + FCH Spectrum enabled = unclear (it's 99.8 MHz)

What is definitely not normal is that the BIOS A70 simply sets the bus clock to 99.8 MHz. I've several HWinfo logs from my RAM testing. With older BIOS versions, it was always 100 MHz, and the bus clock and FCH Spectrum were always set to auto. In any case, something is wrong with the BIOS or default settings when the bus clock and FCH spectrum are set to auto, at least with PCIe 5.0 GPU. With the 4080, it was 100 MHz bus clock, but it was also BIOS A64.

 

[L E G I T]

Solution: Manually set the FCH bus clock to 100.000 MHz and FCH Spread Spectrum back to Auto. Then the bus clock stays at correct 100 MHz and the system boots up without any problems every time. Hopefully it will help other users too.

I set the FCH Base Clock to 100.000 MHz and while HWiNFO does show the Bus Clock as 100.0 MHz, it didn't help and when launching PC, GPU-Z showed PCIe 1.1, which led to a restart.

 

[marceluss19]

It was a bit unclear to me which setting overwrites the other.
Bus clock auto + FCH Spectrum enabled = 99.8 MHz - OK
Bus clock 100 MHz + FCH Spectrum auto = 100 MHz - OK
Bus clock 100 MHz + FCH Spectrum enabled = unclear (it's 99.8 MHz)

What is definitely not normal is that the BIOS A70 simply sets the bus clock to 99.8 MHz. I've several HWinfo logs from my RAM testing. With older BIOS versions, it was always 100 MHz, and the bus clock and FCH Spectrum were always set to auto. In any case, something is wrong with the BIOS or default settings when the bus clock and FCH spectrum are set to auto, at least with PCIe 5.0 GPU. With the 4080, it was 100 MHz bus clock, but it was also BIOS A64.

This whole [***CENSORED***] is weird. On my bios, old 50, I have FCH Spectrum auto, bus 99.8MHz by CPUz but fluctuating, sometimes 100, and perfectly working PC, not a single bad gen after cold boot or any other mentioned in this thread issues.

 

[L E G I T]

It is one big lottery. And MSI has been losing it for a year, unable to release an update that would fix this. But the company is what it is, so are the results.

 

[mr-bcm]

Updated from A64 to the newest 7E49v1A7E.

This appears to be the best version yet for my system. In the past I had to enable "Power Down Enable" so that my computer wouldn't just get stuck booting. Before this version I think A40 was the best as I could restart and not have issues, other versions I would have to just come out of sleep, or shut down and then turn back on. Restarting used to almost always get stuck.

If I notice issues I'll try to update my post, but right now A7E is the best. I'm not doing benchmarking though, so I don't know if it gets the highest performance, but for me it is the most stable.

 

[Anaksagoras]

Updated from A64 to the newest 7E49v1A7E.

This appears to be the best version yet for my system. In the past I had to enable "Power Down Enable" so that my computer wouldn't just get stuck booting. Before this version I think A40 was the best as I could restart and not have issues, other versions I would have to just come out of sleep, or shut down and then turn back on. Restarting used to almost always get stuck.

If I notice issues I'll try to update my post, but right now A7E is the best. I'm not doing benchmarking though, so I don't know if it gets the highest performance, but for me it is the most stable.

I'm glad it's stable. Did the system boot time change?

 

[DarkOmenX]

I'm glad it's stable. Did the system boot time change?

For me, it takes a bit longer than before, especially during the initial phase when the cursor starts blinking.

 

[dlomond]

I flash it last night and change it fch to 100 and will see if it do fix the PCI problem.will see when I get off work.

 

[sko8155902d1]

Solution: Manually set the FCH bus clock to 100.000 MHz and FCH Spread Spectrum back to Auto. Then the bus clock stays at correct 100 MHz and the system boots up without any problems every time. Hopefully it will help other users too.

Unfortunately, I have to correct my statement. After two days, I had the cold start bug again. At least with FCH Spectrum enabled or manual 100 MHz BCLK, the bug occurs less frequently. There's definitely something wrong with A70. I've now updated the BIOS to A7E. So far, the first four cold starts have been error-free. The BCLK is also behaving as usual. BCLK auto and FCH Spectrum auto = 100 MHz according to HWinfo (in CPU-Z 99.98 - 100.01 MHz correctly).