MSI x870 Tomahawk EXPO not working

[derekfoster1971262029b]

Hi all, I recently bought the x870 Tomahawk with Ryzen 9950x and a 64 GB Ram Kit (2x 32gb sticks): Corsair cmk64gx5m2b6000z30. I am on the latest A16 beta bios and whenever I enable the EXPO setting the board fails to boot with A6 motherboard code - the RAM is rated at CL30 6000MHz. I have tried many different settings (EXPO on/off while manually setting 6000Mhz and lower speeds etc.) but nothing works. Also had MCR and Power down off and on but the system will only boot with the RAM at its default 4800 MHz CL40 speeds.
Any suggestions on how to get the RAM working at advertised speeds on this board would be greatly appreciated.
Thanks!

 

[siteregform156f02db]

not true, hypervisor disabled (core isolation disabled) improves latency
It's AIDA that has broken latency measurement, don't trust that software

I see. What software I can use to measure latency and rest of things? By way what number do you see in aida benchmarks. Also could you please share your tweaks as I see you also using DDR 5 6000.Even I should not copy them it can be good start for me to learn.

 

[Michele P]

first, disable core isolation in windows defender, on a gaming pc must be disabled anyway.
If you don't use the pc for gaming and need extra safety, reactivate it after your tests

You want to measure latency? I suggest a quick and easy test
pyprime-2b (download the benchmate suite)

skryabin`s 7 sec 590 ms score: PYPrime - 2b with BenchMate with aDDR5 SDRAM

The DDR5 SDRAM @ 3100.2MHzscores getScoreFormatted in the PYPrime - 2b with BenchMate benchmark. skryabinranks #null worldwide and #null in the hardware class. Find out more at HWBOT.

hwbot.org

it's an 8 seconds test, it should give reasonable constant results (if your system is stable and perfectly tuned) and it's pretty sensitive to system latency
lower (faster) number is better ofc

for the rest aida can be good, except latency
also y-cruncher another very good tool for benchmarking the ram performance/stability (inside the benchmate suite)


You can find few screenshots in the following thread, scroll the pages back and you will find some https://forum-en.msi.com/index.php?threads/x670e-ace-bios-problems.381303/page-76#posts
but don't copy them, I have different cpu/mb/ram

 

[siteregform156f02db]

first, disable core isolation in windows defender, on a gaming pc must be disabled anyway.
If you don't use the pc for gaming and need extra safety, reactivate it after your tests

You want to measure latency? I suggest a quick and easy test
pyprime-2b

skryabin`s 7 sec 590 ms score: PYPrime - 2b with BenchMate with aDDR5 SDRAM

The DDR5 SDRAM @ 3100.2MHzscores getScoreFormatted in the PYPrime - 2b with BenchMate benchmark. skryabinranks #null worldwide and #null in the hardware class. Find out more at HWBOT.

hwbot.org

it's an 8 seconds test, it should give reasonable constant results (if your system is stable and perfectly tuned) and it's pretty sensitive to system latency
for the rest aida can be quite good
also y-cruncher another very good tool for benchmarking the ram performance/stability

You can find some screenshots of my settings in this thread, but don't copy them, I have different cpu/mb/ram https://forum-en.msi.com/index.php?threads/x670e-ace-bios-problems.381303/page-76#posts

Thank you for reply. I will neatly digest that information and come back to you. I appreciate your time and patience.

 

[Ri2dLe]

I would not recommend you to turn off security features. Anyway, I watch the screenshots and you have DDR6000.

Hello dear,

I would like ask question if you dont mind. I currently build a system with x870 tomahawk and 9800x3d. My ram is Teamgroup T Force DDR 5 32GB 6000 MHZ(30-36-36-76) . It is in QVL list. I dont have any issue with turning on EXPO profile which gives me same 6000 at this timings. But the problem is, aida64 benchmark shows weird numbers. Like particularly, the read and copy numbers are not matching the write. They looks way off and to be honest low. My old system DDR4 at 4000 was giving same numbers.((. Also latency number is higher than expected from what people say. In BIOS i did not fine tune anything to be honest. I just made AMD EXPO profile 1 enabled, DRAM speed =Auto, UCLK DIV1 MODE - UCLK=MEMCLK. Sticks are installed to A2 B2 . Bios updated to latest one. Windows 11 latest from website, freshly installed and updated. Is this normal numbers I see or I need to fine tune something over there? I am new to AMD platform. So please don't beat me hard. )) I attached all pictures, please take a look if you can. Any advise is appreciated and thank you in advance.

I will try, not been hard, but don't promise anything. The speed is too high on the voltage you supply on the sticks. Could be... the old bios that is teasing you.
Set the next values from auto to the filled in value(s)... Leave the expo profile enabled.
DRAMSPEED to 6000
DRAM 1.350v
VDDQ 1.350v
VDDIO 1.350v
That's strange! Wondering what an A-XMP is doing in an AMD platform. Are you sure you aren't mixing up two pc's?
If you can't find Infinity Fabric Frequency and Devider, then leave the UCLK DIV1 Mode to auto. I see you did an OC on the core... Maybe you went too low.

 

[siteregform156f02db]

I would not recommend you to turn off security features. Anyway, I watch the screenshots and you have DDR6000.

I will try, not been hard, but don't promise anything. The speed is too high on the voltage you supply on the sticks. Could be... the old bios that is teasing you.
Set the next values from auto to the filled in value(s)... Leave the expo profile enabled.
DRAMSPEED to 6000
DRAM 1.350v
VDDQ 1.350v
VDDIO 1.350v
That's strange! Wondering what an A-XMP is doing in an AMD platform. Are you sure you aren't mixing up two pc's?
If you can't find Infinity Fabric Frequency and Devider, then leave the UCLK DIV1 Mode to auto. I see you did an OC on the core... Maybe you went too low.

Dear Ri2dLe,

Thank you for reply. No I am not, it is one system. I have no idea what A-XMP doing there when we have EXPO. Anyway. I set the values as you told me to. The pictures attached. In first attempt I set values (VDDIO ctrl I did not touch it set by itself to separate mode) , tested, then went back and set UCLK DIV1 Mode to auto. Then again tested. Nothing changing. It feels like whatever I do ram is not responding to that. Regarding OC on core, I did not touch anything. Maybe I set x3d gaming mode on. That maybe. What else we can do?

 

[Michele P]

By the way what number do you see in aida benchmarks

nothing exceptional, I use another antivirus instead of microsoft defender (eset), but before I was using Microsoft Defender with core isolation disabled

I would not recommend you to turn off security features. Anyway, I watch the screenshots and you have DDR6000.

Core isolation it's not a critical component for windows defender, but it adds a layer of virtualization to everything for extra safety but with increased latency
If the pc is for gaming it's advisable to keep it disabled.
If you need the extra security because you think the pc may be exposed to risks, enable again the feature after you are done with your tests (but keep in mind it's just a deterrent that modern malware can bypass anyway)

Set the next values from auto to the filled in value(s)... Leave the expo profile enabled.
DRAMSPEED to 6000
DRAM 1.350v
VDDQ 1.350v
VDDIO 1.350v
That's strange! Wondering what an A-XMP is doing in an AMD platform. Are you sure you aren't mixing up two pc's?

You can also use xmp profiles on amd boards, nothing strange, but in this case he is using expo
Are all the voltages on Auto as it seems? Or did you put manual settings?

From the screens I suggest to change a couple things:
- disable the x3d game mode
- disable memory context restore
- and in the memory advanced options you should also find Power Down Enable, set to disable

All things that can improve the AIDA latency score for sure, for what it counts... because, as I said before, that test has very poor attendibility
If you want to see how much a settings can affect latency see the difference in pyprime time.

 

[Ri2dLe]

Turn the x3d gaming mode off. What are the strange values you are talking about.

Core isolation it's not a critical component for windows defender, but it adds a layer of virtualization to everything (with increased latency)
If the pc is for gaming it's advisable to keep it disabled.
If you need the extra security because you think the pc may be exposed to risks, enable again the feature after you are done with your tests (but keep in mind it's just a deterrent that modern malware can bypass anyway)

Except a hacker can get your password out of your memory.

 

[Michele P]

Except a hacker can get your password out of your memory.

they can extract that even with core isolation enabled
if you don't know there's also another layer of hardware safety, called memory data scramble
but everything is bypassable, everything, so.... it's up to you, if you game a lot and don't access bank accounts too often just disable (use your pc with a working brain): you are not disabling defender entirely, just a single module.

but if you are scared, enable again core isolation after your tests or, maybe better, swap microsoft defender for something else

I don't see strange values, maybe he is using different voltages compared to previous amd cpu series


Everything is on auto right? don't touch anything except the x3d gaming mode, memory context restore, and power down (everything forced on disabled), and you should start see improvements in latency
MCR/PD are options intended for slightly faster training but at the cost of increased latency, and risks of instabilities if you are unlucky

 

[Ri2dLe]

they can extract that even with core isolation enabled
if you don't know there's also another layer of hardware safety, called memory data scramble
but everything is bypassable, everything, so.... it's up to you, if you game a lot and don't access bank accounts too often just disable
but if you are scared, enable again core isolation after your tests or, maybe better, swap microsoft defender for something else

I don't see strange values, the other user saw strange things about xmp, and maybe different voltages compared to previous cpu series
You have everything on auto right? don't touch anything except the x3d gaming mode, memory context restore, and power down (everything forced on disabled), and you should start see improvements in latency

This is strange, you're setting it up for gaming, and you set the voltages of the memory to auto.

Maybe they can... What did you mean by not critical? Windows can block devices been connected when they don't admit the security requirements within the same domain. Don't undervest me wrong, rather like to have a firewall that I can turn on and off when ever I want.

 

[Michele P]

This is strange, you're setting it up for gaming, and you set the voltages of the memory to auto.

I have all my voltages on manual, but in the case of the user having problems it seems he has everything on auto
And he should start from there, the default expo profile is always a good starting point.

The main windows security requirements are probably others: TPM, Secure Boot, ecc
in fact, if you disable TPM for example, you stop receiving windows updates

But Core Isolation I'm pretty sure windows doesn't care at all: it may display a yellow warning from time to time that you can disable too, but it's not mandatory afaik
PS: now that I think better, Windows 11 supports CPUs that lack virtualization capabilities, resulting in the unavailability of core isolation anyway

 

[siteregform156f02db]

I have all my voltages on manual, but in the case of the user having problems it seems he has everything on auto
And he should start from there, the default expo profile is always a good starting point.

The main windows security requirements are probably others: TPM, Secure Boot, ecc
in fact, if you disable TPM for example, you stop receiving windows updates

But Core Isolation I'm pretty sure windows doesn't care at all: it may display a yellow warning from time to time that you can disable too, but it's not mandatory afaik
PS: now that I think better, Windows 11 supports CPUs that lack virtualization capabilities, resulting in the unavailability of core isolation anyway

Hi

nothing exceptional, I use another antivirus instead of microsoft defender (eset), but before I was using Microsoft Defender with core isolation disabled

Core isolation it's not a critical component for windows defender, but it adds a layer of virtualization to everything for extra safety but with increased latency
If the pc is for gaming it's advisable to keep it disabled.
If you need the extra security because you think the pc may be exposed to risks, enable again the feature after you are done with your tests (but keep in mind it's just a deterrent that modern malware can bypass anyway)



You can also use xmp profiles on amd boards, nothing strange, but in this case he is using expo
Are all the voltages on Auto as it seems? Or did you put manual settings?

From the screens I suggest to change a couple things:
- disable the x3d game mode
- disable memory context restore
- and in the memory advanced options you should also find Power Down Enable, set to disable

All things that can improve the AIDA latency score for sure, for what it counts... because, as I said before, that test has very poor attendibility
If you want to see how much a settings can affect latency see the difference in pyprime time.

Hi,

I changed the values mentioned by you. Latency in pyprime dropped from 8.9 to 8.5 in Aida for 1 ms. But copy and read values are the same. I just wonder what is the chance I have got bad ram sticks? I mean, this scenario is possible that ram is bad for copy and read but good for write values? Sound fantastic for me, but it is only for me, as I am not a pro. What would you say about it?

 

[Michele P]

Nah, it’s probably how this new cpu works. Don’t compare your results with different systems
When look at your results compare them with people using same CPU, and at least similar ram settings (ram clock and primary timings)
for example look in the following link, the first AIDA screenshot on this 9800x3d cpu review

AMD Ryzen 7 9800X3D Review - The Best Gaming Processor

The Ryzen 7 9800X3D establishes AMD as the leader in gaming performance. This Zen 5-based X3D chip is not only fast, it also comes with full support for overclocking. Besides gaming, application performance is considerably improved over the 7800X3D, but that comes at a price.

www.techpowerup.com

it's not very different from your screenshot

Half a second improvement in pyprime is huge!
Pls again, don’t trust aida for latency… you can eadily achieve better scores even with ddr4, they way this software measure latency is somewhat arguable.
But Pyprime from 8.9 to 8.5s means your memory latency has improved a lot with the changes you have made, you can be 100% confident about that.

Latency is one thing, copy/read perfomance is another, they are not necessarily linked togheter
Regarding memory bandwidth it’s a combination of cpu and memory timings.
It is what is when you use stock expo values and stock cpu settings (without PBO/OC)

If you want more bandwidth you need to tweak your memory timings or OC your CPU
Typically lower timings should give better results, but it’s not a golden rule.
Not because the ram bad but because ram is a complex game of delays. If the ram is bad, you should have noticed instabilities already.
For example (again, don't copy), the following 2 little timings can impact significantly the bandwidth. Any value lower than 8/12 will result in a bandwidth penalty (it's not uncommon to see MB do that, or people copy/pasting settings from other "overclockers")

 

[siteregform156f02db]

Nah, it’s probably how the cpu works. Don’t compare with different systems

Half a second improvement in pyprime is huge!
Pls again, don’t trust aida for latency… you can eadily achieve better scores even with ddr4, the way it measures latency is arguable.

Latency is one thing, copy/read perfomance is another. Regarding memory bandwidth it’s a combination of cpu and memory timings. It is what is when you use stock expo values, if you want more you need manual timings optimizations, tupically lower values should give better results but it’s not a golden rule. Not because this specific ram is bad but because ram is a complex game of delays

pls compare with people using same CPU, and similar ram settings (ram clock and primary timings)
for example look here, first AIDA screenshot on this 9800x3d cpu review

AMD Ryzen 7 9800X3D Review - The Best Gaming Processor

The Ryzen 7 9800X3D establishes AMD as the leader in gaming performance. This Zen 5-based X3D chip is not only fast, it also comes with full support for overclocking. Besides gaming, application performance is considerably improved over the 7800X3D, but that comes at a price.

www.techpowerup.com

it's not very different from your screenshot

Thank you. This thread you shared give me hope.)) So , seems to be read and copy data is normal for these type cpus. I will fine tune ram further to increase my values then. One more question if dont mind. What is affecting 1 % low FPS. It seems I have huge difference between average and 1 % low. Like average is 144 (I cap it through Nvidia control panel) and 1% low is 70-80. My GPU is 4090.

 

[Michele P]

For 1% lows? In general it's dependent on how much clean your system is.
Keep your windows debloated first.
Again, disable core isolation or use another better light antivirus (if you need one, because antivirus will always be the main offender in this scenario)
Avoid any hardware that requires always running software: kb, mouse, fans with rgb/fan control software, those are cancer (iCUE one of the worst for example)
Do not "over-tune" your system, following online guides
Don't use any monitoring software while testing/gaming (no Afterburner, its power monitoring is always bugged and cause of microstutter)
Then hope for a good gpu driver, this in particular.
Don't overthink, just keep the system clean and avoid problematic drivers, you should be ok.

EDIT: almost forgot the windows power plan
Usually, AMD dual CCD cpus perform better with a balanced power plan instead of the ultimate/high performance one (many online guides are not clear about this)
Do your tests, with a single ccd maybe the ultimate/high performance is better, but don't be blind, ok?
Try the balanced plan or the ultimate/high performance... see which one is better for you
You can also make a couple tweaks on the balanced plan, like selective usb suspend disabled and PCIe power management inactive

 

[siteregform156f02db]

For 1% lows? In general it's dependent on how much clean your system is.
Keep your windows debloated first.
Again, disable core isolation or use another better light antivirus (if you need one, because antivirus will always be the main offender in this scenario)
Avoid any hardware that requires always running software: kb, mouse, fans with rgb/fan control software, those are cancer (iCUE one of the worst for example)
Do not "over-tune" your system, following online guides
Don't use any monitoring software while testing/gaming (no Afterburner, its power monitoring is always bugged and cause of microstutter)
Then hope for a good gpu driver, this in particular.
Don't overthink, just keep the system clean and avoid problematic drivers, you should be ok.

EDIT: almost forgot the windows power plan
Usually, AMD dual CCD cpus perform better with a balanced power plan instead of the ultimate/high performance one (many online guides are not clear about this)
Do your tests, with a single ccd maybe the ultimate/high performance is better, but don't be blind, ok?
Try the balanced plan or the ultimate/high performance... see which one is better for you
You can also make a couple tweaks on the balanced plan, like selective usb suspend disabled and PCIe power management inactive

Appreciate. Will try and report

 

[nun15b602e0]

Same problem here.

After enabling the EXPO profile, my computer fails to boot. It displays the "C5" debug code for a few seconds, followed by the "Ab" debug code. The yellow debug LED (DRAM) also turns on. If I turn the computer off and on again, the same thing happens, but this time, both the red debug LED (CPU) and the yellow debug LED (DRAM) turn on.

I'm using the following components and setup in my build:

Motherboard: MSI MAG X870 TOMAHAWK WIFI (with the latest BIOS and default settings)
BIOS version: E7E51AMSI.1A24
BIOS build date: Jan 20, 2025
CPU: AMD Ryzen 7 9800X3D
Memory kit: CORSAIR VENGEANCE 32GB (2x16GB) DDR5 DRAM 6000MT/s CL28 (AMD EXPO & Intel XMP)
Memory SKU: CMK32GX5M2B6000Z28
Memory installation: 1 stick in A2, 1 stick in B2

Can MSI fix this issue with a BIOS update, or is collaboration with CORSAIR and/or AMD required to resolve the problem?

I also suggest that folks report this issue to MSI, CORSAIR, and AMD. The more reports of these issues they have, the more pressure they have to fix it.

https://account.msi.com

https://help.corsair.com

https://www.amd.com/en/forms/contact-us/support.html

 

[nun15b602e0]

For your information, on March 5th, 2025, MSI released a new BIOS, but the issue remains. It still fails to boot after enabling EXPO.

BIOS version: E7E51AMSI.1A30
BIOS build date: Feb 20, 2025

 

[Ri2dLe]

Please add the numbrification of your sticks on that comment.

 

[hansen.ryan157302e1]

I have the RGB version of the dreaded corsair ram (CMH64GX5M2M6000Z30) in the x870e. I could not get it to work with EXPO at 128gb (A6 eror), but can confirm it does in fact work with 64gb on my machine and passes memtest with no errors.

 

[Frank_BoneBreaker]

Hi everyone,
I’d like to share my experience in getting the Corsair Vengeance DDR5-6000 Kit (CMK32GX5M2B6000Z30, version 5.43.13 – SK Hynix chips) stable on the MSI X870 Tomahawk motherboard with BIOS version 7E51v1A44 (AGESA PI-1.2.0.3c).
My system specs:
Motherboard: MSI X870 Tomahawk WiFi / CPU: AMD Ryzen 9 7900X (12C/24T) / Cooler: Dark Rock Pro 5
RAM: Corsair CMK32GX5M2B6000Z30 (2x16 GB, version 5.43.13, Hynix)
GPU: Sapphire Pulse RX 6750 XT / OS SSD: Samsung 990 PRO 1TB NVMe / PSU: be quiet! Pure Power 12 M 850W

Goal: DDR5-6000 Stable Profile (Manual OC, not EXPO)
The EXPO profile did not work reliably for me – it caused cold boot issues or instability.
After some tuning and testing, I managed to get a fully stable manual configuration at DDR5-6000 (3000 MHz).

Key settings in BIOS (the mentioned timings are manually entered and not left on AUTO)
Memory Settings:
Memory Frequency: DDR5-6000 (3000 MHz)
UCLK = MEMCLK (1:1 ratio)
Gear Down Mode: Enabled
Power Down Enable: Enabled
Memory Context Restore (MCR): Enabled

Primary timings:
tCL / tRCDRD / tRCDWR / tRP: 44 / 44 / 44 / 44
tRAS / tRC: 84/ 128
Command Rate: 1T
Other RAM-related settings:
tREFI: 15600
tRFC: 820
Secondary/tertiary timings (left on AUTO)

Voltages (manual setting):
DRAM VDD & VDDQ: 1.28V / CPU SoC: 1.26V / CPU VDDIO: 1.25V

Stability Testing:
MemTest86: 4 full passes – 0 errors
OCCT (RAM test): 10 min, no errors
RAM temp during stress tests: 47–50°C

Remarks:
Cold and warm boots are now fast and reliable (no more code "Ab" or long POST delays).
This Corsair kit seems picky with EXPO, but with manual tuning, it works perfectly on MSI boards.
Hope this helps someone struggling with similar issues!
Cheers!