Saturday, June 20th 2026

AMD to Restore TSME (Memory Encryption) on Consumer Ryzen Processors After Backlash

AMD backpedaled on its recent decision to remove Transparent Secure Memory Encryption (TSME) feature on its consumer Ryzen processors, which was typically enabled by default. The company had, in April, quietly disabled the feature for Ryzen consumer processors through an AGESA update encapsulated into motherboard firmware updates. The idea probably was to restrict TSME to Ryzen commercial-segment processors, sold under the Ryzen PRO brand, besides Ryzen Threadripper WX workstation processors, and EPYC server processors. For reference, Intel offers a similar feature, TME-MK (Total Memory Encryption-Multi-Key) on even its consumer Core processors, and not just its commercial Core vPro processors.

AMD said that it has decided to restore TSME "based on valuable community feedback," and end-users should expect UEFI firmware updates that restore the feature starting next month (July 2026). TSME is a hardware-based feature that leverages fixed-function hardware on the silicon to encrypt the contents of the memory, providing an additional layer of security. Since AMD shares a common IOD design across its Ryzen consumer and Ryzen PRO commercial product lines, the feature is physically present.
In a statement to Tom's Hardware, AMD said:
AMD Memory Guard (Transparent Secure Memory Encryption, or TSME) is a hardware-based memory encryption technology available on our Ryzen PRO desktop and mobile processors where supported in silicon. It is a foundational security feature, and we have no plans to remove support from our Ryzen PRO lineup. This commitment holds now and in the future.

Regarding certain non-PRO Ryzen 9000-series desktop processors, a BIOS option to enable Memory Guard was previously available but was removed in a recent update. Based on valuable community feedback, we will reinstate this option in an upcoming BIOS release in July."
Source: Tom's Hardware
Add your own comment

90 Comments on AMD to Restore TSME (Memory Encryption) on Consumer Ryzen Processors After Backlash

#26
_roman_
the typical argument, is that CVE even abused in real life, does security matter etc.

When you use 5 or 10 year old software on all of your devices and all of your applications everywhere feel free to argue that way. If not, that just stop.
Posted on Reply
#27
RejZoR
"valuable community feedback" is weird way of saying that community was outraged over the sudden change...
Posted on Reply
#28
tabascosauz
GrimlingI actually was not even aware of this feature before, to be honest.
Was this something you needed to activate in bios, or was it on by default?
GrimlingInteresting. That means it does not have a negative impact on performance.
And that means, it should definitely be on, by default in general.
Good that they reverse this questionable decision!
On AM4, TSME used to demolish your memory latency numbers. Memory performance impact was definitely tangible, something on the order of +10ns. Dunno if AM5 is different.

I'd still rather have the feature present out of principle
Posted on Reply
#29
FR3D1
Ninja Cat
tabascosauzDunno if AM5 is different.
Here's the result; I did a quick comparison test:

Image 1 with TSME


Image 2 without TSME


That would have been too good to be true.

Regards,
Fred.
Posted on Reply
#30
Grimling
FR3D1That would have been too good to be true
well, you use the newest UEFI that do not support TSME.
But nice try
TumbleGeorgeThe right way is monopoly, biiiig oligopolies, cartels to be divided apart and stopped from their actions against consumers.

"Know What? You’re Starting To Remind Me Of Me."

Posted on Reply
#31
FR3D1
Ninja Cat
GrimlingBut nice try
But why am I able to configure that feature "explicitly"? I will double-check this tomorrow using a third-party tool and provide an update afterwards.

Regards,
Fred.
Posted on Reply
#32
Bigshrimp
FR3D1But why am I able to configure that feature "explicitly"? I will double-check this tomorrow using a third-party tool and provide an update afterwards.

Regards,
Fred.
Even if you configure it in the BIOS, it will not work. It was tested as non-functional even if enabled in the BIOS. The only way to re-enable it functionally is to downgrade to a BIOS version that it properly functioned on or wait for a new BIOS version that has its functionality re-enabled.
Posted on Reply
#33
TheinsanegamerN
Another rin a long series of examples that shows that bullying works, especially against AMD.
Posted on Reply
#34
oxrufiioxo
TheinsanegamerNAnother rin a long series of examples that shows that bullying works, especially against AMD.
The only recent time I can remember it working with Nvidia was the 4080 12GB lol... Maybe Nvidia originally planned to show DLSS5 at computex but swept it under the rug till it's in a much better place after it's warm reception. If it's a no show at gamescom I'll start to wonder what's going on with it.

Definitely super effective on AMD they do something almost yearly... Almost like they're doing it on purpose at this point.
Posted on Reply
#35
shuffy_losys
marios15TSME is pretty useless for consumer hardware anyway
enterprise/datacenters...sure i get it, but home users? lol
Tek-CheckIs that true for home users?
Was this a storm in a cup of tea or the feature is genuinely useful?
This was from a comment in Ars Technica, but this is useful in situations where a regular person is the target of someone adversarial. For example, police raiding your home and stealing your computer. And while sure, DRAM data does decay, it's not instantaneous and dunking the modules in liquid nitrogen can retain the data more or less indefinitely (as long as it's cold, of course). Granted, the chances of this specific scenario happening to you are next to zero (probably, I'm assuming you're a good citizen), but I imagine some people would find this useful.

At the end of the day, the bigger issue is AMD removing a feature and not telling anyone about it.
Posted on Reply
#36
Hecate91
Out of all the mega corporations who no longer listen or care about the consumer market, at least AMD still responds, really sad it has to be an outrage for them to listen though.
I never heard of this feature before the news, most home users probably have no need for TSME, but removing it without notice or any reason why is the problem.
oxrufiioxoThe only recent time I can remember it working with Nvidia was the 4080 12GB lol... Maybe Nvidia originally planned to show DLSS5 at computex but swept it under the rug till it's in a much better place after it's warm reception. If it's a no show at gamescom I'll start to wonder what's going on with it.

Definitely super effective on AMD they do somthing almost yearly... Almost like they're doing it on purpose at this point.
The bullying doesn't really work on Nvidia, all they did is lower the price by $100 and charge $800 for a x70Ti card, and if they listened or cared they'd be recalling the 12vhpwr connector.
I know that's a hot take for some people but honestly consumers need to be more critical of companies, or vote with your wallet.
Although all of these companies are testing to see what they can get off with, since the consumer market is no longer the focus.
Posted on Reply
#37
Phayzon
shuffy_losysFor example, police raiding your home and stealing your computer.
How does it even help in this scenario? Isn't this just taking the whole lock with the keys already in it, so to speak?
Posted on Reply
#38
shuffy_losys
PhayzonHow does it even help in this scenario? Isn't this just taking the whole lock with the keys already in it, so to speak?
No, it's more like the police arresting you and confiscating your phone, demanding your password to unlock it, and you can just conveniently forget it.

The encryption key is in the CPU, and it'd be extremely hard to extract it if it's even possible after power is cut, depending on how the key is generated and stored.
Posted on Reply
#39
Dr. Dro
FR3D1Here's the result; I did a quick comparison test:

Image 1 with TSME


Image 2 without TSME


That would have been too good to be true.

Regards,
Fred.
Bad test, you're testing with AGESA 1.3.0.1b, TSME is disabled on both even if you set it in BIOS. This version of AGESA will ignore whatever you set in BIOS because of AMD's restriction.

You need to downgrade to a BIOS with AGESA 1.2.0.3g or earlier series to restore functionality, 1.2.7.0 is also affected
Posted on Reply
#40
FR3D1
Ninja Cat
@Dr. Dro
I’ve been reading here and there that there is confirmation for "some" Ryzen 9000 chips only. However, firstly, this doesn't apply to all of them, and secondly, not to the Ryzen 7000 series—which is what I have.

Attached is a screenshot from HWInfo (focusing on SME!).


Additionally:
Whether TSME is actually active on a running system can currently only be verified under Linux—for instance, by querying the model-specific register MSR_AMD64_SYSCFG using the `rdmsr` command. AMD also provides another tool via GitHub.

It seems odd that verifying the operation of a feature that is nearly ten years old is only possible on Linux, given that it is also intended to enhance the security of business PCs running Windows.

AMD provides information on the various RAM encryption features on the AMD Secure Encrypted Virtualization (SEV) website.
heise(de)

In addition to Heise's quote, you could also use the "fwupdmgr security" command. Anyone currently using Linux—and a Ryzen 7000—could take care of this while they're at it.

Regards,
Fred.
Posted on Reply
#41
Dr. Dro
FR3D1@Dr. Dro
I’ve been reading here and there that there is confirmation for "some" Ryzen 9000 chips only. However, firstly, this doesn't apply to all of them, and secondly, not to the Ryzen 7000 series—which is what I have.

Attached is a screenshot from HWInfo (focusing on SME!).


Additionally:

heise(de)

In addition to Heise's quote, you could also use the "fwupdmgr security" command. Anyone currently using Linux—and a Ryzen 9000—could take care of this while they're at it.

Regards,
Fred.
Yeah, I know, but it should affect 7000 series all the same AFAIK. SME and TSME are slightly different, SME is OS managed, while TSME is firmware managed, encrypts whole memory transparently and this what was disabled.

The same is TME/TME-MK on Intel, but where AMD and Intel differ is that Intel has explicitly supported both on client, business and workstation parts, while AMD just kind of left it enabled on Ryzen and never bothered disabling it, until now. Except that remotely disabling security features purely to enforce market segmentation makes them look like complete and total arses, hence the U turn on it.
Posted on Reply
#42
cvaldes
RejZoR"valuable community feedback" is weird way of saying that community was outraged over the sudden change...
Dude, that is called corpspeak. Nothing new there, humans have been probably doing this since the first company was created. Use clever language to obfuscate a negative situation. Governments do this ALL THE FREAKIN' TIME.

Anyhow, AMD is very thick in the head. Who the hell is going to interpret a step toward less security in 2026 as a good thing?

At least they backtracked quickly rather than let this spiral out of control for weeks/months. Some other companies are not so bright and would just let this thing fester until their bottom lines are negatively impacted (*cough* Sonos *cough*).
Posted on Reply
#43
tabascosauz
FR3D1@Dr. Dro
I’ve been reading here and there that there is confirmation for "some" Ryzen 9000 chips only. However, firstly, this doesn't apply to all of them, and secondly, not to the Ryzen 7000 series—which is what I have.

Attached is a screenshot from HWInfo (focusing on SME!).


Additionally:

heise(de)

In addition to Heise's quote, you could also use the "fwupdmgr security" command. Anyone currently using Linux—and a Ryzen 9000—could take care of this while they're at it.

Regards,
Fred.
I would expect a reduced impact to be a few ns less, not exactly same latency. The variances in bandwidth are normal run to run variance. Exact same result suggests TSME is either both on or both off.

if an AGESA ver. is bugged, it's bugged, I don't see why TSME toggle would only affect Ryzen 9000.

I'm not sure if you meant "too good to be true" for same latency on/off, or with respect to what I said about the latency penalty.
Posted on Reply
#44
Dr. Dro
tabascosauzI would expect a reduced impact to be a few ns less, not exactly same latency. The variances in bandwidth are normal run to run variance. Exact same result suggests TSME is either both on or both off.

if an AGESA ver. is bugged, it's bugged, I don't see why TSME toggle would only affect Ryzen 9000.
It wasn't bugged, this was actually intentional which is why it'd 100% affect all architectures managed by the affected AGESA versions

It was done on the down low, without communication, hoping nobody would notice or care
Posted on Reply
#45
Tomorrow
I tested TSME on my AM4. AGESA 1.2.0.F from October 2025. There's a newer 1.2.0.11 version released April this year, but i haven't updated to it yet.
Since this is a consumer desktop PC i always keep TSME disabled.

No TSME:
With TSME:
Posted on Reply
#46
Dr. Dro
TomorrowI tested TSME on my AM4. AGESA 1.2.0.F from October 2025. There's a newer 1.2.0.11 version released April this year, but i haven't updated to it yet.
Since this is a consumer desktop PC i always keep TSME disabled.

No TSME:
With TSME:
I wonder how does it impact Raptor Lake. I'll have to test it sometime. I'm fairly sure I've always had it enabled, although TME is strictly unsupported when HT is enabled. Not sure about TME-MK.
Posted on Reply
#47
Particle
TSME protects against more than just physical cold boot attacks. Since the memory contents is encrypted at rest, it can make a whole host of side channel attacks like rowhammer and similar more difficult. It's absolutely worth doing. Don't get too caught up on it not being a silver bullet. No single security measure is. It would be like arguing against KASLR or NX. Does either one make your PC invulnerable to attack? No, but they each add a layer of difficulty, and that makes them valuable.
Posted on Reply
#48
kimbentsen
What arrogance from AMD. You paid for a feature, and then they silently remove it. Someone would call that fraud.
Posted on Reply
#49
R-T-B
dgianstefaniIt can have a mild impact I'd imagine, but negligible in most cases. It is a hardware feature after all.
It adds a few ms to memory latency. Nearly all concerned with just performance shut it off, as such. So this is kind of a storm in a teacup tbh.
Particlerowhammer and similar more difficult.
Since full-on ECC is pretty ineffective against rowhammer, I'd be outright shocked if this does anything against it. But yes it does have use cases.
Posted on Reply
#50
TumbleGeorge
FR3D1Here's the result; I did a quick comparison test:

Image 1 with TSME


Image 2 without TSME


That would have been too good to be true.

Regards,
Fred.
Difference is because in tests CPU working on little different frequency in test 1 vs 2. I dint think that TSME is reason.
Posted on Reply
Add your own comment
Nov 11th, 2026 23:50 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts