Friday, April 3rd 2026

New Rowhammer Attack Puts GDDR6-Powered NVIDIA GPUs at Risk

Rowhammer attacks exploit vulnerabilities in DRAM hardware by causing targeted bit-flips, allowing attackers to bypass memory isolation and gain control over a device. These attacks were initially focused on CPUs and their associated DDR memory, such as DDR4. However, recent research indicates that NVIDIA GPUs are also vulnerable due to the fragile nature of the GDDR6 memory they use, which directly compromises the CPU host. Two independent research teams have discovered ways to exploit this decades-old memory vulnerability against modern graphics hardware, with trouble extending beyond the GPU itself. The "GDDRHammer" and "GeForge" groups have each developed functional exploits that use Rowhammer bit-flips in NVIDIA GPUs with GDDR6 memory to gain complete control over the host CPU's memory. This attack can perform bit-flips on some NVIDIA GPU models, ranging from the "Ampere" to "Ada Lovelace" families of cards.

An attacker who succeeds in this can read and write anything stored in the machine's main memory. Both teams have also introduced new Rowhammer techniques specifically designed for GPU architecture, achieving a significantly higher rate of bit-flips on GDDR6 memory than previous methods. The critical step in both exploit chains involves targeting the GPU's memory allocator, using controlled bit-flips to corrupt the GPU's page tables. Once these page tables are compromised, the attacker gains arbitrary read and write access to CPU memory, breaking down the security boundary between the graphics subsystem and the rest of the machine. The end result is a full system compromise, as the attacker can manipulate memory at will and gain root access, achieving total control without interacting with privileged software paths. The affected GPUs include the GeForce RTX 3060, which experienced 1,171 bit-flips, and the RTX 6000 "Ada" GPU, which saw 202 bit-flips from the attack.
More GPUs are being tested, and the official website claims 25 NVIDIA GPUs were tested, with only a few showing signs of vulnerability. However, two mitigations exist. Enabling IOMMU through the BIOS closes the primary attack path by restricting which memory regions the GPU can access on the host system. This technology handles the translation of device-visible virtual addresses to physical host memory addresses and can completely fence off sensitive memory from peripheral devices. Another option is activating Error Correcting Codes on the GPU, which NVIDIA exposes through a command-line setting. However, enabling it reduces the pool of usable GPU memory and adds processing overhead, resulting in a performance penalty. Interestingly, no GPU with GDDR6X and GDDR7 memory is vulnerable, as the exploit only works with GDDR6.
Sources: GDDR.Fail Website, Ars Technica, VideoCardz
Add your own comment

35 Comments on New Rowhammer Attack Puts GDDR6-Powered NVIDIA GPUs at Risk

#1
Solaris17
Super Dainty Moderator
AleksandarKHowever, two mitigations exist. Enabling IOMMU through the BIOS closes the primary attack path by restricting which memory regions the GPU can access on the host system.
In the server world regardless of what manufacturer you are using we enable iommu by default. Shit I do it on my home machines. If you want to do any kind of serious virtualization or in the server space resource sharing it’s a must.
Posted on Reply
#2
JIWIL
Forgive my ignorance, but is this one of those "your system already needs to be compromised for this to compromise it" scenario?
Posted on Reply
#3
kondamin
and what are the practical implications for owners of such a card? is it imperative that I immediately go and turn on IOMMU or is it such a etherial thing that it doesn't matter
Posted on Reply
#4
phanbuey
they can illegaly fetch my compressed textures.
Posted on Reply
#5
Panther_Seraphin
JIWILForgive my ignorance, but is this one of those "your system already needs to be compromised for this to compromise it" scenario?
Nearly all of these attacks requires some point of compromise. the main thing is that it allows things to cross expected barriers into areas that a single compromise shouldnt allow.

So in theory those that run Honeypot VMs for shit'n'giggles with one of these GPUs shared to the honeypot may indirectly lead to complete compromise of the host system and every other VM hosted on it. This is also the cause of concern for hosting platforms that are using these GPUs for Plex hosting etc, leading to the compromise of multiple customers etc.

To everyone else out there? Near 0 realistic impact because if the system is already compromised to do this then there are simpler avenues to gather most of the information they want.
Posted on Reply
#6
TheDeeGee
I mean, whatever.

PCs in general are a swiss cheese of security issues, no doubt thousands which arn't known yet.
Posted on Reply
#7
Hyderz
That is a lot of gpus with gddr6 going back to even gtx16 series… also does it affect AMD ones with gddr6 as well?

If I recall intel also has gpu that uses gddr6 as well
Posted on Reply
#8
LastDudeALive
AleksandarKThe end result is a full system compromise, as the attacker can manipulate memory at will and gain root access, achieving total control without interacting with privileged software paths. The affected GPUs include the GeForce RTX 3060, which experienced 1,171 bit-flips, and the RTX 6000 "Ada" GPU, which saw 202 bit-flips from the attack.
A bit confused as to what these bit-flips numbers indicate. The article says that the attack ultimately gives the attacker full control over CPU host memory. Are these the number of bit-flips necessary to achieve that result? Or is this just the total number of bit-flips that the attackers were able to achieve?
Posted on Reply
#9
Sol_Badguy
Thank Go... I mean Jacket Man for GDDR7. :respect:
All hail the unconquerable Blackwell!
Get ready for price increases for these amazing, and now extra secure GPUs.
Har har!
Posted on Reply
#10
AleksandarK
News Editor
LastDudeALiveA bit confused as to what these bit-flips numbers indicate. The article says that the attack ultimately gives the attacker full control over CPU host memory. Are these the number of bit-flips necessary to achieve that result? Or is this just the total number of bit-flips that the attackers were able to achieve?
Its the number of bit flips required to get the rowhammer working. Past attacks managed a few bitflips which the GPU irons out, but more bit flips is something a GPU cant handle on its own
Posted on Reply
#11
Athena
Solaris17In the server world regardless of what manufacturer you are using we enable iommu by default. Shit I do it on my home machines. If you want to do any kind of serious virtualization or in the server space resource sharing it’s a must.
that is what I was wondering, who in their right mind running servers wouldn't have IOMMU enabled?

I don't see any point in testing any other GPU, since this whole thing is flawed by not having IOMMU enabled
Posted on Reply
#12
MTS-C
AleksandarKways to exploit this decades-old memory vulnerability against modern graphics hardware, with trouble extending beyond the GPU itself.
Did you mean with or without?
Solaris17In the server world regardless of what manufacturer you are using we enable iommu by default. Shit I do it on my home machines. If you want to do any kind of serious virtualization or in the server space resource sharing it’s a must.
I do the same at my house.
Posted on Reply
#13
swaaye
My only interest really is just what performance hit will this have when addressed in drivers. So many exploits in recent years have helped make hardware slower.
Posted on Reply
#14
_roman_
as a consequence maybe
An attacker who succeeds in this can read and write anything stored in the machine's main memory.
still a graphic card uses its own memory.

more an issue with the operating system not keeping software isolated. we talk here about a graphic card which has a vulnerability.

well reading further
Enabling IOMMU through the BIOS closes the primary attack path by restricting which memory regions the GPU can access on the host system
the text contradicts itself. does not change i do not agree with
An attacker who succeeds in this can read and write anything stored in the machine's main memory.
root cause. not dram but graphic card dram which is not directly the cpu dram.

iommu is most likely also a bad fix like secureboot it is. well many will believe it will fix the issue.

so there are no checksums calculated and verified by default
Another option is activating Error Correcting Codes on the GPU, which NVIDIA exposes through a command-line setting.
another bad fix.

prevent running such code should be the prime directive. is the issue closed source binaries or not audited code for such attacks?

-- edit feel free to share your insights. was a gpu ever meant to give any data back? I see ECC, error correction graphic card memory to be specified. hopefully not the fake EEC DDR5 one but a proper one.
Posted on Reply
#15
LastDudeALive
_roman_as a consequence maybe



still a graphic card uses its own memory.

more an issue with the operating system not keeping software isolated. we talk here about a graphic card which has a vulnerability.

well reading further



the text contradicts itself. does not change i do not agree with

root cause. not dram but graphic card dram which is not directly the cpu dram.

iommu is most likely also a bad fix like secureboot it is. well many will believe it will fix the issue.

so there are no checksums calculated and verified by default

another bad fix.

prevent running such code should be the prime directive. is the issue closed source binaries or not audited code for such attacks?

-- edit feel free to share your insights. was a gpu ever meant to give any data back? I see ECC, error correction graphic card memory to be specified. hopefully not the fake EEC DDR5 one but a proper one.
I don't think you understand the attack.

Rowhammer works by manipulating ram cells until it causes nearby bits to flip when they're not supposed to. If you do this on the page table, you move code anywhere you want by remapping memory. You can load malicious code into VRAM, and then manipulate the page table to insert that code into the memory of a legitimate process, like a Windows system process.

IOMMU prevents that by providing a filter between the GPU and CPU page tables, preventing the compromised GPU from mapping the malicious code into the system memory.

ECC prevents the attack by catching and fixing the malicious bit flips that corrupt the page tables in the first place.
Posted on Reply
#16
SiliconSeraph
Maybe I'm a bonehead, but will GDDR7 be unaffected? This seems to be isolated to GDDR6 but maybe I missed something.
Posted on Reply
#17
Wirko
I would be surprised if anyone was running pro applications on a pro card such as the RTX 6000 with ECC off. How much does ECC affect memory bandwidth and available memory in GPUs?
Posted on Reply
#18
mekigos773
Solaris17In the server world regardless of what manufacturer you are using we enable iommu by default. Shit I do it on my home machines. If you want to do any kind of serious virtualization or in the server space resource sharing it’s a must.
There's a new attack disclosed today, GPUBreach that bypasses even IOMMU. So IOMMU is not a sufficient defense.
Posted on Reply
#19
Panther_Seraphin
SiliconSeraphMaybe I'm a bonehead, but will GDDR7 be unaffected? This seems to be isolated to GDDR6 but maybe I missed something.
"GDDR7 adds on-die error correction code, error checking and scrubbing features for chip reliability,"

Similar to DDR5 does vs DDR4 so should mitigate a lot of these attack vectors initially.
mekigos773There's a new attack disclosed today, GPUBreach that bypasses even IOMMU. So IOMMU is not a sufficient defense.
"memory-safety bugs in the NVIDIA driver"
Posted on Reply
#20
Solaris17
Super Dainty Moderator
mekigos773There's a new attack disclosed today, GPUBreach that bypasses even IOMMU. So IOMMU is not a sufficient defense.
Too many everyday users think these issues could affect them because some researcher found a vlun.

It's great that they find them, because you should always build with the smallest piece in mind, but too often it ends in sensationalist threads and random links to blog posts like above.


IOMMU
AMD SME
Intel TME
ECC
TRR
RAM encryption
RAM scrubbing

Thats just the bios settings.

BIOS signature checking
Secure boot
Hardened images
Stripped kernels
User isolation
Service tamper guards

to name a few

Intel SGX
KPTI
TPM

Thats some hardware level stuff.

All of this is used. Literally. That is just the tip of the iceberg, then you get into hypervisor isolation, virtual ram addressing for isolated workloads, a bunch of other software abstration and protection.

Users have absolutely no idea. They never look at this stuff. For some reason people think that what? AWS, Meta? Just install Ubuntu?

No.

From the hardware UP every security checkbox is checked, we flash bios with this stuff pre-configured, software checks during OS install, countless steps in the security chain from boot to workload.

The majority of all major outages and hacks in decades were not hardware chain of custody related, they were breaches of the custom software stack.

I am talking about the DCs I have worked in. A single building 1million you read that right 1 million and up servers per building, all of that surface area no breach yet.

These things should be known, so we can make better architectual decisions for the next iteration, but they should not scare you.
Posted on Reply
#21
R-T-B
TheDeeGeeI mean, whatever.

PCs in general are a swiss cheese of security issues, no doubt thousands which arn't known yet.
I'm glad my industry is completely useless, time to retire I guess.
Solaris17random links to blog posts like above.
To be fair I wouldn't categorize the "University of Toronto" as a random blog post (provided that's actually where that comes from), but I do agree the online "press" sensationalize the shit out of this stuff to a point that users are scaring themselves to death over something that will likely be addressed by some driver tweaks (in this case anyhow).

If you aren't an infested malware farm, this isn't anything to worry about, no.
Posted on Reply
#22
watzupken
SiliconSeraphMaybe I'm a bonehead, but will GDDR7 be unaffected? This seems to be isolated to GDDR6 but maybe I missed something.
This type of news is meant to scare people to upgrade. It mostly affects older tech, and when you upgrade, it will eventually impact GDDR7 or newer hardware. It's a pointless chase.
Posted on Reply
#23
SiliconSeraph
Panther_Seraphin"GDDR7 adds on-die error correction code, error checking and scrubbing features for chip reliability,"
Okay, good to know. Guess that's a upside to upgrading... but a downside given these prices for GPUs....
Posted on Reply
#24
Visible Noise
watzupkenThis type of news is meant to scare people to upgrade. It mostly affects older tech, and when you upgrade, it will eventually impact GDDR7 or newer hardware. It's a pointless chase.
Why would government funded university researchers want people to upgrade?
Posted on Reply
#25
user556
This particular news is a little worse than just scare tactics. DRAM has a looming issue. GDDR not using ECC is why the proof was so easily effected.

As the process nodes get finer, the cell capacitance gets smaller and the proximity to neighboring cells gets closer. Both of these parameters amplify the effects that make the Rowhammer exploit possible. To that end, the problem is a natural limitation of DRAM. It will only get worse with higher density DRAMs. ECC is already borderline effective and will need bolstered further if used as the primary mitigation with increasing densities.

One future option might be reduce array density but start layering instead, not unlike how Flash is done now. Another option is replace DRAM with a different memory type entirely.
Posted on Reply
Add your own comment
Nov 18th, 2026 07:19 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts