So far, only GDDR6 memory seems to be affected by these Rowhammer attacks, and the team tested 25 GPUs with it, finding vulnerabilities in most. More modern cards with GDDR6X or GDDR7 memory were also tested, but couldn't be compromised yet. Nvidia has already suggested enabling ECC to mitigate against Rowhammer attacks, but that reduces total VRAM capacity, introduces performance overhead, and isn't available on every GPU.
The other option is to enable IOMMU (Input–Output Memory Management Unit) in the BIOS, which puts a strict boundary around non-CPU devices when they try to access system memory. Usually, a GPU can read or write system memory without requesting the CPU every time (Direct Memory Access), which is how GDDRHammer or GeForge slip into system RAM. But with IOMMU enabled, corrupted page tables with incorrect mapping couldn't be exploited to gain access to the CPU since they'd be cordoned off.
Think of it as putting the GPU in a sandbox where it can still talk to the CPU, but it's not allowed to go around it. Of course, if the IOMMU itself is misconfigured or targeted in an attack, it can be bypassed as well, but it is the best line of defense for now. The researchers themselves admit that enabling IOMMU closes the vulnerability. By default, it's kept disabled in the BIOS to avoid compatibility problems because it's so stringent.
Rowhammer attacks, while dangerous, still require the attacker to have access to your system in order to run code that will start the process. So, you're not as actively exposed compared to other remote attacks. However, in a network of shared computers, maybe a cluster of GPUs running AI workloads, the risk is increased, so it's important to stay vigilant and informed, nonetheless. Check out the linked research papers if you want to dive deep into the specifics of how either attack works.
Follow Tom's Hardware on Google News , or add us as a preferred source , to get our latest news, analysis, & reviews in your feeds.
Hassam Nasir is a die-hard hardware enthusiast with years of experience as a tech editor and writer, focusing on detailed CPU comparisons and general hardware news. When he\u2019s not working, you\u2019ll find him bending tubes for his ever-evolving custom water-loop gaming rig or benchmarking the latest CPUs and GPUs just for fun. ","collapsible":{"enabled":true,"maxHeight":250,"readMoreText":"Read more","readLessText":"Read less"}}), "https://slice.vanilla.futurecdn.net/13-4-19/js/authorBio.js"); } else { console.error('%c FTE ','background: #9306F9; color: #ffffff','no lazy slice hydration function available'); } Hassam Nasir Social Links Navigation Contributing Writer Hassam Nasir is a die-hard hardware enthusiast with years of experience as a tech editor and writer, focusing on detailed CPU comparisons and general hardware news. When he’s not working, you’ll find him bending tubes for his ever-evolving custom water-loop gaming rig or benchmarking the latest CPUs and GPUs just for fun.
pjmelect This is rubbish, this type of attack does not work. The manufacturers of memory go to great length to ensure that bit flips do not occur in normal operation. Memory that does bit flip is faulty memory. Just because a memory cell is arithmetically next door to another does not mean that it is on the chip (it is not) and is different on every manufacturer and type of chip and the manufacturers do not disclose this information. To get the bits to flip they change the timings of the signals going to the chip which can cause random corruption of the memory. If you have this level of control over a computer what is the point of it? Reply
Key considerations
- Investor positioning can change fast
- Volatility remains possible near catalysts
- Macro rates and liquidity can dominate flows
Reference reading
- https://www.tomshardware.com/pc-components/gpus/SPONSORED_LINK_URL
- https://www.tomshardware.com/pc-components/gpus/new-geforge-and-gddrhammer-attacks-can-fully-infiltrate-your-system-through-nvidias-gpu-memory-rowhammer-attacks-in-gpus-force-bit-flips-in-protected-vram-regions-to-gain-read-write-access#main
- https://www.tomshardware.com
- Efficiency at Scale: NVIDIA, Energy Leaders Accelerating Power‑Flexible AI Factories to Fortify the Grid
- Apple approves drivers that let AMD and Nvidia eGPUs run on Mac — software designed for AI, though, and not built for gaming
- Advancing Open Source AI, NVIDIA Donates Dynamic Resource Allocation Driver for GPUs to Kubernetes Community
- Sequoia Capital shares founder’s 1977 memo for investing in Apple — initial $600,000 investment is worth $26.4 billion today, firm said it was ‘tough to do this
- Nvidia AI tech claims to slash gaming GPU memory usage by 85% with zero quality loss — Neural Texture Compression demo reveals stunning visual parity between 6.
Informational only. No financial advice. Do your own research.