[HN Gopher] GPUHammer: Rowhammer attacks on GPU memories are pra...
       ___________________________________________________________________
        
       GPUHammer: Rowhammer attacks on GPU memories are practical
        
       Author : jonbaer
       Score  : 259 points
       Date   : 2025-07-16 00:05 UTC (22 hours ago)
        
 (HTM) web link (gpuhammer.com)
 (TXT) w3m dump (gpuhammer.com)
        
       | perching_aix wrote:
       | HW noob here, anyone here has insight on how an issue like this
       | passes EM simulation during development? I understand that modern
       | chips are way too complex for full formal verification, but I'd
       | have thought memory modules would be so highly structurally
       | regular that it might be possible there despite it.
        
         | andyferris wrote:
         | I am no expert in the field, but my reading of the original
         | rowhammer issue (and later partial hardware mitigations) was
         | that it was seen as better to design RAM that works fast and is
         | dense and get that to market, than to engineer something
         | provably untamperable with greater tolerances / die size /
         | latency.
         | 
         | GPUs have always been squarely in the "get stuff to consumers
         | ASAP" camp, rather than NASA-like engineering that can
         | withstand cosmic rays and such.
         | 
         | I also presume an EM simulation would be able to spot it, but
         | prior to rowhammer it is also possible no-one ever thought to
         | check for it (or more likely that they'd check the simulation
         | with random or typical data inputs, not a hitherto-unthought-of
         | attack vector, but that doesn't explain more modern hardware).
        
           | privatelypublic wrote:
           | I seem to recall that rowhammer was known- but thought
           | impossible for userland code to implement.
           | 
           | This is a huge theme for vulnerabilities. I almost said
           | "modern" but looking back I've seen the cycle (disregard
           | attacks as strictly hypothetical. Get caught unprepared when
           | somebody publishes something making it practical) happen more
           | than a few times.
        
             | Palomides wrote:
             | someone did a javascript rowhammer in 2015, hardware that's
             | vulnerable today is just manufacturers and customers
             | deciding they don't want to pay for mitigation
             | 
             | (personally I think all RAM in all devices should be ECC)
        
               | grafmax wrote:
               | Manufacturers aren't held liable for negligence like
               | this. It's a classic case of economic externality.
        
               | andyferris wrote:
               | Yes it is - how would you go about fixing that?
        
               | yndoendo wrote:
               | Only means might be cultural. Security conferences such
               | as DefCon or Black Hat create list of insecurely
               | technology that is ubiquitousness and ignored by product
               | designers and OEMs. Vote on ranking their priority and
               | when they should be removed.
               | 
               | News would latch on to "Hacks say all computers without
               | ECC RAM are vulnerable and should not be purchased for
               | their insecurity. Manufacturers like Dell, Asus, Acer,
               | ... are selling products that help hackers steal your
               | information." "DefCon Hackers thank Nvidia for making
               | their jobs easier ..."
               | 
               | Such statements would be refreshed during / after each
               | security conference. There are over 12 conferences a
               | year, about once a month these would be brought back into
               | the public as a reminder. Public might stop purchasing
               | from those manufacturers or choose the secure products to
               | create the change.
        
               | andyferris wrote:
               | > manufacturers and customers deciding they don't want to
               | pay
               | 
               | It's more of a tragedy-of-the-commons problem. Consumers
               | don't know what they don't know and manufacturers need to
               | be competitive with respect to each other. Without some
               | kind of oversight (industry standards bodies or goverment
               | regulation), or a level of shaming that breaks through to
               | consumers (or e.g. class action lawsuits that impact
               | manufacturers), no individual has any incentive to
               | change.
        
               | progmetaldev wrote:
               | Shame is an underrated way towards pushing for better
               | standards. The problem is getting people in the know, and
               | having them vote with their wallet, or at least public
               | sentiment (social media pressure).
        
               | userbinator wrote:
               | The manufacturers tried to sweep it under the rug when
               | the first RowHammer came out. One of the memory testing
               | utilities added tests for it, and then disabled those
               | because they would cause too many failures.
        
               | userbinator wrote:
               | We don't want "mitigation", we want true correctness ---
               | or at least the level of perfection achievable before
               | manufacturers thought they could operate with negative
               | data integrity margins and convinced others that it was
               | fine (one popular memory testing utility made RH tests
               | optional and hidden by default, under the reasoning that
               | "too many DIMMs would fail"!) All DRAM generations before
               | DDR2 and early DDR3 didn't have this problem.
               | 
               | RAM that doesn't behave like RAM is not RAM. It's
               | defective. ECC is merely an attempt at fixing something
               | that shouldn't've made it to the market in the first
               | place. AFAIK there is a RH variant that manages to flip
               | bits undetectably even with ECC RAM.
        
               | nsteel wrote:
               | > AFAIK there is a RH variant that manages to flip bits
               | undetectably even with ECC RAM.
               | 
               | Single Error Correction, Double Error Detection, Tripple
               | Error Chaos.
        
               | ryao wrote:
               | The manufacturers chose this. Most customers were not
               | offered a choice.
               | 
               | It should be considered unethical to sell machines with
               | non-ECC memory in any real volume.
        
               | justincormack wrote:
               | You dont have to buy them.
        
           | userbinator wrote:
           | _but prior to rowhammer it is also possible no-one ever
           | thought to check for it_
           | 
           | It was known as "pattern sensitivity" in the industry for
           | decades, basically ever since the beginning, and considered a
           | blocking defect. Here's a random article from 1989 (don't
           | know why first page is missing, but look at the references):
           | http://web.eecs.umich.edu/~mazum/PAPERS-
           | MAZUM/patternsensiti...
           | 
           | Then some bastards like these came along...
           | 
           | https://research.ece.cmu.edu/safari/thesis/skhan_jobtalk_sli.
           | ..
           | 
           | ...and essentially said "who cares, let someone else be
           | responsible for the imperfections while we can sell more
           | crap", leading to the current mess we're in.
           | 
           | The flash memory industry took a similar dark turn decades
           | ago.
        
         | MadnessASAP wrote:
         | Given that I wasnt surprised by the headlie Inhave to imagine
         | that nvidia engineers were also well aware.
         | 
         | Nothing is perfect, everything has its failure conditions. The
         | question is where do you choose to place the bar? Do you want
         | your component to work at 60, 80, or 100C? Do you want it to
         | work in high radiation environments? Do you want it to
         | withstand pathological access patterns?
         | 
         | So in other words, there isnt a sufficent market for GPUs at
         | double the $/GB RAM but are resilient to rowhammer attacks to
         | justify manufacturing them.
        
         | thijsr wrote:
         | Rowhammer is an inherent problem to the way we design DRAM. It
         | is a known problem to memory manufacturers that is very hard,
         | if not impossible, to fix. In fact, Rowhammer only becomes
         | worse as the memory density increases.
        
           | sroussey wrote:
           | It's a matter of percentages... not all manufacturers fell to
           | the rowhammer attack.
           | 
           | The positive part of the original rowhammer report was that
           | it gave us a new tool to validate memory (it caused failures
           | much faster than other validation methods).
        
       | privatelypublic wrote:
       | This seems predicated on there being significant workloads that
       | split gpu's between tenants for compute purposes.
       | 
       | Anybody have sizable examples? Everything I can think of results
       | in dedicated gpus.
        
         | im3w1l wrote:
         | Webgpu api taking screenshot of full desktop maybe?
        
           | Buttons840 wrote:
           | Do you think WebGPU would be any more of an attack vector
           | than WebGL?
        
           | privatelypublic wrote:
           | Rowhammer itself is a write-only attack vector. It can,
           | however, potentially be chained to change the write address
           | to an incorrect region. Haven't dived into details.
        
             | SnowflakeOnIce wrote:
             | How is it a write-only attack vector?
        
               | privatelypublic wrote:
               | Rowhammer allows you to corrupt/alter memory physically
               | adjacent to memory you have access to. It doesn't let you
               | read the memory you're attacking.
               | 
               | There's PoC's of corrupting memory _that the kernel uses
               | to decide what that process can access_ but the process
               | can't read that memory. It only knows that the kernel
               | says yes where it used to say no. (Assuming it doesn't
               | crash the whole machine first)
        
               | SnowflakeOnIce wrote:
               | Suppose you have access to certain memory. If you
               | repeatedly _read_ from that memory, can 't you still
               | corrupt/alter the physically adjacent memory you _don 't_
               | have access to? Does it really need to be a write
               | operation you repeatedly perform?
        
               | privatelypublic wrote:
               | I probably should have called it "blind" instead.
        
               | extraduder_ire wrote:
               | > Does it really need to be a write operation you
               | repeatedly perform?
               | 
               | Yes. The core of rowhammer attacks is in changing the
               | values in RAM repeatedly, creating a magnetic field,
               | which induces a change in the state of nearby cells of
               | memory. Reading memory doesn't do that as far as I know.
        
         | vlovich123 wrote:
         | Many of the GPU rental companies charge less for shared GPU
         | workloads. So it's a cost/compute tradeoff. It's usually not
         | about the workload itself needing the full GPU unless you
         | really need all the RAM on a single instance.
        
           | privatelypublic wrote:
           | Any examples to check out? The only one i know of is
           | vastai... and there's already a list of security issues a
           | mile long there.
        
             | diggan wrote:
             | I don't think Vast.ai does "shared GPUs", you can only rent
             | full rigs, at least there is no indication the hardware is
             | shared between multiple users at the same time.
             | 
             | But I think services like Runpod and similar lets you rent
             | "1/6 of a GPU per hour" for example, which would be "shared
             | hosting" basically, as there would be multiple users using
             | the same hardware at the same time.
        
         | haiku2077 wrote:
         | GKE can share a single GPU between multiple containers in a
         | partitioned or timeshared scheme:
         | https://cloud.google.com/kubernetes-engine/docs/concepts/tim...
        
           | privatelypublic wrote:
           | Thats the thing... they're all the same tennant. A GKE node
           | is a VM instance, and GCE doesn't have shared GPUs that I can
           | see.
        
         | bluedino wrote:
         | NVIDIA GPU's can run in MIG (Multi-Instance GPU), allowing you
         | to pack more jobs on than you have GPUs. Very common in HPC but
         | I don't about in the cloud.
        
           | privatelypublic wrote:
           | I thought about splitting the GPU between workloads, as well
           | terminal server/virtualized desktop situations.
           | 
           | I'd expect all code to be strongly controlled in the former,
           | and reasonably secured in the latter with software/driver
           | level mitigations possible and the fact that corrupting
           | somebody else's desktop with row-hammer doesn't seem like
           | good investment.
           | 
           | As another person mentioned- and maybe it is a wider usage
           | than I thought- cloud gpu compute running custom code seems
           | to be the only useful item. But, I'm having a hard time
           | coming up with a useful scenario. Maybe corrupting a SIEM's
           | analysis & alerting of an ongoing attack?
        
           | cyberax wrote:
           | No large cloud hoster (AWS, Google, Azure) shares GPUs
           | between tenants.
        
             | shakna wrote:
             | Is that not what AWS is offering here? [0]
             | 
             | "In multi-tenant environments where the goal is to ensure
             | strict isolation."
             | 
             | [0] https://aws.amazon.com/blogs/containers/gpu-sharing-on-
             | amazo...
        
         | SnowflakeOnIce wrote:
         | Example: A workstation or consumer GPU used both for rendering
         | the desktop and running some GPGPU thing (like a deep neural
         | network)
        
           | privatelypublic wrote:
           | Not an issue- thats a single Tennant.
           | 
           | Which is my point.
        
             | spockz wrote:
             | Until the GPU is accessible by the browser and any website
             | can execute code on it. Or the attack can come from a
             | different piece of software on your machine.
        
         | privatelypublic wrote:
         | Update: I thought for a second I had one: Jupyter notebook
         | services with GPUs- but looking at google colab^* even there
         | its a dedicated GPU for that session.
         | 
         | * random aside: how is colab compute credits having a 90 day
         | expiration legal? I thought california outlawed company-
         | currency expiring? (A la gift cards)
        
           | dogma1138 wrote:
           | Colab credits aren't likely a currency equivalent but a
           | service equivalent which is still legal to expire afaik.
           | 
           | Basically Google Colab credits is like buying a seasonal bus
           | pass with X trips or a monthly parking pass with X amount of
           | hours. Rather than getting store cash which can be used for
           | anything.
        
         | huntaub wrote:
         | My (limited) understanding was that the industry previously
         | knew that it was unsafe to share GPUs between tenants, which is
         | why the major cloud providers only sell dedicated GPUs.
        
       | userbinator wrote:
       | No one really cared about the occasional bitflips in VRAM when
       | GPUs were only used for rendering graphics. It's odd that
       | enabling ECC can reduce performance, unless they mean that's only
       | in the presence of ECC errors being corrected, since AFAIK for
       | CPUs there isn't any difference in speed even when correcting
       | errors.
       | 
       |  _In a proof-of-concept, we use these bit flips to tamper with a
       | victim's DNN models and degrade model accuracy from 80% to 0.1%,
       | using a single bit flip_
       | 
       | There is a certain irony in doing this to probabilistic models,
       | designed to mimic an inherently error-prone and imprecise
       | reality.
        
         | hamandcheese wrote:
         | Available ECC dims are often slower than non-ECC dims. Both
         | slower MT/s and higher latency. At least for "prosumer" ECC
         | UDIMMs which are what I'm familiar with.
         | 
         | So it doesn't seem that wild to me that turning on ECC might
         | require running at lower bandwidth.
        
           | ryao wrote:
           | This is incorrect. ECC DIMMs are no slower than regular
           | DIMMs. Instead, they have extra memory and extra memory
           | bandwidth. A 8GB DDR4 ECC DIMM would have 9GB of memory and
           | 9/8 the memory bandwidth. The extra memory is used to store
           | the ECC bits while the extra memory bandwidth is to prevent
           | performance loss when reading/writing ECC alongside the rest
           | of the memory. The memory controller will spend an extra
           | cycle verifying the ECC, which is a negligible performance
           | hit. In reality, there is no noticeable performance
           | difference. However, where you would have 128 traces to a Zen
           | 3 CPU for DDR4 without ECC, you would need 144 traces for
           | DDR4 with ECC.
           | 
           | A similar situation occurs with GDDR6, except Nvidia was too
           | cheap to implement the extra traces and pay for the extra
           | chip, so instead, they emulate ECC using the existing memory
           | and memory bandwidth, rather than adding more memory and
           | memory bandwidth like CPU vendors do. This causes the
           | performance hit when you turn on ECC on most Nvidia cards.
           | The only exception should be the HBM cards, where the HBM
           | includes ECC in the same way it is done on CPU memory, so
           | there should be no real performance difference.
        
             | bilegeek wrote:
             | Their second point is wrong (unless the silicon is buggy),
             | but their first point is true. I researched when buying ECC
             | sticks for my rig; nobody that I've found makes
             | unregistered sticks that go above 5600, while some non-ECC
             | sticks are already at 8200, and 6400 is commonplace.
             | 
             | Frustratingly, it's only unregistered that's stuck in
             | limbo; VCC makes a kit of registered 7200.
        
               | ryao wrote:
               | That is partly due to artificial reasons and partly due
               | to technical reasons. The artificial reasons would be
               | that the 8200 MT/sec UDIMMs are overclocked. Notice how
               | they run much slower if you do not enable XMP/EXPO, which
               | simultaneously over volts and overclocks them. These
               | exist because a large number of people liked overclocking
               | their memory modules to get better performance. This was
               | unreliable and memory manufacturers noticed that there
               | was a market for a premium product where the overclocking
               | results were guaranteed. Early pre-overclocked modules
               | required people to manually enter the manufacturer
               | provided voltage, frequency and timings into the BIOS,
               | but XMP and later EXPO were made to simplify this
               | process. This idea only took off for non-ECC modules,
               | since the market for ECC UDIMMs wants reliability above
               | all else, so there never was quite the same market
               | opportunity to sell ECC DIMMs that were guaranteed to
               | overclock to a certain level outside of the memory IC
               | maker's specifications.
               | 
               | There is no technical reason why ECC UDIMMs cannot be
               | overclocked to the same extent and ECC actually makes
               | them better for overclocking since they can detect when
               | overclocking is starting to cause problems. You might
               | notice that the non-ECC UDIMMs have pads and traces for
               | an additional IC that is present on ECC UDIMMs. This
               | should be because the ECC DIMMs and non-ECC DIMMs are
               | made out of the same things. They use the same PCBs and
               | the same chips. The main differences would be whether the
               | extra chips to store ECC are on the module, what the SPD
               | says it is and what the sticker says. There might also be
               | some minor differences in what resistors are populated.
               | Getting back to the topic of overclocking, if you are
               | willing to go back to the days before the premium pre-
               | overclocked kits existed, you will likely find a number
               | of ECC UDIMMs can and will overclock with similar
               | parameters. There is just no guarantee of that.
               | 
               | As for RDIMMs having higher transfer rates, let us
               | consider the differences between a UDIMM, a CUDIMM and a
               | RDIMM. The UDIMM connects directly to the CPU memory
               | controller for the clock, address, control and data
               | signals, while the RDIMM has a register chip that buffers
               | the clock, address and control signals, although the data
               | signals still connect to the memory controller directly.
               | This improves signal integrity and lets more memory ICs
               | be attached to the memory controller. A recent
               | development is the CUDIMM, which is a hybrid of the two.
               | In the CUDIMM, the clock signal is buffered by a Client
               | Clock Driver, which does exactly what the register chip
               | does to the clock signal in RDIMMs. CUDIMM are able to
               | reach higher transfer rates than UDIMMs without
               | overclocking because of the Client Clock Driver, and
               | since RDIMMs also do what CUDIMMs do, they similarly can
               | reach higher transfer rates.
        
               | bilegeek wrote:
               | Thanks for the explanation on CUDIMM, I never quite
               | grokked the difference besides it being more stable with
               | two sticks per channel. Hopefully they'll make an ECC
               | CUDIMM at some point, but I'm not holding my breath.
        
               | consp wrote:
               | If they don't and you are up for a challenge in bga
               | soldering you can make them yourself if there is pad for
               | the chips. You likely have to buy an extra module to get
               | the chips though.
        
               | ryao wrote:
               | This would also need a SPD programmer and possibly some
               | additional SMT resistors, but it is possible in theory.
        
             | bobmcnamara wrote:
             | Propagation delay is a thing.
             | 
             | Edit: at some point the memory controller gets a chunk from
             | the lowest level write buffer and needs to compute ECC data
             | before writing everything out to RAM.
             | 
             | Without ECC, that computation time isn't there. The ECC
             | computation is done in parallel in hardware, but it's not
             | free.
        
               | ryao wrote:
               | You assume that the ECC is not already calculated when
               | the data is in the cache (and the cache line is marked
               | dirty). Caches in CPUs are often ECC protected,
               | regardless of whether the memory has ECC pritection. The
               | cache should already have the ECC computation done.
               | Writes to ECC memory can simply reuse the existing ECC
               | bytes from the cache, so no additional calculation time
               | is needed on writes. Reads are where additional time is
               | needed in the form of one cycle and if your cache is
               | doing its job, you won't notice this. If you do notice
               | it, your cache hit rate is close to 0 and your CPU is
               | effectively running around 50MHz due to pipeline stalls.
               | 
               | That said, this is tangential to whether the ECC DIMMs
               | themselves run at lower MT/sec ratings with higher
               | latencies, which was the original discussion. The ECC
               | DIMM is simply memory. It has an extra IC and a wider
               | data bus to accommodate that IC in parallel. The chips
               | run at the same MT/sec as the non-ECC DIMM in parallel.
               | The signals reach the CPU at the same in both ECC DIMMs
               | and non-ECC DIMMs, such that latencies are the same (the
               | ECC verification does use an extra cycle in the CPU, but
               | cache hides this). There are simply more data lanes with
               | ECC DIMMs due to the additional parallelism. This means
               | that there is more memory bandwidth in the ECC DIMM, but
               | that additional memory bandwidth is being used by the ECC
               | bytes, so you never see it in benchmarks.
        
               | tverbeure wrote:
               | The kind of ECC that's used for register file and memory
               | protection is trivial to compute and completely in the
               | noise in terms of area. It is essentially free.
               | 
               | The reason people say ECC is not free is because it added
               | area for every storage location, not because of the ECC
               | related logic.
        
         | ryao wrote:
         | Nvidia implements ECC in software since they did not want to
         | add the extra memory chip(s) needed to implement it in hardware
         | to their boards. The only case where they do it in hardware is
         | when they use HBM memory.
         | 
         | That said, GDDR7 does on die ECC, which gives immunity to this
         | in its current form. There is no way to get information on
         | corrected bitflips from on-die ECC, but it is better than
         | nothing.
        
       | iFire wrote:
       | Does the ECC mode on my 4090 Nvidia rtx stop this?
        
         | fc417fc802 wrote:
         | Yes, but it reduces performance, and you don't need to care
         | about this because (presumably) you aren't a cloud provider
         | running multi-tenant workloads.
         | 
         | Worst case scenario someone pulls this off using webgl and a
         | website is able to corrupt your VRAM. They can't actually steal
         | anything in that scenario (AFAIK) making it nothing more than a
         | minor inconvenience.
        
           | perching_aix wrote:
           | Couldn't it possibly lead to arbitrary code execution on the
           | GPU, with that opening the floodgates towards the rest of the
           | system via DMA, or maybe even enabling the dropping of some
           | payload for the kernel mode GPU driver?
        
             | fc417fc802 wrote:
             | Only if an attacker can (1) identify a piece of exploitable
             | data that the GPU firmware or kernel mode portion of the
             | driver relies on, (2) ascertain its location at runtime,
             | and (3) obtain a physically adjacent block of memory.
             | 
             | I'm not certain that something which satisfies 1, let alone
             | 3, necessarily exists. On the CPU you flip some bits
             | related to privilege levels. Are any analogous and
             | similarly exploitable data structures maintained by common
             | GPU firmware? And if so, is such data stored in the bulk
             | VRAM?
             | 
             | It wouldn't surprise me if it was possible but it also
             | seems entirely unrealistic given that either you are
             | restricted to an API such as webgl (gl;hf) or you have
             | native access in which case you have better options
             | available to you seeing as the driver stacks are widely
             | rumored to be security swiss cheese (if you doubt this look
             | at how much effort google has invested in sandboxing the
             | "real" GPU driver away from apps on chromeos).
        
         | keysdev wrote:
         | What about Apple M series?
        
       | PeterStuer wrote:
       | I always found hammering attacks to be extremely satisfying, even
       | from a meta-physical pov.
       | 
       | You escape a closed virtual universe by not "breaking out" in the
       | tradidional sense, exploiting some bug in the VM hypervisor's
       | boundary itself, but by directly manipulating the underlying
       | physics of the universe on wich the virtual universe is founded,
       | just by creating a pattern inside the virtual universe itself.
       | 
       | No matter how many virtual digital layers, as long as you can
       | impact the underlying analog substrate this might work.
       | 
       | Makes you dream there could be an equivalent for our own
       | universe?
        
         | SandmanDP wrote:
         | > Makes you dream there could be an equivalent for our own
         | universe?
         | 
         | I've always considered that to be what's achieved by the LHC:
         | smashing the fundamental building blocks of our universe
         | together at extreme enough energies to briefly cause ripples
         | through the substrate of said universe
        
           | whyowhy3484939 wrote:
           | That's assuming there is a substrate that can be disturbed.
           | That's where the parent's analogy breaks down.
           | 
           | As an example of an alternative analogy: think of how many
           | bombs need to explode in your dreams before the "substrate"
           | is "rippled". How big do the bombs need to be? How fast does
           | the "matter" have to "move"? I think "reality" is more along
           | those lines. If there is a substrate - and that's a big if -
           | IMO it's more likely to be something pliable like
           | "consciousness". Not in the least "disturbed" by anything
           | moving in it.
        
             | rolisz wrote:
             | A nightmare that makes you wake up screaming? I'd say that
             | counts as disturbing the substrate.
        
             | cwillu wrote:
             | It's a pretty exact description: the universe is made of
             | fields, smashing stable excitations of those fields
             | together produces disturbances in other fields ("virtual
             | particles") that sometimes makes (fleetingly) stable
             | excitations in other fields, which then fall apart through
             | the same dance into different stable excitations than we
             | started with, allowing us to prove that the field in the
             | middle exists and start to determine its properties.
             | 
             | https://profmattstrassler.com/articles-and-posts/particle-
             | ph...
             | 
             | https://profmattstrassler.com/articles-and-posts/particle-
             | ph...
        
             | jerf wrote:
             | Another way to think of it. Consider breaking out of
             | Minecraft. Can you do it?
             | 
             | Maybe. There are certainly ways to crash it today. But now
             | let's go through some cycles of fixing those crashes, and
             | we'll run it on a system that can handle the resource usage
             | even if it slows down in the external reality's terms quite
             | a bit. And we'll ignore the slash commands and just stick
             | to the world interactions you can make.
             | 
             | After that, can you forcefully break out of it from the
             | inside?
             | 
             | No.
             | 
             | It is not obligatory for systems to include escape hatches.
             | We're just not great at building complex systems without
             | them. But there's no reason they are necessarily present in
             | all systems.
             | 
             | Another brain bender covering the same idea in a different
             | direction: The current reigning candidate for BB(6) runs an
             | incomprehensible amount of computation [1]. Yet, did it at
             | any point "break out" into our world? Nope. Nor do any of
             | the higher ones. They're completely sealed in their
             | mathematical world, which is fortunate since any of them
             | would sweep aside our entire universe without noticing.
             | 
             | https://scottaaronson.blog/?p=8972
        
           | simiones wrote:
           | The LHC doesn't generate anything like the kind of energy
           | that you get when interstellar particles hit the Earth's
           | upper atmosphere, nevermind what's happening inside the sun -
           | and any of these are many, many orders of magnitude below the
           | energies you get in a supernova, for example.
           | 
           | The LHC is extremely impressive from a human engineering
           | perspective, but it's nowhere close to pushing the boundaries
           | of what's going on every second in the universe at large.
        
             | robotnikman wrote:
             | The closest thing I can thing of is a black hole.
        
           | drcongo wrote:
           | I love that we can switch out LHC for LSD and this comment
           | would still feel perfect.
        
         | hoseja wrote:
         | Well for Earth life there are multiple but evolution just
         | learned to exploit them all.
        
         | arduinomancer wrote:
         | I tried knocking on my wall 100,000 times and it did indeed
         | cause a disturbance in the neighbouring cell of my apartment
         | 
         | Turns out this whole virtualized house abstraction is a sham
        
           | N_Lens wrote:
           | Thanks for the sensible chuckle.
        
         | lukan wrote:
         | "I always found hammering attacks to be extremely satisfying"
         | 
         | On a philosophical level I somewhat agree, but on a practical
         | level I am sad as this likely means reduced performance again.
        
           | MangoToupe wrote:
           | Only for places where you need security. Many types of
           | computation do not need security.
        
             | tuvang wrote:
             | In theory, true. But fixes to issues like this are usually
             | done on hardware level in future generations or very low
             | level software level where most people don't have the
             | knowledge/effort to deal with. Resulting in our
             | editors/games/job tools running slower they can to mitigate
             | security issues irrelevant to our common use cases.
        
             | lukan wrote:
             | Most devices are connected to the internet these days.
             | Anything connected to the internet should be secure.
        
               | fc417fc802 wrote:
               | If a task has been granted native GPU access then it's
               | already on the inside of the security boundary.
               | Conversely, if you don't trust a task then don't let it
               | access the GPU (aside from passing the entire device
               | through to a virtual machine). This attack doesn't change
               | that reality.
        
               | mouse_ wrote:
               | > Conversely, if you don't trust a task then don't let it
               | access the (computer)
               | 
               | Wow you just solved all of cyber security
        
               | aspenmayer wrote:
               | If you liked that I have some art tips:
               | 
               | To draw an owl, start with two circles or ovals for eyes.
               | Then all you have to do is draw the rest of the owl.
               | 
               | https://knowyourmeme.com/memes/how-to-draw-an-owl
        
               | immibis wrote:
               | This is not true. I don't know the details, but GPUs have
               | something similar to page tables, so they can run
               | untrusted tasks. The worst threat is that one could stick
               | in an infinite loop, freezing your display output until
               | it times out - since they don't get timesliced.
        
         | sneak wrote:
         | If we are in a simulation, the number of ways that could be the
         | equivalent of SysRq or control-alt-delete are infinite.
         | 
         | We haven't even tried many of the simple/basic ones like moving
         | objects at 0.9c.
        
           | 0cf8612b2e1e wrote:
           | To ruin a great joke, particle accelerators do get up to
           | 0.9999+ the speed of light.
        
             | sneak wrote:
             | Ahh, I forgot about those.
        
         | amelius wrote:
         | > No matter how many virtual digital layers
         | 
         | So you are saying that a GPU program can find exploits in
         | physics without having access to e.g. high energy physics
         | tools?
         | 
         | Sounds implausible.
        
         | mistercow wrote:
         | > Makes you dream there could be an equivalent for our own
         | universe?
         | 
         | But would we even notice? As far as we were concerned, it would
         | just be more physics.
        
           | LeifCarrotson wrote:
           | The real question is how we'd know about the physics of the
           | parent universe.
           | 
           | I think this short story is interesting to think about in
           | that way:
           | 
           | https://www.lesswrong.com/posts/5wMcKNAwB6X4mp9og/that-
           | alien...
        
         | queuebert wrote:
         | > Makes you dream there could be an equivalent for our own
         | universe?
         | 
         | My idea to attack the simulation is psychological: make our own
         | simulation that then makes its own simulation, and so on all
         | the way down. That will sow doubt in the minds of the
         | simulators that they themselves are a simulation and make them
         | sympathetic to our plight.
        
           | nemomarx wrote:
           | What if the simulators tell you they're also doing this? It
           | could be turtles all the way up perhaps
        
         | mrkstu wrote:
         | I remember as a kid in the 70's I first heard about most
         | physical things being empty space.
         | 
         | Walking into a wall a few hundred times may have damaged my
         | forehead almost as much as my trust in science...
        
       | sylware wrote:
       | On the general case, that's why some optimized assembly written
       | machine code can be an issue compared to the slow compiler
       | generated machine code (not true all the time of course): if this
       | machine code is 'hammering' memory, it is could happen more
       | likely with the optimized assembly machine code than with the
       | "actually tested" compiler genertade machine code.
        
       ___________________________________________________________________
       (page generated 2025-07-16 23:00 UTC)