Video wird geladen...

Video konnte nicht geladen werden

Beim Laden dieses Videos ist ein Problem aufgetreten. Dies könnte an einem vorübergehenden Netzwerkproblem liegen oder das Video ist möglicherweise nicht verfügbar.

We heavily optimized our GLTF files to fit any GPU memory. We built a custom Electron app to reduce polycount & convert textures to GPU-friendly Basis/ETC1S. Compressed textures used ~10-100x less memory & uploaded instantly, preventing decoding lags while scrolling!

Arnaud Tanielian

3,945 subscribers

17,465 Aufrufe • vor 8 Monaten •via X (Twitter)

Anya Rossi• Live Now

Private livecam show

0 Kommentare

Keine Kommentare verfügbar

Kommentare vom Original-Post werden hier angezeigt

Ähnliche Videos

Introducing DuoAttention: Our new framework slashes both memory and latency for long-context LLMs without sacrificing performance! By applying full KV cache only to critical heads, we achieve: ⚡ 2.55x memory reduction ⚡ 2.18x decoding speedup ⚡ 3.3M tokens on a single A100 GPU

Introducing DuoAttention: Our new framework slashes both memory and latency for long-context LLMs without sacrificing performance! By applying full KV cache only to critical heads, we achieve: ⚡ 2.55x memory reduction ⚡ 2.18x decoding speedup ⚡ 3.3M tokens on a single A100 GPU

Guangxuan Xiao

31,085 Aufrufe • vor 1 Jahr

Looking for a fast method to render your large-scale game components? Our lastest High-Perf Graphics paper allows to render terrains, oceans and even entire planets in realtime on any modern GPU using compute shaders. Fun fact: we use a Concurrent Binary Tree as a GPU memory manager.

Looking for a fast method to render your large-scale game components? Our lastest High-Perf Graphics paper allows to render terrains, oceans and even entire planets in realtime on any modern GPU using compute shaders. Fun fact: we use a Concurrent Binary Tree as a GPU memory manager.

Anis Benyoub

50,643 Aufrufe • vor 2 Jahren

Internally at NVIDIA, we use cuOpt based agentic workflows with agent skills to optimize our supply chains. Since it’s open source, you can too. With optimizations ready in minutes instead of weeks, the workflow uses multi-agent LangChain Deep agent orchestration and GPU-accelerated solvers to turn natural language into optimized decisions. Spin it up instantly with a Brev Launchable (preconfigured GPU environment) and grab free developer credits while they last.

Internally at NVIDIA, we use cuOpt based agentic workflows with agent skills to optimize our supply chains. Since it’s open source, you can too. With optimizations ready in minutes instead of weeks, the workflow uses multi-agent LangChain Deep agent orchestration and GPU-accelerated solvers to turn natural language into optimized decisions. Spin it up instantly with a Brev Launchable (preconfigured GPU environment) and grab free developer credits while they last.

NVIDIA AI

22,918 Aufrufe • vor 3 Monaten

The UN estimates we dump a garbage truck’s worth of plastic into our oceans, rivers, & lakes every minute. Less than 10% of plastic is recycled because it’s made too cheaply to be used & reused.📉⬇️The best solution is to reduce, recycle and reuse ♻️ #ClimateAction

The UN estimates we dump a garbage truck’s worth of plastic into our oceans, rivers, & lakes every minute. Less than 10% of plastic is recycled because it’s made too cheaply to be used & reused.📉⬇️The best solution is to reduce, recycle and reuse ♻️ #ClimateAction

Eco Harmony Forum UG

39,515 Aufrufe • vor 7 Monaten

Dylan Patel on the importance of memory and storage Two key quotes: "An $NVDA GPU is faster than an $AMD GPU in most cases, but because AMD GPUs have more memory, they can outperform Nvidia in certain workloads." “It is a difficult, multivariable problem. Generally, you need the best GPU, such as a GB300, but you also need the best storage solutions. I will not spoil who comes out on top, but storage solutions matter a lot, memory solutions matter a lot, and frontend networking also matters significantly" Full Quote: “We have over $80 million of compute: GPUs from $NVDA and $AMD, TPUs from Google, and Trainium from Amazon. We constantly run this benchmark using the newest inference engines, drivers, PyTorch versions, and other software. It runs every day through automated CI across the latest Chinese models from GLM, Zhipu, Moonshot, Kimi, Alibaba, and others. Initially, when we were benchmarking the differences between these chips, inference engines, and parallelism schemes, we used fixed context lengths. But with Agent X, we have now analyzed more than $5 million worth of Claude Code traces. This is real production traffic that users have donated to us, combined with internally generated data, so we now understand what an actual agent workload looks like. When we implement those workloads and run the benchmarks, it turns out that the chip you are using is very important, but how you handle memory offload can be even more important. An Nvidia GPU is faster than an AMD GPU in most cases, but because AMD GPUs have more memory, they can outperform Nvidia in certain workloads. Similarly, you can use a less powerful GPU with a much better storage solution and outperform the best GPU when it lacks those solutions. Simply buying the newest GPU does not necessarily give you the best inference economics. You need to layer in other innovations, including storage and memory.” Interviewer: “Who is the top player on your chart? Can you tell us?” Dylan Patel: “It is a difficult, multivariable problem. Generally, you need the best GPU, such as a GB300, but you also need the best storage solutions. I will not spoil who comes out on top, but storage solutions matter a lot, memory solutions matter a lot, and frontend networking also matters significantly.”

Dylan Patel on the importance of memory and storage Two key quotes: "An $NVDA GPU is faster than an $AMD GPU in most cases, but because AMD GPUs have more memory, they can outperform Nvidia in certain workloads." “It is a difficult, multivariable problem. Generally, you need the best GPU, such as a GB300, but you also need the best storage solutions. I will not spoil who comes out on top, but storage solutions matter a lot, memory solutions matter a lot, and frontend networking also matters significantly" Full Quote: “We have over $80 million of compute: GPUs from $NVDA and $AMD, TPUs from Google, and Trainium from Amazon. We constantly run this benchmark using the newest inference engines, drivers, PyTorch versions, and other software. It runs every day through automated CI across the latest Chinese models from GLM, Zhipu, Moonshot, Kimi, Alibaba, and others. Initially, when we were benchmarking the differences between these chips, inference engines, and parallelism schemes, we used fixed context lengths. But with Agent X, we have now analyzed more than $5 million worth of Claude Code traces. This is real production traffic that users have donated to us, combined with internally generated data, so we now understand what an actual agent workload looks like. When we implement those workloads and run the benchmarks, it turns out that the chip you are using is very important, but how you handle memory offload can be even more important. An Nvidia GPU is faster than an AMD GPU in most cases, but because AMD GPUs have more memory, they can outperform Nvidia in certain workloads. Similarly, you can use a less powerful GPU with a much better storage solution and outperform the best GPU when it lacks those solutions. Simply buying the newest GPU does not necessarily give you the best inference economics. You need to layer in other innovations, including storage and memory.” Interviewer: “Who is the top player on your chart? Can you tell us?” Dylan Patel: “It is a difficult, multivariable problem. Generally, you need the best GPU, such as a GB300, but you also need the best storage solutions. I will not spoil who comes out on top, but storage solutions matter a lot, memory solutions matter a lot, and frontend networking also matters significantly.”

Daniel Romero

38,220 Aufrufe • vor 1 Monat

$LITE CEO Michael Hurlston calls Lumentum the TSMC of the optical industry, because lasers run on indium phosphide rather than the CMOS behind every GPU. "And so for us, we actually operate our own fabs. So in some respect, we're like the TSMC of the optical industry." "Optics depends on a different material. Semiconductors use something called CMOS, and all of the semiconductors that you know well are built on CMOS. Memory built on CMOS. TPUs are built on CMOS. NVIDIA GPU is built on CMOS." "But lasers and fiber optics are built on a different material called indium phosphide. Kind of a mouthful, but it's this indium phosphide, and that's a very boutique thing. It's a little more of a niche." "And so we actually have to make our own indium phosphide products. We have to manufacture those ourselves." ______ Wrote more about it here:

$LITE CEO Michael Hurlston calls Lumentum the TSMC of the optical industry, because lasers run on indium phosphide rather than the CMOS behind every GPU. "And so for us, we actually operate our own fabs. So in some respect, we're like the TSMC of the optical industry." "Optics depends on a different material. Semiconductors use something called CMOS, and all of the semiconductors that you know well are built on CMOS. Memory built on CMOS. TPUs are built on CMOS. NVIDIA GPU is built on CMOS." "But lasers and fiber optics are built on a different material called indium phosphide. Kind of a mouthful, but it's this indium phosphide, and that's a very boutique thing. It's a little more of a niche." "And so we actually have to make our own indium phosphide products. We have to manufacture those ourselves." ______ Wrote more about it here:

Fireside Alpha

55,886 Aufrufe • vor 27 Tagen

Cheney: Harris has built a coalition unlike any other we have seen including the far left & far right, including many people who used to work for Trump. The reason we've come together is not bc we agree on every issue, but we understand our Constitution is at stake this election

Cheney: Harris has built a coalition unlike any other we have seen including the far left & far right, including many people who used to work for Trump. The reason we've come together is not bc we agree on every issue, but we understand our Constitution is at stake this election

Aaron Rupar

530,441 Aufrufe • vor 1 Jahr

researchers gave a tiny local model human-style memory and its context limit basically stopped existing a team from MBZUAI, Princeton and Weizmann took a 1B model and rebuilt how it reads. instead of attending to everything at once, the model reads in 1,024 token chunks and passes the important stuff forward through an associative memory, the same way you carry the plot of a book between chapters without rereading them. the design mirrors human memory on purpose. full attention inside a chunk works as short-term memory. the module that carries information between chunks works as long-term memory. they even trained it like a person, starting with short easy texts and raising the difficulty gradually, because memory thrown into the deep end learns nothing. the numbers back it up. the normal model burns 40GB of GPU memory on a long document and collapses hard past its limit, dropping from 0.86 to 0.32 accuracy. the memory version holds 0.71 at double that length while using a flat 12GB no matter how long the input gets. it also needs about 30% fewer FLOPs. the part i keep thinking about is that nobody scaled anything here. they didn't build a bigger model, didn't stretch the window, didn't add compute. they looked at how a brain handles a long day and copied the architecture. a model small enough to run on a consumer gpu now survives documents its own architecture used to choke on. we keep treating intelligence as a compute problem. sometimes it's a memory problem.

researchers gave a tiny local model human-style memory and its context limit basically stopped existing a team from MBZUAI, Princeton and Weizmann took a 1B model and rebuilt how it reads. instead of attending to everything at once, the model reads in 1,024 token chunks and passes the important stuff forward through an associative memory, the same way you carry the plot of a book between chapters without rereading them. the design mirrors human memory on purpose. full attention inside a chunk works as short-term memory. the module that carries information between chunks works as long-term memory. they even trained it like a person, starting with short easy texts and raising the difficulty gradually, because memory thrown into the deep end learns nothing. the numbers back it up. the normal model burns 40GB of GPU memory on a long document and collapses hard past its limit, dropping from 0.86 to 0.32 accuracy. the memory version holds 0.71 at double that length while using a flat 12GB no matter how long the input gets. it also needs about 30% fewer FLOPs. the part i keep thinking about is that nobody scaled anything here. they didn't build a bigger model, didn't stretch the window, didn't add compute. they looked at how a brain handles a long day and copied the architecture. a model small enough to run on a consumer gpu now survives documents its own architecture used to choke on. we keep treating intelligence as a compute problem. sometimes it's a memory problem.

Alex Veremeyenko

16,147 Aufrufe • vor 1 Monat

Etched is deploying two new technologies in chip design: low-voltage inference and cluster-scale memory. CEO Gavin Uberti says they'll make their chips much more power-efficient and way, way faster than today's leading GPUs. He breaks it down: "We looked at a lot of early research directions, and we realized the key things that models need are way more compute and way faster memory." "If you think about inference, there are two key parts: prefill and decode. For prefill, it's a compute-bound problem. You need to have more FLOPS, more operations per second on each of your chips." "On our GPU, the bottleneck's actually thermals. You can't really run a GPU at more than around 50% of what it could theoretically do, or it'll melt." "So we're using a new technology today called low-voltage inference to try to solve this problem. You bring the voltage of the chip down dramatically, which allows us to have way, way better efficiency in terms of how much power is drawn per unit of math, and thus fit way way more flops onto the chip..." "For decode, it's all about bandwidth. Not just bandwidth on a chip, but bandwidth across your cluster. That's why we have this technology we call cluster-scale memory. It reduces the amount of time it takes to communicate from one chip to another dramatically." "As a result we can go use all of our HBM, HBM bandwidth, SRAM, SRAM bandwidth, and our scale-up domain as a single coherent pool. And that means if you're a user, you can go get much faster tokens per second, while still keeping your costs low."

Etched is deploying two new technologies in chip design: low-voltage inference and cluster-scale memory. CEO Gavin Uberti says they'll make their chips much more power-efficient and way, way faster than today's leading GPUs. He breaks it down: "We looked at a lot of early research directions, and we realized the key things that models need are way more compute and way faster memory." "If you think about inference, there are two key parts: prefill and decode. For prefill, it's a compute-bound problem. You need to have more FLOPS, more operations per second on each of your chips." "On our GPU, the bottleneck's actually thermals. You can't really run a GPU at more than around 50% of what it could theoretically do, or it'll melt." "So we're using a new technology today called low-voltage inference to try to solve this problem. You bring the voltage of the chip down dramatically, which allows us to have way, way better efficiency in terms of how much power is drawn per unit of math, and thus fit way way more flops onto the chip..." "For decode, it's all about bandwidth. Not just bandwidth on a chip, but bandwidth across your cluster. That's why we have this technology we call cluster-scale memory. It reduces the amount of time it takes to communicate from one chip to another dramatically." "As a result we can go use all of our HBM, HBM bandwidth, SRAM, SRAM bandwidth, and our scale-up domain as a single coherent pool. And that means if you're a user, you can go get much faster tokens per second, while still keeping your costs low."

TBPN

20,404 Aufrufe • vor 1 Monat

DeepMind CEO Demis Hassabis says the AI bottleneck is already moving from model ideas to physical supply chains. "How much of the restriction at this point is physical? I am talking memory shortage, GPU shortage, and electricity." Demis: "That is constraining a lot of the deployment for sure. We see so much more demand for our models - Gemini and the other models that we have - than we can possibly serve." Then the less obvious part: "It does constrain a little bit the research, because you need a lot of chips to be able to experiment on new ideas at big enough scale that you can actually see if they are going to work." And on memory: "The whole supply chain is kind of strained... It still in the end actually comes down to a few suppliers of a few of the key components. Anywhere where there is some kind of constraint on the capacity, there is a sort of choke point." Important framing: The next AI race is not just who has the best model architecture. It is who can secure memory, chips, power, suppliers, and enough experimental capacity to find the next breakthrough before everyone else.

DeepMind CEO Demis Hassabis says the AI bottleneck is already moving from model ideas to physical supply chains. "How much of the restriction at this point is physical? I am talking memory shortage, GPU shortage, and electricity." Demis: "That is constraining a lot of the deployment for sure. We see so much more demand for our models - Gemini and the other models that we have - than we can possibly serve." Then the less obvious part: "It does constrain a little bit the research, because you need a lot of chips to be able to experiment on new ideas at big enough scale that you can actually see if they are going to work." And on memory: "The whole supply chain is kind of strained... It still in the end actually comes down to a few suppliers of a few of the key components. Anywhere where there is some kind of constraint on the capacity, there is a sort of choke point." Important framing: The next AI race is not just who has the best model architecture. It is who can secure memory, chips, power, suppliers, and enough experimental capacity to find the next breakthrough before everyone else.

Karl Mehta

25,716 Aufrufe • vor 2 Monaten

I gave a talk at GPU MODE workshop last week on llm.c - the origin story of llm.c - being naked in the world without PyTorch and having to re-invent Array, Autograd, Device, Dtype, Compile, Distributed - how to port a PyTorch layer to 1) explicit PyTorch - and then to 2) write the backward pass - 3) port forward & backward pass to C - 4) string all the layers together - achieving one file of C with no dependencies that compiles and runs ~instantly, where all memory is pre-planned and allocated a single time, fully deterministic, portable code that can run on a potato or a von Neumann probe - how most of llm.c was built at 1am-7am in a water villa porch in Maldives and why this is the recommended way to develop software - convert all of it to run in CUDA on GPU in fp32 - port matmul to cuBLAS - port attention to cuDNN flash-attention - introduce bfloat16 mixed precision - introduce many more optimizations and features like kernel fusions, Packed128, stochastic rounding, full determinism - add multi-GPU training, NCCL, sharded optimizer - add multi-node with MPI or file system or socket - reproduce GPT-2 (1.6B) on one 8XH100 node in 24 hours for $672 in llm.c, achieving (at the time) 29% less memory, 19% faster training that PyTorch nightly, and much faster compile & run - how open source development attracts Avengers from the internet - port to training Llama 3 imminent (branch exists) - many other notable forks - last thought: how software abstractions like Python/PyTorch and everything else really exist only because humans are finite in knowledge, IQ and attention, and how with increasing AI capability LLMs may export custom binaries like llm.c for any application directly, tearing apart and refactoring all abstractions as needed. More links in reply

I gave a talk at GPU MODE workshop last week on llm.c - the origin story of llm.c - being naked in the world without PyTorch and having to re-invent Array, Autograd, Device, Dtype, Compile, Distributed - how to port a PyTorch layer to 1) explicit PyTorch - and then to 2) write the backward pass - 3) port forward & backward pass to C - 4) string all the layers together - achieving one file of C with no dependencies that compiles and runs ~instantly, where all memory is pre-planned and allocated a single time, fully deterministic, portable code that can run on a potato or a von Neumann probe - how most of llm.c was built at 1am-7am in a water villa porch in Maldives and why this is the recommended way to develop software - convert all of it to run in CUDA on GPU in fp32 - port matmul to cuBLAS - port attention to cuDNN flash-attention - introduce bfloat16 mixed precision - introduce many more optimizations and features like kernel fusions, Packed128, stochastic rounding, full determinism - add multi-GPU training, NCCL, sharded optimizer - add multi-node with MPI or file system or socket - reproduce GPT-2 (1.6B) on one 8XH100 node in 24 hours for $672 in llm.c, achieving (at the time) 29% less memory, 19% faster training that PyTorch nightly, and much faster compile & run - how open source development attracts Avengers from the internet - port to training Llama 3 imminent (branch exists) - many other notable forks - last thought: how software abstractions like Python/PyTorch and everything else really exist only because humans are finite in knowledge, IQ and attention, and how with increasing AI capability LLMs may export custom binaries like llm.c for any application directly, tearing apart and refactoring all abstractions as needed. More links in reply

Andrej Karpathy

337,263 Aufrufe • vor 1 Jahr

Ghostty is getting automatic scrollback compression, resulting in 70 to 90% less physical memory usage. It happens incrementally when idle, so it had no measurable effect on IO throughput. I'm not aware of any other mainstream terminal that does this. Demo video below! The gains let us increase the default scrollback limit from 10MB to 50MB, because on average a full scrollback will still compress smaller than the prior limit. More history, for free. ("Unlimited", disk-paged history is on the roadmap too) Let's talk about cool implementation details, cause this was fun. First, the data structure and memory layout ("PageList") I wrote two years ago finally pays off! One of its traits is that screen memory is backed by a linked list of page-aligned, page-sized (or page-multiple-sized) blocks. Because each block is page-aligned and page-sized, we can use madvise to discard its physical backing while keeping the virtual address space reserved. Compressed pages therefore disappear from resident memory, but decompression is still guaranteed because the address space remains valid and we simply fault new pages back in as needed. We use the same trick for our memory pools, too. Unallocated pool pages don't count as resident memory, saving another couple of MB per terminal. This functionality is also available to libghostty-vt consumers via new `ghostty_terminal_compress` APIs. The consumer decides when the appropriate time to compress is and the APIs advise on compressability.

Ghostty is getting automatic scrollback compression, resulting in 70 to 90% less physical memory usage. It happens incrementally when idle, so it had no measurable effect on IO throughput. I'm not aware of any other mainstream terminal that does this. Demo video below! The gains let us increase the default scrollback limit from 10MB to 50MB, because on average a full scrollback will still compress smaller than the prior limit. More history, for free. ("Unlimited", disk-paged history is on the roadmap too) Let's talk about cool implementation details, cause this was fun. First, the data structure and memory layout ("PageList") I wrote two years ago finally pays off! One of its traits is that screen memory is backed by a linked list of page-aligned, page-sized (or page-multiple-sized) blocks. Because each block is page-aligned and page-sized, we can use madvise to discard its physical backing while keeping the virtual address space reserved. Compressed pages therefore disappear from resident memory, but decompression is still guaranteed because the address space remains valid and we simply fault new pages back in as needed. We use the same trick for our memory pools, too. Unallocated pool pages don't count as resident memory, saving another couple of MB per terminal. This functionality is also available to libghostty-vt consumers via new `ghostty_terminal_compress` APIs. The consumer decides when the appropriate time to compress is and the APIs advise on compressability.

Mitchell Hashimoto

166,748 Aufrufe • vor 1 Monat

Amika, our AI waifu, costs about ~$25K to develop up to today. She runs on our in-house R&D’d STT and TTS to achieve the sub-800ms response speed. Her brain is running on custom dynamic prompting system that we built ourselves. Running local LLM models only. Her initial development started around August 2025. We are on our third remake of architecture. She is powered by three RTX5090s. We do not have plans to open source her at this moment. We are planning on releasing her around September 2026. She is not a fork of any publicly available projects. Many confuses her with “Amica,” “Project AIRI,” or “Riko Project.” We are unrelated. We are a small team mainly based in the San Francisco Bay Area. While being one of the most advanced AI waifu, she is still heavily a work-in-progress. Follow our social media or join our discord (link in bio) for ongoing updates.

Amika, our AI waifu, costs about ~$25K to develop up to today. She runs on our in-house R&D’d STT and TTS to achieve the sub-800ms response speed. Her brain is running on custom dynamic prompting system that we built ourselves. Running local LLM models only. Her initial development started around August 2025. We are on our third remake of architecture. She is powered by three RTX5090s. We do not have plans to open source her at this moment. We are planning on releasing her around September 2026. She is not a fork of any publicly available projects. Many confuses her with “Amica,” “Project AIRI,” or “Riko Project.” We are unrelated. We are a small team mainly based in the San Francisco Bay Area. While being one of the most advanced AI waifu, she is still heavily a work-in-progress. Follow our social media or join our discord (link in bio) for ongoing updates.

MekaHime

44,001 Aufrufe • vor 4 Monaten

Stratosphere was our biggest and heaviest character on #Transformers ROTB. He was a real challenge to deal with on the Modeling, Texturing, and Rendering side. He had 1434 UDIMs due to his immense scale and was made up of tens of thousands of objects. His level of detail was truly something else. His vehicle form was also a behemoth. We had also built a mortar for his robot form and a cannon for his vehicle form but neither of these made it into the film as his part in the final battle was cut out. He was originally supposed to help out in the final battle, using both of these weapons against the Sweepers and Predacons. The decision to cut him from the final act came very late as all of his assets were final at that point. We were never told the reason why his role was cut from the film. I had to re-design my texturing system in Mari when working on him to ensure that it could handle such a complex and large character. I had to simplify a lot of the procedural systems and even had to branch off some of them into their own files to ensure that artists could work with somewhat decent performance. We ran into a lot of issues pushing him through our pipeline, both as a model as well as rendering him due to the sheer amount of objects and textures that had to be processed. For shots, it was requested by Lighting that we reduce his memory footprint as Google had complained he was taking up too much memory and their I/O for rendering on the cloud was being affected. The leadership team and I decided that the best way to do this was to half the texel density on the parts that wouldn't be visible in our approved shots and to also half the resolution of the textures in the parts not visible to camera. This allowed us to optimize his memory footprint both in shots as well as storage and I/O. It was massive technical undertaking working on Stratosphere, so much so that we had to delay getting him into shots as we simply couldn't push him through with our normal tooling. I really enjoyed the challenge that he posed for us as a team. CREDITS: Primary Modeling by Oscar Lowe Support Modeling by Arthur Grandjean and his team Final Texturing & Lookdev by Yaz Raji Video credits: Breakdown shot by MPC Film footage by Paramount

Stratosphere was our biggest and heaviest character on #Transformers ROTB. He was a real challenge to deal with on the Modeling, Texturing, and Rendering side. He had 1434 UDIMs due to his immense scale and was made up of tens of thousands of objects. His level of detail was truly something else. His vehicle form was also a behemoth. We had also built a mortar for his robot form and a cannon for his vehicle form but neither of these made it into the film as his part in the final battle was cut out. He was originally supposed to help out in the final battle, using both of these weapons against the Sweepers and Predacons. The decision to cut him from the final act came very late as all of his assets were final at that point. We were never told the reason why his role was cut from the film. I had to re-design my texturing system in Mari when working on him to ensure that it could handle such a complex and large character. I had to simplify a lot of the procedural systems and even had to branch off some of them into their own files to ensure that artists could work with somewhat decent performance. We ran into a lot of issues pushing him through our pipeline, both as a model as well as rendering him due to the sheer amount of objects and textures that had to be processed. For shots, it was requested by Lighting that we reduce his memory footprint as Google had complained he was taking up too much memory and their I/O for rendering on the cloud was being affected. The leadership team and I decided that the best way to do this was to half the texel density on the parts that wouldn't be visible in our approved shots and to also half the resolution of the textures in the parts not visible to camera. This allowed us to optimize his memory footprint both in shots as well as storage and I/O. It was massive technical undertaking working on Stratosphere, so much so that we had to delay getting him into shots as we simply couldn't push him through with our normal tooling. I really enjoyed the challenge that he posed for us as a team. CREDITS: Primary Modeling by Oscar Lowe Support Modeling by Arthur Grandjean and his team Final Texturing & Lookdev by Yaz Raji Video credits: Breakdown shot by MPC Film footage by Paramount

Rassoul Edji

268,448 Aufrufe • vor 1 Jahr

The #GTA 3 port to the Sega #Dreamcast has been progressing at an incredible pace. It's been amazing to see the whole DC community come together to tag-team this "impossible" project... Here it is running on a stock DC, no longer requiring the 32MB RAM hardware modification, only a few weeks into development. Since I finally got the time to sit down, build the codebase, and look into some of what I think is the critical path for performance, let's talk about some technical shit, and some of the future steps I think can be taken to further improve performance. First of all, I want everyone to take note that this is NOT a port of the PS2 version. This is a port of the PC version, which has extra content, increased draw distance, improved textures, and other things that have actually increased the challenge here... Whether the DC version will ultimately have these additions or not will remain to be seen, but we're running into plenty of shit that the PS2 didn't have to worry about (like these big-ass PC replay saves won't fit onto a Visual Memory Unit!) Secondly, lets talk about what is and isn't currently optimized, because it's absolutely vital that the DC's hardware is fully utilized here for the sake of performance and achieving a competitive polygon count. Unlike with modern devices, where the whole graphics pipeline is handled by the GPU, both the PS2 and Dreamcast were responsible for transforming and doing lighting calculations for each vertex BEFORE they got submitted to the GPU. The PS2 had a vector coprocessor to do this, while the Dreamcast had a few extremely important SIMD and fast math assembly instructions on its CPU to do these computations. Up until literally just a few hours ago (not shown in this footage), the Dreamcast's SH4 was doing 100% of these operations in slow-ass plain C and C++ code, which is absolutely sub-optimal and is immediately bogging down its CPU with just transforming vertices, bottlenecking the entire graphics pipeline on the fist T&L stage, and also leaving less CPU time for handling other gameplay logic... this is going to absolutely have to be addressed (and already has begun to be). Another issue that is crippling performance here is the fact that the models are all using individual triangles rather than triangle strips, which the Dreamcast's PVR GPU was designed to handle better... Converting these models to use strips rather than individual triangles will result in MANY different gainz for the DC, as you're going from 3N to N+2 vertices per triangle. Converting the models to triangle strips will 1) reduce load times, since model assets will be smaller 2) reduce the amount of video memory required to hold these vertices on the GPU 3) reduce the amount of shit that must be transferred from the CPU to the GPU and 4) give us back a bunch of CPU time, since the SH4 will be less bogged down transforming redundant vertices! TL;DR: This is still EXTREMELY suboptimal in terms of fully utilizing the graphical potential of the Dreamcast. There is going to be a LOT that can be done still to both improve performance and polygon counts, so stay tuned! FINALLY: Mad respect and love to Stefanos Kornilios Mitsis Poiitidis, for doing an amazing job leading this project, and to Frogbull , Esppiral, and everyone else who is helping us stick it to the PS2 by making this happen! #gamedev #retrogaming #cplusplus

The #GTA 3 port to the Sega #Dreamcast has been progressing at an incredible pace. It's been amazing to see the whole DC community come together to tag-team this "impossible" project... Here it is running on a stock DC, no longer requiring the 32MB RAM hardware modification, only a few weeks into development. Since I finally got the time to sit down, build the codebase, and look into some of what I think is the critical path for performance, let's talk about some technical shit, and some of the future steps I think can be taken to further improve performance. First of all, I want everyone to take note that this is NOT a port of the PS2 version. This is a port of the PC version, which has extra content, increased draw distance, improved textures, and other things that have actually increased the challenge here... Whether the DC version will ultimately have these additions or not will remain to be seen, but we're running into plenty of shit that the PS2 didn't have to worry about (like these big-ass PC replay saves won't fit onto a Visual Memory Unit!) Secondly, lets talk about what is and isn't currently optimized, because it's absolutely vital that the DC's hardware is fully utilized here for the sake of performance and achieving a competitive polygon count. Unlike with modern devices, where the whole graphics pipeline is handled by the GPU, both the PS2 and Dreamcast were responsible for transforming and doing lighting calculations for each vertex BEFORE they got submitted to the GPU. The PS2 had a vector coprocessor to do this, while the Dreamcast had a few extremely important SIMD and fast math assembly instructions on its CPU to do these computations. Up until literally just a few hours ago (not shown in this footage), the Dreamcast's SH4 was doing 100% of these operations in slow-ass plain C and C++ code, which is absolutely sub-optimal and is immediately bogging down its CPU with just transforming vertices, bottlenecking the entire graphics pipeline on the fist T&L stage, and also leaving less CPU time for handling other gameplay logic... this is going to absolutely have to be addressed (and already has begun to be). Another issue that is crippling performance here is the fact that the models are all using individual triangles rather than triangle strips, which the Dreamcast's PVR GPU was designed to handle better... Converting these models to use strips rather than individual triangles will result in MANY different gainz for the DC, as you're going from 3N to N+2 vertices per triangle. Converting the models to triangle strips will 1) reduce load times, since model assets will be smaller 2) reduce the amount of video memory required to hold these vertices on the GPU 3) reduce the amount of shit that must be transferred from the CPU to the GPU and 4) give us back a bunch of CPU time, since the SH4 will be less bogged down transforming redundant vertices! TL;DR: This is still EXTREMELY suboptimal in terms of fully utilizing the graphical potential of the Dreamcast. There is going to be a LOT that can be done still to both improve performance and polygon counts, so stay tuned! FINALLY: Mad respect and love to Stefanos Kornilios Mitsis Poiitidis, for doing an amazing job leading this project, and to Frogbull , Esppiral, and everyone else who is helping us stick it to the PS2 by making this happen! #gamedev #retrogaming #cplusplus

Falco Girgis

88,356 Aufrufe • vor 2 Jahren

Vibe Coding with Codex - Complete Guide Build a Web App, Desktop App & iOS App with Codex + GPT‑5.5 (No Coding Needed, Beginner Friendly) In this video you will learn: > Vibe Coding Basics + Vocab > How to build a web app using Codex > Add db, auth + storage with Firebase > Github Basics > Add AI Features (API's) > Deploy to internet (Vercel) > Convert Web App into Desktop app & iOS App Chapters 00:00 Intro 01:12 Setting up Codex 01:57 The basics of Vibe Coding and Codex 02:20 Projects, Files, App 03:45 Example App - Microsoft Paint 04:25 Running app locally 06:39 Save My Code - Use Github 10:37 Quick Review before building app 12:24 Building a web app - The Prompt 15:32 Creating Web App Project 16:34 Explaining Firebase (Database, Storage, Auth) 18:42 Setting up Firebase Project 22:54 Prompting Codex to build our app 24:49 Inspect Element - Console 26:24 Verify Data being Stored in Database 27:51 Making Changes to App 31:29 Fixing Storage Permissions with Codex 32:20 GPT API - Adding AI to our app 36:20 Making more changes using screenshots 39:02 Queuing vs Steering 40:23 Deploying our app to Vercel (App on Internet) 43:09 Convert web app to desktop app and iOS app 47:15 Web App and Desktop App work Now 48:36 Now let's run the iOS app 50:27 All three apps work! 51:11 Making Changes to iOS app 52:51 Testing agent skill feature of our app 53:55 Summary of what we did

Vibe Coding with Codex - Complete Guide Build a Web App, Desktop App & iOS App with Codex + GPT‑5.5 (No Coding Needed, Beginner Friendly) In this video you will learn: > Vibe Coding Basics + Vocab > How to build a web app using Codex > Add db, auth + storage with Firebase > Github Basics > Add AI Features (API's) > Deploy to internet (Vercel) > Convert Web App into Desktop app & iOS App Chapters 00:00 Intro 01:12 Setting up Codex 01:57 The basics of Vibe Coding and Codex 02:20 Projects, Files, App 03:45 Example App - Microsoft Paint 04:25 Running app locally 06:39 Save My Code - Use Github 10:37 Quick Review before building app 12:24 Building a web app - The Prompt 15:32 Creating Web App Project 16:34 Explaining Firebase (Database, Storage, Auth) 18:42 Setting up Firebase Project 22:54 Prompting Codex to build our app 24:49 Inspect Element - Console 26:24 Verify Data being Stored in Database 27:51 Making Changes to App 31:29 Fixing Storage Permissions with Codex 32:20 GPT API - Adding AI to our app 36:20 Making more changes using screenshots 39:02 Queuing vs Steering 40:23 Deploying our app to Vercel (App on Internet) 43:09 Convert web app to desktop app and iOS app 47:15 Web App and Desktop App work Now 48:36 Now let's run the iOS app 50:27 All three apps work! 51:11 Making Changes to iOS app 52:51 Testing agent skill feature of our app 53:55 Summary of what we did

Riley Brown

147,144 Aufrufe • vor 3 Monaten

In our research lab, we are building “real-time dreaming” - the ability to generate fully playable video worlds prompted from any text or image. Our real-time, action conditioned world model (currently running internally at 16fps at 832x480p) is trained on a combination of data, including proprietary Roblox 3D avatar/world interaction data. World models are different from multiplayer engines in that they store state and memory in video latents. Roblox is multiplayer, and we are actively researching optimal ways to simultaneously store state for thousands of players, and keep them in sync with their environment. Our world model leverages database technology which stores all user interactions on Roblox in a vector format that can be used to re-render video and interaction from any camera angle. We see several immediate uses for our Roblox world model. We will use it side-by-side text, image and video prompts as a way to launch auto-generation of immersive worlds. In Roblox Studio, a creator could walk around and use prompts to “paint” a world and then convert it into a 3D representation or direct to Roblox native as a way for many people to play simultaneously. All of this comes alive as we explore the notion of a “Dream Theater” - where one user is dreaming, while others watch and prompt them. 2/4

Roblox

4,404,098 Aufrufe • vor 6 Monaten

Today I am introducing Eragon , your AI Operating System for work. Our goal is simple: Build the AI that can help you run your company from idea to IPO. In order to truly fix work, we need to solve the root problem. We all use tons of business apps from ERP to CRM to Product Analytics but we still have to endlessly switch, search, and export just to answer a single question. The bottleneck is not having more tools, it's the lack of a unified layer that connects and understands them all. That's why we built Eragon -Through MCP we connect to every business app. -Through custom agents, each tool gets its own intelligence. -And inside one workspace all you have to do is ask Eragon for any business insight: - "Walk me through our sales forecast and what deals are at risk." - "Which users hit paywall but didn't convert?" - "Where's the bottleneck in our factories this week?" Already, teams are cutting days of manual work into minutes. We're working with the fastest growing startups and some of the worlds largest enterprises. And we would love to work with you! Book a demo with us today and usher in the new era of vibe working.

Today I am introducing Eragon , your AI Operating System for work. Our goal is simple: Build the AI that can help you run your company from idea to IPO. In order to truly fix work, we need to solve the root problem. We all use tons of business apps from ERP to CRM to Product Analytics but we still have to endlessly switch, search, and export just to answer a single question. The bottleneck is not having more tools, it's the lack of a unified layer that connects and understands them all. That's why we built Eragon -Through MCP we connect to every business app. -Through custom agents, each tool gets its own intelligence. -And inside one workspace all you have to do is ask Eragon for any business insight: - "Walk me through our sales forecast and what deals are at risk." - "Which users hit paywall but didn't convert?" - "Where's the bottleneck in our factories this week?" Already, teams are cutting days of manual work into minutes. We're working with the fastest growing startups and some of the worlds largest enterprises. And we would love to work with you! Book a demo with us today and usher in the new era of vibe working.

Josh Sirota

99,959 Aufrufe • vor 11 Monaten