GPU

China already has a GPU that competes with Nvidia’s RTX 3060. The bad thing is that it arrives five years late and worse

by

The china crusade for achieving the complete independence in the field of semiconductors has taken a new step. The problem is that this step has not been as promising as we expected, and in fact it makes it clear that today the Asian giant is still far away of the semiconductor manufacturers that dominate the market. The alternative for gamers that promised. Lisuan Tech (砺算科技), a Chinese company dedicated to manufacturing semiconductors and solutions such as graphics cards for the end-user market, has launched its new GPU for the consumer market, the LX-7G100. The price and expectations. The official starting price is 3,299 yuan (about 420 euros at the exchange rate), and at that price the equivalent graphics card should be at least an RTX5060 Ti, which is usually below 400 euros. What we get in performance is far from that. Performance tests of the LX-7G100 typically fell well short of the RTX 3060.Source: NotebookCheck. Worse than the RTX 3060. The problem is that those who have had access to this graphics card and have evaluated their benefits They have realized that this manufacturer’s GPU is very far from that price/performance estimate. In fact, it usually competes more with the RTX 3060 of 2021, but even with it it loses: it offers approximately 65% of the performance from its rival NVIDIA. Good specifications. On paper, the LX-7G100 should offer more performance. It has a 7G106 GPU, 12 GB of GDDR6 memory and decent bandwidth, for example. However, it does not have truly mature support for DX12 and does not offer an alternative to Nvidia’s DSLL or AMD’s FSR. When used in modern games, performance plummets due to rendering glitches and code translation bottlenecks. Not even for AI. At Lisuan Tech they have also tried to bet on their ability to run local and private AI models. However, most of the development of AI projects is linked to Nvidia’s CUDA architecture. It is true that the Chinese company has its own compatibility layer to translate PyTorch and CUDA code to its native architecture, but the loss of efficiency is notable, which makes inference or local model training tasks become too slow compared to those allowed by Nvidia graphics. difficult to compete. Lisuan Technology announced the first milestones of this launch a year ago. The rumors they indicated that its G100 graphics processor is manufactured by SMIC with a 6nm photolithographic process that complies with US restrictions. An attempt was made to launch in 2023, but Lisuan had financial problems and a capital injection of $27.7 million managed to keep the project going. It remains to be seen if sales ultimately follow through, although certainly its price/performance ratio makes it attractive only to audiences like the Chinese, who may have more difficulties accessing models like the Nvidia RTX. In Xataka | The end of Nvidia in China seems to be very near: its current market share is 0%

It already has permission to sell its H200 GPU to 10 Chinese companies

by

Alibaba, Tencent, ByteDance and JD.com are four of the ten Chinese companies that already have access to the GPU for artificial intelligence (AI) NVIDIA H200. According to Reutersthe US Department of Commerce, which is the institution that grants or denies export licenses, has authorized at least ten Chinese companies and several distributors, including Lenovo and Foxconn, to acquire Nvidia’s second most powerful AI chip. This news comes almost two months after the US Government confirmed which was going to allow the company led by Jensen Huang to deliver its H200 chip to its Chinese customers. Nvidia announced in mid-March during its annual developer conference that the US and Chinese Administrations had unlocked the sale of this GPU in the nation led by Xi Jinping. However, so far not a single delivery has been made. In practice, the blockade continues despite the March announcement. In all likelihood this is why Jensen Huang has joined the White House delegation participating in a summit with Chinese President Xi Jinping this week. Nvidia is caught between the opposing interests of the US and China, and Huang is going to try to recover a market, the Chinese one, valued at 50 billion dollars in 2026 and which has come to represent 13% of its income. Now the problem is the Chinese Government Earlier this May, Jensen Huang confirmed that he is currently Its market share in China is 0%. Nvidia has not sold its AI chips in this country for several months because US regulations require Chinese buyers to demonstrate that they have implemented sufficient security procedures and that they will not use the GPUs for military purposes. In addition, Nvidia must also certify that it has sufficient inventory in the US. And all this bureaucracy is not being resolved quickly at all. Currently the greatest reluctance to sell Nvidia chips in China comes from Beijing However, currently the greatest reluctance to sell Nvidia chips in China comes from Beijing. The Chinese Government wants to promote developing your own GPUs for AI at any price, which in October 2024 led him to send a recommendation to Chinese AI companies in which he asked them to use chips produced in China as much as possible. Ten months later this recommendation became a requirement. And the Chinese Government is already forcing state-owned data centers throughout the country to use at least 50% Chinese integrated circuits in their servers. The Administration led by Xi Jinping has made this decision because it can afford it. And it is that It already has three very clear alternatives to Nvidia: Cambricon Technologies, Huawei and Moore Threads. On the other hand, in the US there is also a pressure group that opposes the sale of advanced US AI chips in China. Chris McGuire, senior fellow on China and emerging technologies at the Council on Foreign Relations, holds that “any deal that allows Nvidia to sell more chips to China means fewer Nvidia chips for US companies and a minor US advantage over China in AI“. Besides, McGuire argues that “it is surprising that President Trump continues to allow himself to be convinced to put Nvidia’s interests before those of America.” Image | Nvidia More information | Reuters In Xataka | The US remains committed to stopping China. Now it has targeted the second largest Chinese chip manufacturer

China is preparing the most powerful and rare exascale supercomputer on the planet. No GPU: only Chinese CPUs

by

An exascale supercomputer is one capable of performing at least 1 exaflop (10¹⁸) of floating point operations per second. These machines are the most powerful currently available if we stick to classic computers and leave aside the prototypes of quantum computers. The classification TOP500 identifies the most capable supercomputers on the planetand, as expected, four exascale machines appear at the top of this list: The Captain, FrontierAurora and Jupiter. The first three reside in the United States and the fourth in Germany. Curiously, no Chinese supercomputer appears in the top ten positions of this classification, although we know that some of its most powerful machines are not officially reported to the TOP500 for geopolitical reasons. Be that as it may, the Government led by Xi Jinping is determined to change this scenario. And the Shenzhen National Supercomputing Center has announced that is going to build a supercomputer called Lingshen that, according to this institution, will have a sustained performance of more than 2 exaflops and will integrate only components designed and manufactured in China. Lingshen supercomputer architecture is very unusual The supercomputer ‘The Captain’ from the Lawrence Livermore National Laboratory (USA) is a real beast. This machine exceeds 1.8 exaflops, making it currently the most powerful on the planet. The APUs are responsible for its brute force. Instinct MI300A from AMD, which work hand in hand with the EPYC 9005 processors. However, the most surprising thing is that it brings together no less than 11,340,000 cores and delivers 1,809 PFlops/s Rmax and 2,821.10 PFlops/s Rpeak. Lingshen will bring together 47,000 processors of Chinese origin that will be distributed in Huawei Kunpeng servers The architecture of ‘El Capitan’ is very similar to that of the other supercomputers in the TOP500 classification, but the machine being prepared by the Shenzhen National Supercomputing Center is going to take different paths. And it is that according to Lu Yutongthe director of this center, the Lingshen supercomputer will use only general purpose processors (CPU), and will not use GPU. Not a single one. It is a very unusual decision, and it is surprising that in theory it will exceed 2 exaflops only with this type of chips. Be that as it may, this is not the only thing we know. Lingshen will bring together 47,000 processors of Chinese origin that will be distributed in servers Huawei Kunpeng equipped with Taishan cores with ARM architecture. Lu Yutong has also confirmed that this machine will have 650PB of storage and a million-port interconnection. Everything that the Shenzhen National Supercomputing Center has announced sounds great, but this project also leaves us with some very reasonable doubts. The most obvious is that Lingshen is just a project at the moment. It has not yet been built, so its theoretical maximum performance comes from an estimate and not from a measurement provided by a real test bench. On the other hand, it is very surprising that the Shenzhen National Supercomputing Center has chosen to integrate only CPU. Huawei, Moore Threads and Cambricon Technologies are three of the chinese companies which have domestically made GPUs that could presumably fit into this machine. In any case, it is worth keeping track of this project to see if Lingshen finally lives up to the expectations it has raised. Image | TOP500.org More information | Shenzhen National Supercomputing Center In Xataka | The Frontier supercomputer is the second most powerful exascale machine on the planet. And it has a mission: nuclear fusion

We already know what happens to the GPU hourly price when OpenAI or Anthropic launch a new model: it doubles

by

This week, an analyst named Tomasz Tunguz published in X two revealing graphs. They show the evolution of what it costs AI startups to access cloud computing, and there is bad news. The cost of renting the NVIDIA B200 GPUs with Blackwell architecture has gone from $2.31 per hour in early March to $4.95 per hour this week. It is an increase of 114% in just six weeks and it has a clear cause: the arrival of new models from Anthropic and OpenAI. What the graphs show clearly. Those charts focus on the price index of Ornna cloud computing trading marketplace. The first of them covers the price of renting the B200 chips from the end of 2025 until today, and there are vertical lines showing each release of the latest models from OpenAI and Anthropic. The correlation is almost perfect: GPT-5 Codex, Claude 4.5, GPT-5.3 Codex, Claude Opus 4.7 and GPT-5.5 coincide with a jump in price indices. Every time these companies announce a new version of their frontier models, demand skyrockets, and so does the cost. If you want the best, pay (much more). The second graph shows the price difference between renting the previous generation of chips, H200 with Hopper architecture, and the new B200. The historical average of that “spread” is $1.06, but now it stands at $2.09, practically double. That means buyers—startups and AI companies—are paying a record premium for the extra memory and superior computing power of Blackwell architecture chips. Accessing the latest of the latest was already expensive. Now it is even more so. This also makes the H200 in a second class option for the most demanding models of 2026. Action and reaction. There is overwhelming logic here. When OpenAI or Anthropic release a new model, there is an explosion in inference. Developers and companies want to test them as soon as possible and integrate these models into their products (or compete with them). To do this, they need computing quickly, and a simultaneous demand is caused that unbalances the available inventory in the market for renting AI chips by the hour. The problem is that the supply of B200 does not grow at the same rate. Some companies have wanted to anticipate, and we have the perfect example in Google. He has bought all the B200s he can, and that has made these GPUs around now the 500,000 dollars on the secondary market according to analyst Jack Minor. The irony of efficiency. The curious thing is that the more efficient these chips are – and the B200s are – the more companies want to rent them at the same time to take advantage of those efficiency advantages that should lead to cost savings. What actually happens is that the scarcity of these advanced chips cancels out any theoretical savings. Long term contracts. Startups and companies that think in the short term are especially harmed in this area, because they face price jumps that are increasingly difficult to assume. Companies that signed computer rental contracts at the price then can now operate at less than half the cost of their competitors. Thinking in the medium or long term seems reasonable, although once again those who win are the hyperscalers and those companies that have managed to get hold of many B200s. And who wins even more is of course NVIDIA, which cannot cope. Few alternatives. In other markets such as energy or metals there is usually room for maneuver, Tunguz points out, but the same is not happening at the moment in the AI ​​segment. In the oil market, for example, if the price rises 114% in six weeks, companies can buy futures, options or fixed-price supply contracts to protect their margins. In cloud computing rental, those options are much more limited. And the result is a much more volatile segment. This will go further. We are probably facing a peak in demand that will be followed by a correction: the new batch of B200 chips that arrive in the second half of 2026 are expected to cause a drop in current prices. However, that $4.95 is now the new floor, not a peak, because demand for AI computing will continue to grow faster than TSMC’s production capacity. In the absence of the supply of AI chips growing significantly – and there are certainly movements that are trying to achieve this, such as those of Google with its TPUsAmazon with its Trainium or Huawei with its Ascend—, the problem will still be there. In Xataka | Europe is taking its technological independence so seriously that it is aiming for the most ambitious goal: NVIDIA

You can make money with your GPU when you don’t use it. It is enough that you lend it to those who train AI models

by

To execute and offer tools based on generative artificial intelligence, a lot of calculation power is needed (and that leads to a lot of energy). Therefore, the most powerful market cards and specific processors for Datacenters They are so quoted today, hence companies such as Nvidia, which specializes in this market, are reaping such an overwhelming success. And since not everyone can afford a powerful graphics card to experiment with AI, there is a service that we see more and more common: to rent a graphics card to remove an extra money. There are several platforms to get it and under these lines we tell you everything you need to know. How the business works. The model consists of acting as a host in a Marketplace where clients are looking for GPU instances for their AI projects. You set the price per hour, the platform manages payments and the client executes their work in an isolated container on your machine. You could say that it is like an Airbnb, but focused on computer hardware. Instances with an rtx 4090 in vast ai Numbers that we must take into account. An RTX 4090 is usually Between about 0.20 and 0.60 dollars per hour in these marketplaces, depending on the demand. In the best theoretical scenario, operating 24 hours a day for a full month, a high -end GPU could invoice around 240 gross dollars monthly (considering that we put it for rent 24 hours a day). But reality is usually more modest, since we have to discount what we pay on our electrical bill, the platform commissions (which can reach 24% in Platforms like Runpod) and, above all, that real occupation is rarely 100%. Expanding market. The price difference between traditional cloud giants (AWS, Google Cloud) and these P2P marketplaces is considerable. While renting a GPU on AWS can cost three or six times more, platforms such as Runpod or Vast AI offer access to very powerful graphics cards, as is the case of RTX 4090, for a few cents the time. And of course, these prices are really attractive to developers who want to experiment with artificial intelligence but do not have means to have a team comparable to the projects they work on. What you should know before starting. Turning your PC into a rental server is not plug-and-play. In most cases you need Install Linuxconfigure updated NVIDIA drivers, open network ports And keep your team working for the hours for which you have committed to offer it, together with adequate refrigeration, which will be necessary if your GPU is going to start working much more and for much longer. In addition, your customers expect the machine to be available when they hire it, which means that you will not be able to use it for gaming or personal work. It should also be noted that the income generated is also subject to taxation and it is possible that it is required to register as an economic activity in cases where income exceeds a certain threshold. There are certain risks. Beyond the wear that the hardware can receive for being constantly working, there are maximum performance, there are some security concerns. Although platforms use containers to isolate workloads, some experts warn about possible Vulnerabilities in multi-tean environments (those environments that serve several users) that could compromise our data or use the GPU to improper purposes. Is it worth it? For most users with a single GPU, the benefits are modest once all expenses and others are discounted. Now, the business makes more sense if you already have the amortized hardware, do not pay too much on your electrical bill and accounts with certain technical knowledge to maintain the stable system. Even more if you have a potential graphics card or level for datacentes. As an experiment or complementary income experiment it can be interesting, but do not expect it to make you rich. First steps. If you want to try it, start with offers “interruptibles“, that is, cheaper but that can be canceled, in order to know the real demand. Vast.ai and Runpod They offer detailed documentation to become host, including step -by -step guides and preconfigured templates. Of course, it is advisable to always control real electrical consumption and establish operation limits to prevent your equipment from becoming a slave to the background processes. Cover image | She Don In Xataka | Nvidia, TSMC and SK Hynix are the most powerful chip companies on the planet. None can allow any of the others to fall

When you are openai and you can’t buy enough GPU, the solution is obvious: manufacture yours

by

Openai will create its own artificial intelligence chips. It is a crucial decision for the future of your business, but the ally you have chosen to do it: Broadcom. When the river sounds. The runrún has been listening Since the beginning of 2024. Nvidia, owner and lady of the segment of the AI ​​accelerators, was an ally too powerful for OpenAi. The solution was clear: develop its own chips with which to minimize that dependence. Broadcom takes chest. Hock Tan, Broadcom CEO, yesterday told investors that the company had closed an agreement with a mysterious client that would invest 10,000 million dollars in AI chips. Although Broadcom did not reveal the client’s name, sources close to those agreements indicated In Financial Times that this client is none other than OpenAi. Neither Broadcom nor this last company have confirmed the data. Xpus to power. Those chips, to those who referred as Xpus, are a kind of specialized and personalized variant of the NVIDIA or AMD accelerators. We have the perfect example in the TPUS (Tensioner Processor Units) that Google presented almost a decade ago And that has been improving generation after generation (we are already going for the seventh generation, called Ironwood). Broadcom, by the way, has collaborated in the development of these Google chips, so it has overdue experience in that area. Own chips for internal use. According to sources close to this collaboration, Openai aims to use these internal AI chips, and there are no plans to offer them to external clients. That reinforces the theory that Openai wants to create data centers with these own chips to avoid (or at least mitigate) the dependence of Nvidia. Nvidia will have (a lot) competition. Nvidia dominates Iron fist This segment, but has long for the rivals – both in the West as in the East – work to make their monopoly in this sector disappear. Microsoft He has MaiaAmazon His trainiumGoogle its aforementioned TPU and AMD of course Your instinct. To goal It is about it. But Cuda remains the “Moat” of Nvidia. Of course the true key to Nvidia overwhelming success is not so much in its chips and in the fact that its architecture CUDA is de facto standard In this market and all AI systems developers usually base their projects on that platform. It is the “Moat” of Nvidia, that “pit” that allows you to protect its “castle” from the rivals and continue dominating the market. And here there are also attempts to avoid the dependence of the company, and among them Those from China stand out. And TSMC, what? The funny thing is that for months it seemed that The ally that Openai had sought To carry out this project was the most important semiconductor manufacturer in the world, TSMC. Earl this year that collaboration It seemed to go on the right track and several sources pointed out that we would have the first OpenAI GPUS for 2026. It may simply have chosen to have a plan B (TSMC) to avoid its dependence on NVIDIA, but also prepare a C (Broadcom) plan. Image | Qualcomm In Xataka | China’s self -sufficiency test in chips for AI is already here: it has not bought Nvidia or a single H20 GPU in the last quarter

Nvidia world leadership in chips for AI is brutal. In GPU for games directly has fulminated the competition

by

Nvidia dominates the global chips market for artificial intelligence (AI) with a fee that during the last three years has oscillated between 80 and 94%, according to Fourweekmba. Your leadership is supported by A very competitive hardware and a software ecosystem in which CUDA (Compute Unified Device Architecture) It has an essential role. This technology brings together the compiler and development tools used by programmers to develop their software for NVIDIA GPUs. Most of the artificial intelligence projects that are currently being developed are implemented on CUDA, and replace it with another option in the projects that are already underway it is a problem. Huawei, who aspires to an important portion From this market in China, it has Cann (Compute Architecture for Neural Networks), which is its alternative to CUDA. AND Moore Threads and Cambricon Technologies They have muse and neuware respectively. Even so, the competitors of Nvidia will cost them a lot to break the leadership of Cuda. Nvidia has distributed 94% of GPUs for market games During the second quarter of 2025, 11.6 million graphics cards for PC and 21.7 million processors for desktop computers have been distributed throughout the planet, according to the US consultant Jon Pedie Research. By themselves these figures do not tell us just anything, but they acquire the relevance they deserve if we consider that they indicate that the distribution of graphics cards has grown by 27% and that of CPUs 21.6% compared to the first quarter of 2025. Distributed units allow us to train a very precise idea about market behavior It is important that we do not overlook that these figures quantify the distributed units, and not the units sold. However, there is a direct correlation between them, so the distributed units allow us to form a very precise idea About market behavior. Anyway there is a fact that is even more shocking than all we have collected so far in this article: NVIDIA has distributed no less than 94% of GPUs for market games during the second quarter of 2025, again according to Jon Pedie Research. AMD has been forced to settle for 6% of the distributed units, and Intel does not even appear in the report of this consultant because its presence is anecdotal. So are things in the graphic hardware market for PC. One more note to conclude: the rebound of distributed units of graphics and processors for PC during the second quarter of 2025 against the first responds in all likelihood to the need for stores and users to supply before Tariffs approved by the US government They entered into force. Image | Xataka More information | Jon Pedie Research In Xataka | Nvidia is ready for the chip for the need to survive in China. Who is not ready to let him sell is the US government

He has not bought Nvidia or a single H20 GPU in the last quarter

by

The future of Nvidia in China is every day that spends more gloomy. In early October 2024 Chinese administration arrived to the companies of artificial intelligence (AI) A recommendation in which I asked them to use chips produced in China as much as possible. Ten months later This recommendation has been transformed into a requirement. And is that the Chinese government is already forcing data centers that belong to the State throughout the country to use at least 50% of Chinese integrated circuits on their servers. On the other hand, Nvidia has achieved the export license you need to sell in China Your GPU for IA H20but the Chinese government has vetoed this chip. And he has done so that the administration of the cyberspace of China, which is the main Internet regulatory body in this country, This GPU is thoroughly investigating Because he suspects that he could incorporate a rear door of difficult location by Chinese experts. If so, the possibility of China to use this chip. The direct consequence of this unfavorable scenario for Nvidia is that during the last quarter it has not sold a single H20 GPU in China, As Shaun Rein holdsan expert in the Chinese economy and founder of the consultant The China Market Research Group (CMR), which is housed in Shanghai. This statement is true, but it has a small trick. For a good part of the last quarter Nvidia did not have the export license that he needed to deliver this chip to his Chinese clients, but he already has it. And it could have sold thousands of these GPUs during the last weeks. China has alternatives designed to compete with Nvidia chips Despite the efforts of the US government to avoid it, the avant -garde chips for ia They have continued arriving in China. And they have done it mainly through secondary markets and parallel import roads that run in India, Malaysia or Singapore, in which The US action It is very limited. In addition, the developers of great AI models that have projects with projects with CUDA They have found the appropriate place to get these GPU: The international second -hand market. Cambricon Technologies is one of the companies specialized in the design of GPU for AI with greater growth potential Anyway China already has three alternatives Very clear to Nvidia. Although it is not as well known as Huawei or Moore Threads, Cambricon Technologies is one of the companies specialized in the design of GPU for AI with greater growth potential. In fact, he has received the approval of the Shanghai bag (China) to raise 560 million dollars. It will allocate them to the design of four chips for training and inference of AI models, and also to the development of an alternative to CUDA. On the other hand, Moore Threads He has developed several GPU for AI applications that, on paper, rivaize some of the advanced solutions that have placed in the Nvidia, AMD or Huawei market. The MTT S4000 and MTT S3000 cards are its most interesting proposals right now, although, curiously, in its porpholio the MTT S80 card, a proposal for games and content creation that, according to Moore Threads itself, has a 14.4 TFLOPS calculation capacity also appears in Floating Coma operations of simple precision. The other indispensable actor in the Chinese chips industry for IA is Huawei. His most ambitious proposal right now is the chip Ascend 910dwho seeks to overcome the performance of the GPU NVIDIA H100. However, this Chinese company has also recently presented its chip Ascend 920a solution that is clearly destined to occupy in the Chinese market The gaps that the NVIDIA H20 GPU is going to leave. This proposal will enter large -scale production during the second half of 2025 using 6 NM integration technology that have presumably developed elbow with Huawei elbow and SMIC (Semiconductor manufacturing international corp). Image | Nvidia | Zhang Kaiyv More information | Shaun Rein In Xataka | The US gives Huawei a great opportunity: to get its new chip for AI with the Nvidia market in China

presumably already has its first 6 -nm GPU for full games China

by

In China there are dozens of GPU designers for games and artificial intelligence (AI). The Chinese government has supported the proliferation of these companies with Very juicy subsidies in answer to US sanctions. Huawei, Metax, Biren Technology, Moore ThreadsInnosilicon, Zhaoxin, Iluvatar Corex, Deglinai or Vast Ai Tech are some of the most important, but there are more. Many more. And among all of them, it has just attracted the attention of the semiconductor industry: Lisuan Technology. This emerging company was born in 2021, in the same breeding ground triggered by The conflict held by the US and China which has also resulted in the constitution of other best -known graphic hardware companies, such as Moore Threads or Biren Technology. Although, as we have just seen, it is a very young firm, it is supported by veteran engineers who have developed a good part of their professional career in the US. It is exactly the same that happens with Moore Threads or Bire Technology, who tell in their ranks with former Nvidia employees. Lisuan has announced that it has a GPU as powerful as the GeForce RTX 4060 This week this company has revealed In your Wechat account Something important: it already has a GPU for games as powerful as the GeForce RTX 4060 of Nvidia. This is what Lisuan says, so the most prudent is that we collect it with reservations. When this graphic hardware is finally available we will check if this statement is reliable or not. However, this is not all. The most relevant thing is that this company ensures that this GPU, which it has baptized as G100, is the first manufactured entirely in China using 6 Nm integration technology. SMIC is already manufacturing in its nodes of 6 Nm the GPU for the Ascend 920 of Huawei Lisuan has not confirmed which Chinese manufacturer of semiconductors is producing this chip, but in all likelihood it is about SMIC (Semiconductor manufacturing international corp), The largest Chinese manufacturer of integrated circuits. This company You are already manufacturing in its 6 Nm nodes THE GPU FOR IA Ascend 920 From Huawei, so it is perfectly credible that this chip for Lisuan Technology produces in those same nodes. To manufacture integrated 6 and 7 nm circuits using equipment deep ultraviolet photolithography (UVP) of the Dutch company ASML SMIC uses a technique known as Multiple patterning. This strategy consists in transferring the pattern to the wafer in several passes with the purpose of increasing the resolution of the lithographic process. It works, but is responsible for the performance by wafer Be clearly improvable. In any case, Lisuan has revealed some more data about its G100 GPU. Apparently this graphic processor will work side by side with a generous vram memory map (it is only a conjecture, but it will possibly be 16 GB), will have moderate energy consumption And it will be compatible with the API Directx 12, Vulkan 1.3, OpenGL 4.6 and OpenGL 3.0, so it should be able to deal with current video games without problem as long as their controllers are up to it. Whatever the first graphics cards equipped with this GPU will be available during the third quarter of this year, although it is likely that large -scale manufacturing arrives at the beginning of 2026. Image | Lisuan Technology More information | Lisuan Technology In Xataka | We can forget an AI without hallucinations for now. The general director of Nvidia explains why

Huawei plans to advise Nvidia in China. It has a new GPU for theory that in theory is extremely powerful

by

Huawei is putting all the meat on the grill to absorb so much share in the Chinese GPU market for artificial intelligence (AI) as I can. And it is that the entry into force of the last US sanctions package is compromising with all probability Nvidia leadership in China. The US Department of Commerce It has imposed restrictions to the export to the country led by Xi Jinping of The H20 GPUand this in practice means that this chip presumably will not reach the Chinese clients of Nvidia. This last company has announced that this ban will cause a hole in its accounts of 5,500 million dollars Due to the commitments linked to the H20 GPU that had already acquired the reserves of this chip that it will not finally satisfy. Some of the Chinese companies that have bought large amounts from the H20 chip to NVIDIA and who presumably planned to continue doing them are Tencent, Alibaba or Bytedance, but at the current situation they will have to resort to an alternative. And Huawei has put it on a tray. The GPU Ascend 910D aspires to snatch the leadership in performance from Nvidia Huawei reacted immediately to US sanctions. And is that just a few hours after the entry into force of the new regulation of the Department of Commerce He presented his chip for the ascend 920a solution that is clearly destined to occupy in the Chinese market the gaps that the NVIDIA H20 GPU is going to leave. This proposal will enter large -scale production during the second half of 2025 using 6 NM integration technology that have presumably developed elbow with Huawei elbow and SMIC. Until now Huawei wanted to get his hardware to dominate the inference processes in AI However, this is not the only asset that Huawei has to increase its market share both in China and beyond its country of origin. And is that, According to Reutersthis company is preparing to start the testing and validation phase of a new GPU for AI: The Ascend 910D chip. Unlike the GPU Ascend 920 that, as we have seen, presumably aspires to compete with the NVIDIA H20 chip, the GPU Ascend 910D seeks to overcome the performance of the chip NVIDIA H100. If this movement is confirmed, already priori this information is reliable, it will be evident that Huawei will have chosen to fight in all hardware market segments for the nvidia. Until now this Chinese company wanted to get its hardware dominate the inference processes in AIand not the training of the models, as Georgios Zacharopoulos, a senior researcher of AI who works on the acceleration of inference in the Huawei laboratory in Zurich (Switzerland) points out in this statement. “The training is important, but it only happens a few times. Huawei focuses mainly on inference, which will ultimately give us access to more customers,” says Zacharopoulos. Inference is broadly the computational process carried out by language models with the purpose of Generate the answers which correspond to the requests they receive. In any case, the information we have reflects that the GPU Ascend 910D will allow Huawei to compete with the chips for the most advanced NVIDIA both in inference and in training. Image | Huawei More information | Reuters In Xataka | In a low voice, China has begun to remove some tariffs from US products. Your concern: the chips

Log In

Forgot password?

Forgot password?

Enter your account data and we will send you a link to reset your password.

Your password reset link appears to be invalid or expired.

Log in

Privacy Policy

Add to Collection

No Collections

Here you'll find all collections you've created before.