#High_Performance_Computing

[aresdifesa]

Partnership tra Leonardo e Cineca per High Performance Computing ed Intelligenza Artificiale Strategica la sinergia nel campo del calcolo ad alte prestazioni per tradurre la ricerca scientifica in applicazioni industriali concrete

[phonemantra-blog]

RTX 3060 is in the lead, but only by a small margin Valve once again messed up something in its Steam statistics and now it had to fix it.  Last month we said that the share of the RTX 3060 increased one and a half times over the month , reaching almost 10%, and this is a huge indicator of both growth and absolute numbers. Now another month has passed, and the share of the RTX 3060 has sharply dropped to 5.04%, falling below the September figure. This has already happened with Steam statistics, some attribute it to some kind of correction due to data from China, but now the RTX 3060 figure looks more plausible.   [caption id="attachment_85263" align="aligncenter" width="756"] GeForce RTX 3060[/caption] GeForce RTX 3060 is not that popular at all. At the same time, the share of the GTX 1650 increased by 1.13 percentage points, which is a lot. The card is installed on 4.76% of all Steam players. The third place with a share of 4.32% is occupied by the GTX 1060.   In general, the performance of many adapters has been adjusted. For example, the share of the RTX 4060 decreased from 1.749% to 0.83%.  Similar changes have occurred with processors. The sharp jump in Intel's share a month ago apparently was some kind of mistake, so everything was corrected, and now Intel occupies 66.08%, and AMD has 33.88%.

[jobrxiv]

Postdoctoral Position at the Intersection of AI and Cloud Computing Vanderbilt University Vanderbilt University: hiring a postdoctoral position for a research project on using AI/deep learning to optimize cloud computing resource usage! See the full job description on jobRxiv: https://jobrxiv.org/job/vanderbilt-university-27778-postdoctoral-position-at-the-intersection-of-ai-and-cloud-computing/?feed_id=88908 #AI #artificial_intelligence #cloud_computing #deep_learning #high_performance_computing #ScienceJobs #hiring #research

[phonemantra-blog]

It occupies four expansion slots Another video card with a Noctua cooler has appeared on the market. And she's incredibly fat again. [caption id="attachment_85159" align="aligncenter" width="780"] GeForce RTX 4070 Ti[/caption] Sycom introduced the GeForce RTX 4070 Ti Silent Master, which received Noctua fans. As you can see in the photo, without fans this is a two-slot adapter, but with them it already takes up four slots. Actually, there are simply no high-performance models with Noctua fans (at least for three expansion slots), although the same Sycom previously released a two-slot RTX 4060 Ti . GeForce RTX 4070 Ti Silent Master presented with Noctua fans Like the lower-end cards, the new product uses Noctua NF-A12x25 PWM fans and a special frame for mounting them, developed by Nagao. As the name implies, the new product should stand out primarily for its quiet operation. Compared to the MSI model, according to Sycom's own measurements, the noise level is much lower: 43.6 versus 55.1 dB. Temperatures are also lower, but not as significantly: 74.5 versus 79.3 degrees.

[phonemantra-blog]

They will be shown at CES 2024 Promotional materials have appeared on the Internet with photos of Dell and Alienware laptops, and some of them will be equipped with the latest Intel Core Ultra processors. [caption id="attachment_85043" align="aligncenter" width="780"] Intel Core Ultra processors[/caption] This is what laptops with new Intel Core Ultra processors look like. The images show the XPS 13, 14, 16, as well as the Alienware m16 R2 and m18 R2. All new products will receive a 14th generation Intel Core CPU, but we know that this same 14th generation will include both Raptor Lake Refresh and the latest Meteor Lake. But only the latter will bear the names Core Ultra. Also, some laptops will have discrete graphics cards of the GeForce RTX 40 line. There are no other technical details yet, but these are all fairly high-end laptops. They should be announced at CES 2024, like many other laptops based on the new Intel processors.

[phonemantra-blog]

This is one and a half times faster than the overclocked Xeon W9-3495X The monstrous 96-core Threadripper Pro 7995WX processor continues to tear old records to shreds. As soon as we had time to talk about  1.1 million points in 7zip and 184,000 in Cinebench R23 , the processor immediately reached a completely different height: more than 200,000 points in Cinebench R23! [caption id="attachment_84145" align="aligncenter" width="780"] Threadripper Pro 7995WX[/caption] More precisely, the result was 201,501 points. The Elmor OC team, which set the record, used liquid nitrogen and overclocked the CPU to 6 GHz. [caption id="attachment_84146" align="aligncenter" width="780"] Threadripper Pro 7995WX[/caption] Threadripper Pro 7995WX sets new records At the same time, the processor consumed a monstrous 1.58 kW, however, at one time the Xeon W9-3495X showed a much lower result, consuming almost 1.9 kW. https://youtu.be/CP6vuvOVTcc It is likely that many records will still be achieved with the participation of the Threadripper Pro 7995WX. But whether Intel will be able to release something of the same level in this segment in the coming years is still an open question. The upcoming Xeon Scalable Emerald Rapids generation will not offer a significant increase in the number of cores, nor any completely new architecture.

[phonemantra-blog]

New ones are clearly just around the corner We've already seen that the Threadripper 7980X and Threadripper Pro 7995WX processors offer incredible performance, but there are a number of new records worth noting. [caption id="attachment_83835" align="aligncenter" width="780"] Threadripper Pro 7995WX[/caption] For example, back in October, at a closed event, AMD itself said that the 96-core Threadripper Pro 7995WX can be overclocked in such a way that it will score more than 184,000 points in Cinebench R23. So far no one has achieved such a result, but there are already first results using liquid nitrogen: 177,050 points. Before the advent of the Threadripper Pro 7995WX, the best result was the record of two 96-core Epyc 9654, which was 147,668 points. Threadripper Pro 7995WX sets a number of new records [caption id="attachment_83837" align="aligncenter" width="780"] Threadripper Pro 7995WX[/caption] In Cinebench R24, the result was 8230 points, while the previous record was held by the Intel Xeon W9-3945X and it was 6159 points. The result in 7zip is even more impressive. If previously Threadripper Pro 5995WX held the title of record holder with a score of almost 514,000 points, then Threadripper Pro 7995WX almost doubled it, scoring almost 1.1 million points.

[phonemantra-blog]

There's more liquid nitrogen ahead The day before yesterday, overclocker Sampson managed to overclock the 96-core Ryzen Threadripper PRO 7995WX to 4.8 GHz for all cores using air CO and setting many records, and yesterday he set new records by overclocking the same CPU using a liquid cooling system. [caption id="attachment_83316" align="aligncenter" width="780"] Ryzen Threadripper PRO 7995WX[/caption] Ryzen Threadripper PRO 7995WX overclocked using liquid CO The CO was created from a HeatKiller IV processor unit, a ThermalTake radiator and an Enermax pump. This allowed us to increase the frequency of all cores to 5.265 GHz! This was the maximum frequency without any load and with a consumption of about 600 W. In applications like Cinebench, the processor clocked at about 5 GHz, while consuming up to 1 kW of power! Of course, all this allowed us to set new records. For example, 167,309 points in Cinebench R23. But in the near future we should definitely expect new achievements when the Threadripper PRO 7995WX is overclocked using liquid nitrogen.

[phonemantra-blog]

True, subject to effective use US sanctions that prohibit the supply of Nvidia's latest (and not-so-newest) accelerators to China don't seem to bother everyone. Tencent reported that it has managed to accumulate so many adapters that it will now last for a long time.  The company's president said that the reserves are such that Tencent will have enough to continue developing its Hunyuan artificial intelligence model. Moreover, it will be enough, as the head of the Chinese giant put it, for at least another couple of generations. That is, for several years.  Chinese Tencent has accumulated so many Nvidia accelerators that it will last for a couple of generations to come [caption id="attachment_82888" align="aligncenter" width="600"] Nvidia accelerators[/caption] True, even if the company now thinks so, how correct their assessments are, given the pace of industry development, is an open question.  But the company notes that it needs to find ways to use existing accelerators more efficiently. In addition, Tencent will be looking for or waiting for the emergence of worthy domestic accelerators. 

[phonemantra-blog]

IceGiant ProSiphon Elite cooler used The monstrous 96-core Ryzen Threadripper PRO 7995WX processor has set another record. More precisely, several at once.  [caption id="attachment_82766" align="aligncenter" width="780"] Ryzen Threadripper PRO 7995WX[/caption] The author of three records was overclocker Sampson. He was able to achieve record results in three versions of Cinebench, but not just like that, but using air cooling! Yes, there are coolers that can tame AMD's 350-watt monster. In this case, the giant IceGiant ProSiphon Elite was used , which weighs more than 2 kg and is not just a radiator, heat pipes and fans. The cooler is based on a two-phase cooling system with a dielectric liquid without any pumps or pumps. The principle of operation is exactly the same as in the case of conventional heat pipes: near the CPU, the liquid heats up, evaporates and rises up to the radiator, where it cools, condenses and falls back to the processor. Everything happens under the influence of gravity, so apart from replaceable fans there is nothing that could fail. Ryzen Threadripper PRO 7995WX, consuming almost 1 kW, set three records, and with an air cooling system [caption id="attachment_82768" align="aligncenter" width="780"] Ryzen Threadripper PRO 7995WX[/caption] 350 W, which appears in the characteristics of the Ryzen Threadripper PRO 7995WX, is not even close to the values ​​​​that were achieved during the overclocking process. After increasing the frequencies and voltage, the processor consumed an incredible 980 W! And here the performance of IceGiant ProSiphon Elite can really be called impressive. At the same time, the processor heated up to 102 degrees at its peak.   The results that were achieved by the enthusiast significantly surpass anything that was available before the release of the new Threadripper. For example, in Cinebench R23 in multi-threaded mode the result was more than 161,000 points, and before that the record was more than 123,000 points, and it was installed on a system with two 48-core Epyc 9474F. 

[phonemantra-blog]

Based on RISC-V cores The Chinese company SophGo is developing a processor based on the RISC-V architecture, using solutions from the American company SiFive. [caption id="attachment_82463" align="aligncenter" width="600"] 64-core processor[/caption] The processor, called the SG2380, has 16 SiFive P670 cores, each roughly equivalent in performance to a Cortex-A77, a SiFive X280 AI/ML accelerator, and an AXT-16-512 GPU from Imagination Technologies. The Chinese company SophGo is creating a 64-core processor based on American SiFive technologies In addition, SophGo has already announced the SG2044 SoC, which will be released next year. It will already offer 64 processor cores, support for PCIe 5.0 and LPDDR5x. Both solutions will be produced using the 12 nm process technology, apparently at the facilities of the Chinese SMIC, since initially the SiFive cores, which SophGo uses, are focused on much more modern technology processes. Thus, SophGo is not subject to the latest US sanctions, since they do not apply to the RISC-V architecture. The popularity of this architecture has been growing recently, and American sanctions may make RISC-V an even more popular solution specifically in China.

[phonemantra-blog]

The retailer's top 3 most popular video cards included models from MSI, Gigabyte and PowerColor Statistics on video card sales by the largest German retailer Mindfactory have been published for the 45th week of this year. The Radeon RX 7800 XT is in the lead in the overall top: 755 such accelerators were sold in a week. In second place with a large lag is the GeForce RTX 4070 with a result of 430 units sold. In third place is the Radeon RX 7900 XTX (270 units sold). [caption id="attachment_82146" align="aligncenter" width="780"] GeForce RTX 4070[/caption] The Radeon RX 7800 XT sells almost twice as much in Germany as the GeForce RTX 4070. In just a week, the retailer sold 1,990 AMD video cards, 1,585 Nvidia video cards and only 40 Intel video cards. Despite the meager sales of the “blue”, Intel secured 1.11% of the market - for the first time the share exceeded 1%. It is interesting that Intel does not have a clear leader: 20 copies of Arc A770 and Arc A380 were sold in a week. [caption id="attachment_82147" align="aligncenter" width="174"] GeForce RTX 4070[/caption] The top 3 specific models look like this: first place - MSI GeForce RTX 3060 Ventus 2X OC, second place - Gigabyte GeForce RTX 4070 Windforce OC, third place - PowerColor Radeon RX 7800 XT Red Devil.

[phonemantra-blog]

As if the PC is already ready with them Ryzen 8000G processors, as we have already reported , may appear instead of the Ryzen 7000G at the end of this year or at the beginning of next year. Perhaps Ryzen 9000 is also on the way. [caption id="attachment_80751" align="aligncenter" width="780"] Ryzen 9000[/caption] A photo of a Dell advertising brochure has leaked online, showing an Alienware gaming PC allegedly equipped with a processor from the Ryzen 9000 line. Ryzen 9000 is closer than everyone thought? One of the buyers of the current Alienware PC received such an advertising brochure, but regular customers of the company also received similar advertising. It is likely that this is just a mistake, since next year we should be waiting for the Ryzen 8000 line in all its manifestations. However, given yesterday's news about the Ryzen 8000G, perhaps AMD for some reason decided to use names that we initially expected to see later. Moreover, in this case, the company will not be the first to skip generations. It is worth remembering that the Ryzen 6000 line is completely absent in the desktop segment.

[phonemantra-blog]

It's unclear exactly which CPUs Naver uses Over the past year, Nvidia has become the undisputed leader in the artificial intelligence accelerator market. However, it is not the only one present in this market. Naver has reportedly replaced Nvidia accelerators with similar Intel solutions. [caption id="attachment_78879" align="aligncenter" width="600"] Nvidia GPU to Intel CPU[/caption] Korean company Naver switched from Nvidia GPU to Intel CPU for its artificial intelligence server Naver previously used an Nvidia GPU-based server, but has now replaced it with an Intel-based server. Naver is the largest Internet portal and the most popular search engine in South Korea. In this case, we are talking about a server that is responsible for the AI ​​for the Naver Place mapping service. Global information technology companies are reportedly becoming increasingly frustrated with rising prices for Nvidia GPUs and a global shortage of such GPUs. Analysts say the deal between Naver and Intel will likely reduce Nvidia's influence in the artificial intelligence processor market. Before putting the new server into operation, Naver tested it for about a month to understand whether, in general, Intel CPUs could compete with Nvidia GPUs in the tasks the company needed. Of course, there are no details about this, but since the deal was concluded, it means Naver were satisfied with the results. Additionally, the company says it will likely continue to use Intel's next-generation processors, such as Sapphire Rapids, to upgrade its system.

[phonemantra-blog]

There is a place in the lineup for them Yesterday the Internet started talking again about GeForce RTX 40 Super video cards, and now the characteristics of one of the models have appeared.  [caption id="attachment_71729" align="aligncenter" width="780"] Nvidia[/caption] The RTX 4070 Super, according to insider MEGAsizeGPU, will receive an AD103 GPU and 16 GB of memory with a 256-bit bus. This means that the new product will offer more memory than the RTX 4070 and RTX 4070 Ti, which, we recall, have 12 GB each. In addition, the use of the AD103 GPU hints that the RTX 4070 Super will be a step above the RTX 4070 Ti since the latter is based on the smaller AD104.   Nvidia is preparing two new video cards The insider also says that Nvidia may release a simplified version of the RTX 4070, which will use GDDR6 instead of GDDR6X memory, that is, the memory frequency will be lower. Whether there will be any other differences is still unclear, but such a model will close a rather large hole in terms of performance between the RTX 4070 and RTX 4060 Ti. True, let us remind you that the first costs $600, and the second in the 16 GB version costs $500, so it is not very clear how much a simplified RTX 4070 could cost so as not to destroy sales of the RTX 4060 Ti 16GB.  Insider MEGAsizeGPU previously spoke about the launch of the RTX 4060 and RTX 4060 Ti before the announcement, as well as the release dates for these models.

[phonemantra-blog]

It has 80 Arm cores While Chinese video cards are very far from modern solutions from AMD, Nvidia, and Intel, Chinese processors are developing at a very impressive pace. Phytium has shown off its new processor, which is capable of competing with AMD CPUs based on the Zen 3 architecture. [caption id="attachment_71250" align="aligncenter" width="780"] AMD Epyc on Zen 3[/caption] Chinese Arm processor capable of competing with 24-core AMD Epyc on Zen 3. More precisely, the company compares its FTC870 with the server AMD Epyc 7443. Let us recall that the latter has 24 cores, a TDP of 200 W, and operates at frequencies up to 4 GHz. The FTC870, in turn, is based on the Arm architecture and operates at a frequency of only 3 GHz. The new product has 80 cores and 64 MB of third-level cache. There is also an eight-channel DDR5-4800 controller and 64 PCIe 5.0 lanes. [caption id="attachment_71251" align="aligncenter" width="780"] AMD Epyc on Zen 3[/caption] Interestingly, the company planned to release such a CPU back in 2021, but due to US sanctions the process has slowed down, so such CPUs will only be released now.