A Apple tem a TSMC a fabricar os seus CPUs. A Apple usa ARM.
Por algum motivo a Apple abandonou a Intel e dedicou-se a fazer os seus próprios CPUs parea desktop e laptop.
A Intel anda com problemas nos 10nm, há quase uma década. Parece que os yields já estão normais.
Mas ainda temos de ver como anda de eficiência. A velocidades mais elevadas, parece que vai chupar muita energia.
According to Chinese publication UDN, production using the 3nm node is expected to start in the second quarter of next year, with mass production set to begin in the middle of 2022. Production capacity is expected to reach 4,000 wafers by May 2022, eventually reaching 10,000 wafers per month.The report adds that Intel will tap TSMC's 18b Fab on the 3nm process node for four of its upcoming products: three CPUs and a GPU. In the case of the former, these are non-consumer processors, likely next-gen Xeon, designed for the server and data center markets.
https://static.techspot.com/images2/...-4-j_1100.webp 1100w" data-sizes="100w" sizes="100w" srcset="https://static.techspot.com/images2/news/bigimage/2021/07/2021-07-27-image-4-j.webp 2560w, https://static.techspot.com/images2/...-4-j_1100.webp 1100w">Intel already has its own fabs and is allowing other companies, including Qualcomm and Amazon, to use its current and planned manufacturing capacity as part of its IDM 2.0 initiative. CEO Pat Gelsinger recently announced that the firm would be building a $120 billion mega-fab comparable to a "little city" in a yet to be decided US location. But it's likely that for the multi-tile-based design in Intel's upcoming chips, the company will utilize both its own fabs and TSMC's for production of different tiled dies.
Another advantage from Intel's point of view is that grabbing TSMC's 3nm production capacity could hamstring AMD's and Apple's plans for their own 3nm chips. AMD has already been struggling to secure enough 7nm chips from TSMC for its graphics cards and Ryzen CPUs—much of the production capacity on that node has been allocated to the PS5/XBSX consoles—and it won't want a repeat of the situation with 3nm.
A ser verdade, está tudo lixado. Apple, AMD, nVidia, etc...
Boas!
O que acham destes novos cpus e arquitectura?
A AMD vai continuar a explorar o conceito do Chiplet, a Intel prefere andar com Big Cores e Little Cores.
Gosto da ideia de ambos os conceitos, mas penso que o da Intel poderá ser um caminho mais espinhoso, já que vai ter que andar a optimizar o que é para ser usado em E-Cores e P-Cores...
Boas!
Esse já seria uma bomba em especial com B660 + DDR4.
Existe um 12490f (apenas na china), que tem mais cache já que é um cutdown de um die acima desse.
Actualmente no entanto penso que se queremos bang pelo buck é 10400f ou 11400f com b560 que é mais barata como plataforma e tem praticamente tudo o que se quer...
A AMD simplesmente não tem nada nesta parte do mercado mais budget...
Falta-lhe um "1600AF" mas baseado em Zen 3 é o que é...
A new Spectre class speculative execution vulnerability, called Branch History Injection (BHI) or Spectre-BHB, was jointly disclosed on Tuesday by VUSec security research group and Intel.BHI is a proof of concept re-implementation of the Spectre V2 (or Spectre-BTI) type of attack. It affects any CPU that is also vulnerable to Spectre V2, even if mitigations for Spectre V2 have already been implemented; it can circumvent Intel's eIBRS and Arm's CSV2 mitigations. These mitigations protect from branch target injection, whereas the new exploit allows attackers to inject predictor entries into the global branch history. BHI can be used to leak arbitrary kernel memory, which means sensitive information like passwords can be compromised.
VUSec explained it as follows: "BHI essentially is an extension of Spectre v2, where we leverage the global history to re-introduce the exploitation of cross-privilege BTI. Therefore the attacker primitive is still Spectre v2, but by injecting the history across privilege boundaries (BHI), we can exploit systems that deploy new in-hardware mitigations (i.e., Intel eIBRS and Arm CSV2)."
BHI exploit leaking arbitrary kernel memory in action
The vulnerability affects any Intel CPU launched since Haswell, including Ice Lake-SP and Alder Lake. Affected Arm CPUs include Cortex A15/A57/A65/A72/A73/A75/A76/A77/A78/X1/X2/A710, Neoverse N2 / N1 / V1 and the Broadcom Brahma B15.
CVE ID for Arm is CVE-2022-23960 and Intel is using the IDs CVE-2022-0001 and CVE-2022-0002. Both companies have posted more details about their affected CPUs here (Intel) and here (Arm).
Intel has released the following statement regarding the BHI exploit: "The attack, as demonstrated by researchers, was previously mitigated by default in most Linux distributions. The Linux community has implemented Intel's recommendations starting in Linux kernel version 5.16 and is in the process of backporting the mitigation to earlier versions of the Linux kernel. Intel released technical papers describing further mitigation options for those using non-default configurations and why the LFENCE; JMP mitigation is not sufficient in all cases."
AMD CPUs seem to be immune to BHI. According to Phoronix, team red processors that have defaulted to using Retpolines for Spectre V2 mitigations should be safe.
Security patches from vendors should be released soon. In addition to installing them, researchers recommend disabling unprivileged eBPF support as an additional precautionary measure. Linux has already merged the security updates into its mainline kernel. Whether these security mitigations will impact performance is not yet known.
Source code for VUSec's exploit can be found here.
Likes: 4
Winjer Gostou do Post
Post Original de Winjer
Responder com Citação
Regras
User Alert System provided by
Advanced User Tagging v3.3.0 (Lite) -
vBulletin Mods & Addons Copyright © 2025 DragonByte Technologies Ltd.
Bookmarks