Video Memory Discussion

[FP91G]

Samsung Memory’s DRAM and NAND Roadmap Explained
https://ebics.net/samsung-memorys-dram-and-nand-roadmap-explained/


80 GB of memory, 700 W and 60 TFLOPS. Nvidia H100 accelerator based on Hopper generation GPU introduced
https://gadgettendency.com/80-gb-of-memory-700-w-and-60-tflops-nvidia-h100-accelerator-based-on-hopper-generation-gpu-introduced/
Both versions have 80 GB of memory, and this is HBM3 with a bandwidth of 2 or 3 TB / s. The TDP of the older version is an incredible 700W, while the PCIe 5.0 version has half that figure.

Will we see gaming graphics cards with HBM3 memory?

[FP91G]

Samsung Electronics is ready to mass produce its third-generation 4-nm chips, which is the main item in the foundry ultra-micro fabrication process sector, according to a report. The Korean company has apparently resolved early-stage yield issues and achieved advancements in performance, power consumption, and area improvement. Manufacturing should start during the first half of this year.
https://www.gizmochina.com/2023/03/14/samsung-mass-production-third-generation-4nm-chips/

[swogerino]

Samsung Electronics is ready to mass produce its third-generation 4-nm chips, which is the main item in the foundry ultra-micro fabrication process sector, according to a report. The Korean company has apparently resolved early-stage yield issues and achieved advancements in performance, power consumption, and area improvement. Manufacturing should start during the first half of this year.
https://www.gizmochina.com/2023/03/14/samsung-mass-production-third-generation-4nm-chips/

I wonder what capacity the new graphic cards will have then,as the lower the nm the more transistors can fit in so in theory they should be much more powerful than the 7 nm I know the Rx 6000 series were and maybe Nvidia 3000 series but I am not sure about Nvidia.

Who knows if they improve the graphic cards up to the level described it will surely make mining more enticing again,not for everyone but for the majority who are enthusiasts about crypto and mining in general,so can't wait until we see mass production of 4 nm graphic cards.

[Pterosaur]

Good, the lesser the nm the better laptops and small form factor products will run cooler, graphic cards will benefit from this, I am a fan of PC gaming and the only reason why I haven't looked into handheld gaming PC is because of cooling performance, the intel 11 gen found in latest Ayaneo and the CPUs in onexplayer2 or gpd win 4 isn't good enough, 4nm will be a banger. Maybe mining will be now possible on laptops without worrying too much about burning the motherboard.

[FP91G]

Samsung Electronics is ready to mass produce its third-generation 4-nm chips, which is the main item in the foundry ultra-micro fabrication process sector, according to a report. The Korean company has apparently resolved early-stage yield issues and achieved advancements in performance, power consumption, and area improvement. Manufacturing should start during the first half of this year.
https://www.gizmochina.com/2023/03/14/samsung-mass-production-third-generation-4nm-chips/

I wonder what capacity the new graphic cards will have then,as the lower the nm the more transistors can fit in so in theory they should be much more powerful than the 7 nm I know the Rx 6000 series were and maybe Nvidia 3000 series but I am not sure about Nvidia.

Who knows if they improve the graphic cards up to the level described it will surely make mining more enticing again,not for everyone but for the majority who are enthusiasts about crypto and mining in general,so can't wait until we see mass production of 4 nm graphic cards.

In mining, there are several important points to consider. Not all coins need a lot of RAM. There are a few coins that require more than 8 GB of memory, but 8 GB of memory is still relevant. But I would take a GPU with 12-16 GB of memory.
Not all algorithms are memory dependent. The Ethereum algorithm loved fast memory, and new coins love powerful GPU cores. Therefore, old generations of video cards now give a very small profit, and new GPUs pay off for a very long time.

[safar1980]

SK hynix Develops Industry's First 12-Layer HBM3, Provides Samples To Customers
Develops HBM3 product with industry's largest 24GB memory capacity; customers' performance evaluation of samples underway
Features high-capacity and high-performance through stacking of 12 DRAM chips
Plans to complete preparation for mass production by first half of 2023, aimed at solidifying company's leadership in cutting-edge DRAM market

[FP91G]

some plans about technologies
TSMC

Samsung

https://t.me/HomishTG/5177
https://t.me/HomishTG/5178

There is an opinion that new models of processors and video cards will cost more, because the construction of new factories is a very expensive investment. And earlier we were told that it is difficult to make 3 nm, but today 4 nm is already a common technology.

[FP91G]

Samsung Develops Industry’s First GDDR7 DRAM To Unlock the Next Generation of Graphics Performance
Samsung’s latest 32Gbps GDDR7 to further expand capabilities in applications for AI, HPC and automotive vehicles

Enhancements in GDDR7 include 1.4 times boost in performance and 20% improvement in power efficiency versus existing 24Gbps GDDR6 DRA
https://videocardz.com/press-release/samsung-introduces-worlds-first-gddr7-memory-up-to-32-gbps
Samsung’s GDDR7 achieves an impressive bandwidth of 1.5-terabytes-per-second (TBps), which is 1.4 times that of GDDR6’s 1.1TBps and features a boosted speed per pin of up to 32Gbps. The enhancements are made possible by the Pulse Amplitude Modulation (PAM3) signaling method adopted for the new memory standard instead of the Non Return to Zero (NRZ) from previous generations. PAM3 allows 50% more data to be transmitted than NRZ within the same signaling cycle.

[FP91G]

Micron Delivers HBM3 Gen2 Memory Samples To NVIDIA, Customers Blown Away By Performance & Efficiency
Micron's next-gen HBM3 technology is named "HBM3 Gen2". Regarding specifications, it features a 1β process node, which brings in much faster speeds and denser capacities. The first batch of 2nd-generation HBM3 features an 8-Hi design, offering up to 24 GB capacities and bandwidth exceeding 1.2 TB/s with up to 2.5x the performance per watt of the prior generation. The DRAM will operate at 9.2 Gb/s, a 50% boost over the standard "HBM3", which runs around 4.6 Gb/s speeds.

HBM3 Gen2 memory is expected to be used in Nvidia Blackwell compute accelerators, which could appear as early as next year. Such memory is unlikely to appear in gaming video cards, although GDDR7 will also offer an impressive increase in performance compared to GDDR6X.

https://wccftech.com/micron-hbm3-gen2-memory-samples-nvidia-customers-blown-away-performance-efficiency/

[fmz89]

its been stuck for decades now, when HBM implemented in 2nd generation amd gpu = vega is the first proper high bandwith memory without backed high computational power,
in todays world seems memory getting less bandwith in term pair computational power, while mining need ton of bandwith for asic resistant. while amd seem abandoned HBM series for gaming card
gddr7 seems interesting, but most interesting part was lpddr6 which may take huge leap portable/mobile device and metaverse become more common thing

[FP91G]

In mining, fast memory has lost its meaning because it is not needed in all algorithms. And ASICs now work on coins with a similar Ethereum algorithm. New coins require a more powerful video processor for higher hashrate and the new HBM memory is expensive for regular video cards.

[FP91G]

Samsung Unveils HBM3E Memory and Provides Details on HBM4, GDDR7, and Detachable AutoSSD

One of the key features of Samsung’s HBM3E memory is its faster data transfer speed compared to similar chips from Micron and SK Hynix. With data transfer speeds of 9.8Gbps per pin, it is capable of reaching transfer rates of up to 1.2TBps. Samsung has also enhanced its NCF (Non-Conductive Film) technology to improve thermal conductivity by reducing the gaps between chip layers.

The HBM3E memory is currently in the sampling phase and being sent to clients for testing. Mass production is expected to begin in 2024. Additionally, Samsung unveiled plans for the development of HBM4 memory, which will utilize more advanced chip manufacturing and packaging technologies. The company aims to use a wider 2,048-bit memory interface and FinFET transistors to reduce power consumption.

Samsung also provided more details on its GDDR7 memory, which offers a 50% reduction in standby power consumption compared to previous generations. The company plans to start shipping 16Gbit modules that can operate at up to 32Gbps per pin. This new memory type improves power efficiency in both active and standby states.
https://isp.page/news/samsung-unveils-shinebolt-hbm3e-memory-updates-hbm4-development-2/

[FP91G]

Micron announces 32 Gbit FeRAM Chip: A Leap Forward in Memory Technology
In December 2023, US-based semiconductor company Micron announced at the International Electron Devices Meeting (IEDM) the development of a high-capacity FeRAM memory chip boasting a staggering capacity of 32 Gbits.

FeRAM memory technology has seen around two decades of existence, with usual chip capacities ranging between 8 to 128 Mbits. Micron’s new development, in contrast, offers significantly increased storage capacity that can potentially revolutionize non-volatile memory technology.

https://davesipaq.com/micron-secretly-develops-32gb-feram-chip-a-durable-energy-independent-memory-with-speed-comparable-to-dram/

[safar1980]

GDDR7 at 37 Gbps to enable up to 1.7 TB/s bandwidth for 384-bit cards
Samsung is preparing 37 GB/s graphics memory based on GDDR7 technology.
The upcoming 2024 ISSCC program outlines plans for high-speed memory, featuring 37 Gb/s and 35.4 Gb/s variants based on Samsung and SK Hynix’s GDDR7 specifications. Both companies intend to delve into advancements in graphics memory, utilizing the innovative PAM3 and NRZ signaling technologies.

While GDDR7 memory, with speeds reaching up to 32 Gb/s, has already been officially announced, Samsung and Hynix are pushing the boundaries further with even faster memory developments. Micron is also in the race, announcing its commitment to developing 36 Gb/s GDDR7 memory, expected to hit the market no earlier than 2026. In this competitive landscape, Samsung and Hynix are set to introduce first modules with slightly lower speeds, most likely gradually making 35 Gb/s modules available at a later stage.

[FP91G]

GDDR7 memory to come in 2GB to 8GB capacity
According to NVIDIA hardware leaker “Kopite7kimi”, the first products using GDDR7 memory will be sticking to the minimum density of 16 Gb (2GB) per module, the same capacity as currently utilized by GDDR6X models.

At its introduction, the 2GB GDDR7 capacity will already be an upgrade over GDDR6, which debuted with 8 Gb (1GB) density and was extensively used by first graphics cards using this technology, such as GeForce RTX 20 and later RTX 30 series. Worth reminding that the lack of 2GB GDDR6(X) modules was the reason RTX 3090 required 24 modules on both sides of the PCB.
https://videocardz.com/newz/first-generation-of-gddr7-graphics-cards-sticking-to-16gbit-2gb-modules-3gb-on-roadmaps

[safar1980]

NVIDIA Reportedly Sampling SK Hynix 12-layer HBM3E
SK Hynix is believed to be leading the pack—insiders believe that yield rates are good enough to pass early NVIDIA certification, and advanced 12-layer samples are expected to be approved in the near future. ZDNET reckons that SK Hynix's forward momentum has placed it an advantageous position: "(They) supplied 8-layer HBM3E samples in the second half of last year and passed recent testing. Although the official schedule has not been revealed, mass production is expected to begin as early as this month. Furthermore, SK Hynix supplied 12-layer HBM3E samples to NVIDIA last month. This sample is an extremely early version and is mainly used to establish standards and characteristics of new products. SK Hynix calls it UTV (Universal Test Vehicle)... Since Hynix has already completed the performance verification of the 8-layer HBM3E, it is expected that the 12-layer HBM3E test will not take much time." SK Hynix's Vice President recently revealed that his company's 2024 HBM production volumes for were already sold out, and leadership is already preparing innovations for 2025 and beyond.

[FP91G]

Product Pages of Samsung 28 Gbps and 32 Gbps GDDR7 Chips Go Live

Samsung is ready with a GDDR7 memory chip rated at an oddly-specific 28 Gbps. This speed aligns with the reported default memory speeds of next-generation NVIDIA GeForce RTX "Blackwell" GPUs. The Samsung GDDR7 memory chip bearing model number K4VAF325ZC-SC28, pictured below, ticks at 3500 MHz, yielding 28 Gbps (GDDR7-effective) memory speeds, and comes with a density of 16 Gbit (2 GB). This isn't Samsung's only GDDR7 chip at launch, the company also has a 32 Gbps high performance part that it built in hopes that certain high-end SKUs or professional graphics cards may implement it. The 32 Gbps GDDR7 chip, bearing the chip model number K4VAF325ZC-SC32, offers the same 16 Gbit density, but at a higher 4000 MHz clock. The Samsung website part-identification pages for both chips say that the parts are sampling to customers, which is usually just before it enters mass-production, and is marked "shipping."
https://www.techpowerup.com/320822/product-pages-of-samsung-28-gbps-and-32-gbps-gddr7-chips-go-live

[safar1980]

SK hynix teases GDDR7 memory with 40 Gbps speed and 3GB capacity
During GTC 2024, NVIDIA unveiled its plan for the upcoming Blackwell GPU architecture. Alongside NVIDIA’s presentations, attendees could explore the latest offerings from memory companies. Samsung, for instance, is gearing up to introduce its 28 Gbps and 32 Gbps memory modules tailored for next-generation gaming GPUs. Meanwhile, Hynix is taking a step ahead by confirming the development of modules supporting even faster speeds.

[FP91G]

Minimal Initial Damage to DRAM and Foundry Production Lines Following Major Earthquake in Taiwan on April 3; Details of Losses Still Under Assessment, Says TrendForce
In the wake of a 7.2-magnitude earthquake off the eastern coast of Taiwan at 7:58 AM on April 3rd, TrendForce immediately investigated the damage and operational status of various manufacturers. The DRAM industry, primarily located in the northern and central parts of Taiwan, and the foundry industry, spread across the north, central, and southern regions of Taiwan, appear to have sustained minimal initial damage.

The strongest tremors felt in the northern Linkou area ranged from intensity scale 4 to 5, with other regions experiencing around intensity scale 4 tremors. Manufacturers have begun sequential shutdowns for inspections. Despite the ongoing inspections, no significant equipment damages have been reported so far.
https://www.trendforce.com/presscenter/news/20240403-12104.html

[FP91G]

Samsung Electronics’ operating profit jumps 933% in first quarter, beats expectations
Samsung Electronics, the world’s largest memory chip maker, said operating profit jumped 932.8% in the first quarter ending March.
Citi analysts on Apr. 5 said they see upside of Samsung’s NAND flash memory business from AI computing demand, reiterating their “buy” rating on the firm.
The South Korean electronics giant saw record losses in 2023 as the industry reeled from a post-Covid slump in demand.
https://www.cnbc.com/2024/04/30/samsung-electronics-first-quarter-earnings.html