This article is sponsored by the Semiconductor Industry Versatility (ID: ICVIEWS). HDD still has a place in modern storage. HDDs are now almost obsolete for most users. Whether it's a laptop or a desktop, SSDs have more or less replaced mechanical hard drives as the only real sto

This article is synthesized by the semiconductor industry vertical and horizontal (id: icviews)

hdd still has a place in modern storage.

hdd is now almost obsolete for most users. Whether it's a laptop or a desktop, SSDs have more or less replaced mechanical hard drives as the only real storage option for the average user. Performance, reliability, durability, noise, size and even cost considerations are not favorable to HDDs.

However, even in 2024, there are still reasons to consider using HDDs for your system. They're still a better choice for users who prioritize large capacity and cold storage over performance.

hdd cost advantage

Although most users will not save much by buying a 1tb or even 2tb SSD than choosing an HDD of the same capacity, a larger HDD can still beat an SSD in price. The storage needs of a gaming or work pc are very different from those of a nas, home server and professional setup.

The most suitable ssd for playing games may be the 1tb gen4 nvme ssd, but for people who need 4tb, 6tb or larger capacity hard drives, from a cost perspective, they are still more inclined to use traditional hard drives. For example, a 4tb seagate barracuda hdd is only $85, but a 4tb samsung 870 qv is about $300, which happens to be a sata ssd only.

In storage intensive setups value is more important than performance

Even if you think people building nas might have to buy more expensive nas hdds, even the 4tb seagate ironwolf pro is only $155 - just for samsung sata half of an ssd. When you consider larger drives and raid setups the ssd pricing makes no sense. Considering that value is more important than performance in these storage intensive setups, HDD wins easily.

hdd is more suitable for archival storage

hdd and ssd can reliably store data for years without any issues. While modern SSDs typically last longer (~10 years) than HDDs (~5 years), this doesn't take into account the drive being unplugged. Because of the way ssds store data, a power outage (or voltage outage) for an extended period of time (usually a year or so) can result in data loss.

For most people, having an SSD lose power for a year or more may not sound realistic. However, users who deal with large amounts of data often store filled drives in cold storage (unplugged and saved for later use). Using SSDs for this purpose may compromise data integrity, which is why many people still prefer using HDDs for archival storage.

Not every PC needs fast storage

The responsiveness of modern SSDs is unparalleled for gamers, creators, and professionals who can truly take advantage of blazing-fast read and write speeds and random IOPS. You wouldn't expect an HDD-only setup on your gaming PC or productivity machine. However, there are still systems out there that don't really need fast boot times, load times, or transfer speeds.

This refers to home computers, htpcs, media servers, computers used by the elderly or children, and occasionally used systems that just work fine with mechanical drives. Depending on the frequency of use, type of user, and type of application, an HDD can perfectly suit the needs of many computers. Additionally, many of these systems will also benefit from the superior price/performance of larger HDDs.

The era of SSD is coming

As we saw above, hard drives still have their place in modern storage. However, in a few years, ssds may finally surpass hdds in cost per gb and other aspects. We continue to see developments in the SSD space, such as nvme-over-fabrics (nvme-of) and nvme-over-cxl (nvme-oc).Who knows what the nvme world will look like in another ten years? The large model

ai has created a demand for massive storage, from which both SSD and HDD are expected to benefit. Among them, based on the advantages of speed, low latency, durability and energy saving, the industry is generally optimistic about the development of SSD in the AI ​​era.

Whether it is AI training or inference, SSD has the opportunity to take on important tasks.

The latest survey by TrendForce, a global market research organization, shows that during the AI ​​model training process, SSD is not only responsible for storing model parameters, including constantly updated weights and biases, but also can create check points to save the AI ​​model regularly. Training progress, allowing you to resume from a specific point even if training is interrupted. The above functions all rely heavily on high-speed transmission and write endurance, so customers mainly choose to use 4TB/8TB TLC (triple-layer cell flash memory) SSD products to meet the demanding training process requirements.

ai In the inference process, SSD can assist in adjusting and optimizing the AI ​​model during the inference process. In particular, SSD can update data in real time to fine-tune the results of the inference model. AI reasoning mainly provides retrieval-enhanced generation (rag, retrieval-augmented generation) and large language model (llm, large language model) services, while SSD can store relevant documents and knowledge bases referenced by rag and llm to generate richer information. response.

As more generated information is displayed in videos or photos, the amount of data storage also increases accordingly. Therefore, large-capacity SSDs such as TLC/QLC (four-layer unit flash memory) 16TB or above have become the main products used for AI inference.

trendforce TrendForce predicts that AI-related SSD procurement capacity will exceed 45EB this year. In the next few years, AI servers are expected to drive SSD demand to an average annual growth rate of more than 60%, and AI SSD demand accounts for 10% of the entire NAND flash (flash memory). The opportunity increases from 5% in 2024 to 9% in 2025.

This article is synthesized by the semiconductor industry vertical and horizontal (id: icviews)

hdd still has a place in modern storage.

hdd is now almost obsolete for most users. Whether it's a laptop or a desktop, SSDs have more or less replaced mechanical hard drives as the only real storage option for the average user. Performance, reliability, durability, noise, size and even cost considerations are not favorable to HDDs.

However, even in 2024, there are still reasons to consider using HDDs for your system. They're still a better choice for users who prioritize large capacity and cold storage over performance.

hdd cost advantage

Although most users will not save much by buying a 1tb or even 2tb SSD than choosing an HDD of the same capacity, a larger HDD can still beat an SSD in price. The storage needs of a gaming or work pc are very different from those of a nas, home server and professional setup.

The most suitable ssd for playing games may be the 1tb gen4 nvme ssd, but for people who need 4tb, 6tb or larger capacity hard drives, from a cost perspective, they are still more inclined to use traditional hard drives. For example, a 4tb seagate barracuda hdd is only $85, but a 4tb samsung 870 qv is about $300, which happens to be a sata ssd only.

In storage intensive setups value is more important than performance

Even if you think people building nas might have to buy more expensive nas hdds, even the 4tb seagate ironwolf pro is only $155 - just for samsung sata half of an ssd. When you consider larger drives and raid setups the ssd pricing makes no sense. Considering that value is more important than performance in these storage intensive setups, HDD wins easily.

hdd is more suitable for archival storage

hdd and ssd can reliably store data for years without any issues. While modern SSDs typically last longer (~10 years) than HDDs (~5 years), this doesn't take into account the drive being unplugged. Because of the way ssds store data, a power outage (or voltage outage) for an extended period of time (usually a year or so) can result in data loss.

For most people, having an SSD lose power for a year or more may not sound realistic. However, users who deal with large amounts of data often store filled drives in cold storage (unplugged and saved for later use). Using SSDs for this purpose may compromise data integrity, which is why many people still prefer using HDDs for archival storage.

Not every PC needs fast storage

The responsiveness of modern SSDs is unparalleled for gamers, creators, and professionals who can truly take advantage of blazing-fast read and write speeds and random IOPS. You wouldn't expect an HDD-only setup on your gaming PC or productivity machine. However, there are still systems out there that don't really need fast boot times, load times, or transfer speeds.

This refers to home computers, htpcs, media servers, computers used by the elderly or children, and occasionally used systems that just work fine with mechanical drives. Depending on the frequency of use, type of user, and type of application, an HDD can perfectly suit the needs of many computers. Additionally, many of these systems will also benefit from the superior price/performance of larger HDDs.

The era of SSD is coming

As we saw above, hard drives still have their place in modern storage. However, in a few years, ssds may finally surpass hdds in cost per gb and other aspects. We continue to see developments in the SSD space, such as nvme-over-fabrics (nvme-of) and nvme-over-cxl (nvme-oc).Who knows what the nvme world will look like in another ten years? The large model

ai has created a demand for massive storage, from which both SSD and HDD are expected to benefit. Among them, based on the advantages of speed, low latency, durability and energy saving, the industry is generally optimistic about the development of SSD in the AI ​​era.

Whether it is AI training or inference, SSD has the opportunity to take on important tasks.

The latest survey by TrendForce, a global market research organization, shows that during the AI ​​model training process, SSD is not only responsible for storing model parameters, including constantly updated weights and biases, but also can create check points to save the AI ​​model regularly. Training progress, allowing you to resume from a specific point even if training is interrupted. The above functions all rely heavily on high-speed transmission and write endurance, so customers mainly choose to use 4TB/8TB TLC (triple-layer cell flash memory) SSD products to meet the demanding training process requirements.

ai In the inference process, SSD can assist in adjusting and optimizing the AI ​​model during the inference process. In particular, SSD can update data in real time to fine-tune the results of the inference model. AI reasoning mainly provides retrieval-enhanced generation (rag, retrieval-augmented generation) and large language model (llm, large language model) services, while SSD can store relevant documents and knowledge bases referenced by rag and llm to generate richer information. response.

As more generated information is displayed in videos or photos, the amount of data storage also increases accordingly. Therefore, large-capacity SSDs such as TLC/QLC (four-layer unit flash memory) 16TB or above have become the main products used for AI inference.

trendforce TrendForce predicts that AI-related SSD procurement capacity will exceed 45EB this year. In the next few years, AI servers are expected to drive SSD demand to an average annual growth rate of more than 60%, and AI SSD demand accounts for 10% of the entire NAND flash (flash memory). The opportunity increases from 5% in 2024 to 9% in 2025.