Video cards can get faster with Samsung’s new GDDR6W memories

this week the samsung announced a new type of memory for video cards. I’m talking about the memories GDDR6W🇧🇷 It is an evolution of GDDR6 memories. The new memory modules for GPUs bring twice the density and performance, promising to make top-of-the-line GPU models even more powerful.

Read too
What is the difference between integrated and dedicated graphics card?
What is RAM memory used for?

Increased capacity on each memory chip

What determines the power of a video card is not just the graphics chip. It’s actually a balance between graphics processor and memory. A super powerful video processor will not be able to extract the maximum of its capacity if the memory chips are of bad quality. The reverse is also true, that is, a set of quality memory modules cannot increase the performance of an average graphics chip.

That’s why Samsung’s GDDR6W memories will be used in high-end video cards. The main advantage of these new memory chips is that the Samsung managed to double their capacity🇧🇷 That is, in GDDR6 memories each chip can have a maximum of 16 Gb (gigabits). As for GDDR6W memories, the capacity of each chip increases to 32 Gb.

To achieve doubling capacity, Samsung used a manufacturing technique known as FOWLP 🇧🇷Fan-Out Wafer-Level Packaging🇧🇷 This manufacturing technique stacks two memory layers on chip🇧🇷 But this stacking is done in such a way as not to increase the thickness of the memory module. In fact, the height of the module even decreased, going from 1.1 to 0.7 mm.

Memory chips have thinned out

Greater heat dissipation

Another advantage of FOWLP manufacturing technology is the Improved heat dissipation🇧🇷 That is, stacking memory layers does not cause the module to overheat. This approach gives manufacturers more flexibility in building graphics cards.

For example, the GPU NVIDIA GeForce RTX 4090 It has 24 GB of GDDR6X memory. These 24 GB are divided into 12 chips of 2 GB each.

If the board used GDDR6W memories, the same 24 GB could be distributed in only 6 memory chips, reducing its size. Or, NVIDIA could keep the 12 more chips with a total capacity of 48 GB. Have you thought!?

Greater Bandwidth

New GDDR6W memories have higher bandwidth

GDDR6W memories not only doubled the density but also the bandwidth of the chips. This translates, in practice, into more performance🇧🇷 A GDDR6W memory module has a bandwidth of 48 Gb/s (gigabits per second). That’s twice as much as GDDR6 memory, which only has 24 Gb/s of bandwidth.

This is possible due to doubling the number of pins on each memory unit. To give you an idea, at the system level, the bandwidth of all memory can reach 1.4 TB/s (terabytes per second). This is possible on a 512-bit interface and at a baud rate of 22 Gb/s per pin.

Availability

It is important to make it clear that Samsung just announced this new memory standard for video cards. However, it did not give any timeframe for the release of GDDR6W memories.

However, it is very likely that at some point in 2023 companies like NVIDIA will adopt the standard on their video cards. The company currently uses GDDR6X type memories produced by Micron in its products. They are faster than GDDR6 memories in some respects.

It is also likely that the first video cards with GDDR6W memory were for specific uses, such as artificial intelligence or virtual reality. In this type of application, the volume of data tends to be very large.

Source: pc gamer

Source link

About Admin

Check Also

Sony accused of lying to European regulators

Microsoft communications chief claims Sony lied to regulators The soap opera involving the acquisition of …

Leave a Reply

Your email address will not be published. Required fields are marked *