It’s 2020 and very high graphic games are being developed in the markets that come with the pretty high-end graphics requirement.
These requirements can’t be handled with the integrated CPU’s Graphics card along with the fact that nowadays these integrated Graphics cards are not enough to fulfill the needs of many video editing and other graphic designing software.
As to fulfil this need high end graphics cards are being released nowadays in the market. As lots of the options are available in the market you might be confused which one is the best graphics card to go with and what are the factors that affect the graphics card’s performance?
Although there are many factors that affect GPU’s performance while memory is the more affecting part of a GPUs that plays a direct role in the GPU’s performance. As power consumption, speed, and applications mainly depends upon memory.
When it comes to what technology is used by these memories then the HBM2 and GDDR6 is the most popular option in the market.
Not limiting it to the gamings purposes only these memories are mainly being used for other purposes like Networking Application, AI, Blockchains and more.
But which one is the best and what makes it the best option? Let’s compare both of these memories and figure out which of the one is the best option for you.
Short Introduction of HBM2 and GDDR6
The abbreviation of HBM2 is the “High Bandwidth Memory”. Basically it is a high performance RAM that is mostly being used in graphics cards, 3D-Stacked DRAM Chips, and Network Devices as well.
GDDR6 stands for “Graphics Double Data Rate Type 6”. It is a Synchronous graphics RAM that with pretty high Bandwidth that offers double data rate which is mainly designed to be used in graphics card, gaming consoles and high-end computing
That was the short introduction. Let’s dive into a detailed comparison of GDDR6 and HBM2.
The 2nd generation High Bandwidth Memory is the upgraded version of HBM which is manufactured by Samsung, AMD, and SK Hynix. It is mainly being used in the graphics cards and network devices.
The first generation HBM chip was manufactured by SK Hynix back in 2013 which was later used in AMD devices for the first time in 2015 after that the 2nd generation HBM2 was released back in 2016 and this technology was initially adopted by JEDEC.
Applications are something that can just make a clear cut difference between both of the available options, As HBM offers high bandwidth so it comprises applications like Artificial Intelligence, Blockchain mining and other non-cost effective applications as well.
It is a 256 GB/s based memory system that offers lots of bandwidth at the speed of 2GBbps if we consider it from a system point of view then this is pretty nice to deal. It comes with 1 single DDR with 2 gbps speed that completely utilizes bandwidth of 256 Gigs per second at one single point.
HBM comes with a much efficient power system. As it offers speed upto 2gbps that makes it more power efficient as less speed requires less power.
It has replaced the high clocked memory with low clocked memory by using Silicon Vias (TSVs) and Silicon Interposers. Although HBM offers similar technology but yet it is incompatible with Hybrid Memory Cube.
Bus memory of HBM is very wide if we compare it to GDDR5. The HBM stack comes with 8 dies DRAM (8-Hi) comes with two 256-bit per channel. White retaining 1024 Wide access HBM managed to achieve 2 Gigs per second speed with the high bandwidth of 256 GB. HBM2 technology allows us to achieve 8GB per Package.
To connect HBM memory with the processor a large number of connections are required to be made. That’s why companies like Nvidia and AMD have used silicon chips that are called Interposers to connect memory with GPU.
If you are not familiar with the term Interposer.
Interposer: Basically it is an electrical/electronic interface that is used to perform routing between one socket/connection to another socket/connection.
By decreasing memory paths Interposer became pretty efficient as it made the processor and memory to be physically more close. However, these semiconductor devices are much more expensive as compared to PCBs.
It is Synchronous graphics random access memory (SGRAM) which is developed by samsung that offers pretty High bandwidth the double date rate that makes it more suitable to be used in game consoles and graphics cards along with high performance computing.
GDDR6 is considered to be the perfect combination for both of its previous models which are GDDR5 and GDDR5x, it’s not only considered as a combined version of both of these but yet a very well improved and fast version.
The GDDR5X SGRAM’s Single pin supports data rate upto 13 gigs per second. With the help of Micron technology GDDR6 was able to achieve 16 Gb/s data rate per pin. If we consider GDDR6 with GDDR5 the improvement rate is almost double.
The lacking feature of GDDR5x is overcomed in the new version as it is supported with 32 byte access granularity.
Granularity: is referred to the amount of data that can be read and written from the memory by using a single Read / write command.
GDDR6 is built with a 10nm manufacturing standard that offers you three high densities of 8GB, 16GB, and 32 GB along with 2 Channel count that supports the data rate upto 16 Gb/s. 2 channels support brusht length upto 16-Cycles that enables us to achieve transaction upto 64 bytes. With GDDR6 you have 1 chip but it gives you a result of 2 fully independent DRAMs with their own bus address and 16-bit data bus.
The Quad Data Range signaling the memory bus which makes it to achieve the higher data rate. While it comes with the mode that shares the total pins of two channels that reduces the number of pins on one side of the memory controller.
The Side by Side Comparison of HBM2 vs GDDR6
Both of these technologies come with 256 GB/s Memory systems. HBM2 offers this fast speed with and only 1 Single DRAM used in it means, a complete bandwidth of 256 Gig will be offered through one DRAM while GDDR6 uses 4 DRAM which enables it to achieve pretty high speed and the bandwidth of 256 Gig is further divided into these 4 DRAMs.
GDDR6 uses the old 2D technology by implementing the 4 DDR as just it was used in GDDR.
In terms of power consumption the HBM2 has an advantage over here as it is using a much more power efficient system. As in HBM2 we are running singles quite slower so by inheriting the speed it is able to achieve quite a low power consumption system. If we see it for GDDR6 then we will be little disappointed here as its power consumption is too high just because of its speed. That’s sure speed is always required to utilize more power.
If we compare both of these technologies by setting the price as a factor then it might be a very tough question to decide. As we discussed both of these technologies are different on their own and price can’t be the factor here in this case. If you are looking for something that offers you high bandwidth just like it is required in Artificial Intelligence and Blockchain which are not necessarily cost sensitive.
While GDDR6 is being used in quite high cost sensitive and non-traditional applications that require lots of speed such as Network Application where you would prefer a mid network router even it is being used in wireless networking applications where it acts as an alternative memory system.
Although HBM is using pretty new technology by utilizing an interposer which could be pretty expensive if we compare it with GDDR technology. That’s why in terms of new technology we see HBM with an edge over GDDR but it’s all about the application if you need something with high bandwidth then you must go with HBM otherwise for high speeds you can go with GDDR6.
If we talk about the structure so HBM2 has a Cuboid or Cube structure while the Square structure is the something which we see as GDDR6.
Talking about the memory capacity we see a big edge for HBM2 here as it offers 4 to 8 gigabytes per stack while on other hand we have 2 GB per stack which is half and even less than the half of HBM.
We all know how important bus speed is in terms of performance so HBM2 offers 3092bit bus speed while GDDR6 offers 256 to 384 bit speed.
If we compare both of these in terms of data handling capability as the world is moving toward AI and Big Data is being very popular day by day so considering data capability won’t be a bad factor. Although these technologies are quite enough to last some years but yet its dependent upon its combination with the processor whether that processor is able to achieve high bandwidth or high speed successfully.
Some devices which are using HBM2:
This Artitecture is being used by Nvidia Titan V, Nvidia Quadro GP100, Radeon RX Vega 56, Radeon Vega Frontier Edition, Radeon Vega RX 64 and yet more.
Some Devices which are using GDDR6:
Although there are lots of applications using this technology some of those applications are, GeForce GTX 1650 Super, Geforce GTX 2090 Ti, Nvidia Titan RTX, RTX 2070, Radeon RX5500 XT, Radeon RX 5700, and RX 5700 XT.
If you are still confused which one could be best for you? So here is the final call. If you are looking for something with modern technologies and a Power efficient system and speed isn’t the most important factor for you then you must select a device that has HBM2 memory.
If you are really conscious about the speed of memory and bandwidth and power consumption isn’t something you really bother about then for sure you should go with GDDR6.
But If you want to hear our choice then definitely we would prefer HBM2.
If you found our content anywhere helpful and finally you managed to decide your option then don’t forget to leave your review in the comment section down below. We appreciate every single comment and we try to reply to it as soon as possible.