You’ve probably heard the buzz: Nvidia is steaming ahead this holiday with their launch of the 4000 series of graphics cards.
With the 4090 releasing last month and the 4080 releasing just this week, it seems Nvidia are hungry to make the most of not just Christmas, but it released just in time to benefit from the ravenous Black Friday and Cyber Monday crowd. And they’d be right too: technology is some of the most sought after products during Black Friday, and it’s an ideal time to grab a good deal before Christmas.
So let’s take a deep dive into the 4000 series, and why people are eager to get its technology into their PCs.
Overview
One of the biggest reasons to be excited is the brand-new architecture embedded inside. Ada Lovelace technology is ultra-efficient, 3rd generation architecture, named after one of the most famous computer programmers in the world, and who could also rock frills and petticoat better than her co-worker, Charles Babbage (Although I’m sure only Babbage knows the answer to that one). This new architecture provides a quantum leap in performance and AI powered graphics and even faster creation workflows. With games coming out becoming ever more demanding, the technology required to run it at the maximum settings, the 4000 series seems like the next step of the process to making games look as they did in our memories as a kid. Things such as boosting performance by up to 4X in fully ray-traced titles. In programs that specialise in 3D rendering, people can see up to 2X the performance in rendering, video export speed and AI tools.
The main differences
Okay, it’s time to get a little bit geeky.
CUDA cores
Nvidia GPU chips’ primary computational power comes from a particular processing technique known as CUDA cores. This stands for the Compute Unified Device Architecture and is the core processors and parallel processing, which allows for the most efficient and fast graphics work. Taking a look at the difference between the CUDA cores, we start to see the difference: there is a 68% drop between the 4090 and the 4080, down from 16,000 to just under 10,000. This is the main difference between the two performances.
Frequency
The frequency is how fast the core can process instructions and creates the graphics. This is further split into two options: the base and the boost. The base is your average speed, whereas your boost is when you’re pushing your card and making the most out of it. This varying of utilisation allows the card to run at varying speed and power so it doesn’t end up trying to overclock itself when it doesn’t need it. Of course, if you want to overclock it, there are models that will boost higher, but for now, we’re looking at the regular offerings. In this case, the 4090 leads ahead with 30MHz high base clock, allowing for higher processing of instructions and better performance.
VRAM
You’ve heard of RAM, but this is video memory, the buffer for the graphics card, which can store the frames for displaying your images. This helps lower the latency between the components and prepping things for when they are needed. Video memory is important for higher resolutions as the high pixel count demands more processing and send out. The 4090 has 8GB more GDDR6X memory so it can fit more in there while using the same VRAM. Although the memory clock does increase with the lower tier card from 21.2 Gbps to 23. However, the smaller bus width does limit the bandwidth. So, it goes down from 384-bit to 256 and so the throughput drops from 1,018 GB/s to 737.7 GB/s.
TBP
Total Board Power is the amount of energy that will be used by the card at maximum utilisation. Cards are getting larger because the power consumption is increasing, and the heat conversion means the size is for keeping the cards cool. The statistics speak for themselves: the 4090 reaches a rating of 450 W, leaving the 4080 trailing at 320 W, making a 130 W difference. The RTX 4080 will run cooler with a lower power consumption, for less performance, which might factor into your choice.
|
|
4090 |
4080 |
|
GPU |
AD102-300 |
AD103-300 |
|
GPU process |
TSMC 5nm |
TSMC 5nm |
|
CUDA cores |
16,384 |
9,728 |
|
Base clock |
2,235 MHz |
2,205 MHz |
|
Boost clock |
2,520 MHz |
2,505 MHz |
|
Memory |
24GB GDDR6X |
16GB GDDR6X |
|
Memory clock/bandwidth |
21.2 Gbps/1,018 GB/s |
23 Gbps/735.7 GB/s |
|
Memory bus |
384-bit |
256-bit |
|
TBP |
450 W |
320 W |
Conclusion
In any world, the 4000 series has blown the 3000 series right out of the water, but that of course, comes with a hefty price tag. The 4090 is retailing around £1649, and the 4080 around £1,269. With the 3080 ti’s price floating just over £1000, it’s a significant price hike. However, from what we’ve seen, customers seem happy to pay, and scalpers are happy to take advantage of people wanting the new technology. The 4090 is undoubtedly the better performing card with a 42%/32% higher MSRP than the 4080. The shortage as we’ve seen in the past may happen again, and so it might be worth to make a decision earlier than you might mean to.
