Finding desktop GPUs to buy was a huge problem during the pandemic. The shortages in supply chains due to shut-down fabs and businesses in addition to the crypto mining madness left the gamers without GPUs in RTX 3000 / RX 6000 era. But now, a new season of GPUs is about the begin.
The end of the GPU shortage
The overall GPU shortage which began with the pandemic has finally ended. Because fabs/businesses are again fully operational and crypto miners have lost interest in mining; it is no longer profitable enough. In addition to the lower profits, Ethereum is now killing GPU-based mining operations.
All those combined, we now have more GPU output from the fabs, and crypto miners are rush-selling their rigs; there is a flood of used GPUs right now. They might not be as “shiny” as normal used GPUs but hey, we are just out of GPU shortage. It is finally possible to buy high-end GPUs in the second-hand market below MSRP, which is great news.
Meet the RTX 4000 series
Nvidia on the other hand has just announced its RTX 4000 series GPUs this week, promising great performance improvements compared to the RTX 3000 series. If their claims are true, RTX 4080 and RTX 4090 are absolute beasts. They offer almost up to 300% performance uplift while keeping the TDP at a reasonable level. But, there are some “but”s, as always.
First of all, Nvidia is developing its products with super handy side products/components. Don’t be mistaken, I have no complaints about those. DLSS, RTX Ray Tracing Cores, Tensor Cores, and Nvenc are all great additions to pure GPUs. But, they are also super handy tools for confusing the customers before third-party benchmarkers get the actual hardware and test them.
I have personally watched Nvidia’s event to see the performance claims; because I intend to buy an RTX 4080 when it is available. However, most of the performance claims were based on tests that are DLSS and RTX ray tracing enabled.
Ray tracing cores causes confusion
Nvidia includes dedicated ray tracing calculation hardware to their GPUs which vastly improves the performance where RTX ray tracing is available. However, the number of games that support RTX ray tracing is still very limited after four years of launch. In addition to that, in current titles, generally, the ray tracing calculations create the bottleneck.
Let’s say our RTX 3080 can generate 150 fps on an X title with rasterization. When we turn on RTX ray tracing, the fps dips to 60. This is because the ray tracing cores can’t handle more than 60 fps while the actual GPU can deliver 150 fps. If we magically give our RTX 3080 extra ray tracing horsepower to get rid of the bottleneck and change nothing in actual GPU cores, it could start delivering 150 fps.
In this case, does it mean our RTX 3080 is now 1.5x more powerful? No, it is not.
We need to see what really DLSS 3.0 is
We have DLSS in those performance metrics as well and it makes the whole situation much more complicated. DLSS is an Nvidia technology that takes low-resolution frames in games and upscales them by utilizing its tensor cores. In that way, the GPU cores will generate low-resolution images for higher performance and they will be shown in the native resolution of your screen.
With the RTX 4000 series cards announcement, this incredibly nice-to-have technology has also received an update. With DLSS 3.0, the tensor cores can now generate whole new frames based on the information from previous frames, which is quite similar to the TV’s “Enhanced Motion” features that make 24 fps movies run at 60 fps. However, Nvidia’s approach is expected to be much more successful at this, because it also uses the data from the game engine such as motion vectors and depth. And of course, hardware-wise, Nvidia should be delivering vastly superior computing power to generate more accurate frames compared to the TVs.
We haven’t closely seen the quality of DLSS 3.0 yet; which can end up being a super new technology or a hard fail with blurs (just like DLSS versions until 2.0) and artifacts. But most importantly, it is included in the performance metrics that Nvidia has shown during its announcement, which causes additional confusion. Personally, I don’t expect more than a 50% performance uplift between GPU generations; but Nvidia claims “up to 4x”.
A massive increase in Nvidia GPU prices
The prices of Nvidia GPUs have also uplifted; RTX 4090 costs $100 more than RTX 3090’s MSRP, and RTX 4080 16 GB (which is the real 4080 IMO) has received a massive $500 uplift; going from $699 to $1,199. RTX 4080 12 GB, which needs to be the “xx70” model of the family, is priced at $899. For comparison, RTX 3070’s MSRP was $499.
The new, 4 nanometer GPUs of the Nvidia RTX 4000 series’ is manufactured by TSMC and they have announced that they have increased the prices of chip manufacturing up to 20% this year. It would of course affect the prices of GPUs. However, it does not justify the $500 gap between RTX 3080 and RTX 4080 16 GB. In my opinion, this could be because of any of those three reasons:
- Nvidia might be thinking that people are now used to paying massive amounts of money for GPUs because the prices of them were skyrocketed during the pandemic. That’s why the company decided to push the prices higher.
- Nvidia might also be trying to quickly get rid of the flooded used RTX 3000 series cards as well as the new ones that are currently lying on the shelves. People may prefer to buy RTX 3000 series because “the RTX 4000 series is too expensive”; after cleaning the market, they might decrease the prices of RTX 4000 cards.
- Nvidia can simply be bluffing for AMD by announcing higher prices just to drop them later. Because AMD is the main competitor of Nvidia and they will announce their GPUs the day Nvidia’s products reach the shelves. This is actually a bit unlikely but Nvidia might just be trying to troll AMD.
We should wait
In any case, I expect a slash for Nvidia GPU prices until the end of this year. We just need to wait for a little while. We as gamers have been waiting for more than two years for a decent GPU; we sure can wait for a little more!