RTX 4070 sells for 4799 yuan! The “dessert graphics card” went bankrupt?

[ad_1]

It has been more than half a year since NVIDIA released the RTX 40 series graphics card. I wonder if you have used the new graphics card and experienced the huge performance improvement brought by DLSS 3?

Although Xiaolei has brought you a lot of graphics card evaluations in the past six months, due to budget problems, Xiaolei is still using the first generation of light-chasing graphics cards – RTX 2080. This “milestone” equipped with the first-generation ray tracing core is still glowing in Xiaolei’s gaming computer today.

However, the new graphics card that NVIDIA will release next may change the embarrassment that the RTX 40 “cannot be sold”.

RTX 4070 performance analysis

Judging from the hardware number, the RTX 4070 is undoubtedly the most entry-level graphics card for the desktop RTX 40 at present, and judging from the GPU information released by the famous hardware media VideoCardz,This RTX 4070 seems to be aimed at users who used to use RTX 3070Ti, RTX 3070 or RTX 2080S, mainly running games at QHD resolution.

According to the screenshots of the suspected NVIDIA evaluation guide, the RTX 4070 uses the AD104 core, is equipped with 12GB GDDR6X memory, and has a bandwidth of 192bit. It uses the third-generation optical pursuit core and the fourth-generation Tensor Core.

It is worth noting that the RTX 4070 is recorded in the picture as equipped with 36MB of L2 cache, which is far more than the 4MB used in the RTX 3070Ti and RTX 3070.

However, in terms of memory bandwidth, the 504GB/s RTX 4070 is obviously weaker than the 608GB/s RTX 3070Ti.

Image credit: VideoCardz

In terms of TGP, the reference number given by RX 4070 is 200W, which is significantly lower than the 290W given by RTX 3070Ti and slightly lower than the 220W given by RTX 3070. The power consumption is significantly lower than the other two graphics cards, and the standby power consumption is only 10W.

Careful people should have noticed that I did not mention the number of CUDA units of the RTX 4070. This is because there is no record in the picture of VideoCardz. However, according to another media report, the number of CUDA units of the RTX 4070 is likely to be 5888. 1792 less than RTX 4070Ti.

But what followed was a reduction in the price of the RTX 4070: the suggested selling price (MSPR) of the RTX 4070 is only $599, although this price is $100 higher than the MSPR of the previous generation RTX 3070, but it is also in line with NVIDIA’s pricing overall. curve.

Comparing the MSRP of RTX 4070Ti and the domestic retail price, Xiaolei here boldly estimates that the domestic terminal price of RTX 4070 is likely to be below 5,000 yuan, and some brands can even offer a price of around 4,599 yuan during the 618 period.

Image source: JD.com

Indeed, this price is much more expensive than the RTX 3070, even the price that only flagship graphics cards appeared in the past, but compared with the tens of thousands of RTX 4090, 4080 and the precise RTX 4070Ti, it has DLSS RTX 4070 with a series of new technologies such as 3 has become the cost-effective choice for 40 series graphics cards.

Re-division of mid-range graphics card positioning

But then again, judging from the layout of the RTX 40 series desktop graphics card, NVIDIA seems to want to position the RTX 4070 as a mid-range graphics card:

This graphics card can run mainstream 3A games smoothly at QHD resolution, and also has a certain light-tracing ability. After DLSS is turned on, it can also meet the threshold of 60FPS.

However, due to the limited memory bandwidth, the “bottleneck” of RTX 4070 is also very significant:

Neither the RTX 4070 nor the RTX 4070Ti can cope with the exponentially improved memory throughput at 4K resolution, so the performance of the game at 4K resolution is far inferior to the RTX 3090Ti that was “killed” by itself.

In the RTX 4070Ti graphics card evaluation, we have seen that the RTX 4070Ti has a frame rate performance that greatly exceeds that of the RTX 3090Ti at QHD resolution, but from the test method, these soaring FPS are inseparable from the DLSS 3 in the new graphics card. Sampling techniques and frame generation techniques.

　　In the final analysis, behind the improvement of FPS numbers is the evolution and iteration of DLSS technology and algorithms, which has a limited relationship with the actual performance of the graphics card.

Image credit: NVIDIA

From this point of view, NVIDIA seems to regard the ability of 4K games as the dividing line for the positioning of RTX 40 series graphics cards:

　　High-end and flagship graphics cards have a lot of performance, and mid-range graphics cards maintain a seemingly good FPS performance under limited performance by using AI and other “crazy methods”.

As for why the bottleneck of mid-range graphics cards is spent on 4K resolution?

In addition to the data throughput brought by 4K, 4K resolution is also a dividing line of DLSS technology: even if it is RTX 4090, NVIDIA recommends that DLSS be set to performance mode when using 4K resolution games, and QHD to be balanced. , some games can even be opened to quality priority.

Although the launch of “mid-range graphics cards” such as the RTX 4070 has released NVIDIA’s “goodwill” to a certain extent, considering that there are still many players who only have a budget of 5,000 to 6,000 yuan for the entire computer, a price of 599 US dollars The graphics card is clearly far beyond their affordable price range.

Compared with the illusory QHD performance brought by DLSS 3 and the light-chasing game effect brought by AD104, what they need more is a card under 4,000 yuan, which can meet the basic needs at QHD resolution with “almost” picture quality. The dessert graphics cards required by online games are the RTX 4060 and RTX 4050 graphics cards that were exposed in February.

The good news is that RTX 4060 and RTX 4050 are indeed not far away from us.

Don’t take “dessert” as “staple food”

According to the revelations, Shangke Group, a company that owns Ming? and Soyo graphics card brands, has registered RTX 4060 and RTX 4050 graphics cards. Although Ming? The RTX 4060 and 4050 are already on the way.

According to NVIDIA’s consistent knife method, 4060Ti will use a complete AD106 core, 4060 will be equipped with a castrated version of AD106 core, and 4050 will be equipped with a complete AD107 core.

Although the RTX 4060 that everyone is most concerned about uses the residual blood version of the AD106 core, it is still much stronger than the previous generation of the GA106 core (that is, the core carried by the 3060), even after being castrated. leading performance.

Coupled with TSMC’s new process, DLSS 3, SER and other technical and architectural improvements, the performance of the RTX 4060 will be about 30% stronger than that of the RTX 3060, and the power consumption will drop a part. Generally speaking, it is still a real Dessert graphics card.

Image credit: NVIDIA

As for other configurations, some bloggers have given relevant news: RTX 4060 has 3968 CUDA cores and 8GB video memory, the memory speed is 17Gbps, and the score in the 3D Mark Time Spy Extreme test is 6000 points, which is almost the same as The RTX 3070 series is flat, which is also in line with NVIDIA’s consistent style.

What is certain is that the 40 series will definitely be much more expensive than the 30 series, so Xiaolei guesses that the official price of the RTX 4060 will be around 3,500 yuan, and the 4050 will be around 2,600 yuan. The price increase is inevitable, otherwise there will be no consumers Willing to “help” Lao Huang clean up the remaining RTX30 series inventory.

　　From the price point of view, RTX 4060 is obviously more attractive than RTX 4070, but if you consider the very realistic “CUDA unit number / selling price” ratio, RTX 4070Ti, RTX 4080 or even RTX 4090Ti will be a more cost-effective choice .

Image credit: NVIDIA

It has to be said that although the mine tide that lasted for more than two years has passed, the performance of the game graphics card market is not satisfactory to us. The price/performance ratio of dessert-level graphics cards has declined, and consumers can hardly afford the price of mid-to-high-end graphics cards.

This can’t help but make people worry that the future graphics card market will become more polarized:Although low-end graphics cards are cheap, people are unwilling to spend them. Instead, people are willing to bite the bullet and buy high-end graphics cards, or directly choose old cards to fight for another three years. In fact, this is not a good signal for the development of the entire graphics card market.

After all, PC games are just a form of entertainment for most people. In the face of the real price level, the “waiting party” in the past had no choice but to grit their teeth and save money.

As for the current “waiting party”, if you really can’t wait to play mobile games and game consoles: what NVIDIA loses is the opportunity to clear the inventory, and players who can’t afford a graphics card “lose only the shackles.”

[ad_2]

Source link