There are two things you should know about RTX 3070. First of all, this is an excellent GPU anyway. Nvidia has always claimed that this $500 GPU can match its $1200 RTX 2080 Ti founder version. In Founder’s Edition released earlier this week, this claim has been roughly proven-RTX 2080 Ti can still lead in some 4K games, but in most games, the two go hand in hand. neck.
Second: Like the RTX 3070, the GPU is great, you may still need to wait a little before buying a new card. In any case, it may not be able to buy it-Nvidia warns potential customers that it hopes to have limited supply before the end of this year to 2021. Separately from this, AMD will launch its own Big Navi card in a few weeks. . It might be wise to compare the two before pulling the trigger.
GA102 and MSI RTX 3070 Gaming X Trio
MSI RTX 3070 Gaming X Trio is built around Nvidia’s GA104 GPU, while RTX 3080 and RTX 3090 use larger GA102 processors. Nvidia (Nvidia) has been manufacturing multiple dies on the high-end in its recent products, rather than fusing components of the same design to manufacture low-end components. When Turing starts, RTX 2080 and RTX 2070 each use a different GPU design (TU104 and TU106 respectively). This time, RTX 3090 and RTX 3080 share a common GPU core (GA102), while RTX 3070 is built around its own unique chip (GA104).
Like its Turing and Pascal predecessors, the RTX 3070 retains the same 8GB RAM capacity that Nvidia has used in this price segment since 2016. The company plans to launch high-VRAM variants of the RTX 3070 and RTX 3080, but these plans have been cancelled or at least have to wait until overall availability improves.
RTX 3070 contains 5888 cores, 96 ROPs, 14Gbps GDDR6 and a 1.725GHz enhanced clock. The large number of GPU cores (more than 2.5 times that of RTX 2070) is why Nvidia is so confident that it can match the RTX 2080 Ti with a much cheaper graphics card. The chip is a 17.4B transistor design based on Samsung’s 8nm process. Through calculation, the computing power of RTX 3070 is about 30% lower than that of RTX 3080, and the memory bandwidth is 41% lower. In addition to using slower RAM (reduced from 19Gbps of RTX 3080 to 16Gbps), RTX 3070 also has a smaller 256-bit memory bus.
That is the GA104 GPU itself. What does MSI and Gaming X Trio bring together?
Gaming X Trio uses a three-axis cooler design and includes additional supports to strengthen the GPU and prevent bending during transportation.
In the past few years, I have seen more than one GPU improperly packaged, and I suspect that the underlying problem is that people do not appreciate the weight of these cards. MSI’s reinforced bracket is very good.
MSI’s “mysterious light” system provides RGB support. If you like this kind of thing, the default RGB is very beautiful. RGB lighting is designed to be controllable through MSI’s Dragon Center application, which is used as an all-in-one diaphragm for controlling RGB, overclocking and managing various system functions.
For those concerned about the condition of the power cord, the MSI RTX 3070 Gaming X Trio uses a dual 8-pin connector instead of a dedicated Nvidia cable.
Our competition options for the RTX 3070 are more limited than I want-I don’t have an RTX 2080 Ti to compare with. We will compare the RTX 3070 with the newly released RTX 3080, the older RTX 2080 (not Super) and AMD’s Radeon VII. The performance of RTX 2070 Super and RTX 2080 is very similar, so RTX 2080 assumes the dual responsibility of two SKUs.
I chose Radeon VII instead of 5700 XT because AMD’s highest-end consumer GCN products often outperformed the newer RDNA chips last year, even if it only improved by a few percent.The truth is that AMD does not Have An excellent GPU that can now be compared in this stand. The current price of the 5700 XT is about US$390, which is much lower than the RTX 3070’s basic suggested retail price of US$500. Radeon VII is not currently on the market.
For this review, I decided to give AMD the best overall foothold. In just a few weeks, we will obtain competitive data on Radeon 6800 and 6800 XT, and be able to better estimate the features and functions of Big Navi for you. In any case, AMD’s data is for reference only. Before Big Navi debuted, Nvidia had been competing with itself. If you want to estimate the performance of the RX 5700 XT, please assume that it is close enough to Radeon VII that you never actually noticed it, but if you run these numbers, it will actually be slower.
All tests were performed on Asus Maximus XII Hero Wi-Fi with 32GB DDR4-3600. All tests used Intel Core i9-10900K, and installed Windows 10 2004 and the latest updates and patches.
The singularity: The upgrade shows that the RTX 3070 is 1.2 times ahead of the RTX 2080 at 1080p, although the gap has actually narrowed as we progress. Ashes didn’t let the GPU bend its knees like before.
“Deus Ex: Mankind Divided” has a dubious difference, in terms of its impact on performance, it has the worst MSAA mode I have ever seen. For this reason, I have always used it as a benchmark. Compared with the RTX 2080 of 1080p, the RTX 3070 maintains an improvement of 1.2 times, but the lead has expanded to 1.26 times that of 4K. This is the first time we have seen the average speed of 30fps when the GPU clears 4K with MSAA enabled, so it is not a small achievement for the RTX 3080.
Under 4K, the RTX 3080 is 1.31 times faster than the RTX 3070, but the price is 1.4 times higher. We expect that at these price points, the price performance curve will start to curve from the 1:1 curve, and the ratio is good-provided that both GPUs can be used for MSRP. Radeon VII matches the RTX 2080 exactly here, but it is clearly the last generation card.
The gap between GPUs is smaller here, but the level of detail we use is lower, which means that the CPU usage is more than typical. The RTX 3070 continues to impress. US$500 spends a lot on the GPU, and the RTX 3070’s US$500 high performance hints at the bright future of the GPU, which will go further in the stack.
AMD’s Radeon VII here can get some benefits from DX12, but this is not enough to change the overall situation. Nvidia is not static here, there is a regular bandwidth between each GPU.
I put “Stranger Trip” into the mix to see what the Vulkan title looks like. Apart from the interesting performance of Radeon VII, there is not much to see here. RadeonVII is about 17% behind RTX 2080 at 1080p, behind RTX 2080 at 1440p, and only 3.4% behind at 4K. The performance lost from 1440p-4K per Nvidia GPU is much higher than that of 1080p-1440p, which means this may be a quirk of the engine.
It sounds boring, you have already seen this picture. Remember, RTX 2080 also stands for RTX 2070 Super. The gap with the standard RTX 2070 will be even greater, and the performance per dollar of revenue will be considerable.
No surprises here.
“Final Fantasy XV” did not run well on Radeon VII-this is a test we suspect that RDNA can bring better results. With the upcoming launch of Big Navi, it’s no big deal either way.
Here are three different Metro Exodus diagrams to highlight three different points. First, we have enabled Ultra ray tracing on the Extreme Detail of RTX 3070 versus RTX 2080. This is the worst-case scenario, where 200% of oversampling is active-the RTX 3070 still manages to increase the frame rate by 1.31 times at 1080p. Interestingly, the gap is minimal without ray tracing enabled-here, the RTX 3070 is not much faster than the RTX 2080. Not sure why the RTX 3080 throws strange, lower results in 4K-I confirmed it in the performance log, but I will run the test again and update this graph as needed.
Big Picture Takeaway
The RTX 3070 is priced at $500 and is an objective and excellent GPU. As I expected, it solved all the problems I had initially encountered with Turing. Ray tracing support has begun to appear in games, and the high-end cards in the series are powerful enough to enable them to be implemented in super high quality without having to worry about techniques such as DLSS 1.0. (DLSS 2.0 is better). For Turing, I do not recommend this family as a long-term investment in ray tracing, because the performance of enabling it may be hit by 60-80%.
But most importantly, with Ampere, Nvidia has returned to the Pascal GPU pricing model. When it launched Turing, the RTX 2080 was 1.28 times faster than the GTX 1080 on average, while the price was 1.28 times higher. At the same time, the RTX 3070 is 1.25 times faster than the RTX 2080, while the price is $200 lower than the GPU at launch.
If you know that you will always be the Green Team and have $500 invested in GPU upgrades, then this is a good choice. Nvidia powers most gaming PCs, which means that game developers will build their own products to target any amount of VRAM provided by Nvidia GPUs. It will be interesting to see if AMD’s 16GB cards can provide a performance advantage, but the 8GB cards will not become obsolete in a year or two. The GPU version of MSI doesn’t spend much English, but there is no need to repair undamaged things, and the most important thing about this GPU is not.
The RTX 3070 is the true successor to the value proposition that Nvidia debuted on the GTX 1080, assuming that people can install GPUs on MSRP. Everyone is guessing whether this is possible, and since this year, the influence of bots has been very ugly, and I have to leave a question mark on this proposition.
If you are willing to wait and see what AMD will bring to you, I suggest you do this-it is always a good idea to look at the inventory of your competitors-but only if the $579 is still in your price range. Either way, the RTX 3070 is a huge leap in gaming, and it is a huge value for affordable gamers.