Image: AMD

AMD Vega VII initial impressions – Can AMD challenge Nvidia now?

  • Twitter
  • Facebook
  • Google+
  • Pinterest

Many products were launched at CES 2019. Many of them were conventional products from respective brands. However, many products were surprise releases. Among these today, we are talking about the new high-end Graphics card from AMD. AMD and Nvidia are the two manufacturers of consumer Graphics cards, Intel also announced their arctic sound (the name of their Graphics card) but they are not out yet.

For two generations Nvidia is leading the high-end market of the Graphics cards. AMD was unable to put forth respective competitors for the high-end Pascal or Turing Graphics cards. Last year they released the VEGA GPUs, but they were no match for the GTX 1080 let alone the GTX 1080Ti.

Introduction

On the other hand, since the launch of the new RTX series AMD was determined to produce a Graphics accelerator that can look at the very least rival the tradition rasterization performance. They also released the VEGA 2.0 architecture a few months back. The new architecture was based on TSMC’s 7nm manufacturing process. The manufacturing process could turn out to be the turning point for AMD since Nvidia is still working with the 12nm FinFET process.

We were also expecting a new architecture from AMD that is centered at gaming as VEGA 2.0 was released as a workstation architecture. The Radeon Pro the workstation division of Radeon graphics released a bunch of workstation oriented Graphics cards using the VEGA 2.0 architecture.

Before AMD’s press conference it was not sure whether AMD would use VEGA 2.0 or introduce a new gaming architecture. However, during their press release, they showed their new High-end Graphics card that according to them is capable of beating the RTX 2080.

The AMD Radeon VEGA VII is based on the VEGA 2.0 architecture that is built with TSMC’s 7nm fabrication process. During their keynote, AMD’s CEO Lisa Su announced that AMD would be competing with Nvidia at the high-end level. They have their eyes on the untouchable Nvidia GeForce RTX 2080 Graphics card. You can only find out about it once the Graphics card is launched for the masses on February 7th.

Design

Let’s talk about the design of the sample GPU first. It is huge; we can see the gigantic triple fans on the front of the shroud. The heatsinks too are beefy which means they have to dissipate a lot of heat. There is no specific design or branding here which means it is an early sample of the Graphics card.

Image: AMD
Image: AMD

Architecture

We already know that the GPU they are using in the Radeon VEGA VII is VEGA 20 the same GPU they have announced earlier. It is actually the cutdown version of the VEGA 20. Instead of going all out with the 64CUs AMD decided to go with 60CUs. It should be noted that the previous flagship Graphics card from AMD the VEGA 64 had all 64CUs enabled. AMD is claiming that the incremental benefits of the 7nm manufacturing process.

AMD makes up the loss of compute units by higher clock speeds and much more powerful memory and pixel filtering at the backend. They claimed that the Radeon VII should beat the former flagship anywhere between 20% and 42% depending upon the requirements of the task. On average, the new GPU is around 29% faster than the VEGA 64.

According to AMD, it is barely enough for the Radeon VII to beat the RT 2080. However, RTX 2080 offers a lot more than traditional Rasterization at the same price. On the other hand, the new features put forth by the RTX 2080 have little to no use case at the time being. Despite many efforts from Nvidia most games that promised Raytracing did not bother to release them in the alpha build yet.

Specifications

Let’s talk about the numbers for the new Radeon VEGA VII Graphics card. If you look back at the Radeon Instinct M150, you will notice that the VEGA VII is almost identical to it. There are some changes here that make M150 an enterprise Graphics card and VEGA VII a consumer device.

On paper, however, they look almost identical. Both of these Graphics cards share the same VEGA 20 GPU with 60 enables CUs. The VEGA VII takes complete advantage of these by having better clock speeds throughout. It should be worth noting that the VEGA 20 is not actually an upgrade from VEGA 10. It can be referred to as a side upgrade that took anything wrong with the VEGA 10 and made it right.

Image: AMD
Image: AMD

These side upgrades have their advantages too. Some of them make the architectural jump from VEGA 10 to VEGA 20 slightly better than the jump from Pascal to Turing. First of all the fabrication process that AMD opted was 7nm. We still think it is too early for the consumer market to get the 7nm process because of the added costs and limited benefits.

However, the benefits are here, and they can turn out to be a life-saving strategy for AMD. The other and major benefit comes from the memory department. Nvidia upgraded their memory modules from GDDR5 to GDDR6 which is faster, smaller but costs way more than GDDR5. AMD did not do that they used their existing cost efficient HBM2 memory and increased the speed and bandwidth.

HBM2 Memory

HBM memory is usually better at handling workloads than the GDDR memory because of much higher bandwidth. On the other hand, the GDDR memory is better at processing because of its faster interface. But we have noticed that most games benefit more from the memory interface rather than the speeds. Most pen world games rely on VRAMs to store huge amount of texture data, and the memory with higher bandwidth gets the advantage here because of bigger BUS. In easier terms, the data gets transferred seamlessly.

For Radeon VEGA VII they increased the clock speed of the memory to a reasonable 2Gbps, it is not comparable to the 14Gbps speed of the GDDR6 memory. They also increased the memory interface to a whopping 4096-bit. They also increased the number of memory controllers which resulted in 16GB of HBM2 memory onboard. It makes the overall memory bandwidth equal to 1TB/s. Memory bandwidth on such a scale is not only beneficial for games but also for heavy duty workloads.

The strategy

Even after the improvements that AMD made, it is still the power hungry VEGA architecture. AMD did not talk about the TDP of the Graphics card, but our guess for this card would be around 300 Watts. Since the Instinct M150 that it very closely resembles is rated ta 300 Watts, so it is safe to assume it will also be hitting the same amount of power.

The strategy AMD is employing here is not a good one, but if it works it would be beneficial for the company overall. They are not really worrying about the efficiency of the Graphics card right now the only motive for this Graphics card is to beat or match the performance output of the RTX 2080. AMD claims that it has a 25% performance increase for the same amount of power. It nowhere indicates that it will be as efficient as the RTX 2080.

The other thing they have done is to make sure that the VEGA VII does not cannibalize the sales of the instinct M150. So they tried to differentiate almost similar products. It particularly high output for the FP64 related workloads. In addition, they announced the new low precision modes that will help in neural processing or AI-related tasks. Secondly, they reverted the compatibility with the upcoming PCIe 4.0 standard. So, if you upgrade your motherboard in future for a better PCIe, 4.0 experience Instinct M150 would be a better product for you.

RTX 2080 vs Radeon VEGA VII

Now let us get into the comparison they did between their Radeon VEGA VII and the RTX 2080. It should be noted that these comparisons should be taken with a grain of salt. However, for the moments we would have to work with the numbers provided by AMD itself. If AMD manages to hit the performance output shown by the RTX 2080, then these Graphics cards would be equal to each other at traditional rasterization scenarios.

There is no way AMD can achieve the energy efficiency of the RTX 2080 which makes the RTX 2080 a better deal. The benchmarks they showed during there keynote show the both Graphics cards are almost similar when it comes to performance in DX11. AMD gets ahead when Vulkan API is used in certain games due to better integration with the API. They did not show the DX12 results we can guess who would win those.

Image: AMD
Image: AMD

On top of these AMD has no hardware yet to support the new Direct X Raytracing (DXR). They have plans to get their hands dirty in this field too. But they are waiting to see how the technology develops before implementing it in their Graphics cards. Another point which favors the RTX 2080.

Price

The price is the area where AMD always beat Nvidia. Their Graphics cards usually have better bang for the buck compared to Nvidia’s Graphics cards. However, the case with the Radeon VII is exactly the opposite. It has the same price as the RTX 2080 that is the Graphics card will cost you around 699 dollars. In our early analysis, it seems like the product from Nvidia is better at every stage compared to the AMD’s new “flagship” Graphics card. However, we should save our judgment for the detailed review and comparison of the Graphics card.

They are using the same strategy as Nvidia is using. At this price, you will get the bundle. The bundle includes the Graphics card, and copies of Devil May Cry 5, Resident Evil 2 and The Division2. It should be mentioned the Graphics card can play all these games at 4k max settings. If the product intrigues you, you can learn about ordering the Graphics card at launch here.

Image: AMD
Image: AMD

 

 

Leave a Reply

Your email address will not be published.
Required fields are marked *

error: Content is protected !!