Skip to main content

Why build a $7,000, 2TB graphics card? AMD explains its monster Radeon Pro SSG

It’s rare to find a professional graphics card that makes its consumer counterparts look pathetic. Despite high pricing, cards built for pros rarely put gaming hardware to shame. They’re engineered for durability as much as speed.

The AMD Radeon Pro SSG is different. It uses AMD’s latest Vega core, pairs it with 16GB of second-generation high bandwidth memory (HBM2), and then throws in 2TB of integrated long-term storage made up of four Samsung 512GB SM961 NVMe drives, a newish standard that uses PCI Express lanes to transmit data, providing speeds well in excess of SATA. Two terabytes dwarfs the amount of memory typically found on a graphics card: The Vega RX 64, AMD’s fastest gaming card, has just 8GB of memory.

Stuffing 2TB of solid state storage on a graphics card looks impressive on paper, but why would anyone spend $7,000 – yep, that’s how much it costs — to buy a card that has hundreds of times the available memory of its contemporaries? We spoke with Evan Groenke, head of product management at AMD’s professional graphics business unit, and Gabor Sines, head of video and professional graphics at AMD, to find out.

The Radeon Pro SSG: Evolving a monster

Although AMD is billing the Radeon Pro SSG as the first to “break the terabyte memory barrier,” that honor should technically go to the card’s predecessor, the original Radeon Pro SSG.

Released back in September 2016, it featured a Fiji core (the same found in the Fury line of consumer cards) and a terabyte of additional memory, provided by twin M.2 solid state drives (SSDs) mounted on the card itself. Aimed at producers working with 8K video, it showed impressive performance gains over traditional workstation cards, which are already powerful solutions.

The high-bandwidth cache-controller lets you address lots of memory in a much more efficient manner.”

The new Radeon Pro SSG isn’t distinctly different on the surface but does up the ante in several key areas. Its newer Vega core gives it more processing power and efficiency than its predecessor. That, according to Sines, allowed AMD to improve read and write bandwidth, and reduce the latency between the memory and GPU.

“Year after year, graphics cards have become faster and faster, but as AMD looks at the market and the future, we want to break down the barriers that have existed previously,” Groenke told Digital Trends. “Some of those have been focused around memory, in particular going beyond the gigabyte solutions available on most graphics processing units (GPU) today. We wanted to move into the terabyte range.”

Vega also brings a high-bandwidth cache controller to the table. Through better management of available memory, it handles algorithms that manipulate huge data sets – storing what’s important and ditching data that’s not immediately needed. For gamers, this could result in an end to frame stuttering, but in the entertainment and research sectors, it’s what makes the Radeon Pro SSG’s stack of solid state storage work.

Radeon Pro SSG
Image used with permission by copyright holder

The high-bandwidth cache-controller is what allows you to address lots of memory in a much more efficient manner.” Groenke said. “For all intents and purposes, we’re creating a hardware-based caching solution.”

Despite the improved hardware and capabilities of the new Radeon Pro SSG, it’s decidedly cheaper than its predecessor. The one terabyte SSG cost $10,000 when it first debuted, but the new version is being sold at $7,000.

“The original SSG was a developer kit, so it was a very low volume production run,” explained Groenke. “This is a mass-production product, so its targeted at a different price point and mass professional consumers. We have taken the economy of scale and are pricing it at a place where we believe it’s the right price point in the market.”

Groenke denied that the first-generation SSG was designed to test the water to see if there was a need for such a product. Citing one of his favorite movies, he likened the Radeon Pro SSG to Field of Dreams, stating simply that AMD knew, “if we build it, they will come.”

What’s the point of 2TB of memory?

While 2TB seems like overkill, considering nothing but the SSG’s predecessor comes close to that, Groenke wasn’t wrong when he suggested an interest already exists for such hardware.

“In the medical and entertainment sectors […] we have situations where they are using multiples of our workstation cards,” he said. “In some cases, up to four. They need a lot of GPU power, but they’re also dealing with a lot of data. In those scenarios, they can’t use any of the PCIe slots for storage. The SSG fits into the market very well because it can handle high-end processing and massive data sets.” In a follow-up discussion, and AMD spokesperson clarified the uses, saying “CT, MRI, and anatomical scan visualization would benefit from SSG, particularly for large data sets of full body scans.”

Radeon Pro SSG: Enabling 4K and 8K Content Creation Workflows

8K video editing is another example of resource intensive work that pushes modern hardware to the limit. The first SSG demonstrated enormous gains in this workload, and the second generation will put its foot on the accelerator.

“With uncompressed 8K video editing, you’re looking at gigabytes and terabytes of raw footage that the GPU needs to get access to in a very fast and efficient manner,” explained Groenke. “That was one of the key use cases that we wanted to solve with a product like SSG, by physically co-locating the fast NVMe storage on the graphics board to give it quick access to that storage.”

Sines agreed, stating that “If you are building a visual effect and then render some CG content, you’re rendering in at least 4K. These individual images can be 250 megabytes each. If you do all these computer graphics, and you want to see how your data flows at 30 frames per second, that’s a lot of data to handle.”

By having the storage on the card, the system doesn’t need to wait for data from another drive. Everything is handled right on the card, which helps it manipulate enormous data sets such as real-time raw 8K video editing, and aids users developing high-end 3D models in Maya and Solidworks.

Theory Studios is said to be using AMD hardware for special effects work in the upcoming season of Amazon’s Man in a High Castle, and could end up being one of the first SSG customers.

Scaling for the future

Although the Radeon Pro SSG can already do things that other workstation cards can’t, AMD sees the hardware and its capabilities becoming even more expansive in the future.

“The SSG is more revolutionary than evolutionary at this point,” Sines said. “Two terabytes is our vision of where we see this product fitting. We can produce something that can have more, though.  Two terabytes is just the start.”

When we asked how high he believed the SSG could go on the newly released Vega platform, Sines didn’t seem to think there was any conceivable limit.

“We can produce something that can have more, though.  Two terabytes is just the start.”

“There is no design limit on the storage space. We can go four terabytes or higher, but even if we had eight terabytes, we can always find a scenario where that’s not enough. With the SSG, we are looking to scale in many directions. We are pushing bandwidth, latency, and storage barriers.”

The limits can be pushed further with a tactic hardcore PC gamers will be familiar with — linking multiple graphics cards together in one system. Although it’s not a feature that’s available right now, it’s something AMD is pursuing, and the company thinks it an inevitable future for the hardware.

“We can scale to multiple devices easily,” Sines told us. “Although I can’t disclose the numbers now, from a compute point of view, we are seeing much better scaling in multi-GPU scenarios that you would with gaming.”

AMD is confident in its ability to offer big performance gains because it has a close relationship with its software developers. It’s worked with Adobe in developing the card itself. Internally, AMD has created its own developmental engine called Pro Render. It leverages ray tracing to create photo-realistic images and models which can then be imported into more general game engines.

Radeon Pro SSG: Next Generation Post-Production and Editing

That’s part of AMD’s overall strategy to not only provide the hardware solution, but the software as well. Pro Render supports all manner of 3D applications like Maya, Solidworks and 3DS Max, all of which AMD claims can benefit from the SSG card. Making its own Pro Render work well with them and the card simply completes the package.

As the hardware improves, and software rises to meet its capabilities, we’ll only see greater things from AMD’s Radeon SSG Pro. It’s unlikely to cross over to gamers soon, though it will be worth keeping an eye on. We’re sure AMD will be ready to pounce if game developers find value in the idea.

Jon Martindale
Jon Martindale is a freelance evergreen writer and occasional section coordinator, covering how to guides, best-of lists, and…
Razer lets you stream your PC games on mobile, and you can try it out right now
Razer PC Remote Play visual.

Originally announced at CES in January, Razer's new PC Remote Play platform has now officially launched. Designed to let you borrow the processing power of your desktop PC while playing on a separate device, the service is compatible with smartphones, tablets, Windows PC gaming handhelds, laptops, and even other desktop PCs.

In terms of operating systems, you'll need Windows 10 or 11 on your PCs or on mobile, iOS 18 or Android 14 and up. You can also pair your device with any iOS or Android-compatible controller, or with any mouse and keyboard. If you use one of Razer's own Kishi Ultra controllers, you'll also benefit from some fancy haptic feedback.

Read more
ChatGPT can now remember more details from your past conversations
ChatGPT on a laptop

OpenAI has just announced that ChatGPT received a major upgrade to its memory features. The chatbot will now be able to remember a lot more about you, making it easier to personalize each conversation and adapt its responses. However, the feature won't be available to everyone, and there are a few things to note about the way memory will work now.

The company showed off the new update in a post on X (Twitter), giving a brief demo of how much ChatGPT can remember now. According to OpenAI, ChatGPT can now "reference all of your past chats to provide more personalized responses." Previously, only certain things were saved in memory, but now, ChatGPT can check out every single chat to reference what it knows about you in future conversations.

Read more
Harnessing AI: make Bitrix24’s your sales and marketing MVP
people in a meeting discussing

You’re about halfway through Q2 and your campaigns aren’t landing. Your team is tired. You’re staring down an end-of-quarter push with CRM fields still half-filled, a pile of call recordings no one wanted to transcribe, and one shared doc titled “Q2 Ideas (Pls Delete?).” It’s not that you’re not trying, you’re just tapped.

Enter CoPilot. No fanfare, no flashy onboarding webinars. It’s just there one morning, a new button inside Bitrix24. And somehow, it feels like the only teammate who hasn’t taken a vacation in the past year.
The unexpected power of AI that doesn’t shout “AI”
There’s a lot of hype in the sales and marketing tech world, AI this…neural that…but CoPilot doesn’t posture. It integrates quietly into the Bitrix24 ecosystem: CRM, chat, tasks, feeds, even site-building. It doesn’t try to reinvent your process, it shows up ready to assist with the one you already have.

Read more