Introduction
AMD as a brand that needs no introduction. Be it their powerful and efficient CPU or Processors or their Graphics Cards, AMD has presence wherever computing is required.
In the past few years AMD has seen tremendous success with their CPU business with all their SKUs gaining popularity because of them offering cutting edge performance at the same time being extremely power efficient.
This transformation started with the launch of the very first Ryzen CPU for desktop. It didn’t stop at desktops, it quickly expanded to HEDT, Server, Enterprise, Embedded, Laptops and so on. With every new generation of Ryzen, the domination got more and more stronger.
Many might know that AMD ventured into the GPU business with them taking over Graphics technologies giant, ATI Technologies back in 2006. Since then they tasted success but had their share of lackluster products as well. Even then they had a lot of design wins in the console space where they more or less command a monopoly. Although same couldn’t be said about their Desktop Gaming or Professional graphics products.
While the CPU side of things was very busy and lots was happening, the GPU side of things was comparatively calm at AMD. Once the ship was steadied by Ryzen, AMD focused on the GPU business to work on newer and stronger products which could trade blows with the competition and offer customers a worthy option.
The turning point for the GPU business must be attributed to their brand new RDNA architecture. RDNA marked the end of GCN architecture which was ageing and needed to be scrapped and a fresh start be made if AMD hoped to be competitive again.
Recently we saw the introduction of RDNA3 architecture which saw a big uplift in performance while being quite power efficient. AMD launched RDNA3 with 2 brand new Gaming GPUs, Radeon RX7900XT 20GB and Radeon RX7900XTX 24GB.
Both these GPUs closed the gap substantially against the competition in majority of the games with some exceptions. Not only did they offer excellent performance, they did so at an affordable price.
With RDNA3, it was evident that the fresh approach taken by AMD was giving the expected results and AMD was again on the path to being competitive like in the past.
Now that there is a brand new competitive architecture in hand, it logically makes sense to offer Professional graphics solutions which is a space which was more or less vacant as far as AMD is concerned.
Today we have with us AMD’s latest Professional Graphics solution in the Radeon PRO W7800 32GB which is based on the RDNA3 architecture.
Without wasting much time, let’s get on with getting to know this new GPU better and then take a look at what kind of performance it offers.
Specifications
Architecture – RDNA3
Hardware Ray Tracing – YES
Lithography – 5nm GCD / 6nm MCD
Stream Processors – 4480
Compute Units – 70
Peak Half Precision (FP16) Performance – 90.5 TFLOPS
Peak Single Precision Matrix (FP32) Performance – 45.2 TFLOPS
Power Connectors – 2 x 8 Pin PCIe
Memory – 32GB
Memory Type – GDDR6
AMD Infinity Cache – 64MB
Memory Interface – 256 Bit
Peak memory Bandwidth - ~ 576 GB/s
ECC Support – YES
PCIe Bust – PCIe Gen4.0
Form Factor – Dual Slot
Cooler – Blower Type
Supported Rendering Formats – 1 x Encode & Decode AV1, 2 x Decode (H265/HEVC, 4K H264), 2 x Encode (H265/HEVC, 4K H264)
10Bit Display Color Output – YES
12 Bit Display Color Output – YES
Supported Technologies – AMD Viewport Boost, AMD Software: PRO Edition, AMD Remote Workstation, AMD Radeon VR Ready Creator, AMD Radeon Media Engine, AMD Radeon ProRender
API Support – DX12 Ultimate, Open GL 4.6, OpenCL 2.2, Vulkan 1.2
Closer Look
Most of the Professional Graphics Cards have a Blower Design cooler because it is most ideal orientation when multiple GPUs are installed in a System, especially Rackmount Servers.
3 x Display Ports + 1 x Mini DP Port
2 x 8 Pin Regular PCIe Power Connectors. Good to see AMD utilizing the common 8 Pin PCIe Power Connector. This makes compatibility with older PSUs easier.
The AMD and Radeon Logo is machined in the aluminum shroud rendering an overall premium and top quality finish.
Testing
Just like games, there are countless professional applications which utilize GPUs to accelerate computing. Be it designing, simulations, GPGPU compute, 3D Rendering, etc. The list is endless. It is practically impossible to test a Card for different use case scenarios because the permutation & combinations would be mind numbing.
Therefore, we have chosen a few synthetic tests which should give us a fair idea about this card’s performance. Of course there are various scenarios where it might excel and in some cases lag as compared to the competition. This purely depends on software optimization and of course the hardware architecture.
Test Bench
AMD Ryzen 9 7950X 3D Processor
ASRoc X670E Taichi Motherboard
360mm CPU Liquid Cooler
16GB x 2 Patriot Viper 6800MHz DDR5 Memory (Tuned at 6000MHz 32-38-38-80)
AMD Radeon PRO W7800 32GB Graphics Card
4TB Crucial P3 Plus PCIe Gen4 M.2 NVMe SSD
Corsair RM1000e 1000W 80+ Gold PSU
Software
Windows 11 X64 PRO (Latest updates installed)
Latest Drivers for motherboard, GPU, etc. installed.
Latest Motherboard BIOS installed.
Benchmarks
Geekbench 6 GPU Compute – OpenCL
Geekbench 6 GPU Compute – Vulkan
3DMark Time Spy
3DMark Time Spy Extreme
3DMark Port Royal
3DMark Speedway
Superposition 4K Optimized
Superposition 8K Optimized
AIDA64 GPGPU Benchmark
PCMARK 10 Extended
PugetBench – Adobe Photoshop
PugetBench – Adobe Premiere Pro
PugetBench - Adobe After Effects
PugetBench – Davinci Resolve
The benchmarks results with Pugetbench Adobe Premier Pro were confusing. Most of the Scores of other cards are based on older Premiere Pro version. Maybe the latest version impacts performance and hence the other cards need to be re-tested to validate the scores. We shall share the results of the same in the future.
Similar behavior is observed in the case of After Effects as well. So Premier & After Effects both need to be re-tested to figure how different GPUs stack against each other.
Thermal Performance
We stress tested the GPU for 2 Hours of 8K rendering to monitor the thermal performance.
GPU core temperature maxed out at 82c
GPU Memory Junction temperature maxed out at 92c
GPU hotspot Temperature maxed out at 98c
Ambient temperature recorded by the fan intake temperature sensor was 28c
The cooler fan peaked at ~2200rpm
GPU Total Power Draw hovered around ~325W
GPU Core Clock speed was around 2,500MHz
The temperatures aren’t exactly low as such but that is understandable considering that it uses a blower cooler design which isn’t exactly the most efficient.
It would be interesting to see how the Card performs if it is watercooled. How well the GPU scales with much lower temperatures and how much that impacts the GPU performance would give us an insight as to how much performance is still available to be tapped if thermal limitations are taken care of.
With power draw of ~325W it isn’t power hungry at all for the performance it offers. Here, once again AMD showcases superb power efficiency just like their CPUs. Good Job!
Observation, Conclusion & Final Words
After a long time AMD finally has a GPU architecture which is competitive and can become a solid foundation for their future architecture and products based on them
The RDNA3 architecture has shown serious performance gains over the predecessor and the same can be seen in this case as well.
In addition to the improved performance there are added advantages like ECC Memory, 10 and 12 Bit color display output, Software tweaks and optimizations like ProRender, Viewport Boost, Media Engine, VR Ready Creator, etc. which critical and useful for professional graphics users.
The price for India is not yet disclosed but if we go by international pricing which is 2499 USD, that translates to ~2,05,000/- Ex GST which means ~ 2,41,900/- Inclusive of GST.
Looking at the price, it is evident that it isn’t exactly cheap and it will have to fight it out with established SKU like the A5000, A6000 and so on.
The biggest advantage will be the 32GB frame buffer which is becoming critical nowadays with various workloads demanding more and more frame buffer.
The good thing here is that the architecture is strong enough to exploit this frame buffer.
It would be interesting to see whether AMD plans to introduce cheaper variants in the Radeon Pro lineup because that would see better acceptance and adoption as far as Indian market is concerned.
Thank You!
Comments