'

AMD's on the path to Exascale computing with their new Radeon Instinct MI100 HPC GPU

CDNA is here, and AMD plans to hit Nvidia hard

AMD's on the path to Exascale computing with their new Radeon Instinct MI100 HPC GPU

AMD's on the path to Exascale computing with their new Instinct MI100 HPC GPU

November 2020 is a revolutionary month for AMD's Radeon division, with the company's RDNA 2 and CDNA architectures bringing AMD back into contention within both the gaming and HPC markets. 

Today, AMD has revealed its Instinct MI100, AMD's first CDNA (Compute DNA) product. AMD claims that this GPU is the "World's Fastest HPC GPU", promising up to 2.1x peak performance per $ when compared to Nvidia's Ampere A100.     

CDNA is the graphics architecture which places AMD on the road to exascale computing. AMD has already been tapped to create several exascale supercomputers, and the power of EPYC processors and CDNA graphics is what will bring AMD and their customers to new computational heights.

The Instinct MI100 offers users 120 Compute Units and 7,680 Stream Processors, using TSMC's 7nm FinFET manufacturing processor deliver up to 184.6 TFLOPs of half-precision performance within a TBP of 300W. 

The Instinct MI100 features 32GB of HBM2 memory which offers a total bandwidth of 1.23 TB/s. This HBM memory performance is achieved across four 8GB HBM2 memory stacks. 

  

When compared to the company's older compute architectures, the Instinct MI100 and its CDNA architecture deliver AMD a transformative performance leap. With the company's new Matrix technology, AMD can deliver up to a 7x boost in AI workloads over their older offerings with mixed-precision FP16, and up to 4x fast GPU-to-GPU performance over their new Infinity Fabric architecture. 

With their CDNA architecture, AMD's promising more raw performance than Nvidia in terms of mixed-precision compute at lower power levels. AMD's new ROCm 4.0 software stack should also help deliver strong developer support on the platform. AMD has also been working with DELL, Gigabyte, Hewlett Packard Enterprise and SuperMicro to deliver Instinct MI100-powered systems. 


AMD's on the path to Exascale computing with their new Radeon Instinct MI100 HPC GPU
  

With their CDNA architecture and Instinct MI100 accelerator, AMD plans to compete with Nvidia's HPC lineup and gain market share within this lucrative section of the HPC market. These products should generate notable market share gains for AMD within the HPC market and generate for the company strong revenues. 

You can join the discussion on AMD's Instinct MI100 accelerator on the OC3D Forums.  


AMD's on the path to Exascale computing with their new Radeon Instinct MI100 HPC GPU  

«Prev 1 Next»

Most Recent Comments

16-11-2020, 12:35:55

NeverBackDown
That's a seriously strong card. 120 Compute Units?! That's going to be a pretty big die. Wonder how they managed to alter this architecture compared to RDNA2. Would be cool to see the differences.Quote

16-11-2020, 13:01:34

WYP
Quote:
Originally Posted by NeverBackDown View Post
That's a seriously strong card. 120 Compute Units?! That's going to be a pretty big die. Wonder how they managed to alter this architecture compared to RDNA2. Would be cool to see the differences.
I'm guessing that the Infinity Cache is gone, and that makes space for compute units. Cutting away other gaming features will also give more space for additional compute.

With this market offering richer margins, AMD may also be getting more aggressive with their die sizes.Quote

16-11-2020, 13:26:31

NeverBackDown
You wrote in the article specifically mentioning how the Infinity cache improves performance and the AMD source specifically states it's there in multiple sections. I highly doubt they would alter such a major component of the underlying architecture. Altering cache layout would be a massive and expensive undertaking.

Not sure why you are under the assumption all of a sudden it's gone? Unless I'm reading what you said wrong.

I'm assuming the gaming stuff is out, maybe altering the focus of the RT cores to more compute focused tasks and the support needed for that alongside a large die size increase.. still would like to see a direct comparison just because it's coolQuote

16-11-2020, 13:56:03

looz
Where does the article or AMD page mention Infinity Cache?Quote

16-11-2020, 13:57:35

WYP
Quote:
Originally Posted by NeverBackDown View Post
You wrote in the article specifically mentioning how the Infinity cache improves performance and the AMD source specifically states it's there in multiple sections. I highly doubt they would alter such a major component of the underlying architecture. Altering cache layout would be a massive and expensive undertaking.

Not sure why you are under the assumption all of a sudden it's gone? Unless I'm reading what you said wrong.

I'm assuming the gaming stuff is out, maybe altering the focus of the RT cores to more compute focused tasks and the support needed for that alongside a large die size increase.. still would like to see a direct comparison just because it's cool
RDNA 2 and CDNA are not the same. Both architectures have different requirements and therefore have different features. Infinty Fabric and Infinity Cache are not the same.

I don't know if CDNA has an infinity cache, that's why it is a guess, but with that much HBM2, do they need an on-die cache?

AMD makes no mention of Infinity Cache in their spec page or video.

https://www.amd.com/en/products/serv...instinct-mi100

CDNA and RDNA 2 are designed specifically for certain workloads and price points.

Infinity Cache is a cost-effective solution for AMD and gamers, but when you are using four stacks of HBM2, the cache isn't that necessary. My guess is that AMD saved the die space for additional compute units.Quote
Reply
x

Register for the OC3D Newsletter

Subscribing to the OC3D newsletter will keep you up-to-date on the latest technology reviews, competitions and goings-on at Overclock3D. We won't share your email address with ANYONE, and we will only email you with updates on site news, reviews, and competitions and you can unsubscribe easily at any time.

Simply enter your name and email address into the box below and be sure to click on the links in the confirmation emails that will arrive in your e-mail shortly after to complete the registration.

If you run into any problems, just drop us a message on the forums.