Throughout Dell’s EMC presentation, AMD’s CTO, Mark Papermaster, confirmed that they are going to be introducing the next-generation CDNA architecture-based Radeon Intuition MI100 accelerator through the second half of 2020.
AMD’s Radeon Intuition MI100 CDNA Structure Primarily based Discrete GPU Accelerator Arriving in 2H 2020
The AMD Radeon Intuition MI100 which is internally known as ‘Arcturus‘ will probably be a next-gen HPC half that can characteristic an enhanced model of the 7nm Vega structure. The accelerator has by no means been talked about by AMD formally till now. The GPU appears to be the highest HPC half for 2020 within the AMD first-generation CDNA portfolio. Mark confirmed that the Discrete GPU will probably be launched within the second half of 2020.
Mark Papermaster confirmed MI100 Discrete GPU accelerator for 2H 2020 pic.twitter.com/P6KTrm0B2S
— Hassan Mujtaba (@hms1193) June 17, 2020
Following is the quote from Mark through the Q/A session:
Like our multi-generational dedication to the Zen roadmap in x86 CPU, now we have completed the identical with our DNA architectures for GPU – rDNA for gaming and visualization, and cDNA for compute & AI. The rDNA is driving acquire in AMD share for graphics and deployed within the upcoming Sony and Microsoft new recreation consoles, and for cDNA you will note the MI100 discrete GPU each 2nd half of 2020.
The ROCm software program stack creates an alternate for GPU compute with simple portability and enabling competitors. – AMD CTO, Mark Papermaster
Primarily based on what now we have discovered from varied prototype leaks, the Radeon Intuition MI100 ‘Arcturus’ GPU will characteristic a number of variants. The flagship variant goes within the D34303 SKU which makes use of the XL variant. The information for this half is predicated on a take a look at board so it’s possible that ultimate specs wouldn’t be the identical however listed here are the important thing factors:
- Primarily based on Arcturus XL GPU
- Take a look at Board has a TDP of 200W
- Up To 32 GB HBM2 Reminiscence
- HBM2 Reminiscence Clocks Reported Between 1000-1200 MHz
The Radeon Intuition MI100 take a look at board has a TDP of 200W and is predicated on the XL variant of AMD’s Arcturus GPU. The cardboard additionally options 32 GB of HBM2 reminiscence with pin speeds of 1.0 – 1.2 GHz. The MI60 as compared has 64 CUs with a TDP of 300W whereas clock speeds are reported at 1200 MHz (Base Clock) whereas the reminiscence operates at 1.0 GHz together with a 4096-bit bus interface, pumping out 1 TB/s bandwidth. There is a massive probability that the ultimate design of the Arcturus GPU may very well be that includes Samsung’s newest HBM2E ‘Flashbolt’ reminiscence which provides 3.2 Gbps speeds for as much as 1.5 Tb/s of bandwidth.
AMD Radeon Intuition Accelerators 2020
|Accelerator Identify||AMD Radeon Intuition MI6||AMD Radeon Intuition MI8||AMD Radeon Intuition MI25||AMD Radeon Intuition MI50||AMD Radeon Intuition MI60||AMD Radeon Intuition MI100|
|GPU Structure||Polaris 10||Fiji XT||Vega 10||Vega 20||Vega 20||Arcturus|
|GPU Course of Node||14nm FinFET||28nm||14nm FinFET||7nm FinFET||7nm FinFET||7nm FinFET|
|GPU Clock Velocity||1237 MHz||1000 MHz||1500 MHz||1725 MHz||1800 MHz||~1500 MHz|
|FP16 Compute||5.7 TFLOPs||8.2 TFLOPs||24.6 TFLOPs||26.5 TFLOPs||29.5 TFLOPs||185 TFLOPs|
|FP32 Compute||5.7 TFLOPs||8.2 TFLOPs||12.3 TFLOPs||13.3 TFLOPs||14.7 TFLOPs||23.1 TFLOPs|
|FP64 Compute||384 GFLOPs||512 GFLOPs||768 GFLOPs||6.6 TFLOPs||7.4 TFLOPs||11.5 TFLOPs|
|VRAM||16 GB GDDR5||4 GB HBM1||16 GB HBM2||16 GB HBM2||32 GB HBM2||32 GB HBM2|
|Reminiscence Clock||1750 MHz||500 MHz||945 MHz||1000 MHz||1000 MHz||1200 MHz|
|Reminiscence Bus||256-bit bus||4096-bit bus||2048-bit bus||4096-bit bus||4096-bit bus||4096-bit bus|
|Reminiscence Bandwidth||224 GB/s||512 GB/s||484 GB/s||1 TB/s||1 TB/s||1.23 TB/s|
|Type Issue||Single Slot, Full Size||Twin Slot, Half Size||Twin Slot, Full Size||Twin Slot, Full Size||Twin Slot, Full Size||Twin Slot, Full Size|
|Cooling||Passive Cooling||Passive Cooling||Passive Cooling||Passive Cooling||Passive Cooling||Passive Cooling|
To date, AMD has unveiled that the principle focus of CDNA could be efficiency, effectivity, options, scalability within the information heart market. At the moment, AMD’s GCN structure has served this phase however with CDNA, AMD will probably be creating GPUs particularly optimized for high-performance Compute, Machine Studying, and HPC. The first Gen CDNA GPUs will characteristic a 2nd Gen Infinity Structure and would make the most of the ROCm (Radeon Open Compute Platform) to energy the information heart with key optimizations and enhanced scalability. The 2nd Gen Infinity Structure will permit for 4-8 Approach GPU connectivity in a singular node, permitting the brand new Radeon Intuition boards to run in concord.
AMD has proved that they will provide extra FLOPs at a aggressive worth so perhaps that’s the place Arcturus could be targetting. With 2H 2020 now set for the Radeon Intuition MI100, we’ll quickly understand how AMD’s newest CDNA structure competes in opposition to NVIDIA’s Ampere based A100 GPU accelerator within the HPC phase.