AMD also seems to be operating on its first-technology Exascale APU solution, the Intuition MI300, run by Zen four CPU & CDNA three GPU cores. The aspects of this HPC chip have also been leaked more than in the latest video clip by AdoredTV.
AMD Intuition MI300 To Be Red Team’s 1st Exascale APU Item With Zen four CPU, CDNA three GPU Cores & HBM3 Memory
The to start with references to AMD’s Exascale APU day all the way back again to 2013 with much more bits and items being disclosed in the coming calendar year. Back in 2015, the firm revealed its strategy to present the EHP, an Exascale Heteorgenous Processor, dependent on the then-upcoming Zen x86 cores and Greenland GPU with HBM2 memory on a 2.5D interposer. The unique options were ultimately scrapped & AMD went on to release its EPYC and Instinct lineup in their own CPU and GPU server segments. Now, AMD is bringing EHP or Exascale APUs back again in the form of the following-gen Intuition MI300.
When all over again, the AMD Exascale APU will kind a harmony amongst the company’s CPU & GPU IPs, combining the latest Zen 4 CPU cores with the hottest CDNA three GPU cores. This is mentioned to be the very first technology Exascale & Instinct APU. In the slide posted by AdoredTV, it is talked about that the APU will be taped out by the stop of this thirty day period which signifies we can see a potential launch in 2023, the very same time the company is envisioned to unveil its CDNA three GPU architecture for the HPC segments.
The initially silicon is anticipated to be in AMD’s labs by Q3 2022. The platform itself is regarded as MDC which could mean Multi-Die Chip. In a previous report, it was stated that the APU will carry a new ‘Exascale APU mode’ and feature help on the SH5 socket which is probable coming in the BGA sort variable.
Aside from the CPU and GPU IPs, another vital driver at the rear of the Intuition MI300 APU would be its HBM3 memory help. However we however are not guaranteed of the specific range of dies showcased on the EHP APU, Moore’s Law is Lifeless has formerly exposed die configurations with two, four, and 8 HBM3 dies. The die shot is proven in the slide in the most recent leak and also shows at least six dies which ought to be a manufacturer new configuration. It is doable that there are many Intuition MI300 configurations becoming worked on with some that includes just the CDNA three GPU dies and APU types with both Zen four & CDNA3 IPs.
So it seems like we are surely heading to see the Exascale APUs in motion following nearly a ten years of wait around. The Intuition MI300 is definitely aiming to revolutionize the HPC place with crazy quantities of overall performance by no means viewed in advance of and with main and packaging technologies that are likely to be a revolution for the tech sector.
AMD Radeon Intuition Accelerators 2020
| Accelerator Name | AMD Instinct MI300 | AMD Instinct MI250X | AMD Instinct MI250 | AMD Instinct MI210 | AMD Instinct MI100 | AMD Radeon Instinct MI60 | AMD Radeon Instinct MI50 | AMD Radeon Instinct MI25 | AMD Radeon Intuition MI8 | AMD Radeon Instinct MI6 |
|---|---|---|---|---|---|---|---|---|---|---|
| CPU Architecture | Zen four (Exascale APU) | N/A | N/A | N/A | N/A | N/A | N/A | N/A | N/A | N/A |
| GPU Architecture | TBA (CDNA three) | Aldebaran (CDNA 2) | Aldebaran (CDNA 2) | Aldebaran (CDNA two) | Arcturus (CDNA one) | Vega 20 | Vega 20 | Vega 10 | Fiji XT | Polaris ten |
| GPU Process Node | 5nm+6nm | 6nm | 6nm | 6nm | 7nm FinFET | 7nm FinFET | 7nm FinFET | 14nm FinFET | 28nm | 14nm FinFET |
| GPU Chiplets | 4 (MCM / 3D Stacked) 1 (Per Die) |
two (MCM) 1 (Per Die) |
two (MCM) 1 (For every Die) |
2 (MCM) 1 (Per Die) |
one (Monolithic) | one (Monolithic) | 1 (Monolithic) | one (Monolithic) | one (Monolithic) | 1 (Monolithic) |
| GPU Cores | 28,a hundred and sixty? | fourteen,080 | thirteen,312 | 6656 | 7680 | 4096 | 3840 | 4096 | 4096 | 2304 |
| GPU Clock Speed | TBA | 1700 MHz | 1700 MHz | 1700 MHz | 1500 MHz | 1800 MHz | 1725 MHz | 1500 MHz | a thousand MHz | 1237 MHz |
| FP16 Compute | TBA | 383 TOPs | 362 TOPs | 181 TOPs | 185 TFLOPs | 29.five TFLOPs | 26.5 TFLOPs | 24.six TFLOPs | eight.two TFLOPs | five.7 TFLOPs |
| FP32 Compute | TBA | 95.seven TFLOPs | 90.5 TFLOPs | 45.three TFLOPs | 23.one TFLOPs | fourteen.seven TFLOPs | 13.three TFLOPs | 12.three TFLOPs | eight.2 TFLOPs | 5.7 TFLOPs |
| FP64 Compute | TBA | forty seven.nine TFLOPs | forty five.3 TFLOPs | 22.six TFLOPs | eleven.5 TFLOPs | 7.four TFLOPs | six.6 TFLOPs | 768 GFLOPs | 512 GFLOPs | 384 GFLOPs |
| VRAM | 192 GB HBM3? | 128 GB HBM2e | 128 GB HBM2e | 64 GB HBM2e | 32 GB HBM2 | 32 GB HBM2 | sixteen GB HBM2 | 16 GB HBM2 | four GB HBM1 | sixteen GB GDDR5 |
| Memory Clock | TBA | three.2 Gbps | three.two Gbps | three.two Gbps | 1200 MHz | one thousand MHz | one thousand MHz | 945 MHz | 500 MHz | 1750 MHz |
| Memory Bus | 8192-bit | 8192-bit | 8192-little bit | 4096-bit | 4096-bit bus | 4096-little bit bus | 4096-bit bus | 2048-bit bus | 4096-little bit bus | 256-bit bus |
| Memory Bandwidth | TBA | 3.two TB/s | 3.two TB/s | 1.six TB/s | one.23 TB/s | 1 TB/s | one TB/s | 484 GB/s | 512 GB/s | 224 GB/s |
| Type Component | OAM | OAM | OAM | Twin Slot Card | Twin Slot, Entire Length | Twin Slot, Entire Size | Dual Slot, Entire Length | Twin Slot, Complete Length | Dual Slot, Half Duration | One Slot, Total Duration |
| Cooling | Passive Cooling | Passive Cooling | Passive Cooling | Passive Cooling | Passive Cooling | Passive Cooling | Passive Cooling | Passive Cooling | Passive Cooling | Passive Cooling |
| TDP | ~600W | 560W | 500W | 300W | 300W | 300W | 300W | 300W | 175W | 150W |
The article AMD’s To start with Exascale APU Rumored To Be Instinct MI300: Run By Zen four CPU Cores & CDNA three GPU Cores For Blistering Quickly HPC Functionality by Hassan Mujtaba appeared initial on Wccftech.
