Intel’s upcoming ARC Alchemist GPU line-up is the recent hot-topic of the market and recently many of the Intel executives including Raja Koduri talked about the upcoming GPUs. There were many more things discussed in the interview on topics such as HPC and HPG graphics architectures which include Ponte Vecchio & ARC GPUs.
The first thing here is to understand why Intel chose TSMC over its Fabs to manufacture ARC GPUs and the reason was that during the development of ARC GPUs for the gaming segment, the top priority was given to the process that offered a good manufacturing capacity.
So for this reason Intel’s executives simply decided that intel’s process was not a great fit for ARC GPUs due to the insufficient capacity of the Intel 7nm node process. Other characteristics such as the operating frequency and cost were also factored in & the company decided it was the best to outsource the manufacturing to TSMC as its N6 (6nm) node offered the best balance.
Intel Xe-HPG ARC Alchemist GPUs To Offer More Than 32 Xe Cores?
The executives also went on to further state that its Xe-HPG architecture is very scalable and by this information, we can assume that the 8-slice Xe-HPG ARC Alchemist GPU shown to the press was the flagship 32 Xe-Core config and we can expect to get a higher-end SKU sometime in the future.
Intel ARC Alchemist vs NVIDIA GA104 & AMD Navi 22 GPUs
| GPU Name | Alchemist DG-512 | NVIDIA GA104 | AMD Navi 22 |
| Architecture | Xe-HPG | Ampere | RDNA 2 |
| Process Node | TSMC 6nm | Samsung 8nm | TSMC 7nm |
| Flagship Product | ARC (TBA) | GeForce RTX 3070 Ti | Radeon RX 6700 XT |
| Raster Engine | 8 | 6 | 2 |
| FP32 Cores | 32 Xe Cores | 48 SM Units | 40 Compute Units |
| FP32 Units | 4096 | 6144 | 2560 |
| FP32 Compute | ~16 TFLOPs | 21.7 TFLOPs | 12.4 TFLOPs |
| TMUs | 256 | 192 | 160 |
| ROPs | 128 | 96 | 64 |
| RT Cores | 32 RT Units | 48 RT Cores (V2) | 40 RA Units |
| Tensor Cores | 512 XMX Cores | 192 Tensor Cores (V3) | N/A |
| Tensor Compute | ~131 TFLOPs FP16 ~262 TOPs INT8 | 87 TFLOPs FP16 174 TOPs INT8 | 25 TFLOPs FP16 50 TOPs INT8 |
| L2 Cache | TBA | 4 MB | 3 MB |
| Additional Cache | 16 MB Smart Cache? | N/A | 96 MB Infinity Cache |
| Memory Bus | 256-bit | 256-bit | 192-bit |
| Memory Capacity | 16 GB GDDR6 | 8 GB GDDR6X | 16 GB GDDR6 |
| Launch | Q1 2022 | Q2 2021 | Q1 2021 |
Intel ARC Already Has A Reference Graphics Card Design But Partners are Open To Custom Models
according to information gathered from the executives its made clear that Intel has already shipped its reference design to its partners. As we know from the interview, the company is currently investigating if it’s worth launching a reference-only flavor similar to NVIDIA’s Founders Edition or allows partners (ODMs) to manufacturer their custom variants.
Intel XSS Is Backwards Compatible With DG1 ‘Xe-LP’ & 11th Gen CPUs
XSS has been one of the most interesting technologies of the upcoming Xe-HPG ARC graphics lineup. And the silicon giant has now officially confirmed that the technology will be backward compatible with both Xe-LP-based DG1 GPUs & iGPUs on the 11th Gen Tiger Lake GPUs.
