Tech Instructor

World Best Inspector to Check Tech Reviews

How Will Intel Arc Alchemist Graphics Cards With Xe-HPG Architecture Deliver?
Computers & Laptops

How Will Intel Arc Alchemist Graphics Cards With Xe-HPG Architecture Deliver?

How Will Intel Arc Alchemist Graphics Cards With Xe-HPG Architecture Deliver?

The high-end Arc Alchemist graphics card model can outperform AMD’s Radeon RX 6700 XT and NVIDIA’s RTX 3070 graphics cards, according to estimates.

Intel ARC graphics cards, which will be powered by Alchemist Xe-HPG GPUs, are ready to be released in the first months of next year. We have already talked in detail about all the details about the new Alchemist graphics cards and the Xe-HPG architecture . Now, some predictions have been made about the potential performance that Intel’s new cards will offer.

First of all, let’s mention that TSMC 6nm production technology will be used in the new graphics cards of the blue team . In addition, these GPUs adopt fundamental building blocks that use the Xe-core approach.

RX 6700 XT and RTX 3070

Intel’s ARC Xe-HPG Alchemist flagship model will offer more TMUs and ROPs than NVIDIA and AMD rivals. The total number of cores achieved at 4096 is higher than AMD’s Navi 22 and Navi 21(RX 6800) GPUs, but lower than NVIDIA’s GA104 graphics processor.

Although Intel’s ARC Alchemist GPUs have lower ray tracing units than their competitors, we don’t know exactly how ray tracing applications work. For example, Navi 22 offers more RT cores than GA106 Ampere GPUs. However, NVIDIA’s RT cores are superior compared to AMD, thanks to the architecture and integration within it. Ultimate performance will therefore depend on Intel’s capabilities in this regard and its hardware-level integration for ray tracing applications.

How Will Intel Arc Alchemist Graphics Cards With Xe-HPG Architecture Deliver?

Intel released an impressive demo of its XeSS technology early on. Expectations are that Intel GPUs can outperform NVIDIA’s Tensor Core implementation (DLSS) with XMX architecture. Intel is also expected to have a small but useful game cache on its GPUs. In addition, GDDR6 memory up to 16 GB will be used on the 256-bit bus interface. This means that more powerful memory will be used than NVIDIA’s RTX 3070 and RTX 3070 Ti models.

Finally, the theoretical FP32 computing performance is calculated with an expected peak clock speed of 2 GHz. We have already seen how high frequency speeds can be achieved in TSMC’s 7nm manufacturing technologies. Intel, on the other hand, will benefit from the more advanced TSMC 6nm production. Based on this information, Intel Xe-HPG Alchemist GPUs can offer around 16-17 TFLOPs of computing power. These values ​​are slightly below the computational performance offered by NVIDIA’s GA104 GPUs. However, TFLOP calculations should not be taken as an absolute measure, as gaming architecture works very differently compared to data center chips.


Without further ado, let’s make a comparison according to the available data. Intel ARC graphics cards can easily deliver higher performances than AMD’s Radeon RX 6700 XT and NVIDIA’s RTX 3070 graphics cards. Again, these comparisons are still based on probabilities.

Of course, software optimizations are also important at this point. We know that the blue team has been working on the software side for a long time. Another important point is the prices. Intel can offer competitiveconsumer-grade prices with first-generation graphics chips, and excellent alternatives may emerge.

Comparison of Intel ARC Alchemist, NVIDIA GA104 and AMD Navi 22

GPU Name Alchemist DG-512 NVIDIA GA104 AMD Navi 22
Architectural Xe-HPG Ampere RDNA 2
Production technology TSMC 6nm Samsung 8nm TSMC 7nm
Top Model
(Available GPU)
ARC Alchemist
GeForce RTX 3070 Ti Radeon RX 6700XT
Raster Engine 8 6 2
FP32 Cores 32 Xe Cores 48 SM Units 40 Compute Units
FP32 Units 4096 6144 2560
FP32 Computing Power ~16 TFLOPs? 21.7 TFLOPs 12.4 TFLOPs
TMU 256 192 160
FROCK 128 96 64
RT Cores 32 RT Units 48 RT Cores (V2) 40 RA Units
Tensor Core 512 XMX Cores 192 Tensor Cores (V3)
Tensor Computing Power ~131 TFLOPs FP16?
~262 TOPs INT8?
87 TFLOPs FP16
174 TOPs INT8
25 TFLOPs FP16
50 TOPs INT8
L2 Cache ? 4MB 3MB
Additional Cache Technology 16MB Smart Cache? 96MB Infinity Cache
Memory Bus 256-bit 256-bit 192-bit
Memory Capacity 16GB GDDR6 8GB GDDR6X 16GB GDDR6
Release date 2022 First Quarter 2021 Second Quarter 2021 First Quarter


Your email address will not be published. Required fields are marked *