Raja Koduri, Intel’s Senior VP and Common Supervisor of Intel Accelerated Computing Techniques and Graphics Group (AXG), revealed on Twitter that they do not plan on commercializing the Xe-HP GPU lineup.
We deployed Xe HP in our oneAPI devcloud and leveraged it as a SW growth car for oneAPI and Aurora. We at the moment don’t intend to productize Xe HP commercially, it advanced into HPG and HPC which are on normal market manufacturing path
— Raja Koduri (@Rajaontheedge) October 28, 2022
Koduri defined the plan for Intel is to cease the corporate’s deal with their Xe-HP sequence of server GPUs to be launched to the business market. The reasoning for this seems that Intel has already advanced the unique plans for his or her Xe-HP sequence into the HPC sequence, also referred to as Ponte Vecchio, and the Xe-HPG sequence, also referred to as Intel Arc. Primarily, the plans wouldn’t make sense to have an extra group of GPUs primarily based on the unique Xe-HP structure.
Intel flagship server GPU line, codename Arctic Sound, had been very nicely referred to as being “beneath growth” with the resurgence of the Intel HPC part of the corporate. Koduri even showcased the chips that housed the silicon at Intel Labs as the primary high-performing Xe silicon chipset being developed. Semiconductor big TSMC builds bot Xe-HPC and -HPG dies for Intel.
With the dearth of protection about Intel’s Xe-HP sequence, it was anticipated that there was a change of focus from throughout the firm. It was nonetheless surprising resulting from Intel demoing Xe-HP chips delivering as much as 42 TFLOPS of FP32 throughput, in addition to transport sampling chips to a small group of shoppers.
Throughout this yr, Intel shifted focus to administering and producing Ponte Vecchio for his or her Aurora Supercomputer and their consumer base. The corporate additionally labored towards releasing the Xe-HPG Alchemist GPU household for the primary quarter of subsequent yr. Koduri states that the Xe-HP line has been totally leveraged and utilized as a “growth car for Aurora and Intel’s oneAPI system.
The most important query for now’s will the corporate be capable of retain the competitiveness between themselves and rivals NVIDIA and AMD resulting from this refocusing and basically pushing again their GPU strains, even when it’s only slight. Is it doable that Xe-HP alone was unable to carry out to the identical stage as their rivals? And, in response to the dearth of efficiency, take a look at their projections over the subsequent a number of years to completely combine the Intel Xe-HPC/HPG strains? Proper now, Intel has remained pretty quiet as to their reasoning for the adjustments, however do guarantee their shoppers that they’re trying into the close to future to aim to dominate the GPU market.
Intel Xe HP GPUs
|GPU Identify||Intel Neo 1-Tile||Intel Neo 2-Tile||Intel Neo 4-Tile|
|GPU Household||Intel Xe-HP||Intel Xe-HP||Intel Xe-HP|
|GPU Course of||10nm SuperFin||10nm SuperFin||10nm SuperFin|
|Max Compute Models||512 EUs||1024 EUs||2048 EUs|
|Max Core Rely||4096 Cores||8192 Cores||16,384 Cores|
|Core Clock||1300 MHz (Demo)||1300 MHz (Demo)||1300 MHz (Demo)|
|FP32 Compute||10.5 TFLOPs (Demo)||21.1 TFLOPs (Demo)||42.0 TFLOPs (Demo)|