This week is pretty busy for the GPU (graphics processing unit) market as Advanced Micro Devices (AMD) and NVIDIA (NVDA) fight for the spotlight. There is a third entry in the discrete GPU market, and it’s also taking its share of the limelight. In the last few days, Intel (INTC) has given hints about its most-discussed Xe GPU with ray tracing.
Ray tracing GPU: The new battleground for NVIDIA, AMD, and Intel
On Monday, AMD launched its lower-end, Navi-based, 7nm (nanometer) RX 5500 GPU. Moreover, rumors stated that its next-generation GPU would feature ray tracing and could launch next year. Last Friday, rumors around NVIDIA’s next-generation 7nm Ampere GPU architecture reignited. NVIDIA was the first to introduce ray tracing in its Turing GPUs in August 2018.
Intel’s chief architect, Raja Koduri, leveraged this enthusiasm around GPUs and dropped hints about the company’s upcoming GPU with ray tracing. Last Friday, he tweeted a picture of the license plate on his Tesla Model X and tagged @IntelGraphics. His license plate read “THINKXE,” which means “Think Exponential.” This gives us a hint that Intel’s new Xe GPU architecture could be exponential, scaling from low-end embedded GPUs all the way to high-performance server GPUs.
Intel’s Xe GPU launch date ambiguous
In the tweeted photo of Koduri’s vanity license plate, you can also see stickers reading “June” and “2020” on the plate. Last week, tech blog Tom’s Hardware interpreted these license plate tags as hinting at Intel’s potential June 2020 launch of its Xe GPU. The article read, “The June 2020 date seems too specific to be a coincidence. Intel might have an announcement, and perhaps even a release, planned for Computex in June 2020.”
However, California license plates display registration tags that identify the expiration month and date—in this case, June 2020. This might have been a sheer coincidence, or Tom’s Hardware might have over-interpreted the license plate.
Tom’s Hardware also referred to past tweets from Intel executives that stated that the first Xe GPU would arrive by the end of 2020. Preponing a product like a discrete GPU, which Intel built from scratch, seems unlikely. It’s possible that Intel might offer the first glimpse of its Xe GPU at Computex 2020 and launch it by the end of 2020.
On October 8, at the Intel Developer Conference (or IDC) in Tokyo, the chip giant hinted that it aims to achieve 1080p 60fps (frames per second) performance with Xe. These hints seem to have worked their charm, as Intel stock rose 1.5% on October 9.
What do we know about Intel Xe?
At the IDC Tokyo, Intel released new benchmarks for its Gen 11 Iris Plus GPUs inside its 10th-generation Ice Lake processors for laptops. According to the benchmarks, the Iris Plus delivered 30fps performance at 1080p in modern esports titles.
The chip giant stated that its Gen 12 Xe GPU would deliver double the performance than its Gen 11 GPU. This means that Intel’s Xe aims to hit the desired gaming performance of 60fps at 1080p. This performance level is delivered by NVIDIA and AMD’s discrete GPUs.
It’s too early to draw any conclusion about Xe’s performance, and reality could differ from expectations. New technology comes with complications. NVIDIA’s industry-first ray-tracing Turing GPU and AMD’s industry-first 7nm Ryzen CPUs (central processing units) faced technical issues with the initial batch. There is a possibility that Intel’s initial Xe GPUs could also face some technical issues.
First Xe GPU’s variant
Koduri’s tweet doesn’t indicate whether the first Xe GPU would be a desktop variant or a laptop variant. Wccftech, citing Intel’s director of technology headquarters, Kenichiro Yasu, stated that the new Xe GPU would arrive in Intel’s 10nm Tiger Lake mobile CPUs. These CPUs are scheduled to launch next year. Tom’s Hardware article refers to another tweet from Twitter user @KOMACHI_ENSAKA, which states Intel’s Xe might appear in its 10nm Lakefield-R processors.
First Xe GPU’s process node
Koduri’s tweet also does not specify the process node of the first Xe GPU. At the 2019 Investor Meeting in May, Intel stated that it expects its lead 7nm product to be a Xe-based general-purpose GPU for the data center. This GPU is expected to hit the market in 2021.
This means the first Xe GPU would either be built on the 10nm node or the 14nm node. According to TechQuila, Intel’s first Xe GPU, which is codenamed Arctic Sound, would be built on a refined 14nm process.
What could be Intel’s strategy behind a possible June 2020 launch?
The success of a product depends on three things—its performance, its time of launch, and its price compared to its rivals. Although Intel Xe GPU is behind AMD and NVIDIA in the process node, its launch time is in perfect sync with its rivals.
It would be logical to unveil Xe in June 2020 and launch it by the end of the year. NVIDIA and AMD are touted to release their next-generation 7nm GPUs during the same time period. AMD is expected to launch its custom ray-tracing supported GPU for Sony and Microsoft’s next-generation game consoles in mid-2020.
NVIDIA is expected to launch its ray-tracing supported high-end AIB (add-in-board) GPUs in the first half of 2020. So, Intel’s plans to launch its embedded Xe GPUs in 2020 make business sense.
Intel, AMD, and NVIDIA plan to provide details about their next-generation GPUs at CES (Consumer Electronics Show) in January 2020. A June 2020 launch could mean that Intel might use the Computex event. AMD used this event to unveil its Ryzen 3000 series CPUs.
What are Intel’s plans for the first Xe GPU?
Just after Intel teased the June 2020 launch of Xe GPUs, it confirmed the end-of-life of its Kaby-Lake G CPUs. These CPUs were formed through a collaboration between Intel and AMD. AMD provided the discrete Radeon RX Vega M GPU for Intel’s Kaby Lake mobile CPUs. The final Kaby-Lake G batch is slated to be shipped on July 31, 2020, and should give way to Intel’s embedded Xe GPUs.
Intel’s market leadership in the CPU space should give its embedded Xe GPU an advantage to gain GPU market share. At present, Intel is a leader in the overall graphics market thanks to its integrated GPU found in its CPUs. According to Jon Peddie Research, Intel has more than 65% share in the GPU market.