For the past couple of years, Intel has been uncharacteristically vocal about its plans to enter the discrete GPU market, to take on graphics stalwarts AMD and NVIDIA in both consumer PCs and the data center. Historically, entities like Intel have kept details of unreleased, forward-looking products hush-hush until they are much closer to being introduced. But Intel has engaged with the PC gaming and tech enthusiast communities to answer (and ask) questions regarding its discrete GPU plans, it has partially disclosed go-to-market strategies, discussed next-generation GPU architectures, upcoming features, and performance targets, and even released renderings of futuristic graphics card designs.
One Of Intel's Discrete GPU Concepts.
Intel has said that it plans to enter the discrete GPU space sometime in 2020. To do so, the company would need to have working silicon in-house already to begin validation testing and qualification. And at this point, it appears the company is on track to deliver. In fact, during its most recent earnings call, Intel CEO Bob Swan confirmed that the company’s next-generations discrete GPU (technically, it is not Intel’s first discrete GPU), has already arrived back from the fab and passed initial electrical validation.
During the call, Swan said, “In Q3, we also shipped our first 10-nanometer Agilex FPGAs. And in 2020, we'll continue to expand our 10-nanometer portfolio with exciting new products including an AI Inference Accelerator, 5G base station SOC, Xeon CPUs for server storage, and network and a discrete GPU. This quarter, we've achieved power-on exit for our first discrete GPU, DG1” Swan went on to say the achievement was “an important milestone.”
A few months ago, information from a leaked Intel graphics driver for future Rocket Lake-based processors pointed to an array of Gen 12 Intel GPUs. At the time, the speculation was that these GPUs were on-processor, low and high-power graphics solutions with 128 – 512 execution units. One of those GPUs carried the DG1 designation, so it’s possible that whatever discrete solution Swan referenced in the call is a derivative of one of those designs and may be an entry level part or even just a development vehicle.
Twitter coincidentally lit up with some cryptic Tweets from key members of the Intel graphics team around the time of Swan’s statements. Raja Koduri, senior vice president, chief architect, and general manager of Architecture, Graphics, and Software at Intel Corporation, tweeted “It’s been quite a Dash”. There’s been some speculation that the capital “D” in that tweet has some meaning. To date, most has assumed DG1 stood for “discrete GPU 1”, but it is possible that D references a codename – Dash perhaps? Chris Hook, Graphics and Visual Technologies Marketing Chief at Intel Corporation, tweeted “It’s alive!” which many took as a reference to DG1 powering up as well.
Ari Rauch, Vice President at Intel Corporation for Graphics technologies and Products, made multiple tweets on the subject and erased any ambiguity. One tweet said, “I’m so proud of our engineering teams for reaching the DG1 power on milestone. We still have a very long journey/odyssey ahead of us, but we’re now one important step closer.” And there were a couple of others with similar sentiments.
Intel's Ari Rauch Tweets.
Since Swan made the DG1 disclosure on the Q3 earning call, and Q3 ended September 30, we can assume DG1 completed the power-on electrical validation at some before that date, which means Intel is likely much further along in qualifying DG1 than simply powering it up. It is possible Intel already has run 3D applications on the GPU, but there has been no confirmation of that yet.
Whatever the case may be, Intel’s discrete GPU plans continue their forward momentum and an on-target, 2020 release for an Intel discrete Xe GPU seems all but certain.
