Earlier this week, it was revealed that Intel sold as many discrete graphics processors as AMD in the fourth quarter of 2022. John Peddy Research, felt that something was wrong. Jon Peddie himself was kind enough to warn me that JPR’s shipment estimates are based on his Intel’s recent financial report and his ASP and may be inaccurate.Indeed they were, and Intel Didn’t Sell As Many Standalone GPUs As AMD last quarter.
6%, not 9%
In fact, Intel included revenue from the sale of Ponte Vecchio computing GPUs in the revenue of the division that sells consumer GPUs, resulting in a significant increase in revenue.
On the other hand, JPR assumed that the business unit only sold graphics processors from client PCs, so dividing revenue by average selling price (ASP) was considered accurate for Intel Arc products, The number of client GPUs sold by Intel last quarter was incorrect.
After correction, of the 13 million standalone GPUs for desktops and notebooks sold last quarter, Nvidia shipped 85%, AMD supplied 9%, and Intel sold 6%. turned out to be
In fact, controlling 6% of the discrete GPU market is a big deal for a company that has only been in the market for two years. The majority of Intel’s GPU shipments will likely be cheap notebook GPUs, and its ASP lags behind AMD and Nvidia, but will eventually catch up on this front as well.
this is what happened
Intel is relatively new to the discrete GPU market, but its product portfolio is now fairly broad. The company sells standalone Arc Alchemist GPUs for desktops and notebooks, and Flex series graphics cards tuned for remote rendering and media streaming (but on the same silicon as its Arc offerings for client PCs). ) and previously featured Data Center GPU Max compute GPUs. Codename Ponte Vecchio.
Intel typically included sales of its discrete graphics processors in its Accelerated Computing Systems and Graphics Group (AGX) earnings, but in Q4 2022 that division will split into two separate divisions, with consumer GPUs falling into the client segment. Moving to the Compute Group (CCG), Server GPUs are now: Part of the Data Center and AI (DCAI) division. An analyst at Jon Peddie Research said fourth-quarter computing GPU sales were already factored into his DCAI results, while fourth-quarter consumer GPU sales were part of his AXG results. claimed to be included.
AXG, on the other hand, increased its revenue significantly in the fourth quarter. This is due to increased sales of consumer GPUs, according to JPR. But in Q4 2022, Intel shipped the first big batch of Ponte Vecchio computing GPUs, presumably for the Argonne National Laboratory Aurora supercomputer. JPR estimates the machine will require 60,000 of Intel’s Data Center GPU Max, a good chunk of which shipped in his Q4 2022. Shipment estimates were incorrect as JPR now believes Ponte Vecchio sales are reported in his AXG results.
“Assuming Ponte Vecchio was put together in the Data Center (DCG) group and renamed DCAI, we were using ASP for dGPU from AXG to make the unit shipment,” JPR said in a statement. increase. “The hard part is the split between Xe to computer (DCAI) and his AGX and which bucket was credited for the Aurora shipment.
We believe AXG got credit for all Xe shipments as the GPU group split happened in Q4. This includes over 60,000 Ponte Vecchio dGPUs,” he explained JPR. Intel shipments seem to have increased significantly, and dGPU shipments seem to have increased even more. ”
Jon Peddie Research revamped Intel’s Q4 discrete GPU shipments by subtracting out over 60,000 expensive Ponte Vecchio computing GPUs and got a naturally different result.
“I’ve never counted GPU computing GPUs from AMD or Intel in a quarterly report, and Intel surprised me,” the company wrote. “I don’t think Intel intentionally intended to mislead the industry, they’re simply unfamiliar with the difference between shipping dGPU consumer and data center GPUs. except).”
