To those of us who were dreaming big when Intel teased a photo of a
massive discrete GPU on Twitter, it is time for a reality check—what we saw is not going into a consumer graphics card, no matter how enticing the proposition might seem (from a raw performance standpoint). Intel's graphics boss Raja Koduri, who had been retweeting the photo with
cryptic messages, finally set the record straight.
Right to the point, the large chip depicted is bound for the data center, and not in a product that will compete with NVIDIA's upcoming
GeForce RTX 3080 Ti (assuming such a model comes to fruition, with Ampere underneath the hood).
"Both XE-HPC based Ponte Vecchio (detailed at SC’19) and XE-HP based chip images shared recently are data center targeted GPUs. We need incredible computing solution range from edge to cloud to keep up with the data we are generating in real-time," Koduri said.
We always knew this was a possibility, and a strong one at that. However, given that the industry at large is moving towards an increased focused on artificial intelligence (AI) workloads—such as NVIDIA's use of Tensor cores for DLSS—we held out a sliver of hope that we might have been looking at a high-end graphics card for gaming. No dice, though.
It remains to be seen what Intel will ultimately deliver to gamers, both this year and next. The only real glimpse of performance we have is from the
Xe DG1 Software Development Vehicle (pictured up top) that Intel sampled to ISVs earlier this year. Performance looked underwhelming from a live demo, but according to Intel, those cards were not indicative of what will end up in the hands of consumers.