The professional software industry bears the primary responsibility for the failure to support dual GPUs in the 2013 Mac Pro.
No, Apple bears all responsibility for building a machine that suited them rather than their customers. The choice of dual GPUs was dictated by the need to spread heat output around the core of a small cylinder, not because spreading parallel compute over multiple cards is inherently better. And where parallel compute is a priority, a tower with the space, power and cooling for multiple high-end GPUs makes even more sense.
It wasn't up to the software industry to prop up Apple's hardware preferences.
The industry’s assumption that single GPUs with hundreds of watts of power were necessary led to Apple’s admission of being cornered into a thermal predicament.
Large, powerful GPUs haven't gone away, because parallel workloads haven't and there's a demand for processors that can handle them quickly.
The only one to blame was Apple, for deciding their Pro machine should be a small cylinder, then belatedly admitting it wasn't up to the job of cooling professional GPUs.
However, the demand for such large and cumbersome computers has diminished.
Well, who wants a big, extremely expensive tower with non-expandable RAM, a relatively low memory ceiling, and no ability to add the most popular type of PCIe card - a GPU? It's a bit of a circular argument to make a terrible product, then cite it as evidence that no one is interested in the category anymore.
A tower Mac is something Apple simply don't want to make, for all sorts of reasons. It's an even worse fit in their line up in the Apple Silicon era. They tried refactoring it into a tube, then an iMac; then grudgingly re-launched it as a tower but at a price that made it worth it to them. But their heart clearly isn't in it.
arstechnica.com
