gigapixel ai demo
Enlarge / That is Intel’s DG1 chipset, the guts of the Xe Max GPU.

This weekend, Intel launched preliminary info on its latest laptop computer half—the Xe Max discrete GPU, which features alongside and in tandem with Tiger Lake’s built-in Iris Xe GPU.

We first heard about Xe Max at Acer’s Subsequent 2020 launch occasion, the place it was listed as part of the upcoming Swift 3x laptop computer—which is able to solely be obtainable in China. The brand new GPU can even be obtainable within the Asus VivoBook Flip TP470 and the Dell Inspiron 15 7000 2-in-1.

Intel Xe Max vs. Nvidia MX350

Throughout an prolonged product briefing, Intel harassed to us that the Xe Max beats Nvidia’s entry-level MX 350 chipset in nearly each conceivable metric. In one other 12 months, this is able to have been thrilling—however the Xe Max is simply slated to look in methods that characteristic Tiger Lake processors, whose Iris Xe built-in GPUs already handily outperform the Nvidia MX 350 in each Intel’s exams and our personal.

The confusion right here largely springs from mainstream client expectations of a GPU versus what Intel’s doing with the Xe Max. Our GPU exams largely revolve round gaming, utilizing 3DMark’s well-known benchmark suite, which incorporates gaming, fps-focused exams corresponding to Time Spy and Evening Raid. Intel’s expectations for the Xe Max as an alternative revolve, virtually totally, round content material creation with a facet of machine studying and video encoding.

Xe Max is, roughly talking, the identical 96 Execution Unit (EU) GPU to be discovered within the Tiger Lake i7-1185G7 CPU we have examined already this 12 months—the foremost distinction, past not being on-die with the CPU, is the next clock charge, devoted RAM, and separate TDP funds.

Tiger Lake’s Iris Xe has a peak clock charge of 1.35GHz, and it shares the CPU’s TDP constraints. Iris Xe Max has its personal 25W TDP and the next peak clock charge of 1.65GHz. It additionally has its personal 4GiB of devoted RAM—although that RAM is identical LPDDR4X-4266 that Tiger Lake itself makes use of, which is one thing of a primary for discrete graphics and would possibly result in higher energy effectivity.

An enhancement to Iris Xe, not a substitute

Intel’s advertising and marketing materials promotes the thought of workloads that run on each the built-in Iris Xe GPU and discrete Xe Max GPU concurrently, utilizing the time period “additive AI.” Nevertheless, this should not be confused with conventional GPU interleaving, corresponding to AMD Crossfire or Nvidia SLI—Xe Max and Iris Xe will not be teaming as much as deal with show duties on a single show.

Intel’s model of “additive AI” refers as an alternative to workloads that may be simply divvied as much as run on separate GPUs, with separate reminiscence areas. Such duties will be assigned to each iGPU and dGPU concurrently, and—for the reason that two have very related efficiency profiles—they will achieve this with out a lot complication in doubtlessly slowing the general workload down because of important parts working on the marginally slower built-in GPU.

Sharpening pictures with Gigapixel AI

Considered one of Intel’s extra spectacular demonstrations concerned utilizing Topaz Gigapixel AI to sharpen grainy pictures. In the event you’re not acquainted with Gigapixel AI, it is basically the modern-world, real-life model of Blade Runner‘s notorious improve scene. In fact, Topaz cannot add real info to a photograph that wasn’t already there—however it might probably, and does, use machine studying to provide info that appears prefer it ought to belong.

In Intel’s demonstration, an Xe Max-equipped laptop computer used Gigapixel AI to reinforce a really massive, grainy photograph seven instances quicker than an analogous laptop computer outfitted with an Nvidia MX 350 GPU might. Whereas that was spectacular, we pressed Intel for comparisons to different {hardware}, which an Xe Max-equipped laptop computer would possibly extra moderately compete with “within the wild.”

After a day or so, an Intel engineer received again to us and mentioned we might anticipate an Xe Max-equipped laptop computer to finish Gigapixel AI workloads seven instances as quick as an MX 350, 5 instances as quick as an RTX 1650, and 1.2 instances as quick as a Tiger Lake laptop computer with Iris Xe graphics alone.

The engineer additionally famous that Xe Max has significantly untapped potential to be unlocked in additional optimizations—and that in MLPerf ResNet-50 workloads, utilizing INT8 information, Intel is already seeing 1.4x enhancements over standalone Tiger Lake methods with out Xe Max.

Xe Max vs. Iris Xe

Despite being, effectively, a higher-clocked version of Tiger Lake's integrated Iris Xe GPU, Xe Max doesn't always outperform it—likely due to being a PCIe x4 lane away from the CPU, rather than on-die.
Enlarge / Regardless of being, successfully, a higher-clocked model of Tiger Lake’s built-in Iris Xe GPU, Xe Max does not all the time outperform it—probably because of being a PCIe x4 lane away from the CPU, relatively than on-die.


The true query—for now, no less than—is who will profit sufficient from an Xe Max-equipped laptop computer to justify the extra value. We do not suppose both players or general-purpose computing customers ought to get too labored up about it but—neither workload is more likely to see a lot profit.

Xe Max does not all the time speed up gaming workloads in any respect—because the slide above demonstrates, working Metro Exodus on an Xe Max boosts efficiency roughly 7fps. However working DOTA 2 on it will lower efficiency by roughly the identical quantity. Fortunately, Intel’s implementation of the Xe Max consists of mechanically shifting workloads appropriately. An Xe Max-equipped Tiger Lake system ought to shift every sport to the better-suited GPU, with no need the gamer to know which is which.

For machine studying, the story is extra optimistic—Xe Max does, no less than, appear to persistently outperform the built-in Tiger Lake GPU. However Tiger Lake’s integrated-graphics success story takes a lot of the shine off of Xe Max. A 20- to 40-percent efficiency enhance for a number of hundred {dollars} additional is not a horrible proposition—however it’s a significantly smaller enhance than most individuals anticipate from including a discrete GPU.

For now, customers who encode plenty of video all through their day are probably the very best goal for Xe Max. For this one very particular activity, an Xe Max-equipped laptop computer runs rings round a Tiger Lake laptop computer with out it—or perhaps a laptop computer with an in any other case vastly extra highly effective RTX 2080 Tremendous.


Please enter your comment!
Please enter your name here