Intel EUs are gone! The brand new constructing block of its ‘gaming first’ graphics card is the Xe-core

26
Intel Xe-HPG GPU details

(Picture credit score: Intel)

In a bid to rebrand each little bit of its graphics card structure, Xe, Intel has killed off the common-or-garden Execution Unit. The EU is the basic constructing block of Intel’s graphics architectures, however after years of use, Intel is placing the time period out to pasture to make method for a brand new unit, the Xe-core. 

The change in title comes as Intel says there’s an excessive amount of inside a single EU to actually classify it as an EU these days.

This does not completely change what an EU/Xe-core really does, although. An Intel EU presently incorporates a number of ALUs for floating level and integer operations, and the identical goes for an Xe-core. Nevertheless, the brand new Xe-core nomenclature does play into the numerous shake up incoming with Intel’s first discrete gaming graphics card technology, codenamed Intel Alchemist

The brand new Xe-core inside Alchemist GPUs comes with 16 vector engines and 16 matrix engines. That is really double what’s discovered inside the Ponte Vecchio GPU, which is constructed with the Xe-HPC structure and destined for the Aurora supercomputer.

Intel says that this can be a essential step to scaling up gaming graphics playing cards, and that this chip was specifically optimised for “gaming first”.

Zooming out a contact from the Xe-core and you will find a Render Slice, which incorporates 4 of those Xe-cores clusters, a hard and fast perform unit for DirectX 12 Final help, and 4 new ray tracing acceleration models. Then a set of eight Render Slices will share entry to L2 cache.

These new Intel Ray Tracing Models are designed to speed up ray traversal, bounding field intersection, and triangle intersection. This sounds quite a bit just like the RT Cores inside Nvidia’s Ampere GPUs, and the Ray Accelerators AMD has inside its RDNA 2 structure. Although we do not know the way they get up towards one another in uncooked efficiency phrases.

Picture 1 of 4

Intel Xe-HPG GPU details

(Picture credit score: Intel)

Picture 2 of 4

Intel Xe-HPG GPU details

(Picture credit score: Intel)

Picture 3 of 4

Intel Xe-HPG GPU details

(Picture credit score: Intel)

Picture 4 of 4

Intel Xe-HPG GPU details

(Picture credit score: Intel)

Once more, that general make-up is completely different to Ponte Vecchio on the Xe-HPC structure, which has fewer engines per Xe-core however a larger variety of Xe-cores per Render Slice. There’s 16 Xe-cores in each HPC slice with simply as many ray tracing models. A set of 4 Render Slices then share entry to L2 cache, HBM2e controllers, and a media engine. So on and so fourth. 

There are a lot of ranges to Ponte Vecchio that I dare not cowl right here.

Alchemist might be constructed on TSMC’s 6N course of, nonetheless, whereas Ponte Vecchio will use a mixture of TSMC’s N5, N7, and Intel 7 course of nodes—relying on which tile you have a look at on the Ponte Vecchio bundle. Basically, these are altogether very completely different chips.

What we’re seeing, although, is proof of Intel’s promised tweaks to the Intel Xe structure, depending on how they’re supposed for use. 

The gaming structure is considerably completely different to the datacentre one, and each are estranged from the lower-spec Xe-LP chips in at present’s laptops. This kind of segmentation is not a completely new idea, each Nvidia and AMD’s knowledge centre playing cards look fairly completely different to their gaming ones, nevertheless it’s nonetheless thrilling to see how one other firm’s engineers are tackling GPU design and utilization in a brand new trend.

Intel additionally confirms that past even the configuration of an Xe-core or Render Slice, it has labored on optimisations and methodology to extend the facility effectivity of its upcoming graphics playing cards—these embrace tweaks to the structure itself, reminiscence, and the cardboard’s bodily design.

Intel Xe-HPG GPU details

(Picture credit score: Intel)

Contemplating all of the optimisations Intel has made and the disparities in course of nodes, the Xe-HPG structure reportedly presents roughly 1.5x the efficiency/watt and runs at roughly 1.5x the frequency of Intel’s Xe-LP chips.

Whereas it is felt we have had a reasonably good grip on what was anticipated out of DG2 even previous to Intel’s Structure Day, with reference to core counts and the specifics of the Xe-HPG structure, maybe we did not have fairly nearly as good a grip on it as we suspected. In spite of everything, what we all know now, we had been evaluating apples to oranges between Intel Xe microarchitectures.

Intel’s Alchemist graphics playing cards are coming early subsequent yr, and someday earlier than then we’re prone to hear extra concerning the specs for the precise playing cards themselves and the way they carry out versus the competitors. That is after we’ll discover out if all this tinkering with the Xe structure actually interprets into body charges.

Jacob Ridley

Jacob earned his first byline writing for his personal tech weblog from his hometown in Wales in 2017. From there, he graduated to professionally breaking issues at PCGamesN, the place he would later win command of the equipment cabinet as {hardware} editor. These days, as senior {hardware} editor at PC Gamer, he spends his days reporting on the newest developments within the know-how and gaming trade. When he isn’t writing about GPUs and CPUs, you will discover him making an attempt to get as distant from the fashionable world as doable by wild tenting.