Intel Meteor Lake will include a considerably improve built-in graphics resolution. At its Innovation occasion, Intel offered particulars on the upcoming Meteor Lake product household, which can arrive later this 12 months. Now we have a collection of articles masking the varied points of the brand new structure, which guarantees to be a radical departure from prior processors because of using Foveros chip stacking applied sciences. You’ll find the opposite articles within the checklist beneath; we’ll be specializing in the show and graphics applied sciences for this piece.
Meteor Lake Graphics Structure
Skipping all the architectural particulars for a second, Intel claims the Meteor Lake GPU will double the efficiency per watt of its earlier technology Iris Xe Graphics resolution, which was present in twelfth Gen Intel Core processors. Whereas GPUs usually lay declare to effectivity enhancements, given how energy constrained cellular graphics options can get, there’s an actual chance we might see uncooked efficiency double as properly. That is a probably big gen-over-gen enchancment, and there is loads of different particulars to dissect.
The brand new graphics structure is known as Xe-LPG, the low energy gaming various to Xe-HPG — that is the GPU discovered within the devoted Intel Arc Alchemist GPUs. It is probably double the efficiency of Xe LP, which was the graphics resolution used within the eleventh Gen Tiger Lake, twelfth Gen Alder Lake, and thirteenth Gen Raptor Lake cellular options.
Be aware additionally that the best configurations will not be accessible on all Meteor Lake processors, although the precise particulars of the decrease tier fashions have not been revealed. We might anticipate Intel to disable a few of the GPU processing clusters and maybe a few of the different further options on lesser elements, whereas the Core 5/7/9 chips will provide the whole function set.
The block diagram for Xe-LPG exhibits the utmost configuration, which can include eight Xe-Cores, 128 Vector Engines (previously referred to as Execution Models), two Geometry Pipelines, eight Samplers, 4 Pixel Backends, and eight Ray Tracing Models. These might be divided up between the 2 Render Slices, the primary constructing block for Intel Arc GPUs.
If the above diagram seems to be acquainted on a few of the particulars, that is as a result of it is successfully the identical because the ACM-G11 GPU that is used within the Intel Arc A380 — solely now it is an built-in graphics resolution. Almost the whole lot current within the A380 seems to be right here as properly, apart from GDDR6 reminiscence controllers and XMX models (extra on the latter beneath). As you’d anticipate from an built-in resolution, the Meteor Lake GPU will share system reminiscence with the CPU.
Intel hasn’t offered clock speeds or the L2 cache measurement but, however generally we might anticipate extra conservative clocks than what you will get from a devoted A380. That GPU runs at as much as 2.4 GHz, nevertheless it additionally consumes as much as 75W of energy. The cellular Iris Xe in the meantime clocked at as much as 1.5 GHz in chips just like the Core i9-13905H, with a most TDP for the whole bundle of simply 45W.
We might have a look at this a unique means, although, as Intel’s Xe DG1 graphics card mainly used the identical Xe-LP configuration because the built-in chips. It wasn’t a really potent resolution, and it is little shock that the Arc A380 simply outclassed the Xe DG1. We anticipate an analogous uplift in efficiency and capabilities going from built-in Xe-LP to built-in Xe-LPG.
Picture 1 of 9
Digging a bit deeper into the graphics {hardware}, every Xe-Core comes with 16 Vector Engines. These have the same old assortment of supported quantity codecs, together with FP32, FP16, INT8, and FP64 — the 64-bit assist is a brand new function, by the way. In complete, there are as much as 128 Vector Engines within the Meteor Lake GPU, in comparison with 96 (extra restricted) Execution Models within the earlier built-in GPU.
There’s one exception to Xe-LPG mainly being like an built-in Arc A380, nonetheless: There aren’t any XMX (Xe Matrix eXtensions) models. These had been most likely lower in an effort to assist cut back die measurement, and maybe energy use as properly. It is an attention-grabbing omission, as XeSS upscaling in our expertise seems to be and performs higher on Arc GPUs operating in XMX mode in comparison with XeSS operating in DP4a (8-bit INT) mode on GPU shaders.
Intel did double the shader throughput of INT8 operations in comparison with Iris Xe, nonetheless, and every Vector Engine in Meteor Lake’s GPU has a price of 64 INT8 ops per clock. Relying on clock speeds, that is nonetheless a good quantity of compute efficiency: 1.5 GHz (the identical as Iris Xe) can be 12 teraops, whereas 2 GHz (the identical as Arc A380) can be 16 teraops. We’ll need to see how this performs out in precise XeSS video games.
Ray tracing {hardware} stays, which once more represents an attention-grabbing set of priorities. The Arc A380 does not normally present an excellent expertise in video games with ray tracing, even at decrease settings, and having built-in graphics with ray tracing does not appear to be it could be that essential. However Intel notes there are non-gaming use circumstances for ray tracing as properly, for “Creation” and “Analysis.” Blender rendering on built-in graphics exhibits over 2X the efficiency of the CPU, based on Intel’s slides, however we’re undoubtedly not on the level the place ray tracing on iGPU is essential — or quick sufficient to matter for most individuals who do 3D rendering for a dwelling.
Intel additionally disclosed that the shared L1 cache stays the identical 192KB per Xe-core. What we do not know is how a lot L2 cache might be current. Arc A380 included 4MB of L2 cache. Intel might stick to that quantity in Meteor Lake, or probably even improve the cache measurement — a bigger cache means fewer reminiscence accesses, and the shared system reminiscence could already restrict efficiency a bit.
One other slide (quantity seven within the above gallery) exhibits the low-level enhancements in efficiency for numerous graphics associated computations, evaluating Meteor Lake with Raptor Lake. The Depth Check Fee obtained a significant enhance because of the double price HiZ function. Vertex and triangle processing charges are 2.6x sooner, whereas pixel mix and compute instruction charges are simply over 2x sooner. Does that equate to double the efficiency in video games? Perhaps, however most likely not in all circumstances. We’ll have to attend and see.
Picture 1 of 11
One of many attention-grabbing points of the Meteor Lake “disaggregated structure” is how the varied components that will usually be a part of the GPU get scattered across the different tiles. The graphics tile handles the whole lot mentioned above, whereas the Show Engine and Xe Media Engine are a part of the primary SOC tile, after which the bodily show outputs are a part of the IO tile.
The media capabilities of Meteor Lake are mainly the identical as Arc. It consists of encoding and decoding assist for AVC, HEVC, VP9, and AV1 — the latter two being the brand new additions relative to prior Intel GPUs. It could deal with as much as 8K60 10-bit HDR decoding, and 8K 10-bit HDR encoding, which needs to be greater than sufficient for any video streaming duties.
Placing the show outputs on a unique IO tile is sensible, as exterior interfaces do not scale properly in any respect with newer course of nodes. Intel additionally targeted on bettering energy use on two of the 4 show pipelines, which ought to assist enhance battery life in laptops. There’s additionally a brand new low energy mode for media playback.
The low energy optimizations work to keep away from waking up cores, reminiscence, and show components when it is not needed. For instance, panel self refresh (PSR) in laptops does not must fetch a brand new body if it is only a repeat of the present body. Meteor Lake may also fetch solely the up to date parts of the show buffer — for instance in Home windows, you is perhaps working in a single window whereas the whole lot else on the display remained static.
One of many extra attention-grabbing updates to the show engine is that Intel has added full assist for DisplayPort 2.1 UHBR20 (20 Gbps per lane, 80 Gbps complete). Arc GPUs solely supported UHBR10, whereas AMD’s RDNA 3 GPUs assist UHBR13.5 (aside from the skilled W7900 that has UHBR20 on a single port). As we have famous earlier than, Show Stream Compression (DSC) can already deal with 4K 240Hz, nevertheless it’s nonetheless good to have the newest show requirements accessible.
Meteor Lake Software program and Drivers
Picture 1 of 5
Wrapping issues up, Intel additionally mentioned a few of the software program stack (driver) enhancements to its graphics suite. The entire enhancements which have gone into Arc GPUs over the previous 12 months or so are additionally current with Meteor Lake graphics, together with improved DX9 optimizations in addition to DX11.
We talked about XeSS already, and this has the potential to be fairly helpful on built-in graphics. Even 1080p at medium settings tends to be a bit a lot for built-in graphics, although with probably Arc A380 ranges of efficiency, it turns into much more viable. Toss in XeSS upscaling and a few video games may even attain 60 fps or extra.
For laptops, Intel has additionally targeted on “endurance gaming.” In impact, numerous energy optimizations, focusing on decrease body charges, and tuning of CPU and GPU masses might help deliver improved battery life. For example, the traditional means of operating video games can be to max out efficiency, and in that situation a Meteor Lake processor may use 28W of energy. With Endurance Gaming, which may drop to simply 10W.
Take that with a serving to of salt, nonetheless, because the 10W determine exhibits the CPU utilizing 13%, GPU simply 3% (what sport is that this purported to be?), and Different at 18%. The slide references the appendix for particulars, which can be lacking. Anyway, most 3D video games have a tendency to want much more from the GPU, and the quantity you will profit from Endurance Gaming will differ primarily based on the video games you play.
Closing Ideas
General, there are some clear enhancements to built-in graphics coming with Meteor Lake. Intel hasn’t revealed the varied product SKUs for Meteor Lake but, although we do know it is launching on December 14. Lots of the dialogue clearly focuses on laptops, and we’re undecided what kind of graphics configurations might be current on the desktop chips — these virtually actually will not have the “built-in A380 lite” mentioned above.
We anticipate the full-fat GPU will doubtless solely seem with cellular chips. What is going to Intel use in desktop Meteor Lake? That is much less clear. It might use a lower down model of the above, or extra doubtless simply leverage the present Xe-LP / UHD 770 that is current in Alder Lake and Raptor Lake desktop chips. In spite of everything, if you’d like sooner and extra succesful graphics to your desktop PC, there are many devoted GPUs accessible — together with Arc.
We’ll discover out extra on the precise MTL SKUs within the coming months. In your reference, the total Meteor Lake Graphics and Media slide deck could be discovered beneath.
Picture 1 of 42