A pair of SiSoftware benchmark listings have unveiled two new Intel GPUs sporting 10 and 12 Xe cores, with 12GB of VRAM on each results. It’s doable these two GPUs are future Battlemage graphics playing cards, although the comparatively low clock speeds and restricted core counts leave room for debate. They may be different Alchemist configurations, faked listings, and even potential built-in graphics options. But let’s speak about what we do know proper now.Battlemage is the upcoming successor to Arc Alchemist that shall be focused on the devoted graphics card market. It can even be utilized as a future built-in graphics resolution for Intel’s future CPU architectures (together with Lunar Lake). Intel is rumored to be focusing on a Q3 2024 launch date, which might line up properly with this being a possible Battlemage leak.(*12*) to the SiSoftware listings, the specs for each GPUs are seemingly unimpressive. The least-potent providing sports activities simply 1440 shader ALUs, whereas the stronger SKU has 1728 shader ALUs. To put that in perspective, the Arc A580 has 3456 shaders, whereas the Arc A380 slashes that determine right down to 1024 shaders. But there are another curiosities to debate. Depending on what number of shaders are in every Xe-core, we find yourself with both 10 and 12 Xe-cores, or probably 20 and 24 Xe-cores. Which reply is appropriate issues quite a bit, as we’ll focus on under.Clock speeds are additionally unimpressively low at simply 1.8GHz, which admittedly might be as a result of early silicon and driver checks. The solely considerably larger spec is that each have 12GB of VRAM, which is best than the 8GB discovered on the A580 and A750 midrange GPUs. However, these playing cards use 1GB chips with a 256-bit interface, so it is nonetheless doable that these are Alchemist, simply configured with 2GB chips and a 192-bit interface.And then there’s the 8MB of L2 cache. That’s a little bit of a bizarre quantity for a 12GB card, as Intel’s Alchemist GPUs have had both 4MB or 2MB per 64-bit controller. The A750 and A770 each have 16MB of L2 cache whereas the A580 has 8MB, but it is not clear how chips with presumably a 192-bit interface may arrive at 8MB complete cache — 2.67MB per 64-bit interface simply would not appear possible, and such a small quantity of L2 additionally would not make a lot sense for the longer term Battlemage structure. Of course, the L2 cache might be decoupled from the reminiscence interfaces, and Intel might need L3 cache that is not exhibiting up in these “leaked” results, but we do not know for sure.Image 1 of two(Image credit score: SiSoftware)(Image credit score: SiSoftware)Compared to Intel’s current Arc Alchemist GPUs, these two GPUs provide reasonably low shader core counts — greater than an Arc A380 but far lower than an Arc A580. Note additionally that we won’t rule out the potential for these being early cellular chips for testing and debugging functions. The system used exhibits an Intel Z790 motherboard, but it is doable to check cellular GPUs with a desktop if these leaks got here from a {hardware} lab.In phrases of positioning, these look like finances to midrange graphics options, given the presumed Xe-core counts. We’re additionally accepting SiSoftware’s reported information at face worth, which lists a 160 CU and 192 CU “capability” with 1440 and 1728 “shaders.” Intel would not use CU or Compute Units for its GPUs, as that is AMD, whereas Nvidia makes use of “Streaming Multiprocessors,” but SiSoftware makes use of the identical CU label for all three. These Intel GPUs appear to be utilizing “CU multiplied by 9” for the shader figures, with the bottom CU quantity being 16 occasions the Xe-core rely — or if you wish to be old fashioned, it is the Execution Unit counts from older Intel GPUs.Clear as mud? Let’s give a extra concrete instance with a identified GPU configuration. This Arc A380 itemizing says 128 CU and 1152S. The precise Arc A380 has eight Xe-cores with 1024 shader ALUs unfold amongst sixteen Vector Engines (previously known as Execution Units) per Xe-core. There’s additionally one extra “particular goal” (prolonged math) ALU per Vector Engine. So, 8 * 16 * 9 = 1152 shader ALUs. But the identical calculations from SiSoftware do not appear to use to the bigger Arc A-series GPUs for no matter motive (they use Xe-cores * 16 * 8, with out factoring within the prolonged math ALU). Here’s how issues break down:Swipe to scroll horizontallyFor now, the efficiency information is not terribly useful given we’re not sure on a number of the specs and how excessive Battlemage would possibly go. The core configuration and clock speeds recommend efficiency shall be higher than an A380 — and it’s. It additionally appears as if Intel is getting much more efficiency out of fewer shaders with these unknown GPUs.By approach of comparability, the Arc A580 OpenCL results of 6,121Mpix/s is about the identical because the unknown 10 (or 20?) Xe-core chip’s 6,031Mpix/s, but the A580 has 24 Xe-cores. And this is an Arc A750 OpenCL getting 7,145Mpix/s whereas this Arc A770 OpenCL result’s 8,229Mpix/s — with the supposed 12 (or 24?) Xe-core GPU getting 7,231Mpix/s. Assuming the mathematics is appropriate on shader counts, the unknown chips are simply beating equally configured Alchemist GPUs.These “new” GPUs land on the stage of the A580 and A750, but with far fewer shaders, plus a special reminiscence configuration. If these are Battlemage and can clock 40% larger than the 1.8GHz proven right here, and then if we get chips with as much as 32 Xe-cores and 4096 shaders, we’d find yourself with a good stage of efficiency. Three occasions as many Xe-cores because the presumed 10-core GPU plus larger clocks would probably put such an Intel chip into RTX 4070 Ti territory. But let’s not get too carried away, because the cache and different specs nonetheless look odd, and the Xe-core numbers would possibly really be double what we used.We’ve realized from each AMD and Nvidia {that a} greater cache on a GPU can vastly enhance efficiency. Intel has realized this as properly, and whereas it was most likely too late so as to add an enormous cache to Alchemist, we positively count on Battlemage to return with greater caches. 8MB would not lower it, although it might merely be that Intel has L3 cache and SiSoftware is not querying or reporting that worth.Ultimately, it is too early to inform how briskly Battlemage shall be, but these supposed spec leaks take a look at least considerably promising. If a 12 Xe-core chip at 1.8 GHz can already match the A750, and if Intel goes as excessive as 32 Xe-cores with Battlemage, we would really be fairly aggressive Intel options that might possibly tackle the high-end AMD and Nvidia options. Maybe.Alternatively, that is a 24 Xe-core Battlemage chip working at 1.8 GHz and matching a 2.4 GHz 28 Xe-core Alchemist chip. We already know the place the A750 lands in our GPU benchmarks hierarchy, but whereas boosting clocks and efficiency by 50% will surely assist, that will nonetheless solely end in one thing on the extent of the RTX 3070, and slower than an RTX 4060 Ti. Let’s hope this is not the case.Intel is already two years behind the competitors, and with Intel confirming that Battlemage {hardware} improvement is already finished and 70% of the staff has moved on to (*10*) — its GPU structure after Battlemage — we’re hoping these leaked results are for budget-oriented future GPUs and that considerably bigger and quicker variants are on the way in which. In the interim, the remaining 30% of Intel’s graphics staff is engaged on software program and driver assist for Battlemage, in preparation for its launch later this 12 months.Join the consultants who learn Tom’s Hardware for the within monitor on fanatic PC tech information — and have for over 25 years. We’ll ship breaking information and in-depth critiques of CPUs, GPUs, AI, maker {hardware} and extra straight to your inbox.
https://www.tomshardware.com/pc-components/gpus/mysterious-sisoftware-results-show-two-new-intel-arc-gpus-sporting-10-and-12-xe-cores-possibly-battlemage-but-clocks-and-l2-cache-leave-uncertainties
