Intel Graphics, in its latest teaser video to the Arc A770 Limited Edition "Alchemist" graphics card, posted detailed renders of the card disassembled. The card features a strictly dual-slot cooling solution that uses an aluminium base-plate and a copper vapor-chamber to pull heat from the various hot components of the PCB. This is conveyed by four flat copper heat pipes through an aluminium fin-stack heatsink, which is ventilated by a pair of 80 mm fans. The cooler and its backplate feature four independent RGB lighting zones—the bores of each of the two fans, a light strip running along the top of the card; and toward the tail-end of the backplate, with a total of 90 LEDs. Intel claims that the maximum noise output of the cooler is 39 dBA.

The PCB is shorter in length than the cooler itself, and is full-height (and no taller). It draws power from a combination of 8-pin and 6-pin PCIe power connectors, which combined with slot-power add up to 300 W. A 6-phase VRM powers the "ACM-G10" GPU, while there are three other VRM phases, which could power the eight GDDR6 memory chips, and other power domains of the card. Display outputs include three standard-size DisplayPort 2.0, and one HDMI 2.1. The card's host interface is PCI-Express 4.0 x16, and although not a system requirement, Intel insists that the card be used on a machine with PCI resizable-BAR enabled.

Intel XeSS is a second-generation upscaling technology that's at par with DLSS 2.0 and FSR 2.0, says Digital Foundry, which got an exclusive early-access an Arc A770 graphics card along with XeSS, and the latest version of "Shadow of the Tomb Raider," which features XeSS besides DLSS. The publication compared the A770 + XeSS with a GeForce RTX 3070 + DLSS, and in its testing, found XeSS to offer comparable or better image quality at low frame-times. XeSS uses a AI-ML algorithm to reconstruct details, which is accelerated by the XMX cores on Xe-HPG "Alchemist" GPUs. XeSS hence currently only supports Arc GPUs, however, the company is working on a DP4a (programming model) version so XeSS could work on other GPU architectures, across brands.

An upscaling algorithm by design adds to frame-times (time taken to render a frame to display), and in Digital Foundry's testing, the most aggressive preset of XeSS, Performance, which upscales 720p to 1440p, adds 2 ms to the frame-time. A 1080p to 4K upscaling in the same mode, adds 3.4 ms to the frame-time, which jumps from 8.8 ms to 12.2 ms (an increase of 3.4 ms). The increase in frame-times is a good trade-off when you consider the performance gained—a staggering 88 percent increase in frame-rates for 1080p to 4K upscaling, and 52 percent increase with 720p to 1440p upscaling. Frame-times increase as you move up the presets toward the Quality mode, which renders the game at resolutions closer to native-resolution, so the performance-gained is smaller. XeSS offers a preset it calls "Ultra Quality," which renders the game at a resolution closest to native, while still yielding a 16-23 percent frame-rate gain, with an output that's practically indistinguishable from native-resolution.

NVIDIA is readying two new performance-segment graphics card SKUs to help it clear out inventory in the market-segment, as well as better position itself against the likes of the Intel Arc A770 and AMD Radeon RX 6650 XT. These include the GeForce RTX 3060 8 GB, and the GeForce RTX 3060 Ti G6X. The RTX 3060 originally launched with 12 GB of GDDR6 memory across a 192-bit wide memory interface. The new RTX 3060 8 GB SKU is positioned below it, with a third of its memory sub-system pulled out—8 GB of GDDR6 across a 128-bit wide memory bus. This memory ticks at 15 Gbps (240 GB/s memory bandwidth). At this point we don't know if the CUDA core count is changed from the original RTX 3060 (3,584 CUDA cores).

The second SKU is the GeForce RTX 3060 Ti G6X. The original RTX 3060 Ti had launched with 8 GB of GDDR6 memory running at 14 Gbps (448 GB/s bandwidth). The new SKU has the same amount of memory at 8 GB, but with much faster GDDR6X memory that probably ticks at 19 Gbps (608 GB/s bandwidth). Again, we don't know if the CUDA core count has changed from the original's 4,864 CUDA cores. According to the source of this story, MEGAsizeGPU, NVIDIA could launch these two SKUs in October, which would put it just in time for shopping seasons like Cyber Monday. NVIDIA's next-gen RTX 40-series launch will take a "top-down" sequence, with the high-end SKUs launching first. It could either be late-2022 or early-2023 for NVIDIA to launch performance-segment SKUs, giving these two SKUs some time in market.

Intel on Thursday confirmed that there will be only four Arc "Alchemist" desktop graphics card SKUs in the retail channel, and that it will be led by the A770 Limited Edition, which maxes out the DG2-512 silicon, and features 17.5 Gbps memory across its 256-bit wide memory bus, putting 560 GB/s of memory bandwidth at its disposal. The A750 uses 16 Gbps memory data-rates, and has 512 GB/s of bandwidth. It turns out, that the mid-range A580 features a 256-bit wide memory bus, and not the previously-reported 192-bit, which means it has the same 512 GB/s bandwidth as the A750. The A580 and A750 come with 8 GB of memory, while the A770 tops out with 16 GB.

The PCI resizable-BAR feature is an absolute must for Intel Arc "Alchemist" graphics cards to perform as advertised, to the extent that Intel recommends sticking to "other vendors" (NVIDIA, AMD), for those on machines that lack resizable-BAR or prefer it disabled. In our testing of the Arc A380, we found a big performance gap between resizable-BAR being enabled and disabled. The company said that while it is working on driver optimizations that improve performance for machines lacking resizable-BAR, its general advice for those on older platforms is to stick with other vendors.

Resizable-BAR enables the software to see the entire video memory of a graphics card as a single large addressable block, rather than through 256 MB apertures. AMD and NVIDIA's driver architectures have optimized their memory-management to cope with these apertures through the advent of PCI-Express, but Intel Arc hasn't. Its memory-management model relies on large bursts of memory transfers, for which it needs resizable-BAR. The performance penalty for lacking it could be as high as 40 percent. In related news, Intel Graphics confirmed that the Arc A770 will be available both in 16 GB and 8 GB memory variants at launch—which is still slated for Soonuary, 2022.

Intel on Thursday posted an article that dives deep into the ray tracing architecture of its Arc "Alchemist" GPUs, which are particularly relevant with performance-segment parts such as the Arc A770, which competes with the NVIDIA GeForce RTX 3060. In the article, Intel posted ray tracing performance numbers that put it at-par with, or faster than the RTX 3060, which which it has traditional raster performance parity. In theory, this would make Intel's ray tracing tech superior to that of AMD RDNA2, because while the AMD chips have raster performance parity, their ray tracing performance do not tend to be at par with NVIDIA parts at a price-segment level.

The Arc "Alchemist" GPUs meet the DirectX 12 Ultimate feature-set, and its ray tracing engine supports DXR 1.0, DXR 1.1, and Vulkan RT APIs. The Xe Core is the indivisible subunit of the GPU, and packs its main number-crunching machinery. Each Xe Core features a Thread Sorting Unit (TSU), and a Ray Tracing Unit (RTU). The TSU is responsible for scheduling work among the Xe Core and RTU, and is the core of Intel's "secret sauce." Each RTU has two ray traversal pipelines (fixed function hardware tasked with calculating ray intersections with intersections/BVH. The RTU can calculate 12 box intersections per cycle, 1 triangle intersection per cycle, and features a dedicated cache for BVH data.

Intel Graphics in an interview with PC Gamer on the sidelines of the 2022 IFA Berlin, claimed that the real-time ray tracing architecture of the Xe-HPG graphics architecture in the Arc A770 "Alchemist" graphics card is "competitive or better than" the NVIDIA GeForce RTX 3060, and that the company plans to launch the card at an attractive price-point, to grab a slice of the very top of the gaming graphics market bell-curve. The RTX 3060 is a very successful GPU, especially with graphics card prices on the chill, and AMD is already competing with its Radeon RX 6650 XT, which can be spotted at lower prices. The RTX 3060 has been in the crosshairs of Intel Graphics marketing, in recent performance reveals for the A770.

"When you have a title that is optimized for Intel, in the sense that it runs well on DX12, you're gonna get performance that's significantly above an [RTX] 3060," said Tom Petersen with Intel Graphics. "When you have a title that is optimized for Intel, in the sense that it runs well on DX12, you're gonna get performance that's significantly above an [RTX] 3060. And this is A750 compared to a 3060, so 17%, 14%, 10%. It's going to vary of course based on the title," he said. "We're going to be a little bit faster but depending on your game and depending on your settings, it's trading blows, and that's the A750. Obviously, A770 is going to be a little bit faster. So when you add in DX11, you're gonna see our performance is a little less trading blows, and we're kind of behind in some cases, ahead in some cases, but more losses than wins at DX11," he added. While Intel is still non-committal about a launch date, although it stated that the Arc A770 and A750 will launch with an attractive "introductory pricing."

Intel Arc Alchemist graphics cards span both gamer and creator/professional user market sector, where we witnessed Intel announce gamer and pro-vis GPU SKUs. Today, we are seeing the usage of the flagship Arc Alchemist SKU called A770 in Blender rendering with ray tracing enabled. The GPU is designed to have a DG2-512 GPU with 512 EUs, 4096 Shading Units, 16 GB of GDDR6 memory, and 32 Xe cores for ray tracing, be a powerhouse for games, and handle some professional software as well. At SIGGRAPH 2022, Bob Duffy, Intel's Director of Graphics Community Engagement, showcased a system with Arc A770 GPU running Blender Cycles with ray tracing and denoising.

While we don't have any comparable data to showcase, the system managed to produce a decent rendering in Blender 3.3 LTS release, using Intel's oneAPI. The demo scene had 4,369,466 vertices, 8,702,031 edges, 4,349,606 faces, and 8,682,950 triangles, backed by ray tracing and live denoising. We are yet to see more detailed benchmarks and how the GPU fares against the competition.

Remember that Xe-HPG Scavenger Hunt that Intel hosted last year? If you somehow missed it, Intel was maybe giving away some Arc graphics cards to 300 lucky winners. There were two different tiers of prizes, grand prize and first prize, which later ended up translating to an Arc A770 and an Arc A750 graphics card respectively. Now news via VideoCardz are suggesting that Intel is trying to get out of giving these 300 people their prize, well, at least the promised graphics card, in exchange for an Alder Lake CPU.

Intel has apparently sent out an email to the winners, asking them to accept an Intel Core i7-12700K if they were a grand prize winner or a Core i5-12600K if they were a first prize winner, instead of the promised graphics card. The winners have until Friday the 19th of August to decide if they want a CPU instead of a GPU, although Intel is apparently still allowing them to wait for a GPU, the company just doesn't say how long the wait will be. As the prize has to have a similar retail price, it's also possible to get a ballpark figure of the MSRP of Intel's supposedly upcoming Arc 700-series graphics cards. The Arc A770 should end up at around the $410 mark and the A750 around the $290 mark, as this is the ballpark MSRP for the CPU's that are being offered. It would be interesting to know how many people would be willing to do the trade, but sadly we're unlikely to ever find out.

It's been a busy couple of days when it comes to Intel and its Arc graphics cards, as not only has the company showed off the Arc A770—which looks identical to the Arc A750—but the company has also refuted that it was ever planning to release an Arc A780 card, despite the existence of the A380 and supposedly an A580. A leak with price brackets and performance tiering has also leaked, which gives us a much better understanding of how Intel is planning on positioning its Arc graphics cards versus NVIDIA and AMD and it doesn't look like Intel is as confident as it sounded just a few months ago.

LinusTechTips got the honour to reveal the Arc A770 card, although there appear to be minuscule differences to the physical appearance between it and the Arc A750. The only thing noticeable is a 3-pin header, possibly for some kind of RGB syncing, next to the 8- and 6-pin power connectors, something not present on the A750 card that Gamers Nexus showed off earlier this week. The good news is that the Arc A770 seems to be running cool, as the card was reportedly only hitting 69 degrees C during some hands-on time, although this will apparently be covered in a separate video next week.