NVIDIA on Tuesday announced a roll-out of its implementation of the PCI-SIG resizable base-address register (BAR) feature to select GeForce products. The feature enables your CPU to see the entire video memory of your graphics card as one addressable block, rather than through 256 MB apertures. This should improve certain kinds of 3D rendering workloads, and game engines that are optimized to use it should see a tangible performance boost. AMD earlier introduced the exact same feature under its marketing name "Smart Access Memory," with its Radeon RX 6000 series.

NVIDIA announced that resizable-BAR support will be made available to GeForce RTX 30-series "Ampere" desktop graphics cards, notebooks that have RTX 30-series "Ampere" mobile GPUs, and future products. The support requires not just a compatible graphics card, but also a motherboard that supports the feature. Most leading motherboard- and OEM desktop manufacturers began rolling out resizable-BAR support through UEFI firmware updates. Using the feature requires you to run your machine in native UEFI mode (with CSM disabled).

NVIDIA VP for GeForce, Jeff Fisher hosts the new GeForce RTX: Game On digital event on the sidelines of the 2021 International CES. We expect NVIDIA to unveil its new GeForce RTX 30-series "Ampere" mobile GPUs powering next-gen gaming notebooks; possible additions to its desktop RTX 30-series, including the all-important RTX 3060 and RTX 3050, and maybe some tweaks to the high-end segment. NVIDIA has a knack of surprising us with new gamer-relevant features with such presentations.Update 17:00 UTC: Here we go, with a quick recap of 2020.

Lenovo may have inadvertently disclosed the existence of several upcoming GeForce RTX 30-series graphics cards. The Product Specifications Reference (PSREF) document for a certain Lenovo pre-built gaming desktop model, the Legion R5 28IMB05, lists out all its possible hardware options, covering CPU, graphics cards, and storage. The CPU options cover 10th Gen Core "Comet Lake-S" models that are already out; but things get interesting with the list of graphics options. In addition to certain RTX 20-series, and GTX 16-series SKUs, the list mentions certain RTX 30-series SKUs that haven't yet been announced by NVIDIA.

Among these unreleased GPUs are the GeForce RTX 3050, which is shown featuring 4 GB of GDDR6 memory; the GeForce RTX 3050 Ti with 6 GB of it; and the GeForce RTX 3060 (non-Ti) with 12 GB of it. The already-launched RTX 3070 also finds mention here. It's likely that these are OEM-exclusive SKUs, but if they're not, then we have our first look at how NVIDIA is handling product segmentation between the RTX 3050 Ti and the RTX 3060 (non-Ti), in a possible bid to avoid a repeat of the GTX 1060 3 GB vs. 6 GB confusion (where besides memory, the two SKUs also had different core-configurations). Based on the GA106 silicon, the GeForce RTX 3060 (non-Ti) is expected to feature a 192-bit wide GDDR6 memory interface, which it populates with 12 GB of memory.

Apparently, specifications for NVIDIA's upcoming RTX 30-series mobile solutions have been made public. According to Videocardz via Notebookcheck, NVIDIA will introduce three mobile versions of their RTX 30-series graphics cards in the form of the RTX 3080, RTX 3070 and RTX 3060. Like past NVIDIA mobile solutions, these won't directly correspond, hardware-wise, to their desktop counterparts; NVIDIA has the habit of downgrading their mobile solutions' chips compared to their desktop counterparts. According to the leaked specifications, this means the mobile RTX 3080 will maker use of the company's GA-104 chip, instead of the GA-102 silicon found on desktop versions of the card.

The mobile RTX 3080 should thus feature a total of 6,144 CUDA cores, as present in the fully-enabled GA-104 chip (compare that to the 5,888 CUDA cores available on the desktop RTX 3070, and the 8,704 CUDA cores available on the RTX 3080). These CUDA cores would be clocked at up to 1.7 GHz. The memory bus should also see a cut down to 256-bit, which would allow NVIDIA to distribute as many as 4 versions of the RTX 3080 mobile: Max-Q (TGP 80-90 W), Max-P (TGP 115-150 W), with either 8 GB or 16 GB of GDDR6 memory. The RTX 3070 mobile keeps the GA-104 chip, 256-bit bus and GDDR6 memory subsystem (apparently with only 8 GB memory pool available), but further cuts down CUDA cores to 5,120 (Max-Q TGP 80-90 W, Max-P TGP 115-150 W). Finally, the RTX 3060 mobile should make use of the GA106 chip, set up with 3,072 available CUDA cores and a 192-bit memory bus across its 6 GB of GDDR6 VRAM pool (Max-Q TGP 60-70 W), Max-P (TGP 80-115 W). Expect these specs to be confirmed (or not) come January 12th.

Stock for NVIDIA's latest RTX 30-series graphics cards has been a nightmare for customers across the world, with demand far outstripping supply. This opened up a proverbial can of worms, with scalpers taking to the world wide web, casting their own nets in taking advantage of not only the pandemic (and peoples' refuge in gaming in these uncertain times), but also of said unmet demand. So it has to be nothing short of a Christmas miracle that 500,000 NVIDIA RTX 3080 graphics cards have just been found in an unmarked shipping container in South Korea. The container wasn't registered in the port authority, and was therefore left unopened and unprocessed.

The graphics cards were stored in the container absent of any proper documentation by Samsung, as early as August of this year. Jeff Fisher, vice president of NVIDIA and head of the GeForce division, said in a statement to the company's shareholders that "We've been asking Samsung for this shipment for months. They told us that she had already left the factory, but then they did not present us with any document proving that she had reached her destination". These newfound graphics cards will now be correctly processed and put into the channel.Geeknetic.es made this as a part of the Spanish Fool's Day, which is December 28th. However, considering the current state of the RTX (and AMD RX) market, this is a nice satirical gotcha which I'll keep on TPU. Let's laugh at our misery instead of wallowing in it.

ZOTAC has had a rather spartan custom RTX 30-series lineup in the West, with only its Trinity, Twin-Edge, and Holo board designs covering the RTX 3080 and RTX 3090 series. The company plans to change this, with the introduction of a better endowed custom board series, under the PGF series. ZOTAC debuted the PGF brand with the GeForce GTX 10-series "Pascal" family, mostly in the Greater China region, and targeted at enthusiasts. The RTX 20-series "Turing" didn't see PGF branded cards. It now makes a comeback with the RTX 30-series "Ampere." As with the older cards, these are being launched as China-exclusive. It remains to be seen if they reach Western markets.

Both cards feature a common board design with a large triple-slot, triple-fan cooling solution that's dunked in RGB LED embellishments. Both cards feature exotic VRM components such as multi-phase capacitors for better electrical noise suppression. The RTX 3080 PGF comes with GPU Boost frequencies of 1770 MHz (vs. 1710 MHz reference), while the RTX 3090 PGF does 1755 MHz GPU Boost (vs. 1695 MHz reference). The RTX 3070 PGF runs up to 1785 MHz GPU Boost (vs. 1725 MHz reference), and there's even an RTX 3060 Ti PGF, doing 1725 MHz GPU Boost (vs. 1665 MHz reference).

It seems NVIDIA and Samsung's partnership in bringing to life the green company's semiconductor designs isn't about to end anytime soon. Semiconductor analysts and insiders have said that NVIDIA and Samsung etched a new manufacturing deal on December 17th that still relates to the company's in-high-demand RTX-30 series graphics cards, which should see Samsung increase output - particularly at its Hwaseong plant - to sate the seemingly unquenchable demand from consumers and scalpers alike. The deal, which is roughly valued at "hundreds of billions won" will see Samsung double down on its 8 nm output for NVIDIA's latest gaming chips. This seems to put to rest speculation on an RTX 30-series redesign for TSMC's allegedly better 7 nm process - and according to the industry insiders, NVIDIA looked to Samsung specifically because of the need for "quick delivery of the chips".

This instills new life into Samsung's contract-based foundry business; according to market researcher TrendForce, Samsung's foundry business is expected to post a record $14.05 billion in sales this year, up 17.9% from 2019, as the company expands its client base not only through this and the previous NVIDIA deal, to Qualcomm Technologies Inc., Google, IBM, Cisco and China's Baidu. Samsung is accelerating its investment into EUV lithography in sub-7 nm processes so as to poach more customers and market share from industry behemoth and poster boy TSMC, spending 10 trillion won ($8.6 billion) to both improve technology and increase output on its foundries.

ASUS today introduced White variants of its Republic of Gamers (ROG) Strix custom-design GeForce RTX 30-series graphics cards. These card have essentially the same board design as the original ROG Strix RTX 30-series, but with streaks of matte-white and brushed-metal making up the cooler shrouds, back-plates, and impellers of the three Axial-Tech fans. The PCB underneath is still black. The RGB LED embellishments are identical to those on the original ROG Strix cards. The cards also feature the same factory-overclocked speeds as their original-design siblings.

Among the variants ASUS is launching are the ROG-STRIX-RTX3090-O24G-WHITE and ROG-STRIX-RTX3090-24G-WHITE, based on the RTX 3090; ROG-STRIX-RTX3080-O10G-WHITE and ROG-STRIX-RTX3080-10G-WHITE based on the RTX 3080; and ROG-STRIX-RTX3070-O8G-WHITE and ROG-STRIX-RTX3070-8G-WHITE, based on the RTX 3070. The company didn't reveal pricing, although we expect these cards to be priced at a slight premium over the original ROG Strix RTX 30-series.

IR photographs are one of the best ways to take a look at the inner silicon etchings inside the world's most powerful accelerators. Of course, one could always just "shove off the top", but that would be a crime unto itself. Fritchens Fritz has pointed an IR gun at the heart of NVIDIA's GA102, the crown jewel of the RTX 30-series, and the results are an intimate look at some of the world's most powerful hardware. The 642 mm² chip showcases all of its precision-engineered resources that allow gamers to achieve (almost) 60 FPS in 4K resolution in Cyberpunk 2077, with rows upon rows of transistor building blocks mirroring NVIDIA's (representative) chip resource breakdown in the Ampere whitepaper.

The seven vertical rows showcase the GA102's Graphics Processing Clusters (GPCs) - and in each of those rows, one can count 12 Streaming Multiprocessors (SMs). Image definition goes slightly downhill from there; but if you compare side-by-side with NVIDIA's own Ampere shots, you'll certainly be able to plot the different transistor arrangements that form the Tensor, RT, and CUDA cores.

Michael Driscoll, an Oracle data engineer, has written a data scraper that runs through eBay listings for the latest hardware, comparing products with their sale price. The objective was to see just how pervasive scalping actually is, and to get a (flawed and incomplete, but still extremely interesting) outlook at the scalping ecosystem and their gains with the current hardware and console shortages. Driscoll analyzed sales for the Xbox Series X|S, the PS5 (discless and disc-based) as well as NVIDIA's RTX 30-series, AMD's RX 6000 series, and Zen 3 processors. There are some assumptions on the gathering and analysis of this data, but that is part of the beast.

The results are potentially desperation-inducing. AMD's Zen 3 CPUs have sold for sometimes 240% of their MSRP (looking at the biggest offender, the Ryzen 9 5950X. The RX 6800 XT graphics card has been selling for within an inch of 200% of its MSRP as well, with a median price over the past week set at $1247 (compare that to the $649 MSRP). The RTX 3080 has been selling at 180% of its MSRP for the past week, but it has been moved at 220% of its MSRP before. The case repeats with several degrees of severity for the Xbox family and PS5 consoles.

EK, the leading computer cooling solutions provider, is ready to offer its premium high-performance GPU water block for the MSI Trio and Suprim editions of NVIDIA GeForce RTX 30 Series graphics cards. This new water block is named EK-Quantum Vector Trio RTX 3080/3090 D-RGB and is exclusively engineered for MSI Trio and Suprim RTX 3080 and RTX 3090 GPUs.

The new water block is one of the larger ones in the 3000 series, so make sure your PC case can accommodate it. It's 157 mm wide and 320 mm long and cools all key components on the GPU's printed circuit board. This gives the water block advantage on the market of having a clean design, showing off all of the cooling liquid, and cooling each and every necessary component.

Alphacool presents the Eisblock Aurora Acryl GPX-N block for the EVGA FTW3 graphics cards. The blocks offer outstanding cooling performance thanks to the full cover design. The new backplate, which is included with the coolers, also contributes to this. This stabilizes the graphics card and ensures an even contact pressure of the cooler. The cold plates are made of solid nickel-plated copper. The coolers cover all relevant components such as voltage converters and the graphics memory.

NVIDIA is readying its GeForce RTX 30-series "Ampere" mobile GPUs for the first half of 2021, reports WCCFTech, citing a leaked company slide detailing the upcoming mGPU product stack. The slide mentions a January 2021 launch for the RTX 3080 Mobile and RTX 3070 Mobile, followed by an RTX 3060 Mobile launch later. The same product stack slide compares performance of these parts to desktop RTX 20-series GPUs, with the RTX 3080 Mobile being compared to desktop RTX 2080 Super (or roughly desktop RTX 3060 Ti performance), while the RTX 3070 Mobile is shown offering performance roughly comparable to desktop RTX 2070 Super or RTX 2070. The RTX 3060 Mobile offers roughly desktop RTX 2070 / RTX 2060 performance.

EVGA has launched their halo RTX 30-series product in the form of the RTX 3090 KINGPIN Hybrid graphics card. The pricing is there to match as well, at a full, fat $2,000. For that money, you get a hybrid-cooled graphics card that ships with a 360 mm AIO and has an incorporated OLED display for graphics card monitoring (voltages, temperatures and clocks) or animation display. The card also features some of the highest Boost clocks available in the industry for the RTX 3090 SKU, set at 1,920 MHz Boost.

It looks like NVIDIA will launch its 4th GeForce RTX 30-series product ahead of Holiday 2020, the GeForce RTX 3060 Ti, with VideoCardz unearthing a leaked NVIDIA performance guidance slide, as well as pictures of custom-design RTX 3060 Ti cards surfacing on social media. The RTX 3060 Ti is reportedly based on the same 8 nm "GA104" silicon as the RTX 3070, but cut down further. It features 38 out of 48 streaming multiprocessors physically present on the "GA104," amounting to 4,864 "Ampere" CUDA cores, 152 tensor cores, and 38 "Ampere" RT cores. The memory configuration is unchanged from the RTX 3070, which means you get 8 GB of 14 Gbps GDDR6 memory across a 256-bit wide memory interface, with 448 GB/s of memory bandwidth.

According to a leaked NVIDIA performance guidance slide for the RTX 3060 Ti, the company claims the card to consistently beat the GeForce RTX 2080 SUPER, a $700 high-end SKU from the previous "Turing" generation. The same slide also shows a roughly 40% performance gain over the previous generation RTX 2060 SUPER, which is probably the logical predecessor for this card. In related news, PC Master Race (OfficialPCMR) on its Facebook page posted pictures of boxes of an ASUS TUF Gaming GeForce RTX 3060 Ti OC graphics cards, which confirms the existence of this SKU. The picture of the card on the box reveals a design similar to other TUF Gaming RTX 30-series cards launched by ASUS so far. As for price, VideoCardz predicts a $399 MSRP for the SKU, which should nearly double the price-performance for this card over the RTX 2080 SUPER at NVIDIA's performance numbers.

EVGA over the weekend launched an extensive lineup of GeForce RTX 30-series graphics cards that either feature liquid cooling, or come with preparation for DIY liquid cooling. The EVGA Hydro Copper line of graphics cards include the RTX 3090 FTW3 Hydro Copper, RTX 3090 XC3 Hydro Copper, RTX 3080 FTW3 Hydro Copper, and RTX 3080 XC3 Hydro Copper. These are essentially the same FTW3 or XC3 graphics cards EVGA debuted its RTX 3090 and RTX 3080 lineups with, but come with factory-fitted, full-coverage water blocks. These blocks are made of nickel-plated copper, with clear acrylic tops that have a plastic top-plate with a printed pattern similar to the one found on the back-plates. The tops are studded with addressable RGB LEDs which are connected directly to RGB controllers on the PCB, and can be controlled via the Precision X1 software. These cards have the same factory-overclocked speeds as their air-cooled siblings, but are priced about $150-200 higher.

Next up, are the Hybrid Cooling line of graphics cards, which feature factory-fitted, all-in-one, closed-loop, liquid cooling solutions. Much like the FTW3 and XC3 Hydro Copper series, we see EVGA reuse its RTX 3090 and RTX 3080 FTW3 and XC3 PCBs to carve out their Hybrid Cooling variants. The FTW3 Hybrid Cooling cards feature addressable RGB illumination on the radiator fans, while the XC3 Hybrid Cooling series cards lack illumination on the cards. All four variants feature 240 mm x 120 mm radiators, and a pair of included 120 mm fans. The cooling solution features a pump-block cooling the GPU, while a series of heatsinks and a lateral fan cool the memory and VRM components. These cards have a similar $150-200 premium over the air-cooled FTW3 and XC3 cards.

Alphacool presents three new Eisblock Aurora Acryl GPX-N blocks for the NVIDIA RTX 3070, 3080 and 3090 graphics cards. The blocks offer outstanding cooling performance thanks to the full cover design. The new backplate, which is included with the coolers, also contributes to this. This stabilizes the graphics card and ensures an even contact pressure of the cooler. The cold plates are made of solid nickel-plated copper. The coolers cover all relevant components such as voltage converters and the graphics memory.

INNO3D, a leading manufacturer of pioneering high-end multimedia components and innovations brings you the new INNO3D GeForce RTX 3090 / 3080 iCHILL Frostbite. Following the huge success of its predecessor iCHILL Frostbite from the previous generation RTX 20 Series, we have now also armed our powerhouse RTX 3090 / 3080 graphics cards with an updated version of the iCHILL Frostbite.

Founded in 1998 with the vision of developing pioneering computer hardware products on a global scale. Fast forward to the present day, INNO3D is now well-established in the gaming community known for our innovative and daring approach to design and technology. We are Brutal by Nature in everything we do and are 201% committed to you for the best gaming experience in the world.

Here's the first picture of an MSI GeForce RTX 3080 SUPRIM X graphics card in its retail packaging, courtesy of Harukaze5719. MSI is debuting its SUPRIM line of graphics cards with the GeForce RTX 30-series, as a custom-design lineup possibly positioned between its Gaming/X and Ventus/X lines. So you'll likely get the PCB chops and overclocking capability, along with noise levels and cooler performance, rivaling the Ventus/X series, but with the RGB LED bling and board aesthetics that are closer to the Gaming/X series. It's likely that the SUPRIM/X series will only be launched in selection regions. The picture of the card on the box reveals a triple-slot beast with three fans arranged not unlike the Gaming/X Tri Frozr 2 cooler; but with a lot more RGB illumination than you have on the Ventus/X series. We don't know the cards' clocks, pricing, or availability.

Possibly unsure of the GeForce RTX 3070 tackling AMD's Radeon RX 6000 series parts, NVIDIA is designing a new RTX 30-series SKU positioned between the RTX 3070 and RTX 3080. This is not a 16 GB variant of the RTX 3070, but rather a new SKU based on the 8 nm "GA102" silicon, according to a reliable source with NVIDIA leaks, kopite7kimi. The SKU is based on the GA102 with the ASIC code "GA102-150-KD-A1." The silicon is configured with 7,424 CUDA cores across 58 streaming multiprocessors (29 TPCs), 232 tensor cores, 232 TMUs, 58 RT cores, and an unknown number of ROPs. According to kopite7kimi, the card is configured with a 320-bit wide memory interface, although it's not known if this is conventional GDDR6, like the RTX 3070 has, or faster GDDR6X, like that on the RTX 3080.

NVIDIA recently "cancelled" a future 16 GB variant of the RTX 3070, and 20 GB variant of the RTX 3080, which is possibly the company calibrating its response to the Radeon RX 6000 series. We theorize that doubling in memory amounts may not have hit the desired cost-performance targets; and the company probably believes the competitive outlook of the RTX 3080 10 GB is secure. This explains the need for a SKU with performance halfway between that of the RTX 3070 and RTX 3080. As for pricing, with the RTX 3070 positioned at $500 and the RTX 3080 at $700, the new SKU could be priced somewhere in between. AMD's RDNA2-based Radeon RX 6000 series GPUs are expected to feature DirectX 12 Ultimate logo compliance, meaning that there is a level playing ground between AMD and NVIDIA in the performance segment.

AMD in a letter to its retail partners issued guidelines to prevent scalping of the kind that affected the NVIDIA GeForce RTX 30-series "Ampere" launch. The letter, leaked to the web by RedGamingTech, calls for measures such as real-time bot detection, complex CAPTCHA implementations (such as math, pictures, etc), a queue-based reservation system; manual processing of online orders, minimizing B2B re-selling to ensure sales to private individuals (end-users), and dynamic inventory. The Ryzen 5000 "Zen 3" series desktop processors will start being available from November 5, while AMD will detail Radeon RX 6000 RDNA2 availability in its October 28 public presentation dedicated to the graphics card series.

AMD's RDNA2 graphics architecture features hardware-accelerated decoding of the AV1 video format, according to a Microsoft blog announcing the format's integration with Windows 10. The blog mentions the three latest graphics architectures among those that support accelerated decoding of the format—Intel Gen12 Iris Xe, NVIDIA RTX 30-series "Ampere," and AMD RX 6000-series "RDNA2." The AV1 format is being actively promoted by major hardware vendors to online streaming content providers, as it offers 50% better compression than the prevalent H.264 (translating into that much bandwidth savings), and 20% better compression than VP9. You don't need these GPUs to use AV1, anyone can use it with Windows 10 (version 1909 or later), by installing the AV1 Video Extension from the Microsoft Store. The codec will use software (CPU) decode in the absence of hardware acceleration.

AMD in its October 8 online launch event for the Ryzen 5000 "Zen 3" processors, provided a teaser of the company's next flagship graphics card slotted in the Radeon RX 6000 series. This particular SKU has been referred to by company CEO Lisa Su as "Big Navi," meaning it could be the top part from AMD's upcoming client GPU lineup. As part of the teaser, Su held up the reference design card, and provided three performance numbers of the card as tested on a machine powered by a Ryzen 9 5900X "Zen 3" processor. We compared these performance numbers, obtained at 4K UHD, with our own testing data for the games, to see how the card compares to other current-gen cards in its class. Our testing data for one of the games is from the latest RTX 30-series reviews, find details of our test bed here. We obviously have a different CPU since the 5900X is unreleased, but use the highest presets in our testing.

With "Borderlands 3" at 4K, with "badass" performance preset and DirectX 12 renderer, AMD claims a frame-rate of 61 FPS. We tested the game with its DirectX 12 renderer in our dedicated performance review (test bed details here). AMD's claimed performance ends up 45.9 percent higher than that of the GeForce RTX 2080 Ti as tested by us, which yields 41.8 FPS on our test bed. The RTX 3080 ends up 15.24 percent faster than Big Navi, with 70.3 FPS. It's important to note here that AMD may be using a different/lighter test scene than us, since we don't use internal benchmark tools of games, and design our own test scenes. It's also important to note that we tested Borderlands 3 with DirectX 12 only in the game's launch-day review, and use the DirectX 11 renderer in our regular VGA reviews.

NVIDIA could update the higher end of its GeForce RTX 30-series "Ampere" product stack with two new additions in December 2020. Sources tell VideoCardz that the company is preparing to launch a 20 GB variant of the GeForce RTX 3080, and a 16 GB variant of the RTX 3070. The RTX 3080 20 GB will come with double the memory of the RTX 3080 the company debuted last month, over the same 320-bit wide GDDR6X memory interface, possibly by using two 8 Gbit memory chips per 32-bit path (which is how the RTX 3090 achieves 24 GB, over its 384-bit memory bus). The RTX 3070 16 GB will likely use a similar approach, albeit with GDDR6 memory. Meanwhile, the mid-range "RTX 3060 Ti" could debut in November, following the late-October introduction of the RTX 3070 8 GB. Much of NVIDIA's product stack adjustments could be in preparation for AMD's late-October reveal of the Radeon RX 6000 RDNA2 series.

Recently, it has been brought to light that a particular seller on Ebay was selling price-hiked MSI RTX 30-series graphics cards - such as the RTX 3080 Gaming X Trio - well over MSRP ($1,359 for an RTX 3080, compared to NVIDIA's $699 and MSI's $799). A more attentive look at the seller, identified as Starlit Partner, reveals that they carried only MSI-manufactured SKUs for NVIDIA's RTX 30-series, and was selling them, in most cases, as sealed and brand new. Now, scalping of NVIDIA's latest graphics card series has been thoroughly covered here on TechPowerUp; however, suspicions of ties between Starlit Partner and MSI itself, and the suggestion that this was a coordinated move in order to sell RTX 30-series inventory at marked-up pricing, deserves a longer appraisal.

As soon as the suspicion was posted in the Internet, some users donned their detective cloaks and went digging for more information, and confirmed the ties to MSI. However, MSI has already issued a statement on the issue, clarifying the scenario we were faced with. Indeed, MSI has ties to Starlit Partner, as they themselves acknowledge - the later is an individual sales subsidiary working under MSI. However, MSI further explains in the statement that Starlit Partner is tasked with the sale of refurbished items and excess inventory - and that they should never have had access to NVIDIA's RTX-30 series graphics cards in the first place. MSI launched an investigation that confirmed an error in inventory allocation allowed Starlit Partner to access inventory they shouldn't have had access to (without clarifying the error).