AMD's Radeon RX 9070 XT and RX 9070 "MBA" graphics cards are no longer "best kept secrets"—as demonstrated recently by the "leaking out" of supposed reference models through black market/back alley channels in China. Late last month, a dual-fan non-XT specimen was snapped up by Chiphell forum member—alleged benchmark results were soon shared within that community. A few days later, a "Made-by-AMD" Radeon RX 9070 XT sample was dissected and compared to Sapphire's PULSE Radeon RX 9070 XT 16 GB SKU. An additional MBA XT example emerged last week, courtesy of another in-depth Chiphell thread. A "bored" enthusiast happened upon a 5499 RMB (~$748 USD) when idly browsing through Xianyu listings (Taobao's Ebay equivalent platform). Their buying experience was described as follows: "(I) found a 'public version' Radeon RX 9070 XT in Tianjin. It was said to be manufactured by (an) OEM, so I bought it without hesitation...I made an appointment to meet today and got it successfully. I don't have to wait for a graphics card anymore."

For unknown reasons, AMD decided to launch its first wave of RDNA 4 gaming graphics cards sans first-party designs. Leaked specimens have attracted much attention in China; with owners bragging about their respective ownerships of reasonably priced rarities. The latest back channel customer expressed satisfaction when inspecting Team Red's all-black flagship Navi 48 GPU-based solution: "after I got it, I have to say that it is not very heavy and is quite light. But the appearance is really what I like." After initial tests, they discovered that VRAM temperatures were not up to snuff—as alluded to (pre-launch) by other Chiphell figures. Their personal DIY improvements were described: "(I) took it apart to measure the thickness of the thermal grease pad. Everyone said the temperature of the video memory was high, so I decided to change to something better....(with) original silicone grease FurMark 2K resolution for 20 minutes: maximum core temperature was 62 degrees, maximum hot spot temperature was 84 degrees, maximum memory temperature was 88 degrees, maximum power was consumption 346 W."

At the 2025 International CES, we went hands on with some of the first Schenker XMG enthusiast notebooks that feature a next-generation NVIDIA GeForce RTX 5090 Laptop GPU. These include the Schenker Key 18 PRO, and the XMG Neo 16. The Key 18 Pro is a massive portable workstation featuring an 18-inch 16:10 display with either 2560 x 1600 or 3840 x 2400 pixels resolution, with either 500 nits or 400 nits brightness, respectively. The notebook is powered by an Intel Core Ultra 9 275HX processor (8P+16E, 36 MB cache), with four DDR5 SODIMM slots that you can add up to 192 GB of memory to. The star attraction of course is its NVIDIA GeForce RTX 5090 Laptop GPU, which is configured with 24 GB of video memory, and 175 W of power. A large 98 Wh battery powers the beast.

Meanwhile, the XMG Neo 16, as its name suggests, has a 16-inch display. This is 16:10, with 2560 x 1600 resolution, comes with 300 Hz refresh rate, and 1000 nits brightness. You get the same combination of a Core Ultra 9 275HX processor and GeForce RTX 5090 Laptop GPU, with the GPU getting 175 W of power. There are two DDR5 CSODIMM slots (SODIMMs with CKD). The company also showed off two of its key innovations that power the two notebooks, its mainboard with a proprietary bridge connector between the motherboard and discrete GPU board; and its two new cooling solutions.

The mid-range AMD Radeon RX 5600 XT is not supposed to have 16 GB of video memory, but the same hardware modders from Brazil behind the recent GeForce RTX 2080 16 GB mod, had other ideas for the card. They have not only increased the memory size to 16 GB through memory chip replacement, but also succeeded in widening its memory bus to 256-bit. The RX 5600 XT was launched in 2018 with 6 GB of 14 Gbps GDDR6 memory over a 192-bit memory interface. The card is cut down from the 7 nm "Navi 10" silicon powering the RX 5700 series, by enabling 36 out of 40 compute units (the same count as the RX 5700), but with a truncated 192-bit memory bus wired to 6 GB of memory (and so 25% lower memory bandwidth).

Paulo Gomes and Ronaldo Buassali pulled off the daring Radeon RX 5600 XT memory mod, which involves not just increasing the memory size from 6 GB to 16 GB, but also widening the memory bus from 192-bit to 256-bit. Since the RX 5600 XT is based on the same "Navi 10" GPU as the RX 5700, custom-design graphics cards tend to reuse PCB designs from the RX 5700 series, and have two vacant memory pads that are sometimes exposed and even balled. The mod involves three key stages—to replace the six 8 Gbit GDDR6 memory chips with eight 16 Gbit ones; to add the required electrical SMDs and VRM components for the two additional memory chips; and lastly, to give the card a modified BIOS that can let it play with the new memory configuration. The "Navi 10" silicon also powers certain Radeon Pro graphics cards with 16 GB of memory using 16 Gbit memory chips, so that could be the starting point for the BIOS mod.

February saw community mods bring resizable BAR support to several older platforms; and now we come across a mod that brings it to some older GPUs. The NVStrapsReBAR mod by terminatorul, which is forked out of the ReBarUEFI mod by xCurio, brings resizable BAR support to NVIDIA GeForce RTX 20-series and GTX 16-series GPUs based on the "Turing" graphics architecture. This mod is intended for power users, and can potentially brick your motherboard. NVIDIA officially implemented resizable BAR support since its RTX 30-series "Ampere" GPUs in response to AMD's Radeon RX 6000 RDNA 2 GPUs implementing the tech under the marketing name Smart Access Memory. While AMD would go on to retroactively enable the tech for even the older RX 5000 series RDNA GPUs, NVIDIA didn't do so for "Turing."

NVStrapsReBAR is a motherboard UEFI firmware mod. It modifies the way your system firmware negotiates BAR size with the GPU on boot. There are only two ways to go about modding a platform to enable resizable BAR on an unsupported platform—by modding the motherboard firmware, or the video BIOS. Signature checks by security processors in NVIDIA GPUs make the video BIOS modding route impossible for most users; thankfully motherboard firmware modding isn't as difficult. There is an extensive documentation by the author to go about using this mod. The author has tested the mod to work with "Turing" GPUs, however, it doesn't work with older NVIDIA GPUs, including "Pascal." Resizable BAR enables the CPU (software) to see video memory as a single contiguously addressable block, rather than through 256 MB apertures.

Brazilian tech enthusiast Paulo Gomes, in association with Jefferson Silva, and Ygor Mota, successfully modded an EVGA GeForce RTX 2080 "Turing" graphics card to 16 GB. This was done by replacing each of its 8 Gbit GDDR6 memory chips with ones that have double the density, at 16 Gbit. Over the GPU's 256-bit wide memory bus, eight of these chips add up to 16 GB. The memory speed was unchanged at 14 Gbps reference, as were the GPU clocks.

The process of modding involves de-soldering each of the eight 8 Gbit chips, clearing out the memory pads of any shorted pins, using a GDDR6 stencil to place replacement solder balls, and then soldering the new 16 Gbit chips onto the pad under heat. Besides replacing the memory chips, a series of SMD jumpers need to be adjusted near the BIOS ROM chip, which lets the GPU correctly recognize the 16 GB memory size. The TU104 silicon by default supports higher density memory, as NVIDIA uses this chip on some of its professional graphics cards with 16 GB memory, such as the Quadro RTX 5000.

NVIDIA is preparing the launch of its third and final RTX 4060-series graphics card SKU, the GeForce RTX 4060 16 GB, for July 18, 2023. Going by past convention, reviews of the RTX 4060 Ti 16 GB graphics card priced at its steep $499 MSRP, will go live on June 17, and those priced above the MSRP on July 18, alongside market availability. The RTX 4060 Ti is essentially a memory variant of the RTX 4060 Ti. It offers 16 GB of video memory across the card's 128-bit wide memory interface.

According to the specs-sheet put out by NVIDIA on the May 18 launch date for the RTX 4060 series, besides memory size, there are no other differences between the RTX 4060 Ti 16 GB, and the current RTX 4060 Ti 8 GB. In particular, there is no change in the core-configuration or clock-speed, since the shader compute throughput of both models is listed at the same 22 TFLOPs. Even the memory speed is the same, at 18 Gbps (GDDR6-effective), at which the GPU has 288 GB/s of memory bandwidth at its disposal. It will be interesting to see the performance impact of 16 GB memory.

The price of GDDR6 memory has continued to fall sharply - over recent financial quarters - due to an apparent decrease in demand for graphics cards. Supply shortages are also a thing of the past—industry experts think that manufacturers have been having an easier time acquiring components since late 2021, but that also means that the likes of NVIDIA and AMD have been paying less for VRAM packages. Graphics card enthusiasts will be questioning why these savings have not been passed on swiftly to the customer, as technology news outlets (this week) have been picking up on interesting data—it demonstrates that spot prices of GDDR6 have decreased to less than a quarter of their value from a year and a half ago. 3DCenter.org has presented a case example of 8 GB GDDR6 now costing $27 via the spot market (through DRAMeXchange's tracking system), although manufacturers will be paying less than that due to direct contract agreements with their favored memory chip maker/supplier.

A 3DCenter.org staffer had difficulty sourcing the price of 16 Gb GDDR6 VRAM ICs on the spot market, so it is tricky to paint a comparative picture of how much more expensive it is to equip a "budget friendly" graphics card with a larger allocation of video memory, when the bill-of-materials (BoM) and limits presented by narrow bus widths are taken into account. NVIDIA is releasing a GeForce RTX 4060 Ti 16 GB variant in July, but the latest batch of low to mid-range models (GeForce RTX 4060-series and Radeon RX 7600) are still 8 GB affairs. Tom's Hardware points to GPU makers sticking with traditional specification hierarchy for the most part going forward: "(models) with double the VRAM (two 16 Gb chips per channel on both sides of the PCB) are usually reserved for the more lucrative professional GPU market."

NVIDIA receives a lot of questions about graphics memory, also known as the frame buffer, video memory, or "VRAM", and so with the unveiling of our new GeForce RTX 4060 Family of graphics cards we wanted to share some insights, so gamers can make the best buying decisions for their gaming needs. What Is VRAM? VRAM is high speed memory located on your graphics card.

It's one component of a larger memory subsystem that helps make sure your GPU has access to the data it needs to smoothly process and display images. In this article, we'll describe memory subsystem innovations in our latest generation Ada Lovelace GPU architecture, as well as how the speed and size of GPU cache and VRAM impacts performance and the gameplay experience.