Video Cards & GPUs News - Page 147

GPU shipments are a bit funny right now and we all know it, with analyst firm JPR (Jon Peddie Research) posting a new report that details the global GPU market has hit a milestone of 101 million units in Q3 2021 alone.

JPR also notes that GPU shipments increased by 9% year-over-year while discrete GPU numbers inside of PCs is expected to hit 31% over the next 5 years leading up to 3.24 billion GPUs by the end of 2025. AMD had its GPU market share increase by 1.4% in the three-month period, while NVIDIA GPU market share increased by 4.86% in the same quarter.

AMD GPU shipments dropped by a decent chunk, by 11.4% in Q3 2021 while NVIDIA had its GPU shipments increased by 8% for the quarter. The latest numbers have AMD with 17% of the discrete GPU shipment market share, and NVIDIA with 83% of the dGPU shipment market share.

The price of AMD Radeon RX 6000 series GPUs is about to be a little worse, with AMD reportedly ramping up prices of its RDNA 2 cards by 10%.

AMD's increased pricing on Radeon RX 6000 series GPUs is reportedly starting with the next batch of GPU shipments from AMD to its AIB (add-in board) partners. We should expect the new pricing to be reflected online and in-store in the coming weeks, right as we lead into the holidays.

There was a 9% increase in pricing of AMD's new Radeon RX 6000 series GPUs in November 2021 over October 2021, and now we have another 10% increase moving into December. TSMC wafer costs increasing are being blamed, so AMD has got no other move to make other than take the hit -- which isn't nice -- or boost pricing (also not nice).

NVIDIA has just announced its new RTX A4500 workstation GPU, which uses the GA102 GPU with 7168 CUDA cores and 20GB of GDDR6 memory.

The introduction of the new NVIDIA RTX A4500 workstation GPU offers its 7168 CUDA cores alongside 224 Tensor Cores, and 20GB of GDDR6 memory clocked at 16Gbps on a 320-bit memory bus which will deliver up to 640GB/sec of memory bandwidth.

NVIDIA says the new RTX A4500 workstation GPU will have 23.7 TFLOPs of FP32 compute performance, while it has 46.2 TFLOPs of RT performance. NVIDIA's new RTX A4500 workstation GPU has a 200W TDP, which is 15% lower than the RTX A5000 and 17% slower.

NVIDIA is reportedly preparing a beefed-up GeForce RTX 2060 release in January 2022, which will see the card be released with 12GB of VRAM -- an upgrade from the 6GB of RAM by default.

But before its announcement or release, Russian YouTuber and modder VIK-on has upgraded his GeForce RTX 2060 6GB to a GeForce RTX 2060 12GB. The original GeForce RTX 2060 launched with the cut-down Turing-based TU106 GPU, rocking 1920 CUDA cores, and 6GB of GDDR5 memory on a 192-bit memory bus, clocked at 14Gbps.

NVIDIA is expected to make the GeForce RTX 2060 with 12GB official on December 7, with VIK-on testing the modded RTX 2060 with 12GB of VRAM in crypto mining and Unigine Superposition at first with roughly the same performance as the 6GB model -- duh. We should expect more official news and benchmarks in the coming weeks.

GALAX has just introduced the new GeForce RTX 3060 Metaltop Mini [FG] graphics card, a new GeForce RTX 3060-based graphics card on the super-small Mini-ITX form factor.

The new GALAX GeForce RTX 3060 Metaltop Mini [FG] graphics card rocks a single-fan design, while the "FG" in the name stands for "For Gamers". GALAX thinks by adding "FG" and it being "For Gamers" that the big bad crypto miners won't buy these... hah.

As for the measurements, we're looking at the card measuring in at just 16.8cm in length, 11.6cm in width, and 4cm in height. The single-fan design means it is nice and small and supports 0dB technology which means the single fan will disable when the GPU load is low enough (not gaming).

Meta today announced that the Mercenaries DLC package for Resident Evil 4 is officially in development for the Meta Quest. Along with that announcement, the company revealed that RE4 is the fastest-selling game on the platform. It's almost as if people really want AAA games in VR.

Capcom and Oculus (now Meta) released the Quest version of Resident Evil 4 on October 21st, and in just over a month, the game has become the fastest-selling title in the Quest platform's history. Meta has never released sales numbers for any of its headsets or games, but we know several titles have made millions of dollars in profits. Resident Evil isn't likely in the top-selling ranks yet, but it will be if the high sales volume continues. Beat Saber and Population: One have big head starts, though.

To keep the excitement about RE4 VR going, Capcom is working on a DLC expansion for the Quest title. The Mercenaries expansion will launch next year and let you experience more of Resident Evil in terrifying first-person VR. The Mercenaries expansion adds a hoard-mode experience to the game, where you need to fight off a never-ending hoard of Ganados until your rescue helicopter arrives.

AMD has a couple of new graphics cards are coming: the Radeon RX 6500M and Radeon RX 6300M, which will both feature the Navi 24 GPU.

We have heard about the Navi 24-powered GPUs before from Tom @ Moore's Law is Dead, with the entry-level Navi 24 GPU taking care of at least 6 new graphics cards from AMD. We should see the Radeon PRO W6500M, Radeon PRO 6300M, and then gaming SKUs in the Radeon RX 6500M and Radeon RX 6300M.

These 4 graphics cards are expected to join the already-known Radeon RX 6500 XT and Radeon RX 6400 series desktop graphics cards -- also powered by the same Navi 24 GPU. What we do know about the Navi 24 GPU is that it has a PCI Device ID of 7420-743F which has been spotted in Linux kernel/driver leaks previously.

It's been a couple of hot weeks of rumors and news about Intel's new Arc Alchemist family of GPUs, with Tom @ Moore's Law is Dead has unveiled new details about the Intel DG2-128EU GPU.

We should see Intel's new DG2-128EU GPU in desktop form with the Arc Alchemist, so expect 6GB of GDDR6 memory and a 96-bit memory bus. This beats out previous rumors of a 64-bit memory bus, but Tom also notes that we could see a 4GB GDDR6 variant, as well as a 64-bit mobile Arc Alchemist SKU in the future.

Tom expects the DG2-128EU GPU will have GPU clocks of between 2.2GHz and 2.5GHz, with a 75W TDP while Tom "can't promise they won't overclock it for desktop like Navi 24, but so far everything is low power". With the notes on power consumption, Tom continues: "with such low power, and 6GB of VRAM instead of 8GB, this thing can cost $179 or less with decent margins".

Intel's new flagship Arc Alchemist GPU has been spotted flexing some benchmark muscle on Geekbench, but with low GPU clocks and nowhere near a final driver, this isn't going to beat AMD or NVIDIA right now.

The GPU was spotted with detailed JSON data with 4 benchmark runs, with the GPU itself described as "OpenCL HD Graphics" or "Intel Xe Graphics". This GPU has 512 Execution Units and 12.7GB of VRAM, while the GPU was clocked at 1329MHz which from its max GPU clock of 2.1GHz is 63% lower.

Intel looks like it was testing the Arc Alchemist on an Intel Coffee Lake CPU, with an Intel Core i5-9600K processor, and not their latest 12th Gen Core "Alder Lake" CPU. But rather multiple generations old CPU and motherboard hardware are being used to test the very latest GPU.

Innosilicon Technology has just announced its new "high-performance GPU for servers" with the introduction of the Fenghua No.1 graphics card.

There's no detailed specifications of the Innosilicon Technology Fenghua No.1 GPU, but it will pack high-end GDDR6X memory, PCIe 4.0 support, and HDMI 2.1 connectivity with e-DisplayPort 1.4 thrown in. Fenghua No.1 having GDDR6X memory is a big deal as NVIDIA is the only one using it on their highest-end GPUs, while AMD uses regular GDDR6 memory.

We don't know how much GDDR6X memory the Fenghua No.1 GPU will have, nor what bandwidth it'll be cranking along with but we should expect it to be at 21Gbps. Innosilison Technology explains: "It is reported that "Fenghua No. 1" GPU is the industry's first rendering GPU that adopts the world's top GDDR6X high-bandwidth video memory technology and cutting-edge technologies such as Innolink chiplet".