Graphics Cards - Page 173

NVIDIA's new GeForce RTX 30 series graphics cards will be unveiled on September 1, with the introduction (and requirement on RTX 30 series Founders Edition cards) of a new 12-pin PCIe power adapter.

In a recent finding on Bilibili, the new 12-pin PCIe power connector was spotted -- this time, from Seasonic. But as you can see on the box, the adapter turns 2 x existing 8-pin PCIe power connectors into a single 12-pin PCIe power connector.

Seasonic is recommending that you have an 850W power supply or higher, to use this new 12-pin PCIe power connector. That's pretty crazy, as a GeForce RTX 2080 Ti only "requires" a 650W PSU.

If you had any question that the huge new 12-pin PCIe power connector for NVIDIA's next-gen Ampere-based GeForce RTX 30 series cards -- well, check this out:

Seasonic is preparing a new "NVIDIA 12-pin PCIe Molex Micro-Fit 3.0 Connector", which will convert your 2 x 8-pin PCIe power connectors into a single 12-pin PCIe power connector, to plug into NVIDIA's upcoming GeForce RTX 30 series graphics cards.

According to previous reports it will be NVIDIA's own GeForce RTX 30 series Founders Edition cards that will use the new 12-pin adapter, where AIB partners will be using 2-3 x 8-pin PCIe power connectors. Why? We don't know just yet.

We are getting some teases of pricing on NVIDIA's next-gen GeForce RTX 3000 series -- but I would be taking these with a grain of salt.

Here we have the big drop, NVIDIA will be reportedly pricing its GeForce RTX 3090 graphics card at $1399, the second one down is the GeForce RTX 3080 at $799, the third one in the list is the GeForce RTX 3070 at $599, while the GeForce RTX 3060 would reportedly cost $399.

Here's an easier break down for you:

Can you smell that? I can... it's the smell of fresh next-gen PCBs coming off of the factory lines, with GALAX now teasing its next-gen GeForce RTX 30 series graphics cards.

In some newly-released posters, the new China-focused marketing campaign teases 5 different custom Ampere-based GeForce RTX 30 series graphics cards. We have: Hall of Fame, Master, Boomstar, Metal Master, and Gamer.

As for the text on the posters, our friends at VideoCardz reports that (with machine learning, so this isn't perfect) that they say:

Well, it looks like NVIDIA's next-gen Ampere-based GeForce RTX 3090 is going to be a true new performance beast -- with leaked images teasing a gigantic triple-slot reference GeForce RTX 3090 graphics card.

I mean, just look at it -- it's absolutely huge. We don't know much else from this new leak apart from the physical size comparison between the Ampere-based GeForce RTX 3090 graphics card (which should have 24GB of ultra-fast GDDR6X memory at 1TB/sec) against the Turing-based GeForce RTX 2080.

The second image shows off the back of the card with that fan on the rear of the card, just like the previous leaks teased. Inside, the GeForce RTX 3090 should pack the Ampere-based GA102 GPU with at least 5248 CUDA cores and 24GB of next-gen GDDR6X memory.

Do you remember those leaks from all the way back in January 2020, where we were teased with specs on the purported GeForce RTX 3080 which saw it have up to 20GB of VRAM. Well, it seems those rumors were right -- but there's something strange going on, too.

But now we're receiving some more confirmation from leaker kopite7kimi, who has tweeted that the PG132-SKU20 is the 'RTX 3080 20G from AIC'. So, the custom GeForce RTX 3080 has 20GB of GDDR6X memory... but what about the GeForce RTX 3080 Founders Edition?

He was asked if the 20GB would be for "AIB models only" to which he replied with a simple "yes". This is an interesting turn of events -- NVIDIA will have the only 10GB models of the GeForce RTX 3080, while AIB partners will be 20GB models only.

NVIDIA is a few days closer to the reveal of its next-gen GeForce RTX 3000 series graphics cards, with the PG133A board going through RRA certification.

The new PG133A board was noticed by @kopite7kimi as the board design for NVIDIA's next-gen GA102 GPU-based graphics cards. The same board is used on different SKUs of NVIDIA's graphics cards, including the x080 and x060 series graphics cards.

Normally they have small board changes, where power stages and memory configurations are changed. But now the PG133A board was spotted in the Korean National Research Agency, which provides certification -- the last step before the product is released onto the market.

Intel is entering the GPU space in 2021 in a big way, with a detailed post on that here -- with Intel surprise announcing its new Xe-HPG, a new gaming-focused enthusiast graphics card arriving in 2021.

It seems TSMC -- or Taiwan Semiconductor Manufacturing Company, have secured the contract to make Intel's next-gen Xe-HPG graphics card on its 6nm process node. Intel's other Xe-based designs in the Xe-LP and Xe-HP will be made in-house by Intel, on their just-announced 10nm SuperFin process node.

Intel's new Xe-HPG graphic card is exciting, as the company teased it will have hardware ray tracing technology as well as super-fast GDDR6 memory. This instantly puts Intel's new Xe-HPG right at the (at least technologically) same level as NVIDIA and AMD's current, and soon-to-be-released next-gen GPUs.

Microsoft has detailed a lot of the innards of its next-gen Xbox Series X console, codenamed Project Scarlett, at the Hot Chips 2020 virtual event. The semi-custom APU designed by AMD and Microsoft and made by TSMC on the N7 Enhanced node is a technological marvel. It's a truly exciting chip to power the next-gen Xbox Series X console.

This is it -- the delicious Xbox Series X 'Project Scarlett' SoC. The GPU is so massive that it takes up 47.5% of the entire die, consuming a huge chunk of those 15.3 billion transistors.

We have here the N7 Enhanced node from TSMC packing 15.3 billion transistors into a 360.4mm2 chip, with the chip co-developed between Microsoft and AMD and then made by TSMC. It's nice to see that 'Project Scarlett' nice and raw there on the chip -- which is super small.

It has barely been 72 hours since I wrote about NVIDIA's next-gen GeForce RTX 3090 graphics card packing some serious heat: 24GB of ultra-fast next-gen GDDR6X memory at a rumored price of $1999.

But new rumors have the GeForce RTX 3090 priced at just $1399.

NVIDIA pricing its new GeForce RTX 3090 at $1399 would be an incredible feat, a $200 price hike on the Turing-based current flagship GeForce RTX 2080 Ti that debuted in Founders Edition form at $1199 back in 2018. It packed 11GB of GDDR6 memory, but the new RTX 3090 is said to rock a huge 24GB of GDDR6X memory from Micron.

Alright, switching gears between the flagship NVIDIA GeForce RTX 3090 graphics card and its tease of 24GB of GDDR6X memory that Micron has clocked at a ridiculous 21Gbps -- we have the GeForce RTX 3080 with some juicy specs.

According to a new Device ID listing confirmed by leaker _rogame on Twitter, we have some preliminary specs on NVIDIA's next-gen Ampere-based GeForce RTX 3080. The new GeForce RTX 3080 will have its GPU clocks at a limit of 2.1GHz -- which is identical to the Turing GPUs which have a 2.1GHz GPU clock ceiling for the most part.

NVIDIA's new GeForce RTX 3080 should have 10GB of GDDR6X which according to these leaked specs, will be clocked at 19Gbps. Micron is making the memory for this batch of Ampere-based GeForce RTX 3000 series graphics cards with Micron providing NVIDIA with some blistering fast 21Gbps chips for its flagship GeForce RTX 3090 (which I really, badly want to be called the RTX 3090 ULTRA or RTX 3090 ULTIMATE).

We are getting more and more information on the flagship GeForce RTX 3090 (I hope it's called the GeForce RTX 3090 ULTRA or ULTIMATE) where now we're hearing about price.

According to new rumors we're looking at a price of $1999 for NVIDIA's flagship GeForce RTX 3090 (which is meant to be 40-50% faster than the GeForce RTX 2080 Ti), where if it does indeed feature 24GB of next-gen GDDR6X memory at an absolute blistering 21Gbps from Micron then the price seems about right.

NVIDIA priced its Turing-based TITAN RTX graphics card with 24GB of GDDR6 memory at $2499, so it would make sense to offer a faster card for less money this time around. It will give NVIDIA more wiggle room between the GeForce RTX 3080 and GeForce RTX 3090 pricing, without completely cutting the GeForce RTX 2080 Ti off at the knees.

We first heard about NVIDIA reportedly using a new 12-pin PCIe power connector on its GeForce RTX 30 series graphics cards, and now we have more confirmation.

I've been talking to industry sources of mine who have said that NVIDIA will be using the 12-pin PCIe power connector on its new GeForce RTX 30 series Founders Edition graphic cards, while AIB partners on the other hand will not be using the 12-pin connector, and rather will have multiple 8-pin PCIe power connectors.

This lines up with the purported GeForce RTX 3090 PCB photo that was leaked yesterday, as it was reportedly a COLORFUL iGame GeForce RTX 3090 Vulcan-X with 3 x 8-pin PCIe power connectors and no 12-pin PCIe power connector in sight. That is because it's a custom card -- which I exclusively reported will launch alongside, or very close to the launch of the GeForce RTX 30 series graphics cards.

Micron just finished outing NVIDIA's not-yet-announced GeForce RTX 3090 graphics card, which we now know packs 12GB of ultra-fast GDDR6X memory at 21Gbps.

But they weren't finished with their surprise announcement of the new GDDR6X standard, where Micron has now just announced the development of its next generation memory standard: HBMnext. The new HBMnext memory is most likely what HBM3 was meant to be, but possibly with a new name.

Our friends at VideoCardz report: "Micron expects to produce JEDEC-compliant HBMnext memory in 4-stack 8Gb and 8-stack 16Gb densities. The manufacturer also revealed that the data rate is expected at 3.2 Gb/s. For comparison, the NVIDIA A100 accelerator based on HBM2e technology has a bandwidth of 2.4 Gbps, while AMD Radeon VII based on HBM2 features 2.0 Gbps data rates".

It seems today is another huge day for NVIDIA leaks on the purported new Ampere-based flagship GeForce RTX 3090, with Micron confirming the new card will use its next-gen GDDR6X memory.

Our good friends at VideoCardz have spotted a new technology brief, Micron confirmed it had worked with NVIDIA on the new GeForce RTX 3090 graphics card -- outing it right there in the documents itself. The company says it will have 12GB of GDDR6X on a 384-bit memory interface on the GeForce RTX 3090.

Micron explains: "Micron GDDR6X is the world's fastest graphics memory, delivering legendary performance to help driveray tracing, shadow mapping and silky-smooth animation for an immersive PC gaming experience".

Heavy breathing intensifies.

This could be our first look at NVIDIA's next-gen GeForce RTX 3090 and its beefy PCB, thanks to a new leak over on the NVIDIA subreddit.

NVIDIA's purported GeForce RTX 3090 graphics card and its leaked PCB is coming from a non-reference board from COLORFUL, which would eventually form into the flagship iGame GeForce RTX 3090 Vulcan-X graphics card.

It seems with each passing day the hype train for NVIDIA's next-gen Ampere GeForce RTX series graphics cards grow, with the Ultimate Countdown... well... counting down until the big special event on September 1.

The latest news on NVIDIA's next-gen GeForce RTX series is that the GA102 GPU board design will pack some delicious power delivery, seeing enthusiasts eyebrows raise (well, at least just one of them). NVIDIA has been slowly providing better and better power delivery designs with its high-end Founders Edition cards, but this is taking it to the next level.

NVIDIA's highest-end Ampere GA102 GPU board design should be the one expected in the GeForce RTX 3080 Ti or GeForce RTX 3090, or GeForce RTX 3090 ULTRA or ULTIMATE -- whatever-the-hell NVIDIA end up calling it.

Intel is not so why with its Xe GPU architecture, where it hosted a virtual Architecture Day detailing a mountain of information on its technologies -- including the huge Arctic Sound GPU.

The new Arctic Sound GPU is an MCM-based design with 4 tiles and a huge 16,384 cores -- a multi-chip module with multiple tiles that add huge amounts of performance. The card was benchmarked at 1.3GHz with its 16,384 cores across the 4-tile Arctic Sound Xe GPU with an insane 42 TFLOPs of FP32 compute performance with near-perfect scaling (3.993x better).

Raja Koduri, the chief architect and senior vice president of Intel's discrete graphics division explained that the company is working on a 1-tile version with 512 EUs and 4096 cores, a 2-tile version with 8192 cores, and a 4-tile version that was teased at Architecture Day with a blistering 16,384 cores.

Intel hosted its virtual Architecture Day to the media across the world, with a gigantic debriefing on what Intel has been up to across virtually its entire company -- but the most exciting thing (at least to me) was the introduction of the new Xe-HPG gaming-optimized architecture.

Intel's new Xe-HPG is for enthusiast gamers packing hardware ray tracing support, packing GDDR6 memory. To compare against the datacenter-focused Xe, with the Xe-HP -- which is powered by HBM memory. Intel stepping into the ray tracing ring with NVIDIA, and soon AMD with its RDNA 2 architecture is a big move.

The new Intel Xe-HPG is expected to ship in 2021, where I'm sure we'll hear more details on it early in the New Year. Raja Koduri, the chief architect and senior vice president of Intel's discrete graphics division, explained: "our enthusiast and gamer friends, we heard your requests for Xe for enthusiast gaming. We added a fourth microarchitecture to the Xe family: Xe-HPG optimized for gaming, with many new graphics features including ray tracing support. We expect to ship this microarchitecture in 2021 and I can't wait to get my hands-on this GPU!"

NVIDIA has just announced a new "special event broadcast" for September 1, which will be hosted by NVIDIA CEO and founder Jensen Huang -- where we should be introduced to the new next-gen GeForce graphics cards.

The company has a new post on their official GeForce website, leading up into the Ultimate Countdown. The post is filled with memories of the late 90s and the golden age of PC gaming, which centered around low-res (but high-res at the time) monitors, the original GeForce 256, and dial up internet.

I was 16 at the time and couldn't be more obsessed about technology and games, which is why I went into 10 years of working with and selling computer parts and custom gaming PCs and then now -- TweakTown, reviewing graphics cards and playing games.