Video Cards & GPUs - Page 361

Like it was planned to be teased after NVIDIA's huge unveil of their next-gen Volta GPU on their new Tesla V100 graphics card, news appears through recent Linux drivers of a purported dual-GPU based on Radeon RX Vega. The new info points to specific code:

These codes are new Vega 10 IDs, so we should expect a larger family of Radeon RX Vega graphics cards - with the number 7 floating around in previous reports. The specific code we're looking at here points to a specific PLX chip, which is an ASIC board that splits PCIe lanes to bridge multiple components, like dual GPUs.

There's not much else here, but a dual-GPU Radeon RX Vega is something AMD will need to beat the new Tesla V100 graphics card, that's for sure. A dual-Vega graphics card could also hurt the GeForce GTX 1080 Ti... so we'll have to wait and see.

GTC 2017 - NVIDIA CEO Jensen Huang has come out with some fighting words for its main competitor AMD, and their upcoming Radeon RX Vega graphics card, with the Huang saying he's confident in NVIDIA's position in the gaming GPU market.

NVIDIA has no competition in the high-end market against their GeForce GTX 1070, GTX 1080, GTX 1080 @ 11Gbps, GTX 1080 Ti, Titan X(P) and the new TITAN Xp. AMD has their upcoming Radeon RX Vega graphics card family, but we're still weeks away from its release.

After NVIDIA's unveiling of their new Volta-based Tesla V100 graphics card, Blayne Curtis from Barclays Capital Inc. asked Huang: "I was just wondering if you can just talk about the competitive landscape looking back at the last refresh. And then looking forward into the back half of this year, I think your competitors have a new platform. I'm just curious as to your thoughts as to how the share worked out on the previous refresh and then the competitiveness into the second half of this year".

GTC 2017 - NVIDIA has officially unveiled its new Volta-based Tesla V100 graphics card, the most advanced piece of silicon in the world - on the 12nm node, with 16GB of HBM2 @ 900GB/sec, and so much more - but the AI research systems have also been announced, topping out at $149,000.

NVIDIA's new DGX-1 with Tesla V100 has 960 Tensor TFLOPs of performance, 8 x Tesla V100 graphics cards, NVLink Hybrid Cube technology, and is the equivalent of 400 servers in a box. NVIDIA says that it's capable of pushing down the required time on datasets from 8 days on TITAN X, to just 8 hours on DGX-1 with Tesla V100.

GTC 2017 - NVIDIA removed the wraps off of their next-gen Volta GPU architecture today at GTC 2017, with the new Tesla V100 graphics card and DGX-1 with Tesla V100 system that costs $149,000.

We know that there are over 21 billion transistors on the GV100 GPU, offering 15 TFLOPs of single precision compute performance backed up by 7.5 TFLOPs of double precision compute. But there are two variations of Tesla V100 graphics cards: single-slot, and dual-slot. The single-slot Tesla V100 has a 150W TDP, while the dual-slot Tesla V100 ramps up the TDP to 300W.

NVIDIA hasn't confirmed what the different in performance is going to be for the 150W variant of Tesla V100, and that's a very interesting question. Performance wise, we're expecting a pretty huge 40% performance per watt advantage over the previous Pascal-based Tesla P100, on the new 12nm FinFET process.

GTC 2017 - NVIDIA has unveiled its monsterous new Tesla V100 professional graphics card, the first with their next-gen Volta GPU architecture - and NVIDIA's second graphics card with HBM2 technology.

The new Tesla V100 packs 16GB of ridiculously fast HBM2 on a 4096-bit memory bus that provides a huge 900GB/sec of memory bandwidth. Tesla V100 is capable of a huge 15 TFLOPs of single precision (FP32) performance, while it packs 7.5 TFLOPs of double precision (FP64) performance - enough for the largest of datasets in datacenter/AI/deep learning workloads.

Inside, Tesla V100 rocks the company's next-gen Volta GPU architecture with a huge 5120 CUDA cores at 1455MHz boost clock, on the fresh new 12nm manufacturing process - something we reported on a couple of months ago now. If we compare this to the Pascal-based Tesla P100 that was made on the 16nm FinFET prrocess with 16GB of HBM2 at 720GB/sec, and only 3584 CUDA cores in comparison. NVIDIA's new Tesla V100 is an absolute beast.

GTC 2017 - NVIDIA is preparing to unleash its next-gen Volta GPU architecture at the GPU Technology Conference in San Jose, CA - with Jensen Huang, CEO and founder of NVIDIA saying that the company is building inventory for a new product announcement during his keynote in the morning.

At GTC 2016 the company unveiled its Pascal-based Tesla P100 professional graphics card powered by HBM2 technology, so I'm expecting a Volta GPU architecture announcement alongside the unveiling of the Tesla V100. I'm sure we'll see even more HBM2 on the purported Tesla V100 (The 'P' in Tesla P100 stands for Pascal, so V100 would be Volta, right?) and we could even see a GDDR6-based variant, but I think that'll be the GeForce GTX 20 series announcement for later this year.

Until then, what do you think we'll see unveiled at GTC 2017? Volta in professional form? A new TITAN Xv graphics card on the new Volta GPU architecture, or something completely different?

GTC 2017 - SK Hynix is displaying its next-gen GDDR6 technology at NVIDIA's GPU Technology Conference, their new memory standard that will power NVIDIA's next-gen Volta GPU architecture, with Volta-based graphics cards set to arrive in 2018.

NVIDIA is currently using various different RAM standards on their graphics cards, with their Tesla P100 graphics card using HBM2 (yes, before AMD's upcoming Radeon RX Vega), GDDR5X at 11Gbps on the TITAN Xp, Titan X(P), GeForce GTX 1080 Ti, and the new GTX 1080 11Gbps model, and then GDDR5 for the rest of its graphics cards.

SK Hynix displayed the differences between GDDR5 and GDDR6, with some huge increases in speeds and bandwidth - and up to 16Gb chips, while offering less power consumption.

We've been reporting on AMD's next-gen Radeon RX Vega graphics card family like mad, with our exclusive report that AMD will have less than 16,000 units available in the months post-launch. Now we're hearing rumbles of what Computex will hold for AMD's return to the high-end graphics card market.

Computex 2017 kicks off on May 30, which is only a couple of weeks away - and we'll be there. VideoCardz is reporting that PowerColor have been sending out invitations to the media for meetings during Computex, where they will be showing off "new technologies and other latest products", which could be a tease towards Radeon RX Vega.

I've personally heard from my own industry sources that Radeon RX Vega might show up, depending on whether AMD releases it before Computex - if they do, I've heard that AIB partners will indeed have custom Radeon RX Vega graphics cards on show. Computex 2017 is only a few more weeks away, so we'll have to wait and see what happens.

AMD will be releasing its next-gen Radeon RX Vega family of graphics cards in the next few weeks, and now I've had an exclusive industry source who has told me that AMD will have only a handful of Radeon RX Vega graphics cards at launch.

I've been told that there will be less than 16,000 cards that will ship in the first few months after it launches, something that will come down to the HBM2 used on the card. HBM2 is in extremely limited supply, and is expensive to use - and since there's not enough, that scarcity is driving up the production costs of the card - and will see AMD only having 16,000 cards or so in the months post-launch.

If this is true, AMD could be in a very rough spot with Radeon RX Vega - especially if it was to deliver on performance. There are hundreds of thousands of thirsty Radeon fans that want a next-gen graphics card, and the hype train for Radeon RX Vega is simultaneously withering out - and burning hotter than the sun. Personally, I want AMD to hit a home run with Radeon RX Vega - but at the same time, NVIDIA has completely secured the high-end market with the GTX 1080 and GTX 1080 Ti.

There have been a few leaks on AMD's upcoming Radeon RX Vega graphics card, and now we're seeing the 687F:C1 device ID once again, with some details on its HBM2 speeds and GPU clocks.

First off, I've been reporting for a while now that there will be a family of Radeon RX Vega graphics cards, with a new Vega 10 prototype being used and spotted on 3DMark's Fire Strike database. The card in question had its Vega 10 GPU clocked at 1200MHz, while its 8GB of HBM2 was clocked at 700MHz.

The Radeon RX Vega prototype was benchmarked with AMD's Ryzen 7 1800X processor, but how was the result of its run in Fire Strike?