Video Cards & GPUs - Page 376

EVGA had some issues with its original GTX 1080 graphics card, but the company is looking to announce new ICX-based GTX 1080 graphics cards in both the GTX 1070 and GTX 1080 parts.

EVGA is expected to refresh its GTX 10 series cards with improved thermals, with the "optional" thermal pads now the norm, with a new and improved PCB on the upcoming SuperClocked 2 and FTW2 models.

Jacob Freeman took to Twitter to tease: "I C something eXciting announcing this week", and I don't think he could've been any more clear.

AMD is gearing up for the betas of both Conan Exiles and Tom Clancy's Ghost Recon Wildlands, with the release of the new Crimson ReLive Edition 17.1.2 WHQL drivers. Get the new Crimson ReLive Edition 17.1.2 drivers right here.

The new drivers also have some fixes included, where "Changing memory clocks with Radeon WattMan with more than one display connected may cause memory clocks to intermittently become stuck at their minimum setting or switch between min and max ranges intermittently causing display flickering" is fixed.

Radeon Software Crimson ReLive Edition 17.1.2 WHQL Highlights

NVIDIA has released its new GP100-based Quadro product, joining the Tesla P100 in the full-fat Pascal goodness for professional graphics. The new Quadro GP100 features 16GB of super-fast HBM2, and some huge performance numbers to match.

We're looking at FP64 calculations of 5.3TFLOPs, while FP32 performance hits 10.6TFLOPs, and FP16 with 20.7TFLOPs. The GP100 itself features 16GB of HBM2, 4 x DP 1.4 and a single DVI-D. NVLink makes an appearance, with the flashy new interconnect powering two of NVIDIA's new Quadro GP100 graphics cards.

Throwing two Quadro GP100s together over NVLink offers 32GB of HBM2 pooled, for serious workstation power. NVIDIA will be pushing out their Quadro GP100 graphics cards to HP and Dell customers for workstation configurations next month.

I've just written about SK Hynix's new dates for HBM2 deployment, which will affect AMD's Vega GPU roll out a little, but we're still getting Vega graphics cards in Q2 2017.

AMD confirmed the news during their latest financial report, where they said: "AMD introduced preliminary details of its forthcoming Vega GPU architecture designed to address the most data- and visually-intensive next-generation workloads. Key architecture advancements include a differentiated memory subsystem, next-generation geometry pipeline, new compute engine, and a new pixel engine. GPU products based on the Vega architecture are expected to ship in the second quarter of 2017".

As for the new Vega graphics cards with HBM2, we should expect 8GB of HBM2 with just under 410GB/sec bandwidth - and a release later in the year with the faster 512GB/sec HBM2.

AMD is expected to launch its Vega graphics cards in the next couple of months, with an unveiling in May and a launch in June to be expected - but now SK Hynix have updated their memory product catalog, giving us a better look at the bandwidth numbers around HBM2.

SK Hynix updated their graphics databook, showing off HBM2 for Q4 2016 with a 4-stack 4GB HBM2 product with two different speeds: 2Gbps and 1.6Gbps, with 256GB/sec bandwidth and 204GB/sec bandwidth, respectively. This has now changed, with a shift in availability.

The company is now listing the slower 1.6Gbps (204.8GB/sec) for Q1 2017, while the faster 2Gbps (256GB/sec) HBM2 now stuck with a 'TBD' for availability. This could lead AMD to launch a slower HBM2-based card in Q2, and a faster Vega graphics card later in the year.

NVIDIA launched its super successful GeForce GTX 1080 and GTX 1070 graphics cards back in May, announcing they would be building their own 'Founders Edition' variants at the time. I remember thinking when NVIDIA announced the FE version of the cards, that it would begin hurting AIB partners - turning to YouTuber 'JayzTwoCents' and meeting his eyes with the same conclusion.

Taiwanese tech site DigiTimes has reported that due to 'increasing sales and profits earned by NVIDIA from PC-use graphics cards has prompted the company to actively promote sales of the Founders Edition graphics cards'. In other words, Founders Edition cards will be eating into the orders for the GTX 10 series partners' supply, resulting in price hikes and loss of sales for companies like ASUS, MSI, ZOTAC, GIGABYTE, and others.

Throughout 2016, the six largest graphics card suppliers in ASUS, GIGABYTE, MSI, Palit, ZOTAC, and Colorful - they made $159.4 million in profits. This saw NVIDIA interested in the growing graphics card market, and wanting to take a slice of that pie - even though the pieces of that pie, are selling NVIDIA pie - and it makes a share of those profits as the GPUs and resulting technology and trademarks are owned by NVIDIA.

ASUS, GIGABYTE, and MSI will be increasing retail pricing for its motherboards and graphics cards in China starting next month, right after the 2017 Lunar New Year holidays, according to sources of DigiTimes.

They reported: "The sources expect the motherboard players to raise their product prices by around 5%. Since demand for motherboards has not been strong for the past few years, maintaining sales with the price increases will be a major task for the players".

The same sources said that Intel's new Kaby Lake-capable 200 series motherboards and NVIDIA's latest GeForce GTX 10 series graphics cards will be increased in price in China. DigiTimes continued: "Currently, NVIDIA's GTX 1080 and 1070 GPUs are in shortage due to issues at an upstream foundry partner. Although NVIDIA tried to shift its orders to Taiwan Semiconductor Manufacturing Company (TSMC), TSMC's fully loaded capacity is expected to prolong the shortages. As a result, NVIDIA has been adjusting its GPU supply to graphics card vendors recently and this has caused graphics card prices to rise".

I still remember when AMD launched the Radeon Pro Duo nearly a year ago now, but it was aimed at the prosumer market, and priced rightfully so. Until now.

AMD and its various AIB partners have dropped the price on the Radeon Pro Duo from nearly $1500, to just $799 - not bad considering it's the fastest graphics card from Radeon Technologies Group. The Radeon Pro Duo rocks dual Fiji GPUs with 4GB of HBM1 each, for a total of 8GB of HBM1.

Each of the Fiji GPUs has 4096 stream processors (8196 total) - making the Radeon Pro Duo feature the highest GPU cores of any graphics card, ever. We have 64 ROPs and 256 TMUs per GPU, with each Fiji GPU running at 1000MHz.

GIGABYTE kicked some serious graphics card ass in 2016 with their solid Xtreme Gaming series, but the company is shifting gears into the Aorus brand.

The first Aorus-based graphics card will be the GeForce GTX 1080 Xtreme Edition, slightly modified from the Xtreme Gaming series card. The Aorus GeForce GTX 1080 Xtreme Edition features clocks of 1784/1936MHz for base and boost, respectively.

The usual 8GB of GDDR5X is here at the standard 10GHz, with 2 x 8-pin PCIe power connectors that will ensure the Aorus GTX 1080 Xtreme Edition will smash past 2GHz on a GPU overclock. But now we'll talk about how the Aorus version is different to the Xtreme Gaming version.

Now that I'm back home and at my desktop, I'm sinking my teeth into everything GPU related that has happened over the last couple of weeks. First, we've had the official unveiling of the Vega GPU architecture that will be launching in May. Second, leaked decks of AMD's upcoming GPUs have appeared online, teasing Vega 10 and its dual-GPU brother, Vega 20 - and even Navi 10, oh and the dual-GPU version of Navi 10.

If you thought Vega 10 was going to be it, you'll have to wait for the dual-GPU version expected in late-2017 with 1.5x the performance. Right now, Vega 10 will consume around 225W according to the leaked slides and my industry sources. We might see this reduced if AMD keeps its reference card to under 200W, allowing AIB partner cards to clock the hell out of the Vega 10 GPU and leap up to 250-300W.

The dual-Vega 10 graphics card will most likely arrive as a reference only card from AMD, featuring a 300W TDP and dual Vega 10 GPUs. The clocks will be reduced as all dual-GPU cards experience, so we should see it drop from 1465MHz on the estimated GPU clocks of Vega 10, to 1000-1200MHz. This will let the card scale well, to around 1.5x the performance of a single Vega 10 graphics card, while hitting the 300W ceiling on power consumption. I'm expecting a heftier cooler, and 16GB of HBM2 on the dual Vega 10 graphics card.