Graphics Cards - Page 204

There is another massive crypto mining boom that has rolled through courtesy of the massive spotlight being blasted onto cryptocurrency (with all the alts bleeding right now), and moreso into crypto mining.

Most crypto miners will look at the Radeon RX 570/RX 580 or GeForce GTX 1060/1070/1070 Ti, but with most of those being scooped up by miners, things aren't looking good. The prices of all GTX 10 series cards have risen, with the GTX 1070 Ti going for around $1000 on Amazon while the higher-end GTX 1080 Ti is $1050 at its cheapest, and up to $1600 for some of the custom GTX 1080 Ti cards from EVGA. Unbelievable.

This means you can get yourself NVIDIA's swanky TITAN Xp Star Wars COLLECTORS EDITION graphics cards for $1500, $100 cheaper than a custom GTX 1080 Ti that it most easily beats. Even for miners shelling out cards at this price, you're better off with the TITAN Xp which pushes 42MH/s mining Ethereum, compared to 35-38MH/s from the GTX 1080 Ti.

After going on a Twitter blast a few days ago with VideoCardz where I said I heard early rumbles of Intel working on its own next-gen discrete GPU, rumors are now floating out that Intel is indeed working on their own GPU.

Ashraf Essa from The Motley Fool is reporting that Intel is working on two new GPU solutions with codenames of Arctic Sound and Jupiter Sound, with the former being a 12th generation discrete GPU and the latter a 13th generation discrete GPU. Intel's Gen 9 solution is found on Kaby Lake while the Gen 9.5 solution is on Coffee Lake, while Gen 10 will come with Cannonlake soon.

We should expect them sometime after 2020, but with RTG boss Raja Koduri now on Intel's team, it makes sense. NVIDIA is fighting the good fight with their GeForce line up and virtually no competitor in the GTX 1080 and above space, but Intel could soon enter as a big competitor with their own Radeon-powered discrete GPU that does have me wondering: what's the future of Radeon under AMD if Intel can do better with Radeon GPU tech than AMD can.

NVIDIA is reportedly preparing some new GeForce GTX 1050/GTX 1050 Ti Max-Q products, which would really hit the mainstream gaming laptop market and crush everything AMD is hoping to do in the space with their recently-revealed Radeon RX Vega M GL product.

If you remember, Intel unveiled their new CPUs with Radeon RX Vega M technology inside, so NVIDIA shifting quickly with their GTX 1050 Ti Max-Q makes total sense. NotebookCheck have predicted GPU clock speeds of the purported GTX 1050 Ti Max-Q designs, where they think we'll see GPU boost clocks of 1417MHz on the GTX 1050 Ti Max-Q compared to the 1392MHz on the desktop GTX 1050 Ti, 1620MHz on the full GTX 1050 Ti mobile and 1455MHz of the GTX 1050 on desktop.

NVIDIA has one of the best single graphics cards on the market with the Tesla V100, a card that costs a whopping $8000 and isn't for gamers or even most people on the market. It's a card destined for workstations and servers, for AI and deep learning workloads - and strictly not for mining.

That however doesn't stop people from testing out these graphics cards for mining, with BuriedONE Cryptomining putting NVIDIA's Tesla V100 to work on various crypto mining adventures, with Ethereum mining hitting a 94MH/s. Considering that an overclocked TITAN Xp can achieve somewhere around 40-42MH/s and an overclocked Radeon RX Vega 64 can do anywhere between 38-42MH/s, this is a huge achievement for the Tesla V100.

NVIDIA's super-fast Tesla V100 rocks 16GB of HBM2 that has memory bandwidth of a truly next level 900GB/sec, up from the 547GB/sec available on the TITAN Xp, which costs $1200 in comparison. AMD trails behind with 483GB/sec of bandwidth with its 8GB of HBM2 on the Radeon RX Vega 64.

CES 2018 - AMD hasn't said much more about what's going on with the desktop Radeon GPU side of things since the departure of RTG boss Raja Koduri, but Scott Wasson (the owner and ex-journo of The Tech Report) has worked for AMD for over a year now and was at CES 2018 where he had some things to say about Radeon, and Vega.

During a video interview with PCGamesHardware.de, Wasson said that Vega 10 is capable of being used in a gaming notebook, and seemed to tease the news during CES. Wasson said: "I can't pre-announce products for our partners. It is possible to take a Vega 10 GPU and put it into a notebook. So we'll have to see".

If we take into consideration just how damn hot Radeon RX Vega 56 and 64 get in reference form, I can't see how in the hell laptop manufacturers will be able to use a full Vega 10 + 8GB HBM2 if they require 250-400W of power. Even if its heavily throttled, it would lose to a GTX 1060 or GTX 1070-powered gaming notebook.

Samsung has just announced its latest and greatest advancements in HBM2 technology at CES, something the company is calling "Aquabolt". This new HBM2 is much faster than the first spins of HBM2, where we're looking at bandwidth of an insane 2.4Gbps and in 8-Hi height (8GB) stacks which should see up to 32GB on next-gen graphics cards.

The 8-Hi stacks might sounds weird, but when there's 4 of them on a graphics card we're looking at 32GB of HBM2. As for bandwidth, we're looking at around 300MBps per pin, which on a 1024-bit memory bus should provide around 307GB/sec per package, times 4 bringing us to a crazy 1.2TB/sec of memory bandwidth.

Jaesoo Han, executive vice president, Memory Sales & Marketing team at Samsung Electronics explains: "With our production of the first 2.4Gbps 8GB HBM2, we are further strengthening our technology leadership and market competitiveness. We will continue to reinforce our command of the DRAM market by assuring a stable supply of HBM2 worldwide, in accordance with the timing of anticipated next-generation system launches by our customers".

With all of the hoopla surrounding the Spectre and Meltdown security bugs found in consumer CPUs, setting Intel pretty much on fire, how are the other companies fairing? Well, NVIDIA is doing fine.

NVIDIA's revised security bulletin has provided some insight, with NVIDIA CEO and co-founder Jen-Hsun Huang saying: "Our GPUs are immune, they're not affected by these security issues. What we did is we released driver updates to patch the CPU security vulnerability. We are patching the CPU vulnerability the same way that Amazon , the same way that SAP, the same way that Microsoft, etc are patching, because we have software as well".

Huang added: "I am absolutely certain that our GPU is not affected".

GALAX has surprised with a quick tease of its upcoming GeForce GTX 1070 Ti HOF graphics card, which will include its unique one-click button to increase the OC on the card.

GALAX will be deploying its chunky-but-cool triple-fan RGB cooling solution, with 8+8-pin PCIe power connectors (down from the 8+8+8-pin PCIe connector rig on the GTX 1080 Ti HOF variant).

The company even teased the card in SLI with Hyper Boost OC enabled, which is very impressive - and that SLI bridge, wow.

CES 2018 - NVIDIA might have announced its new 65-inch 4K 120Hz G-Sync TVs, specifically with the new ASUS ROG Swift PG65, but the company is also quick to release a new GeForce 390.65 drivers that includes a security patch for Spectre vulnerability.

NVIDIA's new GeForce 390.65 drivers also include support for the Freestyle beta, which lets gamers apply post-processing filters to your game, while you play. This can be done from an in-game overlay, letting you change the entire style of the game with adjustments to color and saturation and post-process filters.

At launch, NVIDIA has 15 filters with 38 different settings (with some filters packing more than one setting). At launch, NVIDIA is including:

It looks like we are to expect graphics card prises to rise in the next few months, with these inflated prices lasting throughout most of 2018, something that is being amplified by cryptocurrency mining.

According to DigiTimes, graphics card vendors are expected to increase the price of mid/high-end graphics cards by $5-$20 in the "near future to reflect increasing product costs caused by the tight supply of memory devices and GPUs". DigiTimes continued, saying: "Since demand for graphics cards from the cryptocurrency mining segment has not waned as quickly as expected, GPU supply remains tight in the first quarter of 2018".

The site adds that companies like ASUS, GIGABYTE, and MSI all "benefited from the cryptocurrency mining demand and their shipments dropped only mildly in the fourth quarter of 2017", adding: "Meanwhile, graphics card upgrade demand has been picking up in the gaming sector thanks to the recent release of the blockbuster game PlayerUnknown's Battlegrounds. Demand has been especially strong from China due to the game's strong popularity".

NVIDIA has just made some large changes to its GeForce EULA, where the company has now prohibited the use of their GeForce GPUs in datacenter deployment, except for blockchain processing.

This shift will force datacenters into the arms of Quadro and Tesla products from NVIDIA which are much more expensive, instead of chewing up the precious supply of gamer-oriented GeForce GPUs. An extract from the EULA explains it perfectly: "No Datacenter Deployment. The SOFTWARE is not licensed for datacenter deployment, except that blockchain processing in a datacenter is permitted".

Now for the tricky part: NVIDIA's updates to their GeForce EULA will see anyone using GeForce GPUs for anything but gaming and "blockchain processing" breaking the law, and would open them up for potential lawsuits.

NVIDIA's new GeForce GTX 1060 5GB variant was exposed in the last 24 hours, with the new 5GB variant of the GTX 1060 destined for the Chinese internet cafe market. But now, we have a quick look at GIGABYTE's model in the GTX 1060 5GB WindForce OC.

GIGABYTE has confirmed that their new card will have GPU clocks of 1582/1797MHz in OC mode, while Gaming Mode will have the GPU clocked at 1556/1771MHz. The 5GB of GDDR5 will be clocked at 8008MHz on the same 160-bit memory bus that will provide 160GB/sec of memory bandwidth down from the 192GB/sec of memory bandwidth on the 6GB model with its 192-bit memory bus.

GIGABYTE will use their dual-fan WindForce 2X cooling solution that rocks a full cover backplate, while display wise we have 2 x DVI, 1 x HDMI and 1 x DP output.

NVIDIA might be enjoying the holidays a little too much, with news breaking that the company is preparing a new variant of one of their most popular GTX 10 series cards.

News has broken that NVIDIA is preparing a new GeForce GTX 1060, something that will be using the full GP106 with 1280 CUDA cores, but a weirder 5GB of GDDR5 on a smaller 160-bit memory bus (compared to the 192-bit memory bus on the 3GB and 6GB versions of the GTX 1060).

This new GTX 1060 5GB model would be an exclusive model for the Chinese market, where internet cafes are continuing to see major success. These cards would be sold in a very plain way, with generic packaging and most likely no marketing push behind it.

ZOTAC is making moves into the external GPU enclosure business, with the announcement of two new eGPU products: Amp Box, and Amp Box Mini.

Both of ZOTAC's new external GPU enclosures are smaller than competing external GPU enclosures, making them perfect for their own GeForce GTX 1080 Ti Mini graphics card.

Starting with the normal Amp Box which measures in at 5.8 x 10.1 x 10.7 inches (15 x 26 x 27 cm), while the smaller Amp Box Mini is 3.9 x 7.2 x 9.1 inches (10 x 18 x 28 cm). Both the normal Amp Box and Amp Box Mini can house dual-slot graphics cards at up to 7.8 inches (20nm) in length.

I'm waiting with bated breath for next-gen graphics cards and their associated technologies, with one of the cornerstones of 2018's graphics cards to be GDDR6 memory.

Micron has announced today they've finalized design and internal qualification of their upcoming 12Gbps and 14Gbps GDDR6 chips, meaning they are now laser-focused on mass production of this memory while developing even faster GDDR6 that should reach the peaks of 16Gbps.

GDDR5X is the best NVIDIA uses on the consumer GeForce side of things, with speeds of up to 11Gbps offered by the GTX 1080 Ti and TITAN Xp, while they can be overclocked to 12Gbps or more depending on the card, voltages, and cooling. GDDR6 will start off at 12Gbps, ride into 14Gbps, and beyond.

COLORFUL has just announced two new graphics cards in collaboration with professional eSports team RNG, with the first of the two being a GTX 1060 while a higher-end GTX 1080 Ti is offered.

COLORFUL's new iGame GTX 1060 RNG Edition rocks an overlocked GP106-400 GPU with 1620/1847MHz on base/boost, respectively. It has 6GB of GDDR5 RAM at 8Gbps, untouched from its stock frequency.

The higher-end iGAME GTX 1080 Ti RNG Edition rocks an overclocked GP102-350 GPU with base and boost clocks of 1620MHz and 1733MHz, respectively. It rocks the usual 11GB of GDDR5X and is clocked at its stock 11Gbps frequency. Both cards rock COLORFUL's impressive SWORIZER cooler that packs COLORFUL's awesome iGame Status Monitor LCD.

Samsung has announced the smallest 8-gigabit DDR4 RAM chips ever made, with its fresh second-gen 10nm manufacturing technology powering the new chips.

These new DDR4 chips are 15% more power efficient, and are around 10% faster than the previous-generation that was launched only 20 months ago now. Productivity is up 30% on the new DDR4 RAM, meaning Samsung can make more chips, with yields improved above the last-gen DDR4.

Now that Samsung has got its second-gen 10nm tech working beautifully, the South Korean everything giant is working on "accelerating its plans for much faster introductions of next-generation DRAM chips and systems, including DDR5, HBM3, LPDDR5 and GDDR6, for use in enterprise servers, mobile devices, supercomputers, HPC systems and high-speed graphics cards".

AMD might be an entire generation behind with their Vega GPU architecture, but their next-gen Navi GPU architecture is due sometime in late 2018 and has turned up as the elusive "Super Secret" ID in AMD's new driver.

In the latest AMD's drivers, there is code that reads "new_chip.gfx10.mmSUPER_SECRET.enable[0:0]". Someone spotted the new Super Secret ID in the latest Linux release of Radeon drivers, but it could be Team Red having some fun with us.

Navi could be the first GPU from AMD that would see a scalable GPU design being deployed, with many smaller Navi GPU dies interconnected through AMD's super-fast Infinity Fabric technology. This means AMD could drop GPU cores from each die, resulting in less cores per die but multiple dies added together would give us a Threadripper-like GPU design.

Advoli has just announced the launch of the world's first HDBaseT-certified PCIe x16 graphics card, the new TA6 Graphics Card is powered by AMD's Embedded Radeon E9550 Series GPU.

The TA6 is capable of 6 x HDBaseT channels (you know, the Ethernet ports we all know and love) that are each capable of handling 4K videos over a distance of 100 meters (328 feet). Advoli will have a full commercial launch for the TA6 Graphics Card at ISE 2018 in Amsterdam, with the company showing off a live demo at the HDBaseT Alliance Booth (Hall 5, Booth S100).

Advoli's new TA6 Graphics Card has HDBaseT and cable diagnostics, pass-through IR (both in and out), full emulated controls (encoding/decoding of IR, RS232 and CEC signals through the serial command interface). There's also remote diagnostic features that make it easier than ever for system integrators and audio visual installations.

NVIDIA's new TITAN V graphics card is meant to be the flagship AI/deep learning card, but the $2999 product is being used as a mining card in more articles I've seen than anything else.

Any YouTuber or GPU reviewer getting these cards are jumping into mining first with our friends at HotHardware putting the TITAN V through its paces. We have the TITAN V pushing a huge 69MH/s of Ethereum mining performance using Claymore v10.2, an impressive result considering AMD's flagship Radeon RX Vega 64 is only capable of 37MH/s.

But when the new TITAN V is overclocked (thanks to NVIDIA's use of HBM2 on a super-wide memory bus providing masses of memory bandwidth) it cranks out an incredible 82MH/s, beating out even two Radeon RX Vega 64s mining in the same system.