Graphics Cards - Page 211

AMD launched their next-gen Radeon Vega Frontier Edition graphics card yesterday, but you know what? There are absolutely no performance information on Radeon Vega Frontier at all, nothing apart from what AMD has released themselves, or collaborated on with the likes of PCWorld.

VideoCardz wrote something on Radeon Vega Frontier, and said: Vega is quite possibly one of the most anticipated GPU launches in AMD's history. There are many reasons why this launch is so important. The early adopters of 4K monitors are left with no options other than GeForce GTX 1080 and above. AMD has been teasing demos of 4K gaming on Vega for a while, although no one was able to buy Vega card till today. AMD promised that Vega will launch in the first half of 2017, and that promise was fulfilled. What we didn't know is that only Radeon Pro card will be "available".

This triggered me to think about it, and respond. VC are right, it's not something I haven't been saying myself - but the last straw was AMD launching Radeon Vega Frontier Edition without any reviews. I personally reached out to AMD asking for one of these cards, even at my expense ($2000+) so that I could test it with whatever software I have here (and whatever software AMD would want me to test it with) and I would also do some exclusive cryptocurrency testing. Nothing. I've heard nothing in return about any sampling, or even AMD finding a card at retail and letting me take it.

The huge Ethereum mining rush is still going and while difficulty is dropping and the price of ETH itself is falling rapidly, NVIDIA has finally dumped out its new mining-specific graphics cards. Up first, MSI and GALAX models.

YouTuber 'Cryptomined' got his hands-on the new cards, with his video available above.

The new mining GPUs aren't much better than the consumer gaming graphics cards, with 20MH/s of Ethereum mining @ stock clocks, while we have 24MH/s when overclocked. This is what I'm seeing in all of my testing with stock GTX 1060s. MSI however, didn't even remove the label on the box that says 'the most powerful graphics card'... yeah, somehow I don't think so, MSI.

AMD's new Radeon Vega Frontier Edition launches today, with PCWorld being one of the first outlets with the Vega-based graphics card in their hands.

The new AMD Radeon Vega Frontier Edition has been confirmed at $1200, with 16GB of HBM2 memory, and is air-cooled. The liquid-cooled Radeon Vega graphics card will cost $1800, quite a chunk more. The air-cooled version has a 300W TDP, while the liquid-cooled version has a 375W TDP.

Now remember, Radeon Vega Frontier Edition is not a consumer graphics card for gaming... with AMD aiming it at "data scientists, immersion engineers, and product designers". This pushes it into the same sphere as NVIDIA's Quadro and Tesla cards, and not the TITAN Xp so much. PCWorld had the TITAN Xp in their hands for testing, so why not? AMD set up two identical PCs for PCWorld, rocking Ryzen 7 1800X processors, 32GB of DDR4 RAM, SSDs, 4K displays, and Windows 10 Enterprise Edition.

We are living and breathing for everything we can handle on Radeon RX Vega, but now the latest tease is that AMD will have "excellent" pricing on Radeon RX Vega.

According to the Bits And Chips Twitter account, which tweeted: "Forget about VEGA in Enterprise market. Next AMD game changer will be NAVI (EPYC:CPU=NAVI:GPU). However, VEGA price will be excellent". I have personally heard that Navi will be a game changer from my own sources, but haven't heard anything about cheap pricing on Radeon RX Vega. This is because HBM2 isn't cheap, especially compared to GDDR5 that's used on the Radeon RX 500 series.

If AMD could price the GTX 1080 Ti competitor in the new Radeon RX Vega family at $499, this is a game changer. $599 and it better beat the GTX 1080 Ti and have a large amount of stock to sell against the $699+ on GTX 1080 Ti... but if there's only a few thousand, NVIDIA will continue to win because it has massive numbers of AIB partners with multiple GTX 1080 Ti models on the market already. $699 is the price of the GTX 1080 Ti, so AMD really can't price Radeon RX Vega here without shooting themselves in the foot.

I think most people have known for a while that AMD's upcoming Radeon RX Vega will use a lot of power, with the new Radeon Instinct MI25 accelerator delivered with a 300W TDP.

Now we have an MSI marketing director saying: "I've seen the specs of Vega RX. It needs a damn lot of power. We're working on it, which is a start so launch is coming closer :)". We should expect the air-cooled version of Radeon RX Vega to use up to 300W, while a watercooled version could use a little more - but hopefully offer some more OC headroom. Still, I'd rather have a card that uses maximum power, if the performance is warranted, versus GPU Boost 3.0 restricting Pascal-based NVIDIA GeForce GTX 10 series cards.

What do you think?

NVIDIA has already released its Volta-based Tesla V100 graphics card, but they've now detailed their new Volta V100 GPU based on the PCIe standard. This rides the news of AMD announcing its new Radeon Instinct MI25 graphics card, with 16GB of HBM2.

The new NVIDIA Volta V100 is built on the new 12nm FFN (FinFET NVIDIA) node from TSMC, and is custom built for NVIDIA's Volta GPUs. Inside, there's an amazing 21 billion transistors, 5120 CUDA cores, 80 SMs, 40 TPCs, and GPU clocks of 1370/1455MHz for base and boost, respectively. NVIDIA has tapped 16GB of HBM2 with a 4096-bit memory bus that provides a huge 900GB/sec of bandwidth.

NVIDIA's new PCIe-based Volta V100 doesn't support NVLINK of course, but it sports a dual-slot cooler with black/gold color scheme that looks slick as hell. But, how does it compare to the new Radeon Instinct MI25 graphics card? Let's take a look.

AMD has detailed its upcoming Radeon Instinct graphics cards, with their flagship Radeon Instinct MI25 which has 16GB of HBM2 - and 64 next-gen compute units (4096 stream processors) all based on the Vega GPU architecture.

AMD's upcoming Radeon Instinct MI25 features 4096 stream processors, 12.3 TFLOPs of FP32 compute performance (24.6 TFLOPs of FP16 compute perf.), 16GB of HBM2 that provides 484GB/sec of memory bandwidth. All of this fits on a dual-slot card at 10.5 inches long, with a 300W TDP.

Performance wise, Radeon Instinct MI25 beats out NVIDIA's Tesla P100-16 with 1.3x more FP16 and FP32 performance, and far better performance per watt.

After our world exclusive first look of MSI's new GeForce GTX 1080 Ti Lightning Z at Computex 2017 in late-May, MSI has finally detailed its upcoming custom GTX 1080 Ti Lightning Z graphics card.

We are looking at amazing piece of engineering and craftsmanship, with the GTX 1080 Ti Lightning Z rocking a massive 8+8+8 pin PCIe power connector set up for the serious overclockers. MSI has used their new and improved Tri-Frozr cooling technology, that packs 2 x 10cm and 1 x 9cm TORX 2.0 fans. This combines the best of both worlds with a traditional fan blade and dispersion fan blade, that provides great cooling but it won't be loud.

As for clocks, MSI is conservative by saying the GTX 1080 Ti Lightning Z will have three GPU clocks:

AMD is poised to unleash its next-gen Radeon RX Vega graphics cards in the consumer market in early August, but now we're hearing some more news on what the AIB partners will be doing.

The new rumors point to AIB partners shipping custom Radeon RX Vega graphics cards in early August, with AMD set to launch at least two RX Vega graphics cards at first. We should expect a high-end Vega 10 XT part which should compete against the GeForce GTX 1080 Ti and possibly the TITAN Xp, while the cut down Vega 10 Pro will compete against the GTX 1080 and new GTX 1080 with 11Gbps GDDR5X.

There's not much else to report on about the Radeon RX Vega just yet, but I do have some stories that I'll be writing over the next few days that will be world exclusive pieces. Radeon RX Vega is coming soon, and the ramp up towards it over the next 6 weeks is going to be an amazing journey.

Sony is pretty happy with the performance and sales of its PlayStation VR, so much so that they don't have any plans to make a next-gen PSVR headset.

Sony boss Shuhei Yoshida said during an interview at E3 2017 with Jagat Play that the PSVR was made, and is perfectly optimized for the PS4 - so there's no plans for a next-gen PSVR 2.0 headset just yet. But during Computex 2017, I met with someone who had some news on PS5 and said that it would have a discrete GPU... so maybe the PS5 will see the release of the PSVR 2.0 headset.

AMD's new Radeon Vega Frontier Edition graphics cards have been spotted on both Scan UK and Sabre PC websites, in both the air-cooled and liquid cooled versions. Both of the cards have identical specifications, including the super-awesome 16GB of HBM2.

Both the Radeon Vega Frontier AIR and Radeon Vega Frontier LIQUID have 16GB of HBM2, 13.1 TFLOPS of FP32 compute performance, and its Vega GPU clocked at 1600MHz. Both of them rock the new Vega 10 GPU, and have 3 x DP and 1 x HDMI for display connectivity.

As for the pricing, it looks like AMD will have the Radeon Vega Frontier AIR at $1199 while the Radeon Vega Frontier LIQUID will cost $1799... a little crazy, but these are next-gen cards. So now... the speculation begins on pricing of the Radeon RX Vega for consumers. Anything higher than $699 and it's DOA if it can't beat the GeForce GTX 1080 Ti, but if we see a dual GPU version of the card - my expectations of $1199 are blown out of the water if Radeon Vega Frontier AIR is $1199. Interesting. Another interesting note, is that AMD says Radeon Frontier Vega comes with "16GB HBC" and not "16GB HBM2". Hmm.

Globalfoundries has announced that it will have its next-gen 7nm FinFET technology begin hitting the production line in 1H 2018, providing up to 40% more performance to 7nm-based products, and more.

The company said that its new 7nm FinFET design kits are available right now, with "volume production" expected in 2H 2018, an exciting next 12 months ahead of us. The step down to 7nm will provide improvements at the transistor and process levels, with the 7LP tech beating expectations for performance, with over 40% more processing power and twice the area scaling than the current-gen 14nm FinFET node.

Globalfoundries is taking design orders from customers for its 7nm technology, in Saratoga County, N.Y. Senior VP of Globalfoundries, Gregg Bartlett, explains: "Our 7nm FinFET technology development is on track and we are seeing strong customer traction, with multiple product tapeouts planned in the next twelve months. And, while driving to commercialize 7-nm in 2018, we are actively developing next-generation technologies at 5-nm and beyond to ensure our customers have access to a world-class roadmap at the leading edge".

NVIDIA is working on some new crypto-focused graphics cards, with a new GP104-100 and GP106-100 cards that will be offered by NVIDIA's AIB partners, with variations. Note: all of these cards do NOT come with display outputs, so they cannot be used as traditional graphics cards - hence the concentration on cryptomining.

The new GP104-100 card will be a GeForce GTX 1080 for all intents and purposes with 8GB of GDDR5X, with higher efficiency and they're fine tuned for the crazy cryptocurrency phase we're going through. They will ship with different GPU clock speeds with the base model hitting 1607/1733MHz for base/boost, respectively, with 8GB of GDDR5X @ 10Gbps on a 256-bit memory bus.

The upcoming GP104-100 card will have the same single 8-pin PCIe connector and TDP @ 180W, with a price of just $350 - compared to the average price of $470 or so for a GTX 1080. The best bit? These GTX 1080 variants will push a huge 60MH/s, up from the 23MH/s or so that my GTX 1080 is pushing in Ethereum mining right now.

NVIDIA has released their new GeForce 382.56 WHQL drivers, which include support for two new game releases: Dirt 4 and Nex Machina, which both drop later this month. Download the new GeForce 382.56 drivers right here.

There are also some fixes for various games and technical issues, with V-Sync issues alongside NVIDIA's Fast Sync being fixed, so now you shouldn't get any nasty tearing from a FPS limit. NVIDIA also notes that GeForce GTX 1060 owners will have no issues running Dirt 4 at 1080p, while GTX 1070 easily runs 1440p, and the GTX 1080 Ti is reserved to hitting 60FPS at 4K. NVIDIA adds that the GTX 1080 can also run Dirt 4 at 4K 60FPS, if you drop MSAA.

Next up is Nex Machina, which requires the same GTX 1060 for 1080p, GTX 1070 for 1440p, and GTX 1080 Ti for 4K. Added to the new GeForce 352.56 WHQL drivers are updated SLI profiles for Bulletstorm: Full Clip Edition, Little Nightmares, PlayerUnknown's Battlegrounds, and Transformers Online.

The latest on AMD's upcoming Radeon RX Vega is that there are indeed 4 different versions, 3 with 8GB of HBM2 and a higher-end model with 16GB of HBM2.

VideoCardz is reporting that the latest Vega GPU engineering samples have varying clock speeds, starting with the first RX Vega with 8GB of HBM2 having a GPU clock of 1000MHz, an unknown HBM2 clocks. The second model has its Vega CPU clocked at 1.2GHz and its 8GB of HBM2 @ 700MHz, while the highest end 8GB model has its Vega GPU @ 1.5GHz and its 8GB of HBM2 at 925MHz. I think this one will be the GeForce GTX 1080 Ti competitor, while the 8GB @ 700MHz model will be the GTX 1080 11Gbps competitor, while the 8GB @ unknown will be the GTX 1070/1080 competitor.

But it's the 6861:00 card and its 16GB of HBM2 that has me intrigued, with the mysterious piece of technology having its Vega GPU @ 1GHz, while its 16GB of HBM2 are clocked at 700MHz. If I'm right and the 8GB HBM2 @ 700MHz model is the GTX 1080/1080 11Gbps competitor, a dual-GPU with 16GB of HBM2 and both of those cards clocked at 1.2GHz on the GPU and 700MHz on each 8GB cluster of HBM2, we could be in for a very wild ride.

PCI-SIG, the organization that defines the PCIe standard and its specifications with 730+ members across the world in the biggest countries on Earth including Qualcomm, NVIDIA, AMD, Dell, Intel, and many others - has announced the specifications for the PCIe 4.0 and PCIe 5.0 standards.

PCI Express 4.0 is expected to hit 16GT/s, with all of its features locked in, and rocking new performance enhancements over current PCIe 3.0, and will be positioned for the storage market, and cloud computing systems. PCIe 4.0 is capable of determining just how close to "the edge" each lane is operating at, all in real-time with PCIe 4.0's new Lane Margining at the Receiver feature. There are multiple companies already committed to PCIe 4.0, with "major vendors" offering 16GT/s controllers.

This new high-end interconnect technology isn't something we'll see on consumer motherboards anytime soon, even with Intel's newly gimped X299 chipset, or even AMD's beastly ThreadRipper-capable X399 chipset and its 64 PCIe 3.0 lanes. It's an underlined fact for the PCIe 5.0 standard which rocks 32GT/s of bandwidth, and is squarely aimed at the top of the IT market.

Valve has released its Steam hardware survey for May 2017, with NVIDIA enjoying a massive dominance with Steam gamers with 62.41% of people using an NVIDIA graphics card of some sort. This left ATI and AMD with 21.53%, while Intel and its integrated graphics power a surprising 15.65%.

Windows 10 is by far the most used operating system with Steam, powering 77.51% of gamers' PCs. When it came to CPU usage, Intel stands in a sea of blue with 80.22% of Steam gamers using an Intel CPU, leaving AMD with just 19.78% - and it doesn't seem as though Ryzen made a dent just yet, but give it a few more months and we'll take another look at those numbers.

There are still 45.15% of gamers with a dual-core CPU, but 49% of gamers have a quad-core CPU, leaving just 4.25% to the 1%ers who use 6/8/10-core processors.

Manli has officially announced its new GeForce GTX 1080 Ti Gallardo with RGB lights, and for those who are sitting back thinking "what the hell does Gallardo mean?" it represents "elegant" and "bravery" in Italian.

The company has used some slick golden and soldier elements for the packaging, something that'll catch your eye on retailer shelves and online. But we're all here for the cooling and performance, right? Manli has provided two overclocking setups: stock clocks at 4% higher than normal @ 1544/1657MHz, while the super overclocking version hits 7% higher than the GTX 1080 Ti FE @ 1594/1708MHz.

There's no pricing or ETA on Manli's upcoming GeForce GTX 1080 Ti Gallardo, but we'll keep you updated as it breaks.

It looks like the rumors of NVIDIA working on a new Pascal-based cryptocurrency mining solution were right, with VideoCardz posting a new article teasing the GPU mining beast.

The new GPU cryptocurrency mining station rocks 8 x GP106-100 mining cards that are passively cooled, and have no display outputs. This is perfect for mining, as you don't need display connectivity, and the fanless design will keep them nice and quiet versus the huge noise some cards can pump out when the card is pushing full speed 24/7.

We're looking at 200MH/s when mining Ethereum, which is what I'm mining right now - and with a blend of GPUs on 9 systems with 18 graphics cards in total, I'm getting 470MH/s. So to see 200MH/s from a single GPU mining station with 8 x GPUs using less than 1000W, that's impressive indeed. The entire system is powered by an optional 1600W PSU, with 9 x 8cm 5800RPM coolers keeping everything nice and chilled.

HP unveiled their new Omen desktop gaming PC that can be figured with GTX 1080 in SLI or RX 480/580 in CrossFire, but it also teased it was going to be configured with the unreleased Vega 10 GPU from AMD. At the same time, HP announced their new HP Accelerator, their new external GPU enclosure.

HP has stepped it up with their new Accelerator, which can handle graphics cards with up to 300W TDP, meaning virtually everything on the market. It comes with an internal 500W PSU and an integrated charging port that will charge your notebook at up to 60W, while there's also 4 x USB 3.0 Type-C ports, and a single USB 3.1 Type-C port that's Thunderbolt certified. There's even an RJ-45 network port on the back of the new HP Accelerator, in case your wireless-only laptop needed a faster, more stable network connection.

The company has put a lot of design effort into the Accelerator, which includes a 2.5-inch bay for an additional HDD, or super-fast SSD. There's just a single cable required, and it has an easy access door that makes replacing and installing the graphics card and HDD/SDD a simple procedure.