Graphics Cards - Page 207

Until now, you've had to use AMD's own Radeon Software and its pretty damn impressive Wattman overclocking utility, or SAPPHIRE TRIXX, and lastly ASUS' GPU Tweak in order to overclock Vega. We can now add MSI to that list.

MSI's latest Afterburner 4.4.0 Beta 19 has support for low-level access to the SMC microcontroller, providing full voltage control over Vega. This means you can adjust the voltage on Vega up and down, if you want to get into some undervolting action, too. You can't make any per P-state control adjustments, with a slider provided instead.

MSI states in its own changelog: "Added low-level AMD Vega 10 graphics processors family support. This means that now MSI AB can access this GPU directly without AMD ADL API, so more powerful voltage control (no longer limited by AMD ADL API) and extended and more efficient low-level hardware monitoring are available for Vega now. Core voltage control for reference design AMD Vega series cards is now performed via low-level access to on-die SMC microcontroller. Voltage is adjusted in offset form now, applied to all P-states (just like it was on Fiji/Polaris) and full -100..+100 range is available. Added GPU power draw graph to hardware monitoring module for AMD Vega series graphics cards. Added HBM memory temperature graph to hardware monitoring module for AMD Vega series graphics cards".

AMD has had an amazing couple of years, and an even better 2017 with the mega success of Ryzen and Ryzen Threadripper knocking Intel down from its uber-throne. But what about the GPU division?

AMD launched their next-gen Vega GPU architecture, the first consumer graphics cards to feature HBM2 technology. It was back in May this year where I exclusively revealed there would be less than 16,000 Vega graphics cards available in the months post-launch... and here we are, with shortages and only a single custom RX Vega graphics card in the wild. But what about Navi?

Once again I have an exclusive story that AMD will have Navi ready to go sometime in July-August 2018, with a Navi-based professional card being launched at SIGGRAPH 2018. We're still waiting for AMD to launch Radeon Pro SSG, something they unveiled during SIGGRAPH 2017 that hasn't yet materialized. In the meantime, Radeon Technologies Group boss Raja Koduri has taken a sabbatical from the company until early-2018.

I awoke to an email this morning from Futuremark that announced they have a new entry into the 3DMark family of benchmarks, with Time Spy Extreme launching on October 11. What is Time Spy Extreme?

Futuremark's new 3DMark Time Spy Extreme benchmark run is a 4K DX12 benchmark run of the current Time Spy benchmark that will stress multi-core processors, countless GPU combinations, and more. Futuremark explains: "A decade or so ago, hardware sites were asking, "Will dual-core processors enhance your gaming experience?" Today, we're looking at new processors from AMD and Intel that can have up to 18 cores in some cases".

The company continues: "The CPU test has been redesigned to let processors with 8 or more cores perform to their full potential. Compared with Time Spy, the Extreme CPU test is three times more demanding. It also lets processors use more advanced instructions sets up to AVX2 when supported".

It was reported last week that GIGABYTE and MSI weren't working on custom AMD Radeon RX Vega graphics cards, but that rumor was wrong. I quickly reached out to my contacts at GIGABYTE which confirmed they were indeed working on custom RX Vega.

Now we have some photos to show courtesy of our friends at VideoCardz, with the GIGABYTE Radeon RX Vega 64 GAMING OC graphics card. We can see that it rocks what will surely be an updated Windforce 2X cooler, but it is pretty damn massive compared to AMD's sleek Radeon RX Vega 64 reference board.

GIGABYTE will be throwing DVI away (yay) and using just 1 x HDMI 2.0 and 3 x DP ports, while GIGABYTE seems to be using a custom PCB on their RX Vega 64 GAMING OC variant.

AMD's new Radeon RX Vega 64 can beat NVIDIA's best GeForce GTX 1080 Ti in some scenarios, where in multi-GPU configurations the RX Vega 64 begins competing against the GTX 1080 Ti in 8K... and now the same can be said for DX12 and Forza 7.

The lead that RX Vega 64 has is short lived, as it only beats the GTX 1080 Ti in Forza 7 in DX12 at 1080p and 1440p... because at 4K the GTX 1080 Ti reigns supreme once again. ComputerBase.de used an Intel Core i7-6850K overclocked to 4.3GHz, 16GB of DDR4-3000, and Forza 7 in DX12. As for drivers, the site used Crimson ReLive 17.9.3 and the GeForce 385.69 drivers.

Forza 7 was run in DX12 mode with 8x MSAA enabled, which really stresses out the cards - so it's interesting to see Radeon RX Vega 56 beating the GTX 1070/1080 and even the GTX 1080 Ti at 1080p. Even at 1440p, the RX Vega 56 competes against the GTX 1080 and beats it, while it's just 3FPS from the GTX 1080 Ti.

AMD doesn't have enough Radeon RX Vega graphics cards available to fight the good fight against NVIDIA's onslaught of GTX 10 series graphics cards as it is... and the pain will only get worse with the purported GeForce GTX 1070 Ti release that's meant to happen on October 26.

NVIDIA will reportedly be launching a new mid-range GeForce GTX 1070 Ti graphics card that will be a little faster than the GTX 1070, but a little slower than the GTX 1080, and will fight the Radeon RX Vega 56. GTX 1070 Ti will rock 8GB of GDDR5 memory, and higher GPU clock speeds than the GTX 1070. We should expect the GTX 1070 Ti to be a great cryptocurrency mining card thanks to the superior GDDR5 on-board, too.

AIB partners will be rolling out custom GTX 1070 Ti graphics cards next month, so expect a launch on October 26 and a release on November 2.

AMD has just released its new Radeon Software Crimson ReLive Edition 17.9.3 drivers, something that includes multi-GPU support for Total War: WARHAMMER II, and more.

The new 17.9.3 drivers also include support for Forza Motorsport 7, as well as some bug fixes that fix issues like bad multi-GPU scaling in F1 2017 on Radeon RX 580 series graphics cards. ReLive pushing GPU clocks higher on a second RX Vega graphics card in multi-GPU configurations on certain AMD Ryzen systems has also been fixed.

Fixed issues:

Update: I've reached out to GIGABYTE/AORUS which have confirmed with TweakTown that they are indeed making custom Radeon RX Vega graphics cards, there's just "no news on the horizon" yet.

If you've been waiting on GIGABYTE or MSI's custom Radeon RX Vega graphics cards, you won't have to wait any longer - or wait at all, because the companies aren't working on them according to fresh rumors.

AMD has been reportedly having troubles getting Vega GPU production in large numbers, which is something I exclusively reported earlier this year when I said AMD would have less than 16,000 available post-launch. Here we are, post-launch, with no official sales numbers to go by.

NVIDIA have released their new GeForce 385.69 WHQL drivers, just in time for the recently released of Project Cars 2 and the upcoming betas for Call of Duty: WWII and Star Wars Battlefront 2. In the upcoming weeks, we also see the releases of FIFA 18, Forza 7, Total War: Warhammer 2 and EVE: Valkyrie - Warzone, all which have day one support.

NVIDIA have also added 3D Vision profiles for numerous games and SLI support for following games:

Included in this release is a number of bug fixes listed below.

Call of Duty: WWII's upcoming beta will be unleashed on September 29 for PC gamers, with the return to the World War II setting to start pre-loading on Steam... right now.

Sledgehammer Games and Raven Software, the developers behind COD: WWII are recommending NVIDIA's GeForce GTX 1060 as the GPU of choice for gamers running COD: WWII @ 1080p.

Call of Duty: WWII PC Open Beta Recommended Spec

Now that the release of the Radeon RX Vega is over, the next part of the journey for AMD is to unleash custom RX Vega graphics cards with AIB partners. This is meant to be happening in late-October, and should lead through the holidays and into 2018. But what about Vega 20?

Vega 20 will be a refresh of Vega 10, which we should see AMD use the 12nm LP node, but what about performance? We don't know anything performance-wise just yet, but according to leaked slides from Informatica Cero, Vega 20 is coming in Q3 2018, and on the next-gen PCIe 4.0 standard.

PCIe 4.0 has a huge 16GT/s of bandwidth, up from the 8GT/s available on the current-gen PCIe 3.0 standard - something that was released in 2010.

NVIDIA hasn't been able to keep up with the non-stop demand of their GeForce GTX 10 series graphics cards, thanks to cryptocurrency miners across the world gobbling them up for mining.

We reported earlier this month that GTX 10 series pricing was on its way up, with an average of 25% higher pricing according to Mizuho's chief semiconductor analyst. Mizuho reports: "Our checks with the leading GPU and motherboard OEMs indicate SepQ GPU card trends are very strong, with card shipments coming in ~30-50% ahead of flat q/q expectations on strength from cryptocurrency mining. Cryptocurrency demand is driving strength in NVDA's GTX 1060/1070 cards. The GPU/motherboard OEMs also noted GPU pricing was up ~25% in the last six months".

It's not just miners either, but DRAM shortages are going to push us to see a purported decline of GTX 10 series for miners, with China banning cryptocurrencies being a massive driving factor. Mizuho continued: "The OEMs also noted zero inventory of GPUs in the channel and constrained short DRAM supply and pricing also affecting GPU shipments. Coming off a very strong SepQ, there are also expectations in the supply chain that DecQ GPU sales could be muted in pricing and demand on recent cryptocurrency bans and DRAM shortages".

AMD just unleashed their new Radeon Software Crimson ReLive Edition 17.9.2 drivers which add 2-way multi-GPU support to Radeon RX Vega graphics cards, but now the latest rumor has the company preparing 13 new graphics card on their Vega 11 GPU.

The Vega 11-based graphics cards should replace AMD's current-gen Radeon RX 400 and Radeon RX 500 series cards, which are based on the Polaris 10 and Polaris 20 GPUs. The naming system that AMD is reportedly going to use makes sense, as it continues the way that they launched RX Vega: using the compute unit count (in this case 32 and 28 CUs) to name the card.

This is where Radeon RX Vega 32 and Radeon RX Vega 28 nomenclatures come from, with the Vega 11 XT expected to rock 2048 stream processors, with 4GB of HBM2 and a 1024-bit memory bus. Vega 11 Pro should feature 1792 stream processors, with the same 4GB HBM2 on a 1024-bit memory bus.

AMD has released their latest Radeon Software Crimson ReLive Edition 17.9.2 drivers, with a surprise feature: 2-way multi-GPU support for Radeon RX Vega graphics card.

CrossFire support isn't center focus for AMD, something that the company has been very forward with for a while now. They will be pushing out multi-GPU support for their Radeon graphics cards in the future, but they're definitely pulling out of sheer beasty multi-GPU gaming PC rigs.

However, the new 17.9.2 drivers only support 2-way CF setups, with 3- and 4-way RX Vega setups simply not possible right now. I'm just installing the new drivers on my Ryzen Threadripper 1950X machine, with my 2 x RX Vega graphics cards and 8K display... we'll have results of RX Vega 64 CF over the weekend.

EVGA just announced their new GeForce GTX 1080 Ti FTW3 ELITE graphics card with its 11GB of GDDR5X clocked at 12Gbps, but now MSI has outed its upcoming GeForce GTX 1080 Ti GAMING X TRIO graphics card.

What's the difference between the MSI GeForce GTX 1080 Ti GAMING X and the new TRIO variant? It rocks a triple-fan cooler over the dual-fan cooler on the normal GTX 1080 Ti GAMING X, coming in as a 2.5-slot graphic card with an RGB LED bar on the backplate.

MSI has the clocks on the GTX 1080 Ti GAMING X TRIO at 1569MHz base, 1683MHz boost, and the 11GB of GDDR5X clocked at 11124MHz. MSI requires 2 x 8-pin PCIe power connectors, but we don't have any solid photos of the PCB just yet.

It looks like AMD might have scored a seriously large contract through electric car maker Tesla Motors, with the companies collaborating on a new AI accelerator for self-driving cars.

Jim Keller, the legendary engineer in AMD's ranks is reportedly involved in the deal, along with 50 engineers at Tesla. Keller has been involved with Apple's A4 and A5 processors, the original Athlon 64 processors, and AMD's latest Zen architecture.

AMD has reportedly already sent Tesla early samples of the new semi-custom chip, with testing already underway. For Tesla, this is the firs ttime they've opted for a purpose built AI processor, which will be using both AMD and Tesla IP.

AMD can't get enough Radeon RX Vega cards into gamers' hands, or any Vega 10 GPUs into AIB partner hands at all, but we're already looking to the future with Vega 20 and the upcoming shift to the 12nm FinFET process.

AMD has confirmed it will be moving to the 12nm FinFET process for both Ryzen and Vega parts, withi the upcoming Zen+ parts to launch in the second half of 2018. We should expect naming systems like Ryzen 7 1850X (up from 1800X), and changes to Threadripper.

But for the new Vega 20 GPU, we should see it pushed onto 12nm FinFET with "more than 10% improvement in performance over industry 16nm FinFET solutions", with "up to 15% improvement in circuit density".

It looks like we won't be seeing any custom Radeon RX Vega graphics cards until at least the end of October, and probably more into November.

According to Hardware.fr who reached out to AIB partners making custom Radeon graphics cards, and said that no manufacturers are ready apart from one unnamed manufacturer who said that they're waiting for Vega GPUs to arrive.

If AIB partners are still waiting for Vega GPUs, then the custom Radeon RX Vega graphics cards are still many months away, with a November launch making sense. Remember that I posted a world exclusive scoop on there only being less than 16,000 x Radeon RX Vega graphics cards in the months post-launch... and here we are.

EVGA continues to lead the pack with overclocking products, and underlines that fact with the release of their new EPOWER V card. EVGA's new standalone VRM board provides additional power to graphics cards and motherboards, allowing you to break out of the voltage shackles on even the best graphics cards and motherboards on the market.

The EPOWER V board features two fully-independent voltage outputs, as well as a built-in EVBot MKII, which lets you adjust voltage control on-the-fly. The EPOWER V board is powered by 3 x 6-pin PCIe power connectors, while input is provided through a 12+2 phase design that provides a massive injection of VCORE and VMEM into your graphics card, letting it break through those ridiciulously low voltages, and into an entire new level.

EVGA even provides USB 3.1 Type-C and softwre controls, which lets you connect the USB cable up to your PC and use software to control the EPOWER V board. A fully detailed rundown on the amazing new EPOWER V board can be found here.

AMD might have just launched their new Vega GPU architecture with a slew of Vega-based products (Radeon Vega, Radeon RX Vega, Radeon Pro WX, and Radeon Instinct) but the real king is NVIDIA's now months-old Volta GPU architecture.

We don't hear much about NVIDIA's Volta GPU architecture because it's still a while out from finding its way into consumer GeForce graphics cards, but the supercomputer/AI/deep learning markets are now receiving their new Volta-based Tesla V100 accelerators which means... BENCHMARK TIME!

First off, let's look at the difference between the previous-gen Pascal-based Tesla P100 and the new Volta-based Tesla V100. Starting off with 12x more deep learning training performance, with 10 TFLOPs on P100 up to a freakin' is-it-real 120 TFLOPs of 'DL training' on V100.