Graphics Cards - Page 234

We know it's coming, and we can't wait - NVIDIA begins showing off more and more of its new Pascal architecture, teasing the HBM-powered Pascal GPU at GTC Japan 2015 recently.

The last time we saw Pascal was when NVIDIA CEO Jen-Hsun Huang showed off a prototype board at GTC 2015 in March, but now we're seeing a slightly different Pascal GPU board than we saw earlier this year. The board features an actual Pascal GPU now that it has taped out, with the consumer GeForce card to feature a huge 16GB of HBM2 while the professional side of things will bump things up to 32GB of HBM2.

The design of the board is very similar to the Fiji-powered cards from AMD, as NVIDIA's Pascal chip is on a similar interposer to AMD. It looks slightly bigger, but we should expect the Pascal-powered cards to blow the doors off of the Fiji-powered cards from AMD. We will see DirectX 12 support, around 17 billion transistors, a 4096-bit memory interface, and so much more. Expect more news on this as we get closer to the New Year.

AMD has released its Catalyst 15.9.1 beta driver, a slight update over the recently released 15.9 beta driver.

The latter improved performance in the Fable: Legends benchmark and The Star Wars: Battlefront beta, but also gave rise to a severe memory leak bug that triggered when resizing windows. That bug is fixed in 15.9.1, so download and install via the source link to your heart's content.

Driver notes are below.

Exclusive: AMD has enjoyed being the first to market with a HBM-powered video card, with the Radeon R9 Nano, R9 Fury and R9 Fury X all powered by High Bandwidth Memory. But it looks like the company might run into troubles sourcing HBM2 in 2016 according to our industry sources.

Our source reached out to us today, saying that they "wouldn't count on [AMD] using HBM2 next year", but wouldn't elaborate further. This is an interesting rumor, because if it were true, it would mean that the use of HBM2 would shift primarily to NVIDIA. NVIDIA's next-gen Pascal architecture is already being tested internally by the company according to the latest rumors, and will rock HBM2. But AMD's next-gen GPU is rumored to rock between 8GB and 16GB of HBM2, is something we don't know too much about yet.

If AMD is in trouble with HBM2 next year, they might be stuck with short end of HBM2 yields. If so, we could see AMD utilizing HBM2 for their highest end GPU, which will most likely be a Fury X successor. Under that, we could see the company possibly using HBM1, but personally, I think HBM1 is something that should be used on mid-range cards for AMD's next-generation product stack.

AMD has added six new discrete, embedded Radeon GPUs to its lineup. The company says these are being released to meet increased demand for "rich, vibrant graphics in embedded systems", and that the additions will "help designers build mesmerizing user experiences with 4K multi-screen installations and 3-D and interactive displays" as well as "address the toughest parallel compute challenges."

The E8950MXM module is intended for power users, where the E8870 series (MXM and PCIe) is for those a little less so, and then there's the balanced, power-conscious option in the E6465 series (MCM, MXM and PCIe). The full breakdown of each from AMD is below.those a little less so, and then there's the balanced, power-conscious option in the E6465 series (MCM, MXM and PCIe). The full breakdown of each from AMD is below.

E8950MXM

AMD released its Catalyst 15.9 beta drivers earlier today. While beta drivers are usually smooth sailing, this was not the case with this release, which contained a memory leak bug. According to the aptly named Reddit user "MemoryLeakBug", it triggers when resizing any window, at which point RAM usage skyrockets to about 1GB per few seconds of resizing.

"AMDJoe" confirmed the issue and says a fix is in the works. In the meantime, roll back to an earlier set of drivers while you wait on a new release. That or just don't resize any windows...

According to the latest rumors, NVIDIA is working on a new dual-GPU video card that would use two of the company's GM200 GPUs, the same GPUs that power the GeForce GTX 980 Ti and Titan X, on a single card.

WCCFTech is reporting that NVIDIA held "secret" meetings in NYC last week, but they failed to report that there were meetings in LA, too. We were in LA and NVIDIA said nothing to us about a dual GPU card coming soon, as the event was all about the GTX 980 going into laptops. We might have been left out of the loop of course, but we are throwing a lot of salt across our shoulders on this one folks.

But, that doesn't mean it's not credible. NVIDIA is more than capable of releasing a new dual GPU card thanks to its super-efficient Maxwell architecture. Better yet, the "secret" meetings that were held did concentrate on how the company had shrunken down the board that the new GTX 980 goes onto, so what I will add to these rumors is that with two of these boards, a dual GTX 980 could be easily possible and not break through a 300W TDP.

TSMC has confirmed that it has made the first samples of NVIDIA's next generation GP100 GPU, with NVIDIA now testing the new Pascal architecture internally. When should we expect the consumer release? Sometime mid next year.

NVIDIA has been sent the first GP100 samples to its subsidiary in India, but we don't know anything other than this. The company taped out its GP100 GPU in June, which is going to be made on the 16nm FinFET+ process, and we should see it utilizing HBM2, providing 1TB/sec of memory bandwidth. This will be quite the increase on the 334GB/sec that the GTX 980 Ti and Titan X are capable of, and is double the 512GB/sec that AMD's Radeon R9 Fury X can muster with its HBM1.

The new Pascal architecture should also deliver its own upgrades over the Maxwell architecture that has been so good to NVIDIA across 2015. The GP100 rocks 17 billion transistors, and should see NVIDIA pack in some 6000 or so CUDA cores, up from the 2816 CUDA cores found in the GM200-powered GTX 980 Ti. Personally, I think we might see NVIDIA release multiple Pascal products - where I would love to see a GDDR5-based offering for the cheaper mainstream side, and a HBM2-powered GeForce 1000 series card for the enthusiasts.

The GeForce GTX 980M is already quite the mobile GPU, but as usual, NVIDIA isn't just sitting around doing nothing on the mobile GPU front. The company is reportedly preparing a new, very powerful mobile GPU that would pack the full GM204 GPU.

What's the 'full GM204' GPU you ask? Well, it's the identical GPU that is found on the GeForce GTX 980, the desktop video card from NVIDIA that was released last year. This means we can expect the full 2048 CUDA cores, 256-bit memory bus, and 4GB of GDDR5. We did expect a GTX 990M according to the rumors that we last reported on, but it looks like this new GPU is going to be the GM204-based GTX 980, this time inside of a gaming laptop.

The mobile GTX 980 would feature 2048 CUDA cores, 128 TMUs, and 64 ROPS - just like its desktop counterpart. If we compare this to the GTX 980M, found inside of gaming laptops like the ASUS ROG G751 that we reviewed here, it has 1536 CUDA cores, 96 TMUs and 64 ROPs. Moving over to clock speeds, the desktop GTX 980 has a GPU clock of 1126MHz and a Boost clock of 1218MHz while the GTX 980M comes in at 1038MHz, while the new GM204-based mobile GTX 980 will reportedly clock in at 1190MHz.

We have had one of AMD's new Radeon R9 Nano cards for around a week now, and we've only just gotten into testing it after testing out 4-way Fury X cards in a bunch of different games and resolutions.

Well, we just so happened to throw our R9 Nano into our system and straight away went to GPU-z, which said it was an 'AMD Radeon R9 Fury series' video card. But what happens when you throw in the full-blown Radeon R9 Fury X in? It works in Crossfire with the R9 Nano, that's what. The same goes for the normal R9 Fury, which also works in Crossfire with the super-small R9 Nano.

We can't share performance numbers on the R9 Nano until later this week, but we think that our review is going to be something worth reading, that's for sure.

PowerColor has just unveiled its new Devil 13 R9 390 video card, which features two R9 390 GPUs for some serious horsepower. The card is absolutely huge, with a triple-slot design and triple-fan cooler.

The new Devil 13 R9 390 requires an insane 4 x 8-pin PCIe power connectors so that it can draw between 1000W and 1275W of power. But all of that power won't be wasted, as it packs dual Grenada PRO GPUs, each with 2560 stream processors, 160 TMUs, 64 ROPs and 8GB of GDDR5. Each GPU has 345GB/sec available memory bandwidth, so the card has 691GB/sec memory bandwidth in total.

PowerColor adds that the Devil 13 R9 390 has been "built with carefully-designed Platinum Power Kit and ultra-efficient thermal design. It consists of massive 15-phase power delivery, PowerIRstage, Super Cap and Ferrite Core Choke that provides the stability and reliability for such high-end graphics solution. To support maximum performance and to qualify for the Devil 13 cooling system, 3 Double Blades Fans are attached on top of the enormous surface of aluminum fins heat sink connected with total of 10 pieces of heat pipes and 2 pieces of large die-cast panels. This superb cooling solution achieves a perfect balance between thermal solution and noise reduction. The PowerColor Devil 13 Dual Core R9 390 has the LED backlighting that glows a bright red color, pulsating slowly on the Devil 13 logo".

It looks like NVIDIA is really ramping things up for GRID 2.0, so that it can power an insane amount of virtual desktops thanks to its Maxwell GPU architecture.

GRID 2.0 was announced at the VMworld conference yesterday, where NVIDIA unveiled two new video cards based on their Maxwell architecture. The first is a dual-GPU, high-end card in the form of the Tesla M60 while the other is a single-GPU, high-end offering in the Tesla M6. Starting with the Tesla M60, we have 4096 CUDA cores, 16GB of GDDR5 and 7.4 TFLOPS of single precision performance. The Tesla M60 is capable of handling 36 simultaneous H.264 1080p30 streams at once and uses up to 300W of power.

The Tesla M6 features just 1536 CUDA cores, 8GB of GDDR5 and it can handle 18 simultaneous H.264 1080p30 streams at once. It arrives on a bare board on its own, using up to 100W of power. NVIDIA is coy on pricing at the moment, with both of the new Maxwell-based Tesla offerings being made available on September 15.

According to the latest data from JPR, NVIDIA has pushed into owning 81.9% of the GPU AIB market. AMD on the other hand, dropped from 22% to just 18% this quarter, and when looking at the year-over-year data, AMD has dropped down from 38%.

JPR's report finds that total AIB shipments decreased this quarter, down to 9.4 million units. AMD's quarter-to-quarter total desktop AIB unit shipments decreased by a sharp 33.3%, while NVIDIA's dropped 12%. JPR also noted that the "demand for high-end PCs and associated hardware from the enthusiast and overclocking segments has bucked the downward trend and given AIB vendors a needed prospect to offset declining sales in the mainstream consumer space".

With MSI enjoying the fruits of its labor with the just-released MSI GeForce GTX 980 Ti Lightning, we should all turn our attention to ASUS which just teased that they've got their ROG MATRIX GeForce GTX 980 Ti and ROG MATRIX Radeon R9 Fury on their way.

The company will reportedly unveil the two new enthusiast video cards during IFA 2015 on September 2. The new ROG MATRIX cards look like they'll feature the same cooler shroud as previous cards, but the color theme has changed a little - and we think it looks great. It looks like both ROG MATRIX cards could be dual-slot cards, boasting improved PCBs, much higher clock speeds, and much more.

We should definitely expect a new ROG MATRIX card based on the NVIDIA GeForce GTX 980 Ti, but it shouldn't be too far out of our expectations to see a new ROG MATRIX card based around the AMD Radeon R9 Fury as well as the Radeon R9 390X. We will be reporting on any new video card launches from ASUS as they happen.

It looks like we're just weeks away from the official release of AMD's Radeon R9 Nano, with DGLee from IYD.kr posting up some of the best shots of the Radeon R9 Nano yet.

Thanks to the tear down on the card, we get a better look at the small PCB that the Radeon R9 Nano features, with its single 8-pin PCIe power connector powering the card. We have a full Fiji GPU inside, the same chip that powers the R9 Fury and R9 Fury X. The entire card measures in at just 15cm, sitting just over the PCIe connector itself.

As for the price, we are expecting it to fall under $499. AMD's Radeon R9 Nano has a TDP of 175W, and should offer performance similar to that of the Radeon R9 290X (so around 5-10% less than the rebadged R9 390X and less than the R9 Fury and R9 Fury X, obviously. But for its size, this is one of the more exciting cards in AMD's Fiji-powered line up.

One of the more surprising video cards in AMD's new Fiji-powered lineup was the super-small Radeon R9 Nano, which is only as long as the PCI Express x16 port that it gets installed into.

The Radeon R9 Nano uses the same Fiji GPU that the R9 Fury and R9 Fury X are based on, with 4GB of HBM on-board. AMD is teasing that the R9 Nano has 2x the performance-per-watt compared to the Hawaii-based R9 290X, which is quite the claim. Especially when you consider that the R9 Nano will be using just 175W of power.

As for the performance compared to the Radeon R9 Fury X, it will reportedly have up to 85-90% of the performance that the Fury X does, leaving it toe-and-toe with the R9 Fury. If this is true, the R9 Nano will become AMD's most exciting video card released out of all of the Fury and 300 series cards.

We've been hearing about AMD's next generation GPUs for a while now, but it looks like they're beginning to take more form. It's being reported that AMD's upcoming 'Greenland' GPU will be an entirely new microarchitecture, with development on Greenland starting some two years ago.

But more interestingly, rumor has it that it's not just going to be yet another Graphics Core Next architecture (GCN), but it'll feature a new ISA (instruction set architecture) that will be so different to GCN, that it will be very exciting. On the surface, it'll be like previous generation Radeon products, with so much more happening underneath.

Greenland will reportedly usher in twice the power efficiency of GCN, where we can expect these GPUs to be made on the 16nm FinFET or 14nm process, with a serious jump on the number of stream processors when compared to the Fiji architecture. Fiji is the GPU behind the Radeon R9 Fury and R9 Fury X cards, powered by High Bandwidth Memory, or HBM. More importantly, it's being reported that Greenland will pack HBM2, with enthusiast level and professional cards packing up to 32GB of HBM2. The consumer orientated Greenland-powered Radeon cards will come in two flavors: 8GB and 16GB, both of the next-gen HBM2 technology.

Today NVIDIA unleashed the GeForce GTX 950 card aimed at gamers who want to play PC games at 60fps 1080p without breaking their wallets.

The Maxwell-based GTX 950 has an entry-level price point of $159 and packs in 768 CUDA cores clocked in at 1021 MHz that can be boosted to 1188 MHz, and sports 2GB of GDDR5 VRAM operating on a 128-bit bus clocked at 6.6GHz. Power will be supplied via 6-pin connector, and the card will draw 90 Watts of power.

Standard NVIDIA performance tech is compatible with the GTX 950 including SLI, V-Sync adaptive sync, and PhysX. This budget model will fully support the new DirectX 12 API and will be able to play DX12 games when they become available.

The last time we had GPU market share numbers, NVIDIA was dominating AMD with 76% of the discrete GPU market, leaving AMD with scraps. This was back in Q4 2014 (with our article released in February 2015), where NVIDIA's best video card was the GeForce GTX 980.

This was before the release of the Titan X in March, and before the GTX 980 Ti in June. At the time, AMD had its Hawaii architecture inside of the R9 290X, and the dual-GPU in the form of the R9 295X2. At the time, all signs pointed to the R9 390X turning things around, but the R9 390X ended up being yet another rebrand, while the R9 Fury X was discovered with our world exclusive during Computex 2015 in June, powered by High Bandwidth Memory.

Fast forward to now, where we're in Q3 2015, and AMD has multiple new products on the market: the R9 Fury X, R9 Fury, R9 390X and a bunch of rebranded 300 series video cards. According to Mercury Research's latest data, NVIDIA has jumped from 76% of the discrete GPU market in Q4 2014 to 82% in Q2 2015. This leaves AMD with just 18% of the dGPU market share, even after the release of multiple new products from Team Red.

According to the latest rumors, NVIDIA is preparing a new mobile GPU in the form of the GeForce GTX 990M. With the GTX 980M already being quite powerful, it still isn't as fast as a desktop GTX 980, but the new GTX 990M will fix that.

NVIDIA's purported GeForce GTX 990M would reportedly be as fast as current GTX 980M SLI configurations, with the GTX 980M being based off of the GM204 die with 1536 CUDA cores. A laptop powered by GTX 980M SLI beats a desktop GTX 980 by around 10%, so for a GTX 990M to beat both of those setups, we're talking about some serious mobile performance.

One of the big problems that NVIDIA will have is power consumption, with the GTX 980M SLI using around 180W TDP. Either NVIDIA is going to have to work some magic on the purported GTX 990M, or we're looking at around 150-180W TDP on the mobile GPU. After using the ASUS ROG G751 gaming laptop and its included GeForce GTX 980M for a few months now, it doesn't get too hot or loud at all. All NVIDIA would need is some half decent cooling, which all gaming laptops feature these days, and we could be looking at mobile 4K 60FPS gaming on the GTX 990M.

It looks like MSI is preparing what could be the best GeForce GTX 980 Ti on the market with its 980 Ti Lightning card, a beast that the company was teasing at Gamescom in Cologne, Germany over the last few days.

MSI confirmed they were making their GTX 980 Ti Lighting, with no prototype card on display. We should expect the MSI GeForce GTX 980 Ti Lightning card to sport the usual custom PCB, beautiful design and kick-ass cooling. Rumor has it MSI is going to have a stock GPU clock of 1217MHz resulting in a Boost clock of 1342MHz, but I would dare say we should expect more than that considering the ZOTAC GeForce GTX 980 Ti AMP! Extreme Edition is clocked higher than that, without a custom PCB.

We should hopefully have the MSI GeForce GTX 980 Ti Lightning in for review as soon as its released, so we'll keep you updated on what could be the most exciting GM200-based card yet.