gpu
Latest
More details leak on AMD's Fusion platform, Fusion now officially the Palm OS II of CPUs
In yet more news about the perpetually forthcoming Fusion CPU / GPU combos from AMD, new details have been leaked regarding the chips' sizes, configurations, and other bits of insufferably tedious details that only Professor Frink could truly love. According to this latest report, the "first" Fusion processor will be called "Shrike," sport a dual-core Phenom CPU, an ATI RV800 GPU, and a footprint of just 40nm. Another chip, codenamed "Falcon," will debut in 2010 at a 32nm size, based around a platform known as "Bulldozer," and will be launched to combat Intel's 32nm offering. Of course, we've been hearing about these chips since 2006, so we're not exactly on the edge of our seats hearing new details of what essentially amounts to vaporware.
Intel discusses ATI and Nvidia killing Larrabee, launching as early as 2009
While Intel's Larrabee might not be a household name for consumers just yet, it's certainly at the table where Nvidia and AMD/ATI eat. The many-core (8 to 48, at least, according to that Intel graphic) x86 chip runs all your existing apps while tossing in support for OpenGL and DirectX thus eliminating the need for a discrete graphics chip. At least that's the plan. While the exact number of cores remains a secret as does the performance of each core compared to current GPUs, given the importance Intel places on Larrabee, it's reasonable to assume that an 8-core chip will launch in 2009 or 2010 with comparable performance to GPUs on the market at that time. Intel does say that Larrabee cores will scale "almost linearly" (read: within 10%) in games; that means that a 16-core chip will offer nearly twice the performance of an 8-core chip, 32-cores twice that of 16, and so on. Apparently this has already been proven in-house with Intel name-dropping Larrabee-coded titles such as Gears of War, FEAR, and Half-Life 2, Episode 2. It's no coincidence then to hear that Intel's first Larrabee product will target PC gamers. Click through if you're just dying to read about Larrabee's 1024 bits-wide bi-directional ring network and other bits of technical wonderment sure to create at least the hint of a silicon malaise. [Via CNET and Washington Post, Thanks Dan R.]
NVIDIA gets official with GeForce 9800 GT / 9800 GTX+ GPUs
We already had a glance at NVIDIA's newest low-ender, the GeForce 9500 GT, but the outfit clearly wasn't done. In case the previously mentioned card was just a tad too weak for your needs, you can also check out the GeForce 9800 GT and / or 9800 GTX+, which feature 112 / 128 stream processors and support for HybridPower and PhysX. Both cards can be found just about everywhere right now for $149.99 and $199.99, respectively, with 512MB of RAM on board. [Via Hardware Revolution]
Dell explains NVIDIA GPU issues, throws out BIOS updates to help
So yeah, quite a few NVIDIA GPUs have been acting up. Nothing new there. However, Dell's attempting to help its laptop owners out by making a few notable BIOS updates readily available for download. Apparently the issue "is a weak die / packaging material set, which may fail with GPU temperature fluctuations." In order to combat the problem, Dell's BIOS flashes "modify the fan profile to help regulate GPU temperature fluctuations," though the Round Rock powerhouse only promises that the updates will "help reduce the likelihood of GPU issues." Hit the read link and give it a go -- it can't hurt... we hope.[Via Electronista]
NVIDIA's GeForce 9700M / 9800M laptop GPUs get quasi-official
So it seems that NVIDIA has been sneaking its 9700M / 9800M GPUs into potent gaming laptops for a little while now, but we're finally getting some official-ish verbiage on the new-ish families. The GeForce 9700M GT, 9700M GTS, 9800M GT, 9800M GTS and 9800M GTX should all be available as we speak in some of your favorite portable LAN machines, with the latter capable of a mind-melting 420 gigaflops. For more on the graphical powerhouse that could be sitting within your machine right now, check the read link.
Diamond Multimedia announces two ATI All-in-Wonder HD cards
We told you that ATI's reinvigorated All-in-Wonder would be hitting the streets in late July, and sure enough, Diamond Multimedia is first out of the gate with versions of its own. The All-In-Wonder and the All-In-Wonder Premium iterations are both slated to hit reputable e-tailers "soon," with the latter unit including a bundled remote and an input / output board with an array of easy access connections. As expected, each card will boast 512MB of DDR2 memory, a clock speed of 725MHz and a single hybrid tuner for capturing local NTSC / ATSC broadcasts and ClearQAM signals. Unfortunately, the most pressing piece (ahem, price) was omitted.
Lucid Logix HYDRA tech brings together any GPUs for powerful matrimony
There's a lot of tech speak to cut through, but if Lucid Logix's claims are true, we might be looking at a revolution in the high-end gaming segment. Lucid is building a new "real-time distributed processing engine" system on a chip called HYDRA, which can mix and match any GPU from any manufacturer and work with any chipset, and piles it all together for performance scaling that Lucid claims is "near-linear" or "above-linear." It's the above-linear part that particularly makes no sense -- how can you squeeze more power out of cards than is there to begin with? -- but we're gonna give Lucid the benefit of the doubt for the moment and wait for the benchmarks. HYDRA is slated to come to market in the first half of 2009.
PowerColor PCS HD4850 graphics card packs 2GB of memory
Remember how we all swooned over Diamond's ATI Radeon HD 2900 XT in June of last year? At the time, that was pretty much your only shot at seeing 1GB of memory on a single graphics card. Fast forward to now, and we've got what's widely believe to be this planet's very first 2GB card. The PowerColor PCS HD4850, which is based on the RV770 core, comes with 800 stream processors and two whole gigabytes of GDDR3 memory. And just think, next summer you'll be sticking your nose up to find GPUs emerging with "only" this much memory. So fickle, we are.[Via PCLaunches]
All NVIDIA 8400M / 8600M chips faulty?
NVIDIA's stock took a pretty big hit last week when it announced that "significant quantities" of "previous-generation" GPUs and mobile and communications processors were defective and that it would take a $250M charge against earnings to repair and replace the affected chips, but the company didn't say which chips specifically were faulty, nor how many. That might be because the problem is much worse than it even sounds -- according to a report in The Inquirer, every single G84 and G86 GPU in the 8400M and 8600M series of cards is affected. Apparently both chips share an ASIC, and the core design suffers from the same heat-related issues. That certainly implicates a "significant quantity" of chips, all right, but this is just a rumor for now -- one that's probably best handled by NVIDIA stepping up and letting its customers know exactly how big the problem is.[Thanks, Rich]
Fujitsu Siemens' Lasso external graphics card get spotted, should hit retail soon
An external graphics card that's actually nearing release? Be still our hearts! Fujitsu Siemens' recently leaked external unit, which will be based on ATI XGP technology, is actually really, really close to hitting retail -- according to a piece over at Fudzilla, anyway. Slated to be titled Lasso, the RV670-based device should be more than enough to handle basic tasks, but unfortunately, benchmarks were off limits. On the upside, digital photography was not, so head on down to the read link for a few more snaps.
NVIDIA says "significant quantities" of laptop GPUs are defective, stock tumbles
If you're the type to watch the late stock tickers, you might have noticed that NVIDIA's stock just took a pretty big hit, down 24 percent to $13.56 -- that's because the company just informed investors that "significant quantities" of previous-generation graphics chips have been failing at "higher than normal rates," and that it's lowering its Q2 estimates due to pricing pressure. NVIDIA will be taking a $150M to $250M charge against earnings next quarter to cover the cost of repairing and replacing the affected chips, but didn't specifically announce what products were defective, just that they include GPUs and "media and communications processors." Laptop makers have apparently already been given an updated GPU driver which kicks in fans sooner to reduce "thermal stress" on the GPU, and NVIDIA says it's talking to its suppliers about being reimbursed for the faulty parts. That's great and all, but we'd really rather know which chips specifically are failing -- if you're serious about playing in the big leagues, you better come clean, guys.
Intel exec says NVIDIA's CUDA will be a "footnote" in history
NVIDIA execs have been talking smack about general-purpose CPUs for a while now, and it looks like Intel's ready to do some talking of its own -- speaking to CustomPC, Intel SVP Pat Gelsinger said that general-purpose GPU computing initiatives like NVIDIA's CUDA would be nothing more than "interesting footnotes in the history of computing annals." According to Gelsinger, the lack of a viable new programming model has held back similarly different architectures like the PS3's Cell because "years later the application programmers have barely been able to comprehend how to write applications for it." That's certainly an interesting point, but we'd say Gelsinger's not really taking stock of the big picture here -- fully utilizing the power of the GPU is the whole point of CUDA, after all, and OS developers like Apple are pushing OpenCL as a way to make GPU acceleration easier to for developers to access. Still, Intel has already said that discrete graphics are on their way out as hybrid tech like Larrabee enters the scene, and Gelsinger basically repeated the party line, saying that and "evolutionary compatible computing model" will be the "right answer long term." Those are some fightin' words -- it's going to be interesting to see how these competing chip strategies play out as other entrants like AMD's Fusion slowly make the scene as well.
PS3 graphics chip going 65nm this Fall
In addition to talking movie downloads, Sony Computer Entertainment CEO Kaz Hirai confirmed at a business briefing yesterday (via Next-Gen) that the PlayStation 3 consoles produced after last holiday season all have the 65nm Cell processor (down from 90nm at launch). In addition, Sony will be slimming down the RSX graphics chips to 65nm this Fall. The GPU shrinkage will make the PS3s cheaper to produce, use less energy and not heat up as much. That doesn't necessarily mean a price drop is imminent, but it does give the company some more breathing room to make up that $3.32 billion.
ATI Radeon HD 4870 graphics card gets official
Just five days after ATI fessed up with an official release of its Radeon HD 4850, along comes a similar announcement for the just-leaked Radeon HD 4870. This dual-slot beast is finally being outed on the record, and at the very least, Diamond Multimedia's version will come stocked with 512MB of DDR5 memory, a clock speed of 725 MHz, 900MHz memory speed and 800 stream processors. The unit also provides CrossFireX upgradeability and should be available here soon from your favorite online retailer.
Twin GeForce 8800M GTX upgrade now available for Dell's XPS M1730
If you pulled the trigger early on an XPS M1730, you may have gotten "stuck" with "just" a GeForce 8700M GT or similar. If you just can't be behind the times, Dell is finally offering a solution with the GeForce 8800M GTX upgrade kit. Said package includes dual SLI 1GB GDDR3 PCI Express graphics card designed specifically to boost frame rates and install all sorts of Vista-compatible bragging rights. If you're comfortable with doing the swap yourself, the upgrade price sits at a penny under $1,100. For those seeking professional help, you'll be coughing up $1,218.99. For those of you not frozen in place by sticker shock, you can peek the how-to guide for installation after the break.[Via Laptoping, thanks Danijel]
ATI Radeon HD 4850 gets official: available immediately
Considering that we've already seen AMD's ATI Radeon HD 4850 benchmarked, it's not like we really needed some official verbiage to cement our belief that the unit was real. Nevertheless, said verbiage certainly doesn't hurt, and that's precisely what's been delivered this morning. The HD 4850 is a single-slot PCIe 2.0 card featuring 512MB of DDR3 RAM, a 625MHz clock speed, 993MHz memory speed, 480 stream processors and support for CrossFireX / DirectX 10.1. We're also told that at least Diamond Multimedia's HD 4850 is available as we speak from a number of fine retailers, thus we presume everyone else's version of the card shouldn't be too far behind.
NVIDIA's 512MB GeForce 9800 GTX+ hits the bench
Yeah, that bench. NVIDIA's just-announced-yesterday GeForce 9800 GTX+ has already been used and abused for your satisfaction, and the folks over at PC Perspective have the benchmarks to prove it. The 55nm-based card was pit against an 8800 GT and AMD's extraordinarily fresh Radeon HD 4850, but we're not going to insert any spoilers in this space (okay, so it fared well... really well). All the graphs and screen captures you crave are waiting just down there.[Thanks, Ryan]
RV770-based AMD Radeon HD 4850 gets benchmarked
Judging by the fact that AMD tipped the whole world off to its upcoming RV770-based GPUs earlier this week, we don't suppose it'll be too upset that a 512MB MSI Radeon HD 4850 happened to land a little early in the PC Perspective labs. Design wise, there's nothing too out of the ordinary -- a single-slot cooler design, twin dual-link DVI ports, single 6-pin PCIe power connector and one goofy looking monster that you'll never see again once this thing gets installed. Care to see how it fared when facing the pressures of modern day gaming? You know where to look.
NVIDIA unveils second-gen Tesla GPU-based workstation cards
NVIDIA's Tesla GPU-based high-performance computing workstations and add-in cards have been on the market for a whole year now, and to celebrate, they're getting birthday cake, balloons, and an upgrade to GT200-based chipsets. Like AMD's recently-announced FireStream 9250, the new T10P processing units are capable of breaking the teraflop barrier, up from the first gen's paltry 518 GFlops, and they're up to 240 cores from the first gen's 128. You'll have to shell out to get all that horsepower, though: the entry-level, 900GFlops C1060 PCI card will sell for $1699, while the four-GPU 1U S1070 blade will sell for $7995 for two PCIe-interface version or $8295 for the single PCIe connect model. The standalone Tesla workstation has been discontinued, as customers were increasingly buying the cards, so it looks like those are really fast collectors' items for now. So, who's going to be the first to add one of these bad boys to the Engadget Folding@Home team?[Via Tom's Hardware, thanks Matan]
AMD offers up ATI Radeon HD 3870 Mac & PC Edition
Hey Mac users, tired of looking longingly at that PC-lovin' ATI Radeon HD 3870? AMD has teased you long enough, and at long last, the Mac & PC Edition has arrived. This card is optimized for use within Apple's Mac Pro systems and includes 512MB of GDDR4 RAM, 320 stream processors, PCI Express 2.0 support and twin dual-link DVI ports for treating that pair of 30-inch Cinema HD displays like they ought to be treated. Check the read link for the full list of specifications, and get set to drop $129 when this one lands later this month.[Via TUAW, thanks to everyone who sent this in]
