Video Cards & GPUs - Page 412

AMD have shown off their first working 28nm notebook GPU to the public at a media event held in San Francisco, California. The chip is part of the upcoming AMD Radeon HD 7000 series and was shown off displaying Dirt 3 but resolution, graphic settings and FPS were unknown. The upcoming HD 7000 series dons the codename "Southern Islands" and is being built by TMSC. The new cards are looking to be unleashed onto the market by the end of the year - but there are also rumors TMSC are having issues with the 28nm production.

The new cards are expected to support switchable graphics and continue pushing current technology features such as Eyefinity and H3D stereo 3D gaming. There are reports that AMD are planning to launch four different GPUs based on the 28nm version of the VLIW4 architecture and will use them in three different series: the Radeon HD 7800, HD 7600 and HD 7500. These GPUs will sport between 768 and 1536 stream processors as well as using GDDR5 memory.

Later on, AMD will launch the Radeon HD 7900 series featuring the new Graphics Core Next (GVN) architecture that is has been talking about since June. The idea behind GCN is to create a GPU that can perform well at both graphical and computing tasks.

Following leaked details on AMD's Radeon HD 7000 series graphics cards earlier in the week, some new slides (said to be straight from AMD) have since popped up which add some more meat to the equation.

We are assured that the HD 7000 family will use the first VLIW4 circuits at 28nm. In usual fashion the series will be rolled out in steps, with each sub-series based off different base architectures. Looking at the latest leaked roadmap, we can ascertain that AMD is initially planning the launch of four different GPUs making up three series, said to be scheduled for a Q4 2011 launch.

There will be the Radeon HD 7870 and HD 7850, HD 7670 and HD 7570. They will be equipped with GDDR5 memory across the board and sport 768 to 1536 Radeon cores (stream processors).

A little birdy has leaked out some specs on AMD's next generation Radeon HD 7000 series of graphics cards, though the information is most certainly still questionable at this stage with the lack of confirmation on any of it.

Codenamed Southern Islands, these next-gen chips will be 28nm (TSMC) based and as the chart above denotes, sitting at the top of the hill is the dual GPU wielding HD 7990. As per tid bits of info on the two Tahiti XT/Pro Cores the card will use, the max stream processor count is estimated to be around 4000 ALU, with the card featuring a TDP of >300W.

Head on over to the source link for a better idea of how the complete lineup is likely to shape up, with details on 8 known SKUs in the 7000 series shared. But remember, it's all potentially very loose information just at this point, so do remember to take it with a pinch of salt.

Here's an innovative, yet simply executed piece of tech MSI has come up with for future graphics cards. Word has come out that MSI has been testing a new feature it's looking to help make its already kickass lineup of non-reference graphics cards stand out even more.

MSI calls it "Dust Removal Technology", which is obviously to help rid your card of a negative scenario that affects every card sooner or later - dust buildup, translating to heat buildup. MSI say that enough dust buildup can eventuate to a rise in temperatures by as much as 15c, and from personal experience I can say that's not an exaggeration. Their new founded technology is quite simple in principle really.

The way it works is by simply running the fans in an opposite rotation at 100% speed for the first 30 seconds of system startup. This would result in air being drawn from the heatsink itself and pushed back out through the intakes of the cooler. While it certainly wouldn't remove every single bit of dust from the heatsink, MSI say it does work to significantly reduce dust accumulation in those tough to clean areas.

External graphics solutions for the mobile platform has been a talked about subject for several years now, and while we've seen a few forms of these on display at various trade shows and in prototype form around the web, the big players are yet to really make a firm jump into this area.

However, it's still definitely on the cards according to NVIDIA. Fuad of Fudzilla was recently at NV headquarters where he got to have a chat with general manager of NVIDIA's notebook division, Rene Haas and while the finer details of what's planned were retained, he did say that external graphics is something they are working on and the word 'Thunderbolt' was also mentioned, possibly meaning it'll be the connection of choice to ensure enough bandwidth is provided.

Fuad goes on to say that the conversation gave him the impression that external graphics may well surface sometime in 2012, using 28nm NVIDIA mobile parts (initially at least).

No matter how much you try to convince people that "the more memory, the better" when it comes to graphics cards is a somewhat fuzzy statement, the idea of having your card crammed with as much memory as possible is just too hard to pass up for some folks, hence add-in partners sometimes going a little overboard in this respect.

On that note, it seems NVIDIA card partner Point of View is looking to break some kind of record with a new GeForce GTX 550 Ti offering, packing it with not 1.5GB, not 2GB, but a whopping 4GB (!) of memory - equating to 4 times the amount on the reference model.

Such a large increase in memory would be questionably logical on a high end model, but on a mid range card with only a 128-bit memory interface, it does very much seem overboard indeed. It also needs to be mentioned that the memory used isn't the usual GDDR5, PoV took a shortcut here and went for DDR3, clocked in at 1066MHz. Meanwhile, the core runs at 900MHz, with its 192 CUDA cores operating at 1900MHz.

Some slightly fuzzy details on NVIDIA's upcoming 28mm mobile GPU parts have been leaked today; these being Fermi based shrinks, not based on next-gen Kepler.

It should first be noted, the chart provided by semiaccurate that you see above gives a list of the majority, but not all of the upcoming mobile GPU parts on the way, as there will likely be a few more 40nm parts entering the market in amongst this lot.

The other important attribute of what we can learn from this chart is the mass production periods noted. Sources are indicating that the much anticipated Kepler lineup of GPUs won't start being released until (at least) three months after these begin rolling out. So if we're not seeing 28nm parts hit the market until January or so, chances are Kepler will miss the boat for a Q1'12 release.

A particularly interesting looking HD 6850 graphics card has been found in Japanese stores by a company unheard of to most, AFOX. It's not just the brand name and box design that stands out, though. More interestingly, this card is quite a big step away from the reference design in that it is of low-profile stature and without the need for external power.

The low profile heatsink/fan design is also a standout feature with its dual-fan aluminum setup. AFOX appears to have left the stock clock rates on the card alone, however, which are 775MHz on the core and 4000MHz on the GDDR5 memory.

Due to being a low-profile card, it only has 1 x DVI and 1 x DisplayPort output. AFOX does ship the card with a separate full-profile bracket.

Following the launch of its DisplayPort to dual HDMI adapters yesterday, ZOTAC makes headlines again today with a couple more products.

In collaboration with water cooling specialists CoolIT, ZOTAC has introduced the liquid-cooled GeForce GTX 580 Infinity Edition which has CoolIT's OMNI self-contained water cooling system mounted.

ZOTAC ships the card with increased clock rates of 815/1630MHz on the core and shader, and 4100MHz on the 1536MB of GDDR5 memory. The card itself is otherwise much a standard fare with its 512 CUDA cores, 384-bit memory interface and dual-DVI and single mini HDMI outputs.

There have been rumors flying about NVIDIA's upcoming next-gen Kepler-codenamed GPU for a while now, that it would be delayed until 2012 because of issues with the adaption of the 28nm-based process. But, Chris Malachowsky spoke at their GTC Workshop Japan event, saying that Kepler is still on track for launch this year. Chris was careful with wording, saying only that the parts would begin "shipping" by the end of the year which could mean a paper launch and not an actual card-in-your-hands-giving-you-sexy-times launch.

Kepler is meant to deliver a threefold increase in double precision performance per watt over the previous Fermi, as well as being easier for developers to utilize GPGPU applications. Whatever, NVIDIA, just give me the damn cards now please.