Graphics Cards - Page 254

Render. Rinse. Repeat! That's AMD's new recommended method of hair care. The company could have just very well ushered in a new era of 3D graphics with its new hair processing technique dubbed TressFX. The process showcases more "realistic" and "natural" hair for in game characters.

AMD says that the new technique allows for individual strands of hair to be rendered out, and everything will flow more naturally with unique physics properties and collision detection. Long gone are the days when you have to suffer trough Laura Croft's pony tail bouncing within a seemingly flat plane.

AMD's blog stated:

We have some good news for you, NVIDIA are set to unleash their latest GPU, the GeForce GTX Titan on February 18. This is just days away, so if you had any cash saved from not buying Valentine's Day presents, you should look at getting one of these puppies.

Now, before you get too excited - this isn't the GTX 800 series, nor is it a refresh, nor is it the Maxwell-based GPUs we should expect later this year. According to TechPowerUp's database, this is what we should expect:

What should we expect performance-wise from the GeForce GTX Titan? Well, we should expect 2688 stream processors, 224 texture units, and 48 raster operating units. The Titan should slam its hammer down with 6GB of GDDR5 memory running at 6GHz on a 384-bit bus. This is the GK110 chip, folks, featuring over 7 billion transistors laid down on a 521mm2 of space and will have clock speeds pumping at 875MHz.

During an interview with Japanese site 4Gamer.net, AMD's Product Manager for Desktop Graphics Products, Devon Nekechuk, revealed some startling information: that AMD won't be launching any new GPUs this year.

AMD will instead concentrate on their already great Radeon HD 7000 series, by adjusting the price and performance increases through driver releases. A slide released to 4Gamers.net shows that AMD's Radeon HD 7900 series (high-end), HD 7800 series (performance), and HD 7700 series (mainstream), will carry on the company's mantle "throughout 2013."

It looks like we're seeing a slowing down of GPU releases, which I'm guessing is because there's hardly any games that push GPUs anymore. It's also incredibly expensive in terms of R&D for GPU makers to continue pushing new GPUs out every 6-12 months, but I think the real reason is that a $200-$300 GPU will play most games at 1080p@60 now.

There's a benchmark score floating around the internet that was reportedly earned by NVIDIA's upcoming GeForce Titan. The number shows the new GK110-based GPU besting the dual-GPU GTX 690 by a hefty margin. In 3DMark11, the card reportedly earned a score of X7107. For comparison, a GTX only achieves a score of around X6000.

The details can't be confirmed because the site reporting this won't release their source. The Titan is said to be coming with 6GB of RAM on a 384-bit memory bus. The naming scheme is said to retain the GeForce Titan nomenclature, possibly as a reference to the Cray Titan supercomputer that houses just shy of 19,000 NVIDIA Tesla K20X cards.

Pricing is said to be $899 with availability starting at the end of February.

ZOTAC have just launched a new GPU, named the GeForce GTX 660 Thunderbolt Edition. With 'Thunderbolt' in the name, most people would expect that ZOTAC have baked in some Thunderbolt interface support, but nope - it has nothing to do with the high-speed connectivity called Thunderbolt.

The new ZOTAC GeForce GTX 660 Thunderbolt Edition, does however, arrive with a high-grade non-reference design PCB, and a very cool new cooling solution. The cooling solution sports a massive aluminum heatsink with heat-transfer helps by three copper heat pipes. This gives us a hybrid-type device, somewhere between heatsinks and fin-stacks. The pictures below show off how a hybrid heatsink is better than fin-stacks.

ZOTAC calls the cooler features "EClean", which is a mechanism that allows owners of the GPU to easily detach the cooler shroud on which the fans are mounted by simply pressing a retention notch, and sliding out the shroud that's suspended on rails on the main heatsink. All of this provides the ability to clean the heatsink, very, very nice.

A new rumor is saying that NVIDIA is readying a GK110-based GeForce Titan card to go toe-to-toe with the upcoming AMD HD 8000 series of GPUs. The new card would consist of a single GPU die and would have performance between the current GTX 680 and GTX 690, which is two GTX 680 dies on one card.

The move is likely to steal some thunder from AMD, who will try to take the performance crown back from NVIDIA. Several sources have said the card will materialize late next month with a price tag of $899. The name comes from the Titan supercomputer, which features 18,688 nodes utilizing NVIDIA's Tesla K20X GPU.

The GK110 die powers NVIDIA's enterprise Tesla video cards. The final die configuration will feature 2688 CUDA cores, an impressive number and over 1000 more than the GTX 680. Of course, this is all a rumor, though a lot of it does make sense. The card is likely utilizing chips that weren't quite up to par with the other Tesla chips as these cards have one SMX unit disabled.

I'm not a fan of rebadging GPUs, as you can see here with our AMD Radeon HD 8000M Series news, but it looks like NVIDIA have also pulled out their huge green rubber rebadging stamp, too. Le sigh.

NVIDIA are pushing new parts out to OEMs, with two being unveiled thus far: The GeForce GT 730M and GeForce 710M, there aren't many details on them at all, but AnandTech have used their magic powers to try and figure out what we should expect, as the picture above shows.

The 730M looks like a GK107-like rebadge, which we have now as the 640M, which supports Kepler-based tech such as TXAA, PCIe 3.0 and DisplayPort 1.2. This could end up being a 650M rebadge, but we won't know for a little while yet. The other part listed is the GeForce 710M which should be a GF117 rebadge, and should end up being a rehashed 620M. This chip doesn't support Kepler features such as PCIe 3.0, TXAA or resolutions above 2560x1600.

CES 2013 - AMD's new Radeon HD 8000M is something special for mobile computing, as its the first mobile GPU that is based on the Graphics Core Next (GCN) architecture. These GPUs are built on the 28nm process and feature DirectX 11.1 support, as well as the cool AMD tech found in ZeroCore and Enduro.

The Radeon HD 8000M lineup will include 8500M models (384 stream processors, 650/1125MHz), 8600M series models (384 stream processors, 725/1125MHz), 8700 series models (384 stream processors, 850/1125MHz) and a higher-end 8700M that will feature 640 stream processors, 700MHz core and only uses GDDR5.

AMD have pulled their rubber rebadging stamp out of the 'why aren't we over this yet' cupboard, and are currently stamping rebrands of the HD 7000 family into the HD 8000 family to OEMs from top to bottom. Even the Radeon HD 8970 is a damn rebadge of the Radeon HD 7970 GHz Edition which sports 2048 stream processors, 1GHz core clock and 3GB of GDDR5 clocked at 6GHz, all sprinkled on a 384-bit bus.

CES 2013 - ASUS released the latest edition of its ARES Graphics Card line today and it, as usual, is a beast. The ROG ARES II is the fastest most powerful GPU available to date, and it looks almost as good as it preforms!

The ARES II features exclusive ASUS DIGI+ VRM and Super Alloy Power technologies for greater stability and product longevity. This GPU Monster is powered by two AMD Radeon HD 7970's, 6GB GDDR5, and is water cooled from the factory. Clocks are set at 1050MHz Base, 1100MHZ Boost, and 6600MHz Memory. Max resolution is set at 2560x1600 through a DVI connection.

ASUS has only manufactured 1000 of these beauties and each one is numbered for collect-ability. No word yet on pricing for the ARES II but with the previous ARES and MARS cards selling in the $1500 range, it is not far fetched to think the ARES II will be in that same range, if not higher.

Just days before the world reportedly ends, leaked details of AMD's upcoming Radeon HD 8000M "Solar System" range of mobile GPUs have made their way onto the shores of the Internet.

The new mobile GPUs are set to launch for CES 2013, and will be based on the new Graphics CoreNext (GCN) micro-architecture. GCN supports DirectX 11.1, Enduro Technology and AMD ZeroCore power technology. Enduro is an NVIDIA Optimus-like tech which switches between discrete and integrated GPUs while the ZeroCore tech can completely power down GPUs when they're not needed.

Peering over the slides, we can see there is a new ASIC which covers three product lines - HD 8500M, HD 8600M and the HD 8700M. The sure-to-be-sweet silicon sports 384 GCN stream processors, a 128-bit memory interface that supports both GDDR5 and DDR3. The HD 8500M series sports GPU clock speeds of up to 650MHz, with the 8600M cranking it up to 775MHz. The 8700M series goes one better and sees clock speeds hitting 850MHz.

I'm sure you didn't see this one coming, did you? Neither did I, don't worry. Spanish site BitDreams.se has posted what they are claiming to be the specs on AMD's upcoming next-generation dual-GPU, the Radeon HD 8990.

The leak reports that we should expect the HD 8990 in Q2 2013, meaning it is not far away at all. The leaked roadmap unveils that there are two Radeon HD 8800 series GPUs to be shipped this quarter, and with just weeks left, if it did happen, it's most likely going to be a paper launch - if it happens at all.

Now that we've got the lower-end gear out of the way, let's get straight to that succulent GPU meat - the Radeon HD 8990 "Venus XTX" which will see two Radeon HD 8970s bound onto the single GPU. The card is teased as sporting a whopping 5.1 billion transistors, 5210 stream processors and 160 texture units per core. The HD 8990 is listed as featuring 48 ROPs per core, and two 384-bit buses. Memory should reach the dizzying heights of between 6GB and 12GB and should come factory clocked at 1250MHz. The GPU cores should be set at 950MHz if the leaked roadmap holds truth to it.

We've been sitting on the AMD Radeon HD 7000 series for a while now, and like anything else, that is set to change in the coming months. DigiTimes' latest report is that AMD plans to release their next-gen HD 8000 series in Q2 of 2013.

The company are currently going through a business reorganization in order to help their poor performance from the quarter just gone, according to DigiTimes' sources from the "upstream supply chain". DigiTimes reports that AMD were set to unveil their Radeon HD 8000 series at the end of the year to compete with NVIDIA, but pulled back.

The Chinese market is huge, and both companies want it - obviously. AMD's market share in the country has dropped to just 30%, and they're hoping that the launch of the next-gen GPUs will swell those numbers up considerably.

The thrifty among our readers will be happy to hear that the AMD Radeon HD 7850 has seen another price cut, bringing the retail price for the 2GB version down to $189. Along with that cut, the rest of the product line continues to trend downwards as the series ages, the process matures, and competition heats up between AMD and NVIDIA.

Anandtech has produced the nice chart you see above. The chart shows how the prices of AMD's offerings have declined continually since launch by comparing launch price, spring MSRP, late summer MSRP, and fall retail price. AMD has changed from quoting MSRP to quoting retail price as that usually is $10 lower than MSRP.

While the prices on the HD 7000 series started out really high, as of late they've become extremely reasonable due to the continual price cuts that AMD has been implementing. With the HD 8000 series due out early next year, prices will likely continue to fall on the HD 7000 series, though probably not until the holidays or just past.

NVIDIA have pushed forward something called the Green Light program, which is a certification process that is designed to ensure that GPUs meet specific requirements, reports the Bright Side of News.

The Green Light program requires vendors to send in their board designs to NVIDIA, where they wait for their approval from the GPU maker. NVIDIA will check that the design meets their noise, power, voltage and heat numbers - and if it meets them, the card is approved. If not, it is declined and they'll have to resubmit.

What happens if a partner doesn't submit their card for approval? Well, they can have their GPU warranty and BIOS support removed, or worse - they could have their allocation of GPUs from NVIDIA cut down, or stopped completely.

Here I was thinking we wouldn't see much toward this time of the year on the GPU landscape, but AMD are already close enough to the launch of their Radeon HD 8800 series with sources now leaking out some information on silicon that is codenamed "Oland".

Oland is built on the same 28nm process, throwing in 3.4 billion transistors onto a die-area of around 270 mm². If the news is solid, we're looking at the Radeon HD 8870 "Oland XT" outperforming its predecessor, the HD 7870 in both performance per watt, and cost-performance ratios.

If the chart is right, the HD 8870 could begin offering performance close to that of today's high-end offerings, which would really shake up the market. The HD 8870 clocks in at 1050MHz, with 1100MHz PowerTune Boost frequency, with its little brother the HD 8850 clocked at 925MHz with 975MHz boost frequency. Both SKUs have memory clocking in at 6GHz, providing an insane 192GB/s of memory bandwidth - even at a proposed $199 price for the HD 8850.

NVIDIA has had to push back its launch schedule due to troubles with the manufacturing processes. Originally, Kepler was supposed to launch in 2011, however, it is still being launched to this day and we've almost hit 2013. No new architecture will be produced until Maxwell in 2014.

Maxwell is an interesting animal. It will be a 20nm chip, which isn't a big deal. However, it is being designed so that it can be manufactured at GlobalFoundries, IBM, Samsung and TSMC. This should allow NVIDIA to ensure a large enough supply, or make a change if one of the fabs is having process issues.

The GTX 700 series should be a similar launch schedule to what has so far occurred with the GTX 600 series. This means we won't see anything until March 2013 or later. But the good news is that the new chip should between 25-30% higher performance and power efficiency. The Kepler refresh will have to compete with AMD's Sea Islands, so the real battle will take place in 2014 with Maxwell taking on Sea Islands.

It's about time that someone went and made a 7990. We've been hearing that AMD has them coming for a long time now, but they have failed to deliver. That is, until today. PowerCooler has announced the world's first dual-Tahiti GPU and is calling it the 7990. Does this mean that we may finally start seeing them from other manufacturers?

For right now, PowerCooler is billing this card as the first and only 7990 in the world. The specifications are pretty beefy and really provide what everyone was expecting. The Devil 13 HD7990 has default settings at 925MHz engine clocks and 1375MHz memory clocks and sports 6GB of GDDR5.

If you're looking to buy a new AMD graphics card, the time could not be better. AMD has again dropped prices on its high-end cards so that they are more competitive with NVIDIA's latest offering. With the NVIDIA GTX 660 Ti finally outed by the company, AMD wants to position its cards in places where they can compete.

The street prices have dropped accordingly:

The 7970 does not see a price drop in this latest round, the both the 78XX and 7950 have had their prices lowered. The 3GB 7950 price drop brings its price down to a level similar to the GTX 660. In tests, a 2GB 7950 and 660 Ti were about 50/50. The 3GB version will likely pull out a couple of extra wins, and with the new pricing, should win business.

Contrary to rumors flying around the internet, Sapphire has not cancelled the 6GB 7970 Toxic graphics card they produce. It seems that because one major reseller put "discontinued" instead of "out-of-stock", or something similar, enthusiasts freaked out that they wouldn't be able to get the super fast card.

At Gamescon, however, Sapphire has a large booth and sitting right in the middle as the pride and joy of the booth was the 6GB 7970 Toxic card. KitGuru asked Bill Donnelly for a little more information regarding the rumor flying around the web. The short answer? It's not discontinued, just facing production constraints.

Production is supposedly running smoothly and cards should start to become available again soon. In the meantime, KitGuru has provided us with the pretty lady above.

The red camp fires back. AMD has just announced an updated version of the 7950 in the same vein as the 7970 GHz edition update. What this means is that the card is getting faster base clocks as well as a new feature that boosts the core clock when the TDP isn't fully loaded. With AMD and NVIDIA nearly matched on performance and AMD losing in power efficiency, they've decided to go for the best performance-per-dollar.

So, aside from throwing power consumption out the window, what else have they changed? Not much really. The new base clock is up to 850MHz from the original 800MHz. Boost clock speeds allow the clock speed to jump to 925MHz. The reason power usage will jump is they have increased the core voltage to support these speeds.

With the 7970, it wasn't too bad as they had top bin parts. With the 7950, they are dealing with either broken chips or chips that wouldn't hit clock speed requirements. Because of this, AMD is using 1.25v which is higher than the 7970GHzE's 1.218v. It will be interesting to see how this increase affects benchmark results.