TweakTown NewsRefine News by Category:
SAPPHIRE is a force to be reckoned with when it comes to their Toxic-branded video cards, and it looks as though the SAPPHIRE HD 7970 Toxic is going to be no different, and maybe even step it up a bit when it comes to smashing the ball out of the park. The card comes with a resolution-busting 6GB of memory in 24 GDDR5 memory chips, 12 on each side.
In order to power the factory-overclocked "Tahiti" GPU from AMD, and 24 GDDR5 7 GT/s memory that takes some seriously strong VRM. Sapphire decided to implement an 8+3 phase power supply, which uses solid-state chokes, that don't whine under stress, as well as International Rectifier DirectFETs. The FRM draws power from two 8-pin PCIe connectors. Sapphire went a step further by implementing LEDs for each of the 8 vGPU phases, which gives a real-time indication of their individual loading. These LEDs can be seen through a window on top of the card, as pictured below.
The cooler is built from a large aluminum fin heatsink which draws heat from teh GPU and memory chips on the obverse side of the PCB. The FETs on the obverse side are cooled by additional heatsinks. From here, we have the memory chips on the reverse side of the PCB cooled by a metal back-plate. The heatsink makes use of four 8mm-thick nickel-plated copper heat pipes, as well as a vapor-chamber plate. This is ventilated by two 80mm fans.
With the release of the kick-ass GTX 680 at a competitive price point, many thought, including me, that the HD 7970 would undergo a price cut to be more competitive. Let's be realistic: the GTX 680 beats the HD 7970 in almost every single benchmark, including being more efficient, at a lower price point. It's pretty hard to justify charging $50 more for a worse card.
Or is it? Again, let's be realistic. The GTX 680 can hardly be purchased anywhere and when it is in stock, it's bought out in a matter of hours. This was something the HD 7970 had trouble with as well up until a few weeks before the 680 launch. Then it suddenly got plenty of stock. So while the 680 can hardly be purchased anywhere, the HD 7970 has the luxury of being widely available.
So, there's a few conclusions that can be drawn from this. First, AMD thought Nvidia would continue their practice of pricing higher performing cards at a premium. Second, it could be that AMD can't afford to drop the price on the 7970. The silicon is a pretty complex chip, and as such is expensive.
Additionally, the current pricing scheme places each series $100 apart. A $50 drop in the 7970 pricing would require a similar drop across the board. As of yet, Nvidia doesn't have any competing chips at the lower mainstream price level where the majority of sales are. So AMD can either sacrifice the HD 7970 or they can sacrifice the rest of the line up to save the HD 7970.
Time for another rumor, so go ahead, get the salt out and throw some over your shoulder, and for good measure, find some wood and knock on it. Now that we're done with that, let's move onto the goods. We have what we believe to be the specifications for the Kepler based GK106. This chip should end up being a direct competitor to the HD 7950, HD 7870, GTX 560 Ti, GTX 560 Ti 448 and GTX 560.
The specifications are said to be as follows:
- 28nm TSMC lithography based on Kepler
- 210mm2 die size
- Two GPC
- Four SMX
- 768 CUDA cores
- 64 texture units
- 24 ROPs
- 1.5/2GB RAM
- 192-bit memory interface
- No frequencies have been specified
The card is expected to land in Q3 of 2012 which is still a fair distance off but could be just in time for the summer upgrade season. If it performs anywhere like the GTX 680, we should have another good card on our hands. It will be interesting to see if they price it at a competitive price like they did with the GTX 680.
According to 'Gibbo' on the Overclockers.co.uk forums, we should expect the new NVIDIA GEFORCE GTX 670Ti and GEFORCE GTX 670 to launch late-April, or early-May. These two GPUs will be placed into price segments aimed at AMD's Radeon range of GPUs.
The GTX 670Ti will replace NVIDIA's GeForce GTX 580, and take on the HD 7950 3GB. The GTX 670 on the other hand, replaces the GTX 570 and should be slightly faster, too. The GTX 560Ti and 560 are not expected to be replaced with 600-series GPUs until much later into 2012.
The low-end 520/550 cards will reportedly be rebranded into the 600-series, where we should expect the usual same card, re-boxed as 600-series GPUs. Gibbo also mentions the dual-GPU NVIDIA beast, which we should expect to arrive as the GEFORCE GTX 690, and "can be released when NVIDIA desire to do so". The roadmap that Gibbo is coming from says that the GTX 680 is the fastest single-GPU card, but a faster single-GPU card is expected later this year.
It looks like we're not quite done with news about the Nvidia GeForce GTX 680. As our review showed, the card is pretty amazing and is still efficient while being amazing. Pictures have surfaced of a Galaxy-brand GTX 680 which is cooled by a single-slot cooler. This is made possible by the amazing efficiency of the new Kepler architecture.
195w TDP is pretty high to be handled by a single-slot cooler, so frequencies should be expected to be reference or below. This device certainly will be nice for smaller, mid-tower PCs or for a cramped PC in which you want to run SLI. No release date or specifications have been released. For all we know, this could be a prototype that never sees the market.
Earlier on today we met up with the folks from Palit and Gainward at their Taipei headquarters to discuss all things NVIDIA and GeForce since they now (finally) have a new GPU model to pimp and sell, that of course being the GeForce GTX 680.
During our meeting someone (we never actually got introduced) walked into our meeting and dropped off a second video card, besides the first Palit one we were looking about - but that Palit model needs to stay quiet a little longer yet, we may have already said too much.
Anyway, back on topic - what the man dropped off was a Gainward GTX 680 Phantom video card. On first inspection I asked if it was passively cooled, quickly realizing that would be near on impossible due to the fact that the GTX 680 is a high-end GPU part and generates a lot of heat. On closer inspection we saw that the Phantom actually includes two cooling fans built inside or at least just below the massive radiator / heatpipe design which covers the full top surface of the video card...
Kepler truly was a make-or-break launch for NVIDIA. With their competiton, AMD, squeezing out a full top to bottom line up before NVIDIA could even manage one launch, they had to have something wonderful and able to compete with AMD's offerings. Enter Kepler. This card has been roughly four years in the making. Rumors have been flying rampant regarding its projected performance.
Finally, launch day arrived (today). We were the first to publish a comprehensive review of the new card, and what a card it is. It's faster and more efficient than AMD's top card, the 7970, and it even overclocks better, at least with extreme cooling solutions. The launch was a little bit messy, what, with us not getting a card and stuff, but overall, NVIDIA had a successful launch of the GTX 680. So much of a great launch, that the CEO sent out an email company wide. You can see the email below the line.
Welcome to the future of graphics processing technology. Today, NVIDIA released the next-generation Kepler series with their current flagship card, the GTX 680. Following along with this release, NVIDIA has introduced their new mobile offerings in the form of the GeForce 600M series. It consists of several rebrands along with a bunch of new Kepler goodness.
The top two offerings, the 675M and the 670M, are basically rebrands of the 580M and 570M. It's likely that NVIDIA will release a Kepler based 680M sometime in the future, but currently, the top Kepler based card is the 660M. The 640M has some confusing aspects surrounding it. It is offered in both a 40nm and 28nm variant with differing core counts under the same name. I'm not 100% sure on how these will be told apart in market.
Most of these chips will not see the light of day until manufacturers start shipping notebooks based upon Intel's upcoming Ivy Bridge, but NVIDIA notes that some of the offerings are already available in Sandy Bridge based notebooks. The list of notebook manufacturers using Nvidia graphics include Acer, Alienware, Asus, Dell, Gigabyte, HP, Lenovo, MSI, Samsung, Toshiba, and Vizio.
We've had quite the eventful pre-launch news nuggets for the upcoming 28nm Kepler-named GEFORCE GTX 680 GPUs, but this is a totally new step in a very good direction: Newegg are the latest to jump the gun and offer the new GEFORCE GTX 680 GPU's for sale.
The cards are on sale for anywhere between $499.99 and $534.99, they have ASUS, GALAXY, ZOTAC, EVGA, GIGABYTE, PNY, and MSI cards listed. Availability is of course going to sink like hell on this as this news starts spreading, as everyone will be flocking to Newegg to hit up some of that Kepler lovin'. Get in quick, and remember to send me three of them, ok? The interesting thing I found is, when I've searched for them - they're gone. So it looks like Newegg have pulled them from the site - so I would keep your eyes peeled on this one, folks! They should pop up any second/hour/day now!
To throw in some extra awesome sauce on this news, we also have vice president of NVIDIA, Ujesh Desai presenting the GEFORCE GTX 680, in the video below:
Taiwan Semiconductor Manufacturing Company (TSMC), is filled to the brim with 28nm orders, and is now likely to expand the process capacity later in the year, according to industry sources, reports DigiTimes. The sources have said that TSMC is reportedly running at full capacity at its 12-inch fabs, due to strong orders for 28nm, as well as 40nm and 65-based designs.
TSMC, instead of having to avoid orders and lose sales to United Microelectronics (UMC) and Samsung Electronics, the company will have to speed up the pace of it's leading-edge capacity expansion in particular, its 28nm capacity, according to these sources. The sources have added that a number of fabless and IDM companies have approached UMC and Samsung regarding their capacity for 28nm processes, while finding out that TSMC is unable to satisfy their demand due to its tight supply.
The firm is expected to spend $6 billion on capex in 2012, revising previous numbers. TSMC also stated during their most recent investor meeting that 28nm process technology could account for 5-percent of total wafter revenues in Q1 2012, up from 2-percent in Q4 2011.