One Kepler Later…

And we’re still not quite sure what the big one will look like…

Last week provided a pretty good stream of reviews, but by far the largest number of reviews focused upon Nvidia’s GTX 680. For a semi-complete list of those reviews you can head over to Overclockers.net. Now that that’s out of the way we can take a look at some of the other reviews from last week.

 

Up to this point we’ve seen a greater number motherboards based on AMD’s A75 chipset get reviewed than motherboards based on the lower end A55 chipset. Well last week Anandtech’s Brendan van Varik brought us a review of Gigabyte’s GA-A55M-S2V. This motherboard lacks most of the features that we’ve come to associate with Llano’s desktop incarnation, like USB 3.0 and SATA III. But it compensates for those missing check-box features by hitting the $65 price point. Despite its lack of features this A55 motherboard overclocks comparably to A75 motherboards and generally manages to keep up with them in Anandtech’s benchmark suite. It will be interesting to see if AMD’s upcoming A85 chipset manages to put more distance between itself and the A55 chipset than the A75 chipset could.

The guys over at Hardware.info have brought us the review that we’ve all been looking for since Nvidia’s GTX 680 launch last week, Quad-SLI versus Quad-Crossfire. As it turns out this is a pretty even match up really, with victory depending more on the game than on the graphics cards themselves. Additionally, dual and triple card configurations were benchmarked to assess the scaling of both AMD’s and Nvidia’s top end GPUs. Performance scaling is pretty good for cards from both companies, with Nvidia coming in at an average of 63 percent scaling per card and AMD coming in at an average of 69 percent scaling per card. As always, drivers tend have the biggest impact on the performance of these multi-GPU configurations, thus revisiting this match up in another six months might give us a clearer winner.

MSI LogoGrady McKinney of [H]ard|OCP got his hands on MSI’s recently launched HD 7970 Lightning video card. In the past MSI’s Lightning brand has been used on its top end offerings, which are often designed for extreme overclocking. This card is no different, and comes with MSI’s Twin Frozr IV cooling system, as well as a factory overclock to 1070Mhz Core and 1400Mhz Memory. Of course with all of these goodies MSI is charging a hefty $599 which is a good $50 more than the HD 7970’s MSRP. Unfortunately, in their testing [H]ard|OCP was only about to overclock the Lightning to 1190Mhz Core and 1475Mhz Memory. This overclock, while still in the general range for HD 7970’s, does not compare well to what some reviewers have been able to achieve with only HD 7970 reference boards. Overall though, there’s a lot to like about MSI’s Lightning, especially when it can hit clocks of 1.8Ghz Core under LN2.

AMD logoLegit Review’s Dan Sholtz took a look at CPU scaling with AMD’s HD 7950 last Monday. He tested with chips from Intel’s full desktop line up, from Pentium to i7. While some games showed up to a 30 percent benefit from using an i7-2600K rather than a mere Pentium, for the most part maxing out the settings within the given video game bottlenecked the game on the GPU side, which led to there being little real difference between Intel’s $60 chip and Intel’s $300 chip in terms of gaming performance. I doubt this is really that surprising to anyone, as this is a trend that’s seen often highlighted in CPU reviews like that of Intel’s Core i7-3960X.

Nvidia world iconDavid Kanter of Real World Technologies gave his impressions on Nvidia’s Kepler in a short article last Thursday. While Kepler’s focus on graphics performance over compute is well known, David takes a detailed look at the places where Nvidia chose to use the extra transistors gained in the 28nm process jump, and what places they passed over for improvements. He points to Kepler’s simplified hardware scheduler as an obvious example of Nvidia choosing graphics over compute when designing Kepler. Raising a question that many of us have been wondering, David asks if the Kepler core in the GTX 680 will reappear when Nvidia releases products aimed at the compute market. This is something that seems highly unlikely when you compare the compute performance of the Kepler core to its GCN counterpart. Despite the launch of GK104 the composition of Nvidia’s GK110 is still a bit of a mystery.

The Blog Post of the Week –

This week we turn to a post from AMD’s John Fruhe in support of the recently launched Opteron 3200 series. This post offers probably the best argument for the existence of the 3200 series I have yet to hear, and it’s an argument that’s centered mostly around cost. AMD is attacking the 1P hosting space by offering cheap chips, which is a strategy that it has probably perfected to some degree over the last half decade. Continuing the pattern of percentage based sales points that AMD’s marketing crew has developed upon since the launch of the 6200 series, we’re told that the 3200 series can offer 66 percent lower cost per core, and 38 percent better price/performance than competing solutions. Whether or not those numbers are worth noting, it’s probably for best that AMD can layout the most persuasive argument for its 3200 series of chips.

Down in the Fourms –

In the past week the most active topics were, “Trinity – Enhanced BD“, “Kepler Performance“, and my personal favorite from the week, “The GPU price-war that wasn’t“. Once again, If you’re in the mood for some occasionally heated discussion, and links to all the leaks worth looking at, the forums are the place to be.S|A

The following two tabs change content below.
Thomas Ryan is a freelance technology writer and photographer from Seattle, living in Austin. You can also find his work on SemiAccurate and PCWorld. He has a BA in Geography from the University of Washington with a minor in Urban Design and Planning and specializes in geospatial data science. If you have a hardware performance question or an interesting data set Thomas has you covered.