Donanimhaber looks to have gotten a hold of some more slides for us this week. This time they’re about AMD’s low end 28nm APU’s. The slides show you pretty much what you’d expect from these APU’s, lower power, higher performance, and new form factors. Although the small performance gain (20 percent), moving from a current generation dual core, to a next generation quad core, makes you wonder about the quality of the core to core interconnect. Or, maybe the 20 percent increase they’re quoting is when you do a direct core to core comparison. It’s hard to say, but at this point we can say that we have a very limited understanding of how this chip is going to perform.
Michael Larabel of Phoronix did a compiler performance comparison on AMD’s A8-3850. While the results of this comparison don’t mean much to a Windows user, the article is still extremely interesting because it exemplifies the difference in performance your application’s compiler can make.
AMD’s banging the Open API drum again; this time at Gamescon in a short interview with Bit-tech’s Joe Martin. Sasa Marinkovic, of AMD, is quoted as saying, “If there’s an open standard, they just don’t work. OEM’s don’t like them and developers don’t like them either.” Although, the funniest part of this article is watching to commenters duke it out.
Xtreview is confirming some details about AMD’s upcoming FX-8150 and FX-8120 SKUs. Both chips are based off of the B2 stepping with TDPs of equal to or less than 125 watts. The 8150 is sporting a base clock of 3.6 Ghz and a turbo of 4.2 Ghz, while the 8120 comes in at a 3.1 Ghz base and a 4.0Ghz turbo. Those speeds are a bit better than the 3.2 Ghz base speeds we were hearing a few months back.
When it comes to Intel’s SOCs and graphics products the story is always the same; it’s the drivers. According to a report from LG Nilsson of VR-Zone that’s exactly the reason for the Cedar Trail Delay and why it will be launching with DX 9 drivers, and not with DX 10.1 drivers like it’s hardware supports. I’d like to point out that Intel could of saved itself a whole bunch of money by just not including the DX 10.1 silicon on the chip; instead of building hardware that’s just going to sit there and suck up power until they get their act together. I mean, come on guys, it’s absolutely exasperating to watch a multibillion dollar company with almost complete supremacy in its market get a whole product line tripped up on the drivers. If AMD and Nvidia can do it, it seems like a forgone conclusion that Intel should be able to.
It looks like Intel finally figured out how to fix the bug with its new SSDs that keeps reducing the space available on the drive to a mighty 8MB. Good on them for supporting their products and coming up with a fix. Although, it’s hard to say what the long term impact of an issue like this one will have on the strength of Intel’s brand in the SSD market.
In somewhat surprising news Xtreview is reporting that Nvidia’s upcoming Kepler based chips will not be its first off the 28nm production line from TSMC. But rather a few Fermi chips will be paving the way for the next generation. Nvidia did this previously with its transition to 40nm by producing a few rather small G200 based chips before popping out a Fermi. It will be interesting to see if Nvidia keeps most of these 28nm Fermi for the OEM’s like it did during the 40nm transition.
Fuad Abazovic of Fudzilla is reporting the Nvidia wants to get its GPUs back into Apples products. “In order to do that, you need to have a great GPU and have it on time.” is supposedly what Nvidia’s CEO told Fudzilla’s sources. I guess we were right when we told you their problems at Apple back in June of 2009. It seems that the reports of Nvidia being market share hungry are right on the money. Now the question is, will AMD be able to hold on to the gains they’ve made?
Fudzilla also had a story this week on Kepler’s production schedule, stating that Kepler will go into production late this year and will be on the shelves by early 2012. This lines up with the messaging that’s been coming out of Nvidia, and atleast the sampling date lines up with what others have been speculating.
Nvidia announced the poaching of a Cray executive recently. Steve Scott will be coming to Nvidia as the CTO of the Tesla business unit. It will be interesting to see how he influences the Tesla roadmap, as he has a background in interconnects. Keep an eye out for some interesting announcements coming from Nvidia, about its Tesla product line, after they get Kepler off the ground.
A new report that came out this week is making a few waves in the browser wars. NSS Labs found that in their security testing Internet Explorer 9 detected 99.2 percent of socially engineered malware, as compared to Chrome 12, which picked up 13.2 percent, and Firefox 4, which picked up 7.6 percent.Although, the question of which browser is the safest is still up to debate, despite these findings, but this is good news for the team at Microsoft.
Mozilla released a new Aurora build to much fanfare this week. With tweaks to improve security, more HTML5 support, memory usage, and performance it’s definitely worth a look if you’ve got some time to spare.
Techreport’s Scott Wasson has done it again with detailed exposé on Crysis 2’s DX 11 patch. Finding that quite a few objects are tessellated to the sub pixel level for no apparent improvement in visual quality, he questions the quality of Crysis 2’s DX 11 implementation, and even considers the possibility of Nvidia’s influence on the development of the patch. Quite a few things look better with the Crysis 2 DX 11 patch, but what’s the point of having heavily tessellated water running underneath the game map where you’ll never see it; or tessellating a flat window sill? If you want a perfect example of how console ports are ruining DX 11 look no further than this article. Oh, and pick up the Crysis 2 SDK if you want to experiment with it.
One of the guys over at [H]ard|OCP snagged a rather interesting interview with the CEO of Unlimited Detail. Unlimited Detail is a company that keeps popping up and promising the world, but never actually turning any of their demos into products. Even John Carmack weighed in on the work Unlimited Detail has been advertising in a recent interview. Both are definitely worth a read if you want to see a guy defend himself from some well-conceived criticism.
Techreport has put up a wonderful comic this week. It offers a flash back to Intel’s marketing in the days of Pentium 4, Nvidia’s marketing in the days of the FX series, and then compares it to AMD’s marketing for Fusion. While the specific similarities and dissimilarities are debatable, I’m glad to see that someone still has a sense of humor, and for that matter that Nvidia’s not the only one making cartoons about the market place.
Xtreview seems to have gotten word that Apple’s upcoming quad core A6 chip won’t be coming out of TSMC’s fabs until Q2 2012. It rather difficult to verify this information, but seeing as what we know now about TSMC’s 28nm process, as well as circumstantial evidence from rumors about AMD’s new GPUs slipping into Q1*, and Nvidia outright confirming Q1 as the launch for its new chips, Q2 for Apple chips on the same process doesn’t really sound all that farfetched.S|A
*Editor’s note: Charlie’s sources dispute this timing.