SoylentNews is people
SoylentNews
from the can-i-still-root-for-them-if-they're-no-longer-underdogs? dept.
AMD Ryzen Improves IPC by 52%, not 40%
AMD Ryzen 7 desktop chips are now available for pre-order, and will be released generally on March 2nd. The cheapest "Ryzen 7" 8-core chip will be $329, but has 16 threads just like the $399 and $499 versions (there had been some concern that it might not have multi-threading). Ryzen 3 and 5 series quad-core and hex-core chips will be released later.
AMD held a "Tech Day" to share details about its new chips. Over the past year or two, AMD has said that their goal was to improve instructions per clock (IPC) by 40% with Zen/Ryzen. Now they are saying that they have achieved that... by improving IPC by 52%. It's an impressive number that would not have been possible had Ryzen's predecessor, the Bulldozer architecture, been competitive with Intel's CPUs (on single-threaded performance). However, keep in mind that comparisons between Ryzen and Bulldozer or certain Intel Core i7 CPUs are still based on information provided by AMD.
The 8-core design features 4.8 billion transistors and "200 meters of wiring".
YouTuber removes footage of Ryzen Overclock World Record
YouTube tech pundit Austin Evans uploaded a video earlier today containing footage of AMD's Ryzen launch event. Shortly afterwards, the video was made private, then replaced with a re-edited version of the piece with a couple of changes. A section in which Austin outed the Vega video card by name was edited and overdubbed with the description "a really cool graphics card".
Another section of the video showed an overclocking competition held at the event with a number of well-known LN2 overclocking experts seeing just how far they could push the new Ryzen chips. This whole section was removed, including footage of "a Ryzen chip" running at just over 5.1GHz, followed by the OCers celebrating as a run on Cinebench's R15 multi-threaded performance test set a new world record.
Comments on the video asking why the video was replaced and uploaded confirm that the Vega mention and OC section were removed. Austin Evans did not explain exactly why these changes were made, merely saying "YOU SAW NOTHING" in the comments section.
The 5.1GHz figure is from my memory. I think it might have been 5.16GHz more exactly, but I'm not completely sure about that and obviously can't verify it from the original video.
(Score: 1, Insightful) by Anonymous Coward on Thursday February 23 2017, @11:10PM
Its my understanding that all future AMD chips will have AMD's PSP technology in them. This is their equivilent of Intels AMT/MT/vPro remote control tech. If this is correct, I find it quite disheartening, as now pretty much all new mainstream CPUs are backdoored from the factory. :(
I wonder if CPU manufactures are even allowed to make CPUs without this tech anymore ... There certainly seems to be a niche market for it.
Another option would be to open source the firmware that controls the co-processor (master processor??) and/or allow a user to upload their own keys to it, so that alternative firmware could be loaded by it. It would open the door for a more secure system, and a lot of different alternative uses for that chip.
(Score: 0) by Anonymous Coward on Friday February 24 2017, @02:58AM
There is no reason they couldn't make the X models unkeyed and just use writelocking of the BIOS to ensure security, except for 3rd party DRM implementations. Which is really what it comes down to. It isn't remote management capabilities that are causing the signed firmware fiasco. It is DRM sidechannel concerns from the lame duck media content producers who want to bring us into 1984 just so they can chase these supposed larger profits from preventing piracy.
Given the restrictions on Ryzen, and the new VBIOS signing on the 5k/6k Intel processors, I will probably be buying old G34 or Xeon hardware that either doesn't contain ME provisions, or at least can have it disabled to the maximum extent possible.
Pricing for Ryzen already puts it in the realm of just buying G34 chips, even if they will be much lower performance both IPC and max clock. Honestly though 16-64 threads is far beyond what I really need even today though.
(Score: 2) by RamiK on Friday February 24 2017, @04:20PM
Or just buy a network card along with a new PC so the packets won't go through the micro-controller.
Of course, that's only secure if everything you're running blob-less FOSS without enabling any DRM features.
compiling...