Comparison of the AMD PRO 460 and the Nvidia 960M:
https://www.youtube.com/watch?v=jw7B2x-PN0E
It looks like in the 3DMark tests the PRO 460 is ahead and about the same in GTA V and Nvidia tends to do better in that game AFAIK.
Development cost of FinFETS? Not convinced they are any different than before. You write your VHDL, compile it for the target process, get the masks made. Those masks are expensive and the fab costs are expensive thanks to the multi patterning, but design is hard on any process.
I'm assuming information like this is remotely correct: http://chipdesignmag.com/sld/files/2...ts-Cadence.jpg
It's not a very recent source but it reflects the sort of things companies are saying about increasing development costs. I don't know for sure though, obviously.
Of course, I'm referring to the 14/16nm FinFET nodes as a whole vs e.g. 28nm, not just planar vs FinFET for a given node.
I think the 'compile it for the target process' is perhaps a bit of an oversimplification and an area which has been getting substantially more complex with stricter design rules; features have to be laid out in increasingly specific ways in order to be manufacturable, factoring in things like resolution, interference patterns, multiple patterning and its alignment, etc. E.g. as I understand it, you can't just place features wherever you want them due to being well below the wavelength of the light used in the lithography (and haven't been able to for a long time, it's just getting more specific); things like DRAM and NAND, having mostly repetitive patterns, often make it to volume production on a node first because of this.
Working within those rules from the fabs presumably means designing/purchasing increasingly complex design software/compilers, more processor time, and more bug checking before you even make it to a fab?
Did AMD give up on the idea of getting AM4 motherboard on the market before Zen or something?
Looks like the Smach Z is not vaporwear anymore:
https://www.kickstarter.com/projects.../posts/1735755
Specifications:
CPU: AMD Merlin Falcon RX-421BD (12-15w) SoC at 2.1 GHz
Cores: 4/4
iGPU: Radeon R7 at 800 MHz
RAM: 4GB (SMACH Z) || 8GB (SMACH Z PRO) DDR4 2133 MHz
HD: 64GB (SMACH Z) || 128GB (SMACH Z PRO)
Screen: 6” FULL HD (1920x1080). Capacitive touch-screen.
Battery: 5 hours of gaming.
MicroSD Card Slot.
USB 3.0 type C.
HDMI video output connection.
Wi-Fi connectivity 5.0 Ghz. 4G LTE mobile network connectivity (PRO model only)
Bluetooth connectivity.
Front-camera 5-megapixels (PRO model only)
Edit!!
It also looks like the WX5100 is bus powered:
https://forum.beyond3d.com/posts/1952015/
So a 1792 shader Polaris 10 based card with a single slot cooler.
Last edited by CAT-THE-FIFTH; 21-11-2016 at 03:39 AM.
Possible name for Zen based CPUs:
https://trademarks.justia.com/871/15/amd-87115584.html
??
Edit!!
More performance tidbits:
https://translate.google.co.uk/trans...n-3491050.html
Last edited by CAT-THE-FIFTH; 22-11-2016 at 12:39 AM.
Probably to early to tell but is there any info on the new Zen CPUs if they will have the same temperature monitoring system as the FX CPUs? Or will they revert back to a individual core temp sensors?
Hmm,thats very interesting indeed,the FX8300 and FX8310 are rated for nearly 70C instead of the 61C of the earlier chips:
http://www.cpu-world.com/CPUs/Bulldo...20FX-8300.html
http://www.cpu-world.com/CPUs/Bulldo...20FX-8310.html
It makes me wonder whether these are like the late Phenom II X6 chips like the 1045T which were the best of the line?? Both are 95W TDP chips too.
Looked on YT someone got an FX8300 upto 4.7GHZ!
Rated spec though. Dropping the core clocks can mean the same part can tolerate higher temperatures, allowing a cheaper heatsink. I doubt it is anything specific to these parts, CPU-World is claiming C0 stepping still for all these parts. So yeah you expect process improvements to help, but in a quick look I didn't see any mention of 32nm let alone 32nm SOI on the GlobalFoundaries web site, so I suspect the 83[01]0 parts are old silicon from the back of the warehouse packaged up with different fuses blown.
OTOH, back in the Athlon days AMD remarked a load of end of life Opteron parts as Athlons. If that was happening here, then that could mean top bin silicon that has been through extra testing.
More to the point, the 8300 is a £100 part?? I paid not much more than that for my 8350 years ago, I can't believe how these parts are keeping their price so close to going obsolete.
Like I said someone hit 4.7GHZ with theirs and this sounds awfully like the Phenom II X6 1045T which was around £100 near the end of the Phenom II X6 production run and despite the locked multiplier were very good overclockers. Apparently on the aluminium cooler of fail and a 780G motherboard you can hit 4.2GHZ looking on YT??
Edit!!
The FX6300 is still around £100 too.
Yeah I remember those. Lots of people were talking down the 8350 thanks to the poor 8150 benchmarks, saying the X6 was a better purchase. So glad I didn't listen, the 8350 seems to have gotten better over time.
But still, if this was a year ago I might have gotten a bit excited, but if Zen is out in January and with OEMs already using AM4 it just seems irrelevant.
Sorry if this is old news but Wccftech has a speculative article on the release of Zen chips early next year:
http://wccftech.com/amd-zen-8-core-35ghz-cpu-spotted/
The real winner could be the 6core/12 Thread 65w Zen going possibly for $249 I'm thinking? Although if AMD want to make head way in the i5 arena, $199 might be more reasonable.
There are currently 11 users browsing this thread. (0 members and 11 guests)