Jump to content

Wikipedia:Reference desk/Archives/Computing/2016 August 16

From Wikipedia, the free encyclopedia
Computing desk
< August 15 << Jul | August | Sep >> August 17 >
Welcome to the Wikipedia Computing Reference Desk Archives
The page you are currently viewing is an archive page. While you can leave answers for any questions shown below, please ask new questions on one of the current reference desk pages.


August 16

[edit]

Will AMD Zen architecture mean improvements to their GPUs as well?

[edit]

AMD will release processors with a new architecture that is a great improvement on previous iterations. Will the same transistor technologies also apply to their graphics cards, also producing 40% increase in instructions per clock like with the CPUs? --78.148.98.177 (talk) 05:19, 16 August 2016 (UTC)[reply]

I'm fairly confused what you're referring to. AFAIK, there's no magic transistor technologies resposibility for AMD Zen's improvements, and a quick read of our article and associated article supports this view. AMD never managed to move to the tick-tock cycle used by Intel (which has also been forced to slow it down due to the recent slowing in process improvements) so it's fairly common for themr to launch major microarchitecture changes along with moves to newer process nodes. And they abandonded their own nodes several years ago due to their major financial issues so they have limited ability to make their own improvements anyway, instead having to rely on what's available commercially. And as AMD production levels have gone down and others like ARM vendors have gone up, AMD isn't exactly a vital customer even for GlobalFoundries who had problems of their own anyway, although I guess they were still important enough to I think be the major launch customer for GlobalFoundries's 14 nm FinFET licenced from Samsung. So as is common with AMD, Zen improvements are surely due to a multitude of factors. Actually frankly, AMD would have been hit so badly with Intel's Core 2 Duo if their only problem was process nodes, remembering that the launch Core 2 Duos were 65 nm and AMD was at 65 nm at the time or nearly. (Although to be fair Intel moved to 45 nm a fair amount faster than AMD, but that was clearly just compounding problems AMD already had.)

In other words, the move to 14 nm FinFET obviously helps, but general microarchitecture improvements also help. Notably it's generally accepted that AMD Bulldozer was a bit of a disaster, as the Clustered Multi-Thread idea along with other design choices like the pipeline length simply didn't work as well as intended for a multitude of reasons. Also it's is fairly old by now with only small improvements and the AMD Bobcat (and it's successors like Jaguar) not much better and neither of these were particular stellar even at launch, and AMD needed both of these because the Bulldozer design simply wasn't suited for very low power designs due to the design choices (especially because of the high idle usage) and likewise it's not clear how well Bobcat would scale (similar to the various Atoms). Sure there were some improvements but these were fewer and more limited compared to Intel's general practices. And in terms of proccess nodes used, AMD simply couldn't afford to move much, they were still on fairly old processes compared even to many ARM, let alone Intel.

On the GPU side, AMD has already moved to 14nm FinFET with the AMD Polaris, the AMD Radeon 400 series has already launched down the mid end AMD 460. You can check out reviews for these on many sites. So really, AMD is a bit ahead in terms of process nodes on their GPUs since Zen only has leaked engineering samples AFAIK, definitely not a launched product. (Although designing and fabbing a GPU is a little different from a SOC CPU anyway.) Again though, it would be a mistake to assume all improvements are simply due to the process node change, and ignore the relevance of general microarchitecture changes and differences.

Nil Einne (talk) 08:41, 16 August 2016 (UTC)[reply]

Google maps artifacts

[edit]

When I go to Google Maps directly, it has lots of visual artifacts[1] making it borderline unusable. At this point even Bing maps is better. How do I fix this?

It's nothing wrong with Google, my browser, or my computer because if I go to a third party site that relies on Google Maps e.g. [2] it looks perfectly fine[3].Crudiv1 (talk) 19:41, 16 August 2016 (UTC)[reply]

I don't see anything wrong with that map. What problem are you seeing? Rojomoke (talk) 22:09, 16 August 2016 (UTC)[reply]
Lots of jagged edges like it's an impressionist painting or something. It's especially apparent around the Swiss alps. Crudiv1 (talk) 23:25, 16 August 2016 (UTC)[reply]
Googlemaps are to my mind the more reliable as they display the current political status. If Bing shows unequivocal borders, then it is out of date. Some more info here: Understand country borders and names. If you find the Goolemaps borderline unusable why? For what purpose are you needing to know borders to such précisément? --Aspro (talk) 22:22, 16 August 2016 (UTC)[reply]
In the most common usage, borderline has nothing to do with political borders. ―Mandruss  22:32, 16 August 2016 (UTC)[reply]
Are you using Internet Explorer? Apparently Google maps can have display problems if IE is set to use Compatibility View. This help page describes how to fix this problem. Good luck. Ca2james (talk) 03:49, 17 August 2016 (UTC)[reply]
I'm on Firefox. I just tried Chrome and it has the same problem[4].

To editors who can't see the problem: The border of San Marino is a single straight line here[5]. I don't expect a free service to have beautiful cartography or even have up-to-date political borders, but it should, in the very least, not have basic errors like this. Crudiv1 (talk) 04:50, 18 August 2016 (UTC)[reply]

Just checked the first map I posted and it has the same San Marino border error. Crudiv1 (talk) 04:52, 18 August 2016 (UTC)[reply]

Not sure if this is what's being discussed, but I have experienced something like this too in satellite view, "an impressionist painting or something." per Crudiv1 particularly strikes a chord. It looks like the actual trees and other foliage are replaced by computer generated 'stand-ins'. 220 of Borg 19:22, 20 August 2016 (UTC)[reply]