New CPU Breakthrough Could See a 2000% Performance Boost In 16xCore Processors

[IAEInferno]

Just found this while searching for CPU news.

 

As usual, take this with a grain of salt.

 

 

Source: http://www.game-debate.com/news/21677/new-cpu-breakthrough-could-see-a-2000-performance-boost-in-16xcore-processors

[Tech_Dreamer]

inb4 $80000 consumer cpu

[SCHISCHKA]

this authors claims of improvements for games; he is a dumb ass. the improvement comes for large number of cores not 4 cores

[Guest]

AMD take note

[Energycore]

12 minutes ago, IAEInferno said:

Just found this while searching for CPU news.

 

As usual, take this with a grain of salt.

 

 

Source: http://www.game-debate.com/news/21677/new-cpu-breakthrough-could-see-a-2000-performance-boost-in-16xcore-processors

Source is game-debate. You can basically assume that it's false with little risk of being wrong

[Mira Yurizaki]

On one hand, inter-core communication is a thing you need to watch out for when looking for performance caps in multi-core design. One of the biggest issues for instance being cache coherency.

 

On the other hand, the author of the article sounds like he has no idea what he's talking about.

 

EDIT: I mean the classy part is "if this technology were backwards compatible". It's not going to be. It's a hardware problem. That's like asking if we can implement HyperThreading on a 80486

[Enderman]

Just fyi, if you ever see an article on the internet that says "breakthrough" you can automatically translate that to "not gonna happen in the next decade or probably ever".

[Wolther]

5 minutes ago, Energycore said:

Source is game-debate. You can basically assume that it's false with little risk of being wrong

https://www.extremetech.com/computing/238586-using-hardware-queues-break-multi-core-bottleneck

http://techreport.com/news/30898/researchers-work-their-way-around-multi-core-software-bottlenecks

 

extremetech and techreport also did report outs on it about a ~week ago

 

I feel like this will be something we see far into the future, and who knows where they got the 20 fold numbers from.. and how many cores they correlated that with.

[iamdarkyoshi]

2 minutes ago, Enderman said:

Just fyi, if you ever see an article on the internet that says "breakthrough" you can automatically translate that to "not gonna happen in the next decade or probably ever".

This right here ^^

[ARikozuM]

2 minutes ago, Enderman said:

Just fyi, if you ever see an article on the internet that says "breakthrough" you can automatically translate that to "not gonna happen in the next decade or probably ever".

Like carbon transistors?

 

#ShatteredDreams

[Energycore]

1 minute ago, Wolther said:

https://www.extremetech.com/computing/238586-using-hardware-queues-break-multi-core-bottleneck

http://techreport.com/news/30898/researchers-work-their-way-around-multi-core-software-bottlenecks

 

extremetech and techreport also did report outs on it about a ~week ago

These two, I will read.

[AresKrieger]

This is total BS since modern 4 core processors even in theory would only be able to be out performed by a factor of 4 ie 400% by a 16 core processor if both were used to 100% efficiency

[Guest Darth Revan]

So I guess this isn't AMD, right?

[Enderman]

1 minute ago, ARikozuM said:

Like carbon transistors?

 

#ShatteredDreams

Like "new battery technology increases capacity and less charging time!" and "modular smartphones!" and "super high efficiency CPU!" etc...

[AresKrieger]

Just now, Darth Revan said:

So I guess this isn't AMD, right?

Not only is it not amd it is not likely written by a person who understands how cpus work

[Wolther]

2 minutes ago, AresKrieger said:

This is total BS since modern 4 core processors even in theory would only be able to be out performed by a factor of 4 ie 400% if both were used to 100% efficiency

Author obviously didn't know what he was talking about. I'm thinking the percent increase will be from the diminishing returns from much larger core counts. 

[Herman Mcpootis]

2 minutes ago, Enderman said:

Like "new battery technology increases capacity and charging time!" and "modular smartphones!" and "super high efficiency CPU!" etc...

local university develops batteries that can charge to 70% in 2 minutes http://media.ntu.edu.sg/NewsReleases/Pages/newsdetail.aspx?news=809fbb2f-95f0-4995-b5c0-10ae4c50c934

that was 2 years ago, where's my quick charging battery 

[NeoSk11zle]

6 minutes ago, herman mcpootis said:

local university develops batteries that can charge to 70% in 2 minutes http://media.ntu.edu.sg/NewsReleases/Pages/newsdetail.aspx?news=809fbb2f-95f0-4995-b5c0-10ae4c50c934

that was 2 years ago, where's my quick charging battery 

do you mean qick charge cause that already exists m9

[Enderman]

4 minutes ago, herman mcpootis said:

local university develops batteries that can charge to 70% in 2 minutes http://media.ntu.edu.sg/NewsReleases/Pages/newsdetail.aspx?news=809fbb2f-95f0-4995-b5c0-10ae4c50c934

that was 2 years ago, where's my quick charging battery 

These things never happen for a ton of reasons.

The only time you should expect a technology to actually exist commercially is when a market date is actually announced and that date is less than 6 months.

All the ones that say "coming in one year!" or two years, or stuff like that also just randomly disappear and never happen.

[mrchow19910319]

Another "breakthrough tech found in batteries related fields that can make your phone last 100 years without recharging" topic.

Nope. Not interested at all. 

[IAEInferno]

Welp, I got clickbaited :/.

[GoodBytes]

Moved to CPUs, Motherboards, and Memory, as it doesn't follow the Tech News section posting guidelines.

 

[AresKrieger]

13 minutes ago, gabrielcarvfer said:

Parallel computing is not 16 cores running independent code, but 16 cores running code that exchanges data. That kind of workload get benefited because some cores get stopped to run a protocol if they need to exchange data with other cores. That kind of protocol usually needs to flush caches until the last level cache (the one that is shared with all cores) so that the other core can access the data and allow everyone to continue working.

If they did a harware to replace the protocol that was executed by the cores, of course it speeded up the data exchange between them and the overall performance. Remember when DMA was introduced? It's something similar.

Still isn't going to net more performance than 100% of theoretical values would be it will merely get the numbers closer to the max possible than it was without such tech, efficiency cannot exceed 100%

[AresKrieger]

18 minutes ago, gabrielcarvfer said:

snip

And what I was saying 16 cores can never be more than 400% more performance that 4 cores if the cores are being utilized properly hence the article's title is BS, whether the method causes improvement or not is irrelevant to my point as the wording is incorrect in the article

 

Additionally in games it is more or less impossible to run everything in such an extreme parallel fashion as old calculations affect newer ones and are needed immediately following the old this is why implementations of multiple physical cores if done just for the sake of it often causes stutter

[DocSwag]

I don't really see how this would give a 2000 times performance uplift... From a performance standpoint, it's only increasing the efficiency of having multiple cores, so to have a "2000" times performance uplift, wouldn't you be comparing a 2000 core CPU to a single core CPU?

 

EDIT: Me stupid. It say 2000%, not 2000 times. Still, you would at best be comparing a 20 core CPU with a single core one, which isn't a common scenario...