Intel cannonlake and icelake

[Bananasplit_00]

19 minutes ago, MyName13 said:

I highly doubt your MB is cheap, we are comparing 45-50$ h110 with 80-95$ b350, i was taking about cheaper motherboard like asrock ab350m and Msi gaming pro, maybe the can handle hexa cores now, but what about future ryzen CPUs?I wouldn't overclock more than 4 cores on cheap overclocking motherboards.

Everyone says that everything below 3ghz for ram is too slow for ryzen, not every country has the same availability as yours, I can find only 16 gb 3200 mhz kits , 8 gb sticks are unavailable or too expensive.4% increase in performance is nothing, it's not worth even for a 10$ price increase.Ryzen benefits more than that.

There are less B350 motherboards on the market that can't handle an overclocked 8 core then I have fingers on right hand. The only board I know is really bad is the super shit MSI one with a triple phase VRM, but that one is fine for hexa core chips. You also don't NEED faster RAM, it's nice, but you don't NEED it. 

 

16 minutes ago, TheBeastPC said:

It seems like the 15-30% performance increase on coffee lake is a common feature that intel has promised with each cpu that they have released. If the dyes are shrinking, wouldn't this limit overclocking capabilities by using the 10nm process in terms of cooling the cpu since the dye is so small?

Well they are adding cores so maybe not that much 

[Taf the Ghost]

19 minutes ago, TheBeastPC said:

It seems like the 15-30% performance increase on coffee lake is a common feature that intel has promised with each cpu that they have released. If the dyes are shrinking, wouldn't this limit overclocking capabilities by using the 10nm process in terms of cooling the cpu since the dye is so small?

Coffee Lake is 14nm++ node. This node really wasn't supposed to exist, but 10nm = 14nm+ but more expensive. That's why the first 10nm, which is Cannonlake, was canceled for mainstream. There was one use case, mid-level laptop CPUs, that it is being rolled out for in a few months.

 

Icelake will be a Node Shrink & a uArch upgrade. This is the first time, in a long while, that Intel has upgraded both at the same time. 

[pas008]

for starter people saying cinebench is great at showing raw horsepower

then why is 8350 winning in multi in cinebench but cant win in multitasking benches here

https://www.bit-tech.net/reviews/tech/amd-fx-8350-review/5/

http://techreport.com/review/23750/amd-fx-8350-processor-reviewed/9

 

 

also another thing how can anyone logically state intel wasnt ready for mass multicore cpus in the extreme series when they incorporated the mesh design into skx, isnt mesh used in mass multicore cpus? lol

 

[Guest]

What are the benefits of using the 10nm process?

[XenosTech]

14 minutes ago, TheBeastPC said:

What are the benefits of using the 10nm process?

Better power consumption would be my wild guess and *hopefully* better temps due to the cpu's drinking less power but who knows we could have another kaby lake heat crisis again.

[Mr.Meerkat]

22 minutes ago, pas008 said:

for starter people saying cinebench is great at showing raw horsepower

then why is 8350 winning in multi in cinebench but cant win in multitasking benches here

https://www.bit-tech.net/reviews/tech/amd-fx-8350-review/5/

Oi oi oi, do I see another use for my favourite article?

 

As you can see in the performance conclusion section, the 8320 OCed to 4.2GHz can match a stock 6600K in productivity applications which basically means that your article is too old and you should be looking at newer articles as programs have been optimised for more cores over the years, making cinebench more relevant actually as its scales very nicely with more cores/threads.

 

I mean the article you linked has 7-zip as part of the score and well...the FX trashes the 6600K like there's no tomorrow on it so...ermmmm  

https://www.ocaholic.ch/modules/smartsection/item.php?itemid=4003&page=7

 

29 minutes ago, TheBeastPC said:

What are the benefits of using the 10nm process?

Potential benefits:

- Lower power consumption

- Higher clock speeds

- Better performance (as 'signals' get to travel a smaller distance)

- More dies per wafer(?)

 

49 minutes ago, Bananasplit_00 said:

The only board I know is really bad is the super shit MSI one with a triple phase VRM

Welll..although its bad, its still 'good enough' to run a 1700 on it although I'd recommend zero overclocking  

[Guest]

1 minute ago, Mr.Meerkat said:

Better performance (as 'signals' get to travel a smaller distance)

I guess so, since the dye is smaller and signals have a shorter distance to travel to communicate with each other. 

[Dylanc1500]

23 minutes ago, XenosTech said:

Better power consumption would be my wild guess and *hopefully* better temps due to the cpu's drinking less power but who knows we could have another kaby lake heat crisis again.

Temps are going to be dependent more so on how much intel decides to cram into the die. With the extra space from the shrinkage they may just shove more transistors into the cores.

[AnonymousGuy]

1 hour ago, Taf the Ghost said:

Coffee Lake is 14nm++ node. This node really wasn't supposed to exist, but 10nm = 14nm+ but more expensive. That's why the first 10nm, which is Cannonlake, was canceled for mainstream. There was one use case, mid-level laptop CPUs, that it is being rolled out for in a few months.

 

Icelake will be a Node Shrink & a uArch upgrade. This is the first time, in a long while, that Intel has upgraded both at the same time. 

Source on the cancelled for mainstream?  2,4, and 8 core parts were in the pipeline at one point (and when I say pipeline I mean the silicon exists).

[Bananasplit_00]

17 minutes ago, Mr.Meerkat said:

Oi oi oi, do I see another use for my favourite article?

 

As you can see in the performance conclusion section, the 8320 OCed to 4.2GHz can match a stock 6600K in productivity applications which basically means that your article is too old and you should be looking at newer articles as programs have been optimised for more cores over the years, making cinebench more relevant actually as its scales very nicely with more cores/threads.

 

I mean the article you linked has 7-zip as part of the score and well...the FX trashes the 6600K like there's no tomorrow on it so...ermmmm  

https://www.ocaholic.ch/modules/smartsection/item.php?itemid=4003&page=7

 

Potential benefits:

- Lower power consumption

- Higher clock speeds

- Better performance (as 'signals' get to travel a smaller distance)

- More dies per wafer(?)

 

Welll..although its bad, its still 'good enough' to run a 1700 on it although I'd recommend zero overclocking  

Yes it's fine for everything at stock, but not once you overclock. Also more transistors per wafer would be more accurate then more dies

[Mr.Meerkat]

5 minutes ago, Bananasplit_00 said:

Also more transistors per wafer would be more accurate then more dies

Well technically both are correct as the same amount of transistors on a smaller wafer = more dies per wafer  

 

Its down to how you word it really...

[Bananasplit_00]

2 minutes ago, Mr.Meerkat said:

Well technically both are correct as the same amount of transistors on a smaller wafer = more dies per wafer  

 

Its down to how you word it really...

Well you don't need to keep transistor count, you might shrink the node by 30% but then just make the chips have 30% more transistors or 50% more or you could have less of them. 

[pas008]

49 minutes ago, Mr.Meerkat said:

Oi oi oi, do I see another use for my favourite article?

 

As you can see in the performance conclusion section, the 8320 OCed to 4.2GHz can match a stock 6600K in productivity applications which basically means that your article is too old and you should be looking at newer articles as programs have been optimised for more cores over the years, making cinebench more relevant actually as its scales very nicely with more cores/threads.

 

I mean the article you linked has 7-zip as part of the score and well...the FX trashes the 6600K like there's no tomorrow on it so...ermmmm  

https://www.ocaholic.ch/modules/smartsection/item.php?itemid=4003&page=7

 

think you are missing the point of the multitasking,

article is old but still shows us higher multithreaded doesnt equal multitasking raw power
in real life many of us are using a mixture of single or multi threaded programs

point is cinebench isnt really showing real performance

its one or the other not combination

 

to me its just another benchmark that gives rough idea

because we all dont use one or the other

[XenosTech]

29 minutes ago, Dylanc1500 said:

Temps are going to be dependent more so on how much intel decides to cram into the die. With the extra space from the shrinkage they may just shove more transistors into the cores.

Not necessarily, it comes down to the glue they're using to keep the IHS on the die. They could cram a trillion transistors in it and we'd still be able to cool it, but fuck up the glue and as MageTank and others have noted it'll screw temps by a lot.

[King_of_Oz]

And, as he approached the lake, an EPYC vessel RYZED from the waters, like a moth RIPPING the THREADS in cloth.

[Mr.Meerkat]

Just now, pas008 said:

think you are missing the point of the multitasking,

article is old but still shows us higher multithreaded doesnt equal multitasking raw power
in real life many of us are using a mixture of single or multi threaded programs

point is cinebench isnt really showing real performance

its one or the other not combination

Ah, get your point now and well you've answered your own question already...

 

1 hour ago, pas008 said:

for starter people saying cinebench is great at showing raw horsepower

Cinebench is a benchmark that shows the raw horse power of a certain part, it was never design to show how something would perform in a real life scenario (to a certain point) so...  

[pas008]

2 minutes ago, Mr.Meerkat said:

Ah, get your point now and well you've answered your own question already...

 

Cinebench is a benchmark that shows the raw horse power of a certain part, it was never design to show how something would perform in a real life scenario (to a certain point) so...  

thats my point with links to show otherwise

then, but???

and showed links because if the raw multithreaded power should have showed otherwise in multitasking

[Taf the Ghost]

39 minutes ago, AnonymousGuy said:

Source on the cancelled for mainstream?  2,4, and 8 core parts were in the pipeline at one point (and when I say pipeline I mean the silicon exists).

https://www.digitaltrends.com/computing/intel-coffee-lake-processor-14nm-hexacore/

 

That's Coffelake making the rounds in July, 2016. But with mention of Cannonlake in the ultra-low power Laptops.

 

https://www.kitguru.net/components/cpu/anton-shilov/intel-readies-ice-lake-processors-with-integrated-voltage-regulator/

 

July, 2015. Cannonlake "mainstream" was canceled. Coffee Lake would be added in 2016.

 

It gets messy because there's really not official announcements. It's just roadmap leaks. For stuff like Intel's server CPUs, they actually have working copies at least 2 years out already in the hands of their top-tier partners. 

[ravenshrike]

 

2 hours ago, TheBeastPC said:

It seems like the 15-30% performance increase on coffee lake is a common feature that intel has promised with each cpu that they have released. If the dyes are shrinking, wouldn't this limit overclocking capabilities by using the 10nm process in terms of cooling the cpu since the dye is so small?

The 30% performance increase comes from shifting a mobile part from 2 cores to 4. In terms of IPC it looks like 1-2% at most with maybe a higher overclock when delidded.

[VagabondWraith]

13 hours ago, ARikozuM said:

Everyone also took up video editing lately, as well. 

Zen itself bred 1 million new streamers & video editors  

[MyName13]

18 minutes ago, ravenshrike said:

 

The 30% performance increase comes from shifting a mobile part from 2 cores to 4. In terms of IPC it looks like 1-2% at most with maybe a higher overclock when delidded.

How can you get a 30% performance increase from a 100% core increase? 

 

 

11 minutes ago, VagabondWraith said:

Zen itself bred 1 million new streamers & video editors  

Seriously, what's with that streaming and multitasking meme lately?

[Dylanc1500]

1 hour ago, XenosTech said:

Not necessarily, it comes down to the glue they're using to keep the IHS on the die. They could cram a trillion transistors in it and we'd still be able to cool it, but fuck up the glue and as MageTank and others have noted it'll screw temps by a lot.

I'm basing that off of a "perfect" cooling solution. The more transistors and voltage you pump into a chip the more heat is produced. If they keep everything the same and just shrink the node it will bring the amount of energy needed down producing less heat.  Of course if someone puts a crappy cooling solution on any chip it's going to heat up.

[ravenshrike]

11 minutes ago, MyName13 said:

How can you get a 30% performance increase from a 100% core increase? 

Dropped base clock. It's a low power laptop part.

[MyName13]

7 minutes ago, ravenshrike said:

Dropped base clock. It's a low power laptop part.

Is there any point of these cpus?Don't most laptop users need 2 fast cores instead of 4 slower ones?I'm worried that CL will be only for laptops (why did they even put a laptop on their announcement photo...).

[ARikozuM]

3 minutes ago, MyName13 said:

Is there any point of these cpus?Don't most laptop users need 2 fast cores instead of 4 slower ones?I'm worried that CL will be only for laptops (why did they even put a laptop on their announcement photo...).

Some of us do need quad-cores in our laptops for work and not having to search for HQ is going to make things easier.