CPU Anything to get excited about for from Cannon/Coffee Lake?

[JosephEK]

Intel is supposedly launching two new desktop platforms by the end of this year, Coffee Lake which is yet another 14nm refinement and Cannon Lake which is on a new 10nm process.

I feel Ryzen has trounced Kaby Lake in value aspects at 14nm, but maybe 10nm will give Intel some breathing room as far as having higher efficiency? And I wonder if they will finally solder their heat spreaders or whatever that's causing some people to de-lid ther CPUs for better temps from what I'm hearing. Ryzen doesn't seem to have that problem.

So yeah if the choice is Kaby Lake vs Ryzen, I'm going team Red to upgrade from my i5 3450, but If Intel's going to be replacing Kaby Lake soon I wonder if I should wait and see. I only need quad cores so if they can make one that uses even less energy than an i5 7400 I might be interested.

 

[NateDawg72]

The Pentium thing may just have been a rumor.
http://techreport.com/news/32225/the-pentium-g4560-is-alive-and-well-says-intel

I like more options so I'm looking forward to see what coffee lake brings but I'm not expecting much.

To be entirely honest, my criteria for considering coffee lake is going to come down to 1 thing: socket/chipset compatibility. If I can't drop it into the Z270 board I have a Pentium g4600 in then I won't even consider coffee lake. Most likely a used Skylake/kabylake i7, or just maybe Ryzen refresh since it may be awhile before I retire my LGA 1155 system.

 

[JosephEK]

The Pentium thing may just have been a rumor.
http://techreport.com/news/32225/the-pentium-g4560-is-alive-and-well-says-intel

I like more options so I'm looking forward to see what coffee lake brings but I'm not expecting much.

To be entirely honest, my criteria for considering coffee lake is going to come down to 1 thing: socket/chipset compatibility. If I can't drop it into the Z270 board I have a Pentium g4600 in then I won't even consider coffee lake. Most likely a used Skylake/kabylake i7, or just maybe Ryzen refresh since it may be awhile before I retire my LGA 1155 system.

Oh okay, I heard the Pentium thing from watching RandomGamingInHD, but if it's false or still just a Rumor than I'll delete my reference to it for not being constructive.

I'm still using i5 3450. I wanted to upgrade to Ryzen 3 or 5 1400 because I want to go from ATX to ITX and be compatibe with DDR4 and M.2 NVMe drive. Idon't do anything that justifies more than 4 cores though.

 

[Hifihedgehog]

I feel Ryzen has trounced Kaby Lake in value aspects at 14nm, but maybe 10nm will give Intel some breathing room as far as having higher efficiency?

If I may, this has been an ongoing discussion with several forum members I know at TabletPCReview for the last several months and the conclusion we have reached is Intel Cannonlake will have maybe a few months of relevance before AMD makes it outmoded for a good year or two.

...

Here are the facts:

1) Just as with their 14-nanometer process, Intel's 10-nanometer process is going to be fashionably late. The most recent news points to a Q1 2018 release window and, as history has shown, Intel is well known for missing targets even this far out. In addition, mobile products (Y-class and U-class ULP parts) will undoubtedly be the first to come followed by several more months of waiting before desktop products finally hit the market. Plan on Q2 2018 for the desktop counterpart of Cannonlake.

2) AMD will be releasing desktop products manufactured with 7-nanometer technology in 2018. GlobalFoundaries, AMD's spun foundry partner, already states early 2018 as the go signal for the first commercial products.

3) Intel, meanwhile, has no public roadmap yet for 7-nanometer technology and only has announced a plant as of this year. Current estimates peg Intel at a 2020 to 2022 expected date for commercial production (http://wccftech.com/intel-losing-process-lead-analysis-7nm-2022/). Suffice it to say, Intel is on the verge of being leapfrogged and engulfed by the entire industry, with AMD above and ARM underneath. Unlike what Fool insists in their tech articles, I highly discourage purchasing Intel stock since the company is on a crash course for failure.

...

In short, current market trends show that AMD is on track to become the next new big dog dominant player in the x86 desktop market. It thus would be highly advisable to invest in their platform since developers will be turning their focus from Intel to optimize their products to AMD and thereby tap into the power of these Zen-based products which will be flooding the market.

PS: EPYC smackdown by AMD. Intel had this coming for years after all the underhanded monopolistic practices that landed them a huge fine from the EU and almost sent AMD to their grave. Karma hurts, eh?
http://www.anandtech.com/show/11544/intel-skylake-ep-vs-amd-epyc-7000-cpu-battle-of-the-decade/21

 

[NateDawg72]

I'm still using i5 3450. I wanted to upgrade to Ryzen 3 or 5 1400 because I want to go from ATX to ITX and be compatibe with DDR4 and M.2 NVMe drive. Idon't do anything that justifies more than 4 cores though.

With the performance of ryzen and kabylake I ended up just getting a used i7 3770 for ~$140. Performs similar to the quad core ryzens for games and saved money that would've gone to a motherboard and DDR4 by reusing my lga1155 board which is itx thankfully.

 

[JosephEK]

If I may, this has been an ongoing discussion with several forum members I know at TabletPCReview for the last several months and the conclusion we have reached is Intel Cannonlake will have maybe a few months of relevance before AMD makes it outmoded for a good year or two.

...

Here are the facts:

1) Just as with their 14-nanometer process, Intel's 10-nanometer process is going to be fashionably late. The most recent news points to a Q1 2018 release window and, as history has shown, Intel is well known for missing targets even this far out. In addition, mobile products (Y-class and U-class ULP parts) will undoubtedly be the first to come followed by several more months of waiting before desktop products finally hit the market. Plan on Q2 2018 for the desktop counterpart of Cannonlake.

2) AMD will be releasing desktop products manufactured with 7-nanometer technology in 2018. GlobalFoundaries, AMD's spun foundry partner, already states early 2018 as the go signal for the first commercial products.

3) Intel, meanwhile, has no public roadmap yet for 7-nanometer technology and only has announced a plant as of this year. Current estimates peg Intel at a 2020 to 2022 expected date for commercial production (http://wccftech.com/intel-losing-process-lead-analysis-7nm-2022/). Suffice it to say, Intel is on the verge of being leapfrogged and engulfed by the entire industry, with AMD above and ARM underneath. Unlike what Fool insists in their tech articles, I highly discourage purchasing Intel stock since the company is on a crash course for failure.

...

In short, current market trends show that AMD is on track to become the next new big dog dominant player in the x86 desktop market. It thus would be highly advisable to invest in their platform since developers will be turning their focus from Intel to optimize their products to AMD and thereby tap into the power of these Zen-based products which will be flooding the market.

PS: EPYC smackdown by AMD. Intel had this coming for years after all the underhanded monopolistic practices that landed them a huge fine from the EU and almost sent AMD to their grave. Karma hurts, eh?
http://www.anandtech.com/show/11544/intel-skylake-ep-vs-amd-epyc-7000-cpu-battle-of-the-decade/21

So AMD is bypassing 14nm for 7nm with Zen 2 in 2018-19? That's a lot less time on 14nm than Intel spent. IIRC that's very close to the limit capable on Silicon before weird stuff starts happening. Maybe I should stay on Ivy Bridge just a bit longer and wait for that if it's even more efficient than Ryzen. Would Zen 2 work on AM4? I personally like AMD more than Intel, and I love AMD coming back from the brink the way they did. Intel is going to get punished and they deserve it, but I personally don't like to see someone who loses so bad that they're never able to recover. Fair fights are so much more fun to watch.

With the performance of ryzen and kabylake I ended up just getting a used i7 3770 for ~$140. Performs similar to the quad core ryzens for games and saved money that would've gone to a motherboard and DDR4 by reusing my lga1155 board which is itx thankfully.

My i5 3450 originally came in an Alienware X51, but the ITX motherboard died and it was too difficult/expensive to source an ITX alternative so I went ATX with a 20 dollar case and a space PSU my younger brother had sitting around. Been trapped at "Large Form Factor" ever since until I'm able to make the money needed for this grand general platform upgrade plan I've got. I would intend for this Ryzen system to last me just as long as Ivy Bridge did: ~5+ years if not longer if Moore's Law is grinding to a halt.

 

[Phuncz]

PS: EPYC smackdown by AMD. Intel had this coming for years after all the underhanded monopolistic practices that landed them a huge fine from the EU and almost sent AMD to their grave. Karma hurts, eh?
http://www.anandtech.com/show/11544/intel-skylake-ep-vs-amd-epyc-7000-cpu-battle-of-the-decade/21

I'm really happy EPYC is looking so promising for the future, in a few years we'll be looking at replacing most servers at work and I was hoping for more choice in flavour than plain vanilla.

 

[jeshikat]

AMD will be releasing desktop products manufactured with 7-nanometer technology in 2018. GlobalFoundaries, AMD's spun foundry partner

Intel, meanwhile, has no public roadmap yet for 7-nanometer technology and only has announced a plant as of this year.

Keep in mind they measure differently. GF's 7nm is roughly equivalent to Intel's 10nm.

 

[Hifihedgehog]

Keep in mind they measure differently. GF's 7nm is roughly equivalent to Intel's 10nm.

Also keep in mind Intel has taken liberal efforts in their marketing to exaggerate their processes' size advantages. Even so, Ryzen at 14 nanometers is on par with Broadwell IPC. At 7 nanometers next year, Ryzen 2/Zen 2 should be able to hold its own against Intel's best 14- and 10-nanometer designs.

Maybe I should stay on Ivy Bridge just a bit longer and wait for that if it's even more efficient than Ryzen. Would Zen 2 work on AM4?

Reports are claiming a 15% IPC improvement over Zen 1/Ryzen besides the inherent power savings from the 7-nanometer process. The point being, if you are willing to hold out just that little bit longer, it may be well worthwhile. As for AM4 socket compatibility, that I am not entirely sure of.

Intel is going to get punished and they deserve it, but I personally don't like to see someone who loses so bad that they're never able to recover. Fair fights are so much more fun to watch.

Recall Intel was the one refused Apple's offer to work on the original iPhone project, who exited the ARM mobile market prematurely, and who then chose for a time to wholly devote themselves to the X86 market. It is a self-afflicted fate and a combination of poor tactical errors, but I am certain Intel will always be in the game, albeit never quite again with the form or edge they once had over this last decade.

 

[jeshikat]

Also keep in mind Intel has taken liberal efforts in their marketing to exaggerate their processes' size advantages. Even so, Ryzen at 14 nanometers is on par with Broadwell IPC. At 7 nanometers next year, Ryzen 2/Zen 2 should be able to hold its own against Intel's best 14- and 10-nanometer designs.

Absolutely, but I also think it's a bit disingenuous to say Intel is going to be leapfrogged since GF will be at 7nm while Intel is "only" at 10nm.

Though what I'm really curious is what happens beyond 7nm. Will the fabs eek out another node shrink or two? Will we finally go graphene? Optical? Biological? Exciting stuff!

 

[Kmpkt]

Graphene CPUs to go with graphene heatsinks IMO. Also the one thing that might be appealing about Coffee Lake is they are rumoured to have a 6 core variant for those that can use it. That being said, Ryzen at 8c/16t is going to be pretty hard to beat with that chip.

 

[Hifihedgehog]

Absolutely, but I also think it's a bit disingenuous to say Intel is going to be leapfrogged since GF will be at 7nm while Intel is "only" at 10nm.

I should have also qualified my statements with the tidbit that AMD is giving Intel, who is on a third generation, technically superior 14-nanometer design, a good fight with their "inferior" 14-nanometer technology. With a "smaller" 7-nanometer process than Intel, it is my opinion AMD will remain ahead in the price-to-performance realm and leapfrog Intel in IPC as well.

Though what I'm really curious is what happens beyond 7nm. Will the fabs eek out another node shrink or two? Will we finally go graphene? Optical? Biological? Exciting stuff!

Exciting indeed! I personally think graphene is where we are headed next before quantum computing becomes a reality.

 

[jeshikat]

leapfrog Intel in IPC as well

I don't think that will happen but I'd love to be proven wrong.

 

[Hifihedgehog]

Just saw the livestream and it was quite underwhelming. Here are some photos summarizing its lack of exciting information for wondering minds:

 

[Kmpkt]

They might have to start paying OEMs more to use their CPUs. Bummer Intel.

 

[NateDawg72]

Sounds like coffee lake requiring a 300 series chipset is confirmed at this point.

Time for me to start looking into an AM4 board.

 

[EdZ]

New Sounds like coffee lake requiring a 300 series chipset is confirmed at this point.

It could be assumed long before it even had a codename: Intel have operated on two-gens-per-socket for a good decade now.

The announced Kaby Lake Refresh chips should be very interesting for SFF: taking the existing 4+2 dies we see on desktop chips and kicking them all the way down to 15W should make for some really nice NUCs. Think mating a -T CPU with an on-package chipset.

 

[BirdofPrey]

They also had a new microarchitechture every other generation though. Coffee Lake is another Sylake refresh, which is why people were hoping it would be compatible.

 

[NateDawg72]

It could be assumed long before it even had a codename: Intel have operated on two-gens-per-socket for a good decade now.

Well, it's still LGA 1151. Chipsets have changed but not the socket. Yes, that's how they've usually done things in the past but they also progressed faster with node changes in the past. The 6-8th generations are much more iterative than before and some early samples of coffee lake CPUs were reportedly run on 200 series boards. So while it was a fairly safe assumption that it coffee lake wouldn't work on older chipsets it wasn't completely out of the question.

 

[EdZ]

Well, it's still LGA 1151. Chipsets have changed but not the socket.

Same situation with LGA2011 and LGA2011-3. Same number of pins, different electrical pinout, different socket keying.