Discussion Half of the Core "gens" have all been Skylake

[NTMBK]

In the first five generations we went through Nehalem, Sandy Bridge, Ivy Bridge, Haswell and Broadwell. The following five generations have been... Skylake. Crazy.

 

[Markfw]

In the first five generations we went through Nehalem, Sandy Bridge, Ivy Bridge, Haswell and Broadwell. The following five generations have been... Skylake. Crazy.

I would say they really aren't gens, just skylake, skylake+, skylake++, skylake+++, skylake++++.

But I am sure I am a little off, just seems that way,

 

[mikegg]

The CPU design team must be so frustrated with the node team. Imagine the CPU team asking the node team if 10nm is ready yet every year. And the node team replying "yes! this is definitely the year. No need to backport to 14nm".

4 years later, they might have to backport Rocket Lake to 14nm.

 

[NTMBK]

I would say they really aren't gens, just skylake, skylake+, skylake++, skylake+++, skylake++++.

But I am sure I am a little off, just seems that way,

I was referring to the marketing that calls the latest power guzzler as 10th generation.

 

[Markfw]

I was referring to the marketing that calls the latest power guzzler as 10th generation.

Anandtech even said the 10700k was skylake.....So its Skylake++++

 

[krumme]

In the first five generations we went through Nehalem, Sandy Bridge, Ivy Bridge, Haswell and Broadwell. The following five generations have been... Skylake. Crazy.

Nehalem is perhaps more of a bridge design than a core product.
Perhaps like zen1 and zen2

 

[SPBHM]

it's interesting that it still is relevant and competitive after 5 years

even with all the massive gains AMD and the ARM stuff provided in the last 5 years,

Skylake and 14nm were a pretty amazing product for the time I guess.

but yes, a very unexpected situation to be in given how the first half of the decade was...

 

[SAAA]

Which reminds me, if all went according to plan what was intended to release today? Like even with a full year of delay Cannon and Ice would still have been 2017 and 2018 products, what about late 2020? If Golden Cove was a thing for 2020 (and 7nm) there must be some serious build-up, notwithstanding security issues and all the work to fix them.

That's more like two nodes screw up followed by poor substitutes for years looking from the consumers side, it's indeed amazing Skylake held well enough to this point.

 

[Markfw]

Which reminds me, if all went according to plan what was intended to release today? Like even with a full year of delay Cannon and Ice would still have been 2017 and 2018 products, what about late 2020? If Golden Cove was a thing for 2020 (and 7nm) there must be some serious build-up, notwithstanding security issues and all the work to fix them.

That's more like two nodes screw up followed by poor substitutes for years looking from the consumers side, it's indeed amazing Skylake held well enough to this point.

Well, I would not say "held well enough" exactly. They are power hungry and don't perform that well, and don't have all the hardware fixes for the security flaws. AMD just keeps getting farther away.

 

[Shmee]

You forgot Hawell E/EP and Broadwell E/EP

 

[SAAA]

Well, I would not say "held well enough" exactly. They are power hungry and don't perform that well, and don't have all the hardware fixes for the security flaws. AMD just keeps getting farther away.

I'm mostly speaking about the architecture point of view, certainly even Skylake would fare decently power consumption wise if it were on 7nm with maybe a few 100 mhz less. It would compete also on performance being a good deal smaller so possibly featuring more cores than today 10, like die shrink of it and MCM for 16+ cores aren't far fetched. Security wise there's no excuse of course, there newer arch might have helped by nature of being well, new, with different flaws maybe but still undiscovered.

 

[mikegg]

I'm mostly speaking about the architecture point of view, certainly even Skylake would fare decently power consumption wise if it were on 7nm with maybe a few 100 mhz less. It would compete also on performance being a good deal smaller so possibly featuring more cores than today 10, like die shrink of it and MCM for 16+ cores aren't far fetched. Security wise there's no excuse of course, there newer arch might have helped by nature of being well, new, with different flaws maybe but still undiscovered.

I don't think what you're saying is actually true.

If you look at it from an EPYC vs Xeon perspective, Zen2 runs circles around Skylake variants in raw performance, cost/performance, and performance/watt. EPYC Rome is 4x faster for the price. That's 2-3 huge generations ahead of Skylake.

On the workstation side, AMD has zero competition with Threadripper. Intel simply cannot compete with Threadripper at Threadripper prices. Impossible with Skylake.

On the desktop side, it's still close because Intel has been able to add a few cores here and there and completely ignore power limits. But AMD can always lower prices to beat Intel in performance per/cost. In addition, Intel still doesn't have an answer to the 3950x, which is only a "casual" two chiplet processor to AMD.

Would Intel compete if they used TSMC's 7nm? Probably in raw performance but Intel will most likely be significantly more expensive due to their monolithic designs.

Keep in mind that Intel’s highly optimized 14nm process is likely much cheaper than TSMC’s 7nm already.

It’s hard to imagine Intel being able to manufacture a 7nm Skylake 64 core processor that is cost effective versus a Zen2.

Skylake right now is unable to compete in anything except 1080p gaming.

 

[AnandThenMan]

Sky is the limit.

 

[DrMrLordX]

Which reminds me, if all went according to plan what was intended to release today?

Cannonlake 2017
IceLake 2018
TigerLake 2019
Alder Lake 2020
Meteor Lake 2021

That would be my guess.

 

[NTMBK]

Cannonlake 2017
IceLake 2018
TigerLake 2019
Alder Lake 2020
Meteor Lake 2021

That would be my guess.

So would Meteor Lake be 5nm, in the original plan?

 

[DrMrLordX]

So would Meteor Lake be 5nm, in the original plan?

Possibly. The old pre-Cannonlake Fiasco roadmaps went up to Tiger Lake (I think) and showed little more, so it's all guesswork from Alder Lake on up. It's possible that Alder Lake may have never existed had 10nm worked.

 

[liad]

Pretty sue the skylake engineers are quite proud as their core is on huge chunk of the X86 market and will be for the next 4-5 years (My 8700 wont need replacing anytime soon)

 

[gdansk]

It's more sad than anything but we do have Sunny Cove in 10th gen too.

 

[mikegg]

Pretty sue the skylake engineers are quite proud as their core is on huge chunk of the X86 market and will be for the next 4-5 years (My 8700 wont need replacing anytime soon)

I think they're probably feeling more frustrated that their new designs have been unable to make it to market due to 10nm.

Keep in mind that these engineers are usually given company stock. They probably feel like their stock compensation would have been a lot more if the company didn't screw up 10nm.

 

[lamedude]

What happened to only the paranoid survive? Can you imagine if Intel had stuck with Pentium this long and only release P6 years after K6.

 

[IntelUser2000]

So would Meteor Lake be 5nm, in the original plan?

I still think it would have been 7nm just earlier.

The rumors of needing plusses for process have existed for years, even before we knew about the troubles regarding the 10nm process.

10nm would have been Cannonlake, Icelake, and Tigerlake. If it's Alderlake and Meteorlake for next, then those would be 7nm, and the one after that.

I was referring to the marketing that calls the latest power guzzler as 10th generation.

Skylake sucks. It was OK back in late 2015, but that's about it. I thought the CPU was nothing special back then. Now add 5 years on top of that.

The CPU uses enough power to justify having its own PSU, it has almost half the cores of the competition, there are more bugs than in a rainforest in summertime, and tiny ARM chips are rivaling it in uarch.

 

[mikegg]

Skylake sucks. It was OK back in late 2015, but that's about it. I thought the CPU was nothing special back then. Now add 5 years on top of that.

That's right, Skylake was roughly 5% faster than Haswell. It was not even a good generational leap.

And since then, Intel has been squeezing 1-2% IPC gains each year.

The only reason some people in this thread think Skylake is good is that AMD has been so bad for so long.

But if you compare Skylake's improvements to say the mobile SoC market or A.I. accelerator market, it's crap.

 

[IntelUser2000]

@senttoschool There's zero IPC gains in Skylake cores. They are all same.

 

[scannall]

@senttoschool There's zero IPC gains in Skylake cores. They are all same.

If anything there have been regressions for all the security issues.

 

[TheELF]

I think they're probably feeling more frustrated that their new designs have been unable to make it to market due to 10nm.

Keep in mind that these engineers are usually given company stock. They probably feel like their stock compensation would have been a lot more if the company didn't screw up 10nm.

Oh yeah they certainly thought that they would be getting more than only almost double...

Intel - 43 Year Stock Price History | INTC

Historical daily share price chart and data for Intel since 1980 adjusted for splits and dividends. The latest closing stock price for Intel as of October 16, 2023 is <strong>36.56</strong>. <ul style='margin-top:10px;'> <li>The all-time high Intel stock closing price was...

www.macrotrends.net