Skip to content
 
If you can read this, either the style sheet didn't load or you have an older browser that doesn't support style sheets. Try clearing your browser cache and refreshing the page.

(The Register)   Intel no longer has its senior VP of silicon engineering inside, "effective immediately." Something crashed   (theregister.com) divider line
    More: Followup, X86, Intel's senior veep, Jim Keller, Central processing unit, X86-64, Mr Keller's work, personal reasons, A4  
•       •       •

826 clicks; posted to Business » on 12 Jun 2020 at 9:05 AM (15 weeks ago)   |   Favorite    |   share:  Share on Twitter share via Email Share on Facebook



13 Comments     (+0 »)
 
View Voting Results: Smartest and Funniest
 
2020-06-12 9:58:22 AM  
Probably because they can't get a working 7nm design in general production (especially in consumer parts), AMD on the other hand has no such difficulty.
 
2020-06-12 10:17:49 AM  
ignoring ten-teams and other theories of CEO replacement; 1: a (blue-blood lawyer to fix legal,2:  an accountant to cut costs, 3: a salesman to promote and 4: engineer to actually make a product.

wanna take bets intel goes blue?
 
2020-06-12 10:41:17 AM  
He retired to spend more time with his money.
 
2020-06-12 10:42:50 AM  

DarkSoulNoHope: Probably because they can't get a working 7nm design in general production (especially in consumer parts), AMD on the other hand has no such difficulty.


Indeed.  Back in 2010 to 2012, it felt like Intel was lapping the field.  Now, they have fallen behind the competition, though their pricing doesn't entirely reflect this fact.
 
2020-06-12 10:51:22 AM  
It'd be interesting if he ended up at ARM.

/the dude doesn't seem to be doing so well in x86 land
 
2020-06-12 10:57:31 AM  

DarkSoulNoHope: Probably because they can't get a working 7nm design in general production (especially in consumer parts), AMD on the other hand has no such difficulty.


Because AMD uses TSMC to fabricate their chips, and TSMC is the world leader in cutting-edge wafer fabrication. Intel still manufactures their own chips but hasn't invested in keeping up with the competition, so they're basically a distant third place now.
 
2020-06-12 11:06:22 AM  
When you have George Conway on your legal team, it can take a long time, but the results are worth it.
 
2020-06-12 11:10:12 AM  

Sum Dum Gai: DarkSoulNoHope: Probably because they can't get a working 7nm design in general production (especially in consumer parts), AMD on the other hand has no such difficulty.

Because AMD uses TSMC to fabricate their chips, and TSMC is the world leader in cutting-edge wafer fabrication. Intel still manufactures their own chips but hasn't invested in keeping up with the competition, so they're basically a distant third place now.


Oh, Intel invests heavily in trying to keep up (and surpass) the competition, but they farked up 10nm (and 7nm?) so badly that they completely lost their lead.
 
2020-06-12 12:19:45 PM  
With good virtualization available today, the x86/x64 architecture needs to be thrown out and started over. The chips still boot up in freakin' 16-bit real mode FFS.
 
2020-06-12 12:31:26 PM  

ImpendingCynic: With good virtualization available today, the x86/x64 architecture needs to be thrown out and started over. The chips still boot up in freakin' 16-bit real mode FFS.


As far as server farms go, sure, that may happen, but that's because ARM is much more power efficient, not because of specific performance issues.  At a workstation/local server level, meh.  Itanium was a mess.  x64 is here to stay for quite a while
 
2020-06-12 12:53:57 PM  

ImpendingCynic: With good virtualization available today, the x86/x64 architecture needs to be thrown out and started over. The chips still boot up in freakin' 16-bit real mode FFS.


The underlying architecture hasn't looked anything like x86 in ages - there is a ucode layer that breaks the x86/x64 CISC instructions into micro-ops and feeds the actual HW.
 
2020-06-12 1:19:43 PM  

OptionC: ImpendingCynic: With good virtualization available today, the x86/x64 architecture needs to be thrown out and started over. The chips still boot up in freakin' 16-bit real mode FFS.

The underlying architecture hasn't looked anything like x86 in ages - there is a ucode layer that breaks the x86/x64 CISC instructions into micro-ops and feeds the actual HW.


Is there a mode where you can skip the x86 translation and just talk to the underlying hardware?  Then all you need is a compiler and hey!  Free speedup!

Of course, then you have to keep your x86 stuff separate from your ia or whatever stuff, but seems doable.
 
2020-06-12 1:31:58 PM  

Snotnose: OptionC: ImpendingCynic: With good virtualization available today, the x86/x64 architecture needs to be thrown out and started over. The chips still boot up in freakin' 16-bit real mode FFS.

The underlying architecture hasn't looked anything like x86 in ages - there is a ucode layer that breaks the x86/x64 CISC instructions into micro-ops and feeds the actual HW.

Is there a mode where you can skip the x86 translation and just talk to the underlying hardware?  Then all you need is a compiler and hey!  Free speedup!

Of course, then you have to keep your x86 stuff separate from your ia or whatever stuff, but seems doable.


You are very unlikely to get any sort of speed-up by skipping the ucode layer.  A lot of the pipeline optimizations like superscalar dispatch, out-of-order execution and speculative execution are handled at least partially in there.  (of course this is where all of the specter and meltdown and other security problems originate as well...)
 
Displayed 13 of 13 comments

View Voting Results: Smartest and Funniest

This thread is closed to new comments.

Continue Farking




On Twitter



  1. Links are submitted by members of the Fark community.

  2. When community members submit a link, they also write a custom headline for the story.

  3. Other Farkers comment on the links. This is the number of comments. Click here to read them.

  4. Click here to submit a link.