Intel Trims Cascade Lake Xeon Scalable Lineup

18 Comments

View All Comments

• Holliday75 - Thursday, November 7, 2019 - link

  Looking at Intel's CPU lineup is like looking at General Motors before 2008. Jumbled mess with many seemingly overlapping and competing products. They need to trim the fat and this is a start.

• guycoder - Thursday, November 7, 2019 - link

  That's what you get when you let Sales and Marketing run the show for too long and you squeeze your customers though excessive "segmentation" and restricted/failed innovation to extract maximum profits. Intel overdid this and left a massive hole for AMD to drive though.

• shompa - Thursday, November 7, 2019 - link

  But in real-world: Its not AMD that is Intels treat. Its ARM/RISC V. The only reason X86 exists is that it was cheap enough and had Windows. Never that it was the fastest or best product. The only reason Intel managed to fight off RISC chips was that Intel was 2.5 generations ahead in manufacturing and therefore could eat the 30% CISC tax. But thanks to Apple/Samsung a new manufacturing node has been online since 2011 erasing Intel's lead. That also saved AMD. Ryzen is nothing special. Going from 28nm to 14-16nm FinFET gives 50% performance if you just follow the design guidelines. The 2/4 cores that were standard before Ryzen: 50-80% of the die area was iGPU. AMD removed the iGPU and replaced it with cores. Something Steve Jobs demanded from intel in 2008.
  Intel has lost the low powered market since they can't compete with 25-50 dollar chips and X86 (the 30% larger die area because of CISC) We are talking billions of SoCs each year. Highend servers today don't use windows. All run Linux. We have Amazon/Google/Facebook all trying out own ARM chips. All top 500 computers in the world today use graphic cards for acceleration. X86 is just a node command chip. Highend servers is going away from Intel. I just dont understand mainstream desktop PCs at work that are there for MS Office. The averege high-end phone is faster, so this market should go away. That leaves Intel/AMD a niche market for gamers and creative pros. Especially next year when Apple moves to ARM. And we also have the dark horse: RISC V. Open source, real high-end chips. The rumor is that for example, Samsung is moving to RiscV and therefore can save licensing fees for 100+ million ARMs each year. Yes. Intel is trying to lock in customers by having software optimized för AVX256/512 because CPU to CPU: ARM is already 30% faste per clock and its even real 64bit, something Intel is not. X86 can never remove the 32bit parts in the CPU, something that Apple, for example, did 2 years ago. Huge part of the die area saved. The sad fact is that Windows/X86 is the only 64bit platform that is slower than 32bit. Anyone can check this out. Find a A7 and compile the same code in 32bit and 64bit. Strange how the 64bit code is 30% faster. On Windows/X86: 32 vs 64bit code. 64bit is 3% slower.
  Some historical facts: Late 1990is most large companies had 0 PCs. They had a few fast servers and used Xterminals. Now we call it the "cloud". Year 2000 RISC had 90%+ server revenue. Intel only had Windows Servers for people without a clue (and bribes). 2000-2005. Every 3 month Intel bumped its CPUs. Why? Because Apple bumbed G4/G5 every 3 months. And when Apple could not get faster G4, they started to include extra CPUs for the same price. Something that was easy with a real Unix OS + RISC cores that are great at many instances. Intel stopped innovating the same day Apple moved to X86 and Intel had an exclusive 5-year deal. X86 has held us back for decades. 64bit was introduced in 1990. Internet was 64bit in 1997 (since almost all web services used Sun SPARC stuff. They were the . in .com). Even 2005 RISC had over 50% of server revenue. But what happened? Well. Intel started to sell Xeons for 300 dollars and killed RISC. Now Intel thinks 15000 dollars is normal for a server CPU. SPARCs cost 4K dollars and that's why they died. Now Intel does the same thing.
  And last: for everyone that has worked with IT. Have you ever meet a IT boss that knows anything about IT? I do not know anyone working in IT that knows about a IT cheif that knows IT. That explains the scary stuff I have seen where some of the largest IT compnies in the world moves to MSFT and kills of a working IT infrastructure. These companies had hundreds off real gurus warning them for all problem, but the IT cheif didnt care. LM Ericsson. Top3 telecom companie in 1997.200, Moved to X86/Windows. Share price 50 dollars.Just 4 years later when they tried to make new phones/base stations in Excell the company was on the brink of closing down. They had to print more shares for 30 cents each to get capital and survive. From 110000 employees to maybe 10K today. Nokia was the same. And more. I know. I was there. IT is the area where 90% working with it have no clue. What is a IF/AND/OR gate. You need to know if you want to understand IT, not brand names like MSFT/Apple/Oracle. Intel will die. And I predicted that SPARC/RISC would die in 1997 and I was right. The day with Intels 95% profit margin needs to die. (8 core Intels cost under 10 dollars to manufacture and still stupid "journalists" thinks iPhone is too expensive and that the SoC cost just 25 dollars. Missing the point that today's ARM SoCs are larger than Intels 4 core = more expensive to manufacture. Still its normal for Intel to charge 100 dollars for a motherboard with 3.8GB/s bandwidth 1x4 PCI express3 + 300-600 dollar for a CPU. ARM should save us. Not AMD.

• Jorgp2 - Thursday, November 7, 2019 - link

  Was that a copypasta or an original shitpost?

• Alistair - Thursday, November 7, 2019 - link

  looks like his PHd thesis

• Spunjji - Friday, November 8, 2019 - link

  Dr Derp

• Korguz - Friday, November 8, 2019 - link

  what was that huge ramble about ???

• versesuvius - Friday, November 8, 2019 - link

  It is true that when it comes to money Intel drives it in to the hilt, but:
  
  -When was the last time that the average computer user, i.e. PC(i.e. Personal Computer) user, which by the way comprise the major part of the total computing power in the world, could buy a RISC processor and motherboard that can house that processor and assemble a computer?
  
  -Linux gives you fine control and freedom over how the user wants to work. But that is only in the software. When it comes to hardware, it is again the x86 processors that provide that freedom and choice.
  
  - And last but not least, not every application is better off on a RISC processor, and from experience only closed systems and environments can benefit truly from a RISC processor at a very high cost, as in Apple or server houses and so forth, not the PC user, and even there most supercomputers use x86 processors to build those mighty number crunchers running some flavor of Unix.
  
  So, until there are RISC processors that are widely available to the public to pick and choose from and until when there is clear advantage to writing the software that the millions of people use, for the RISC architecture, who cares. RISC was there before x86 and is still there and will be there for the foreseeable future, but never did manage to become anything more than what it was then and what it is now, open source or not.

• Dr_b_ - Friday, November 8, 2019 - link

  So much wrong with the post above besides pagination, but here a few things that are just factually untrue:
  
  "I just dont understand mainstream desktop PCs at work that are there for MS Office. The averege high-end phone is faster, so this market should go away."
  
  This is even funnier:
  "Some historical facts: Late 1990is most large companies had 0 PCs. They had a few fast servers and used Xterminals."

• rrinker - Friday, November 8, 2019 - link

  The second one is semi-accurate. In those days, I spent a lot of time installing Citrix servers. Companies were even making kits to convert obsolete PCs into Citrix clients (mainly to make them boot right to the client, but also not use the old and likely to fail hard disk, etc). Those didn't sell too well, but thin clients did, embedded machines that in some cases ran multiple types of terminal clients - common ones did 3270 emulation, Citrix, VT100 or some variation of DEC terminal, and some did indeed function as X terminals, but not too many people I ever dealt with ran Unix systems with X terminals.
  But even those companies with huge Citrix deployments didn't have "0 PCs". There were always things that didn't work well in a Citrix environment and needed a dedicated PC. We DID have clients where maybe 95% of the users only had a thin client on their desk.