Intel stock price takes a $20Bil Bath

[WillyW]

Looks like the veil of Intel's lies to the investor market have finally been pulled out from in front of them.

 

If there was anyone paying attention, and placed a short on Intel, they could have made a lot of money today, with information that was generally available to the tech community.

 

".. shares fell 9 percent on Friday, after the chipmaker’s upbeat results were overshadowed by concerns that Advanced Micro Devices Inc may be chiseling off market share from its high-margin data center business. "

https://www.reuters.com/article/us-intel-results-stocks/intel-shares-slip-on-worries-of-data-center-losses-to-rival-amd-idUSKBN1KH1OT

 

And an older article from yesterday afternoon:

" Despite the record revenue, a 78% gain in profits to $5 billion, and increased guidance, the company's stock slipped 6% in after-hours trading due to slower-than-anticipated growth in its data center portfolio. "

https://www.tomshardware.com/news/intel-10nm-ceo-earnings-datacenter,37518.html

 

 

Full disclosure, I own shares in AMD.

[WillyW]

f

[Brooksie359]

Well they didn't really lie to investors though. The former CEO told their investors that they would lose market share to AMD in the server market...............

[WillyW]

Yeah, it's not that they lied, it's just that just like the Facebook thing, the supposed "smart people" on Wall Street were not really paying attention. When I worked for a discount brokerage, you would be surprised at how clueless people can be. Just goes to show that if you have enough discipline and you really know the companies you can make money on tech stocks, because as this and the FB thing shows, the general populace thinks it's all just a bunch of pretty lights.

[KarathKasun]

Intel has been skirting the truth about 10nm progress for years now as well.

[AnonymousGuy]

Intel's stock always drops after earnings are announced.  Every quarter, every year.  Doesn't matter if estimates are beaten, how many bagillion dollars are earned, etc.  It's the juggernaut syndrome, every dipshit institutional investor wants to be the first to predict the beginning of the end.  Q2 2018 didn't have any significant markers in INTC's numbers, certainly not enough to trigger an 8% drop.  AMD benefits from the other side I like to call the death twitch...everyone wants to be the first to guess a miracle is going to happen.

[Stefan Payne]

9 hours ago, WillyW said:

 

".. shares fell 9 percent on Friday, after the chipmaker’s upbeat results were overshadowed by concerns that Advanced Micro Devices Inc may be chiseling off market share from its high-margin data center business. "

 

People should mention that the Spectre and Meltdown Mitigations really hurt in Database and File Server application.

 

Everywhere where you have to access I/O Stuff like the Harddrive. That is something that is put under the carpet by some peoples because it doesn't look like its a Problem with Gaming.

Well it might not be but its a Problem in Datacentres.

 

There just is no alternative to an AMD Epic right now!

[leadeater]

2 minutes ago, Stefan Payne said:

Well it might not be but its a Problem in Datacentres.

It's really not been that big a of a problem performance wise, the security aspect was much more important. Post patching hasn't lead to needing to buy any more servers to cover load increase and hasn't required a sooner capacity expansion.

 

4 minutes ago, Stefan Payne said:

There just is no alternative to an AMD Epic right now!

A microcode updated Xeon.....

[mr moose]

In other news, a slightly under-experienced bunch of forum dwellers claim the death of Intel because they misinterpreted old news about a nothing event.

[arnavvr]

Not really all that surprising imo. Intel has been stagnating for the better part of a decade, and it's catching up to them now.

[leadeater]

7 minutes ago, mr moose said:

In other news, a slightly under-experienced bunch of forum dwellers claim the death of Intel because they misinterpreted old news about a nothing event.

But I read on the internet.... *ignores real life experiences*, yep confirmed Intel is doomed.

[GoldenLag]

4 minutes ago, leadeater said:

But I read on the internet.... *ignores real life experiences*, yep confirmed Intel is doomed.

*AMD was gone from the market for several years

 

Ye intel is gone after 5 years........

[mr moose]

3 minutes ago, leadeater said:

But I read on the internet.... *ignores real life experiences*, yep confirmed Intel is doomed.

you read it here first:

 

Quote

The company’s profit and revenue beat Wall Street targets, but double-digit growth in its data center chip business fell short of analysts’

Forget the stuff in bold, the last 4 words is all I need to read to know that Intel is stagnating with a plummeting bottom line... 

[Stefan Payne]

10 minutes ago, leadeater said:

It's really not been that big a of a problem performance wise, the security aspect was much more important. Post patching hasn't lead to needing to buy any more servers to cover load increase and hasn't required a sooner capacity expansion.

 

A microcode updated Xeon.....

Nope, you remember that?

https://arstechnica.com/gadgets/2018/01/heres-how-and-why-the-spectre-and-meltdown-patches-will-hurt-performance/

 

https://www.theregister.co.uk/2018/01/09/meltdown_spectre_slowdown/

 

https://blog.architecting.it/storage-performance-spectre-meltdown/

Quote

the impact from implementing the initial patches for Spectre/Meltdown has been a report of anything from 30-50% in system performance for I/O intensive workloads.

 

https://www.theregister.co.uk/2018/07/26/meldown_spectre_batters_hpc/

Quote

• Disk access – "With all mitigations enabled, the mainline kernel is slowed down by approximately 50 per cent on local disk and 33 per cent on Lustre. The GRSecurity-enabled kernel is slowed down by 90 per cent on local disk and 33 per cent on Lustre."

 

And that is what I was talking about.

Some other Benchmarks ignore the CPU utilization when Accessing Disks.

 

So your claim that a Xeon with Microcode updates would be an option, is just bogus. That is the Problem right now!

[Emberstone]

I don't expect Intel to crash and burn anytime soon, but if AMD grabs more of the market with better Epyc CPU's and whatever 7nm produces, it will be really interesting to see what they can do with all that revenue considering Ryzen came from very little.

 

2019 and 2020 should be pretty exciting.

[leadeater]

2 minutes ago, Stefan Payne said:

Nope, you remember that?

https://arstechnica.com/gadgets/2018/01/heres-how-and-why-the-spectre-and-meltdown-patches-will-hurt-performance/

 

https://www.theregister.co.uk/2018/01/09/meltdown_spectre_slowdown/

 

https://blog.architecting.it/storage-performance-spectre-meltdown/

 

https://www.theregister.co.uk/2018/07/26/meldown_spectre_batters_hpc/

 

And that is what I was talking about.

Some other Benchmarks ignore the CPU utilization when Accessing Disks.

 

So your claim that a Xeon with Microcode updates would be an option, is just bogus. That is the Problem right now!

Because everyone runs solely HPC clusters which rely on ultra fast ultra low latency interconnects and extreme disk and memory performance. Those people are experts they know what they are doing, they'll just use the Kernel switch to disable certain Spectre patches to get the performance back exactly where it was, it's not like HPC critically need them on but you still need to test it.

 

Coming back to the majority of the server world it has not been a significant performance problem, you can reject reality but it is what it is. You're free to ignore my experiences on this matter, it's not like I haven't asked multiple other places either who share the same, but the difference is this is my day job. As a systems engineer responsible for server hardware platforms and security patches of all our servers, 3 data centers total, with hundreds of large multi socket multi node servers under pinning our VMware clusters and physical database clusters I would have thought listening to my actual experiences would be worth it but I guess not.

[mr moose]

19 minutes ago, Stefan Payne said:

Nope, you remember that?

https://arstechnica.com/gadgets/2018/01/heres-how-and-why-the-spectre-and-meltdown-patches-will-hurt-performance/

 

https://www.theregister.co.uk/2018/01/09/meltdown_spectre_slowdown/

 

https://blog.architecting.it/storage-performance-spectre-meltdown/

 

https://www.theregister.co.uk/2018/07/26/meldown_spectre_batters_hpc/

 

And that is what I was talking about.

Some other Benchmarks ignore the CPU utilization when Accessing Disks.

 

So your claim that a Xeon with Microcode updates would be an option, is just bogus. That is the Problem right now!

I guess the data centre and cloud service industry didn't get the memo.  They are still making lots of money money money.

Quote

 

Intel was at the forefront of selling chips used in making servers where data is stored remotely or in so-called cloud servers. Over the past few years as more companies rushed to the cloud to move data online, Intel enjoyed healthy cloud revenue growth.

 

Amazon (AMZN.O), which beat profit estimates on Thursday due to its cloud business, and Microsoft (MSFT.O) have been the main beneficiaries of the cloud adoption.

 

 

[S w a t s o n]

6 minutes ago, leadeater said:

Because everyone runs solely HPC clusters which rely on ultra fast ultra low latency interconnects and extreme disk and memory performance. Those people are experts they know what they are doing, they'll just use the Kernel switch to disable certain Spectre patches to get the performance back exactly where it was, it's not like HPC critically need them on but you still need to test it.

 

Coming back to the majority of the server world it has not been a significant performance problem, you can reject reality but it is what it is. You're free to ignore my experiences on this matter, it's not like I haven't asked multiple other places either who share the same, but the difference is this is my day job. As a systems engineer responsible for server hardware platforms and security patches of all our servers, 3 data centers total, with hundreds of large multi socket multi node servers under pinning our VMware clusters and physical database clusters I would have thought listening to my actual experiences would be worth it but I guess not.

The reality is that not every use case is yours (just like not everyone is running HPC clusters), and many companies have made public claims of reduced performance well into double digits. For some it doesn't matter, for others it's a shit show.

Personally, I'm more excited to see the shitshow that is 10nm, and that semi accurate article about a company with at least 20 bil in market cap going under because of it.

[mr moose]

13 minutes ago, leadeater said:

Because everyone runs solely HPC clusters which rely on ultra fast ultra low latency interconnects and extreme disk and memory performance. Those people are experts they know what they are doing, they'll just use the Kernel switch to disable certain Spectre patches to get the performance back exactly where it was, it's not like HPC critically need them on but you still need to test it.

 

Coming back to the majority of the server world it has not been a significant performance problem, you can reject reality but it is what it is. You're free to ignore my experiences on this matter, it's not like I haven't asked multiple other places either who share the same, but the difference is this is my day job. As a systems engineer responsible for server hardware platforms and security patches of all our servers, 3 data centers total, with hundreds of large multi socket multi node servers under pinning our VMware clusters and physical database clusters I would have thought listening to my actual experiences would be worth it but I guess not.

I reject your reality in favor of my ideals. 

[leadeater]

Just now, Swatson said:

The reality is that not every use case is yours (just like not everyone is running HPC clusters), and many companies have made public claims of reduced performance well into double digits. For some it doesn't matter, for others it's a shit show.

Personally, I'm more excited to see the shitshow that is 10nm, and that semi accurate article about a company with at least 20 bil in market cap going under because of it.

Yes there have be reports of problems, Epic games for example, but for every report of something there is hundreds of thousands of people not having problems and that just isn't worth reporting on. Have to remember doing specific tests to find performance impacts isn't the same thing as actual operation. In the case of Epic and everyone using Cloud services you can't know what other people are running on the same hardware, your code might not be significantly impacted by Spectre patches but someone elses might be who are also running on the same hardware as you and their slow down effects you.

 

We're an extremely varied environment with hundreds of different applications of all kinds (Windows and Linux), even things like virtual Citrix NetScalers which a network load balancers which if you have followed Spectre is on the list of worst cases. All our critical business applications run through those and it has not been a problem.

 

If it was actually a shit show you'd be hearing far more about it and more real life issues, for a "devastating"  performance impact the resulting outcome just doesn't show it. We even waited 2 months before patching anything specifically to listen out for problems and then did staged patching and performance testing to understand the impact.

 

Lastly something going slower doesn't make it unusable, just slower. It could make it unusable though.

[Speed Weed]

AMD boiiiii

[Stefan Payne]

28 minutes ago, leadeater said:

Because everyone runs solely HPC clusters which rely on ultra fast ultra low latency interconnects and extreme disk and memory performance.

1) I was specificly talking about those Situations!

2) there are other things like a Database Server or other things that just do mostly I/O Stuff. Like a Webserver.

3) HPC Clusters are made of different units.

Some are I/O based solutions (Datastorage), some are the hard Calculating stuff.

 

And the Solution for those is to either disconnect them for the Internet or to not patch them because the Performance Impacts are so huge. And Data Security isn't necessarily Important for those units.

 

So would you pls Stop Saying B when I was talking about A??

 

Quote

Those people are experts they know what they are doing, they'll just use the Kernel switch to disable certain Spectre patches to get the performance back exactly where it was, it's not like HPC critically need them on but you still need to test it.

...and again the Defending of Intel and Bashing everyone who disagrees starts...

 

Why can't you just admit that there is a Problem and there might be some reason to go Full EPYC and throw the Intel Solutions out - ahead of Schedule, because the Performance Impacts of the Spectre/Meltdown Patches hurt the Performance for these Applications.

 

Quote

Coming back to the majority of the server world it has not been a significant performance problem, you can reject reality but it is what it is.

No, I wasn't even talking about that!

You are deflecting what I was trying to say and lead the discussion to other areas I wasn't even thinking about.

 

I specificly mentioned Heavy I/O Operation such as Harddisk Access. 

And your posting here does NOT help the people that are working with those on a daily basis and have the Problems.

 

That they Make Money is a stupid argument because most of those have a bit of romm so they try not to load the Servers too much.

And for Data/Storage Operation, one might assume that the Server isn't really utilized that hard most of the Time and there is some room to spare.

 

But if the utilization of the Server jumps from what you see in the Links I provided, from 20% to 50 or 70%, that is a cause for immediate concern. Because that means that the Servers are in need of upgrades.

 

But that isn't the only Problem. You miss the increased Power Consumption as well. And in an air conditioned Server Room that is double the Problem as it would be for the Consumer. 

 

Quote

You're free to ignore my experiences on this matter,

Oh well, it would be a nice start if you would stop ignoring what I was talking about and adress the specific points I was making.

 

And that is that the Load increased dramatically due to the Spectre/Meltdown Patches. And that it pissed off some administrators because of that.

 

Well, maybe another Solution could be to switch to a Windows based solution...

I don't know what Mitigations they implemented and how much it impacts performance though. Because All the Reports I've seen are Linux Based...

[mr moose]

@leadeater Turns out your education and experience is no match for reading a few internet articles and a blog post.    We see right through your posts and know you are just defending Intel now.   Ha... hahaha...  hahahaha... hahahahaha!!!     

 

 

 

EDIT: just a quick edit to let you know I am wiping a small tear away from laughing so much.

[leadeater]

25 minutes ago, Stefan Payne said:

1) I was specificly talking about those Situations!

You generalized then when challenged gave HPC examples, what am I supposed to do with that other than point out other non HPC uses cases which is the bigger server market segment?

 

25 minutes ago, Stefan Payne said:

2) there are other things like a Database Server or other things that just do mostly I/O Stuff. Like a Webserver.

We have all of those, very large ones too.

 

25 minutes ago, Stefan Payne said:

3) HPC Clusters are made of different units.

We have that too, I'm not in that team though but as an academic institution we most certainly have it.

 

25 minutes ago, Stefan Payne said:

And your posting here does NOT help the people that are working with those on a daily basis and have the Problems.

Got any wide scale examples or just the few reports that first came out?

 

25 minutes ago, Stefan Payne said:

But if the utilization of the Server jumps from what you see in the Links I provided, from 20% to 50 or 70%, that is a cause for immediate concern. Because that means that the Servers are in need of upgrades.

And contained in those articles

 

Quote

Before the weekend, Amazon confirmed the updates will ding AWS virtual-machine performance to some degree, albeit with no "meaningful performance impact for most customer workloads" expected, apparently.

 

Or in that one about storage performance testing where it shows the Netapp FAS 8200, the very network storage vendor and actual models we have, 16 8200's to be exact.

 

Also for our VMware clusters we use Nutanix which is a software solution and all disk access run through a VM that sits on every host, aka Software Defined Storage. Those VMs have direct control of the storage in the servers and present it as a pool of storage back to the ESXi hypervisor over NFS. So every VM we run should be crippled by this performance loss of Spectre as it's the exact case of what you're talking about but it didn't happen.

 

I'd rather not have to keep this back and forward going as it's off topic, just be careful when making wild claims about how bad the Spectre performance impact on the data center world is since some of us actually work in it and might want to give input on it when brought up. You used isolated incidents and targeted performance benchmarks to dismiss the entire range of Intel Xeons in a rather disingenuous way, one that I know doesn't match actual usage in the industry.

[leadeater]

Just now, mr moose said:

@leadeater Turns out your education and experience is no match for reading few internet articles and blog post.    We see right through your posts and know you are just defending Intel now.   Ha... hahaha...  hahahaha... hahahahaha!!!     

Darn and I distinctly remember being a big supporter of EPYC and being very excited about it, fake memories.