Intel on Linux

Linux is not mainstream... by Anonymous Coward · 1999-05-11 22:57 · Score: 2

...until I can buy it pre-installed on low-priced laptops. ($5K VA Research machines are out of the reach of many). And while Linux does run just fine with X on many sub $2K laptops, these machines tend to have built-in winmodems, win-ethernet ports, and on some Compaqs, wierd proprietary CD-ROM drives (that don't behave like ATAPI standard CDROMS), all of these being Linux hostile. While I can stick in some PCMCIA modems and network cards, I want to be able to use the built-in stuff and have free slots for other stuff, consarnit! WinHardware is a machination of purest aevil!

I dont mean to sound like a troll but linux is not by Anonymous Coward · 1999-05-12 00:23 · Score: 2

Lets get one thing straight. I am not anti-linux but the latest fud does prove one thing that is very true of linux. The recent fud is 100% truth when it comes to raid, scsi and multiple etherent adapters. Just look at mindcrafts and NT magazines and now zdnet results. There are 2 articles now pointing to performance problems with linux's sync. i/o.

Problems:
1:) Linux has alot of trouble with raid because of its i/o and hardware raid can kill performance but software raid is better.

2:)Linux has trouble with smp because it doesnt support async i/o and non reepmtive cpu points. THis would bring intels 8-way cpu server to its knees and if NT was stable enough it might even outperform linux because even NT supports async i/o. I think solaris x86 would be much more preferable.

3:)The scsi module in linux has some performance and stability proeblems because 95% of all linux developers use eide because they are hobbyist and dont have the money for expensive scsi and raid controller cards.

Linux is a killer workstation and low end unix server and a ok mid-range server but if all these busssiness put mission critical stuff on them and have awefull performance and maybe some instability then solaris will seem a whole lot more viable. Remember what happened to NT after coporations believed Jess Berst form zdnet and microsoft. Texas instruments and Motorola switched form unix to NT only to switch back again and they regret ever wasting time and money on NT. Linux will have a similiar fate if its used on mission critical systems. Also if you need a high end system like a 8-way server, then its advisable to go risc with alpha or a sun box with solaris or digital vms. If yuour 2-way alpha box can perform the same as an 8-way intel box, then dunp solaris and use linux on the alpha but adding more scpu's puts stress in the i/o which is a little fragile in linux right now. Other then that linux rocks but lets hope these mindcraft and zdnet fud will help bring light and hope to linux by fixing problems and they might actually improve linux and help bring linux out of stuff where it doesn't beling (yet....).

I would trust solaris or vms alot more with an EMS app then linux right now.

What Intel should do... by Anonymous Coward · 1999-05-12 04:27 · Score: 2

Instead of just talkin smack they should commit to using linux as the reference platform for all of their newest technologies. Then by going GPL with all of their latest and grestest "enterprise" hardware designs Linux could match any other OS out there in terms of features within a year.

All I'm saying is that everyone (meaning all of the big money vendors) are looking to linux to shore up margins and keep some of the money they are sending to redmond. Yet you aren't going to see compaq developing drivers for linux.

There is no reason for Intel not to help develop Linux support for evey device they build. From nic's to motherboards.

Everyone should be asking Intel:

Where is the Linux i2o implementation?
You guys invented USB, why do we have to reinvent the wheel? How bout throwing that in the pot?
Where is the VIA implementation?
If you actually want anyone to care about MMX why not patch it into the kernel to demonstrate how powerful it could be?
Standard device drivers among all intel based unices? Why not develop it, then anounnce it, and give away some code?

They released the source to their video chipset which is a good start. But we all know that they have to write drivers for their own purposes internally, how bout making linux the standard reference platform and making that software public when the device ships? Even if the code is in rough shape the free *nix communities would be more than capable of polishing it up.

Hardware vendors need to realize that using a common code base for all of their products (from printers to servers) will shorten their development cycles if they will just work with the public.

I thought... by jtseng · 1999-05-11 22:08 · Score: 2

Notably, the Red Hat* Linux 6.0 distribution pushes symmetric multiprocessor (SMP) support to at least four processors.

I thought the SMP support was for up to sixteen processors. Who is the one who is misinformed here?

--

Sanity.html - Error 404 not found

Re:I thought... by Natedog · 1999-05-11 22:51 · Score: 2

I think this was a misunderstanding on their part. Linux does support up to 16 processors but, it doesn't scale well after 4 processors. A proper benchmark should extend out to a point where a definate maximum has occured. In a load vs performace graph this should create a horizontal asymptote. To check for scalability, you need to check that the interval between these asymptotes remains approx. the same as you attempt to scale up. What this means is that this interval will remain consistent for Linux until you try to scale past 4 processors, at which time the intervals will get smaller and thus show that the performace gain for each new processor after 4 becomes less and less.

--
\forall code \in C, \frac{\Delta readability(code)}{\Delta t} < 0
Re:I thought... by genes · 1999-05-12 00:22 · Score: 2

Its not the fact that Linux SMP scales to 16 ... its the fact that the parties who are going to profit from this scalability,
namely VA Research, have to do some tuning and benchmarks to make it a reality.
I'm tired of seeing MS proxies wave their 4 processor benchmark resultss without a response from VA Research which
has the test hardware capability for 4 processors (and 8 processor in the future) to prove the validity of these well
pubicized tests.
MS is trying to leap-frog over their weak performance in single and dual processors, while setting up the conclusion that if
NT Server is best in the 4 processor config it's best in all other configs. Secondly, I expect to see in the near future some
MS PR about Windows 2000 beating NT 4 Server (in some special contrived situation) and therefor is the top performing
OS server for Intel based systems.
Right now, MS is in the 'discredit Linux' mode of their PR campaign, in an attempt to turn off the medias love affair with
Linux ... and they're succeeding because our 'big guns' are not responding!

Why Intel is so hot about Linux. by mrsam · 1999-05-11 23:57 · Score: 4

One word: Merced.

The Intel/HP contingent has poured a nice chunk of change into the IA-64 chip. Now, what would you do if you were Intel right now, hhmmm?

Here you are. The amount of money you dumped into the chip is bigger than the gross national product of most third world nations. You are several months from shipping it. And now you ask yourself, who's going to use it?

Microsoft? Windows? HAHAHAHAHAHA. Microsoft's too busy getting their crap together with W2K. I don't remember ever hearing when MS intends to add support for the IA-64 chip in their OS. If ever.

Meanwhile, you keep reading that Linux will have IA-64 support right out of the box, for the chip. I recall that either VA Linux or Red Hat (or both) are working on porting the kernel to the 64 bit chip.

Right now, it looks like Linux would be the only OS that can actually support the 64 bit architecture of the Merced chip. I may be wrong, I don't read the trade press much, but for the life of me, I can't think of anything else other than Linux that has announced actual Merced support for the chip, when it ships.

Looks to me like Intel's putting all of its eggs in the Torvalds basket. Linux stands to be the only thing that can actually sell Merced chips for Intel.

we have a number opinion oriented moderators by Wah · 1999-05-12 03:17 · Score: 2

I finally had to change my settings to "-1" to see ANY "Linux isn't the end all be all" comments. C'mon moderators, try to be at least a little objective. If you ignore your faults they will only grow, then one day you'll turn around and see a Grand Canyon.

"News for Nerds, stuff that Matters" != Linux is the best thing ever.(period)

Before you flame think....games!

(my attempt at moderation without actual jurisdiction)

--
+&x

Quite good informative article by SiLmArIlLiOn · 1999-05-11 21:49 · Score: 2

Well, I think it a good introduction to Linux from an entreprise point of view although I think it may be too short.

I think the conclusion is quite realistic : 20 years to get rid of Microsoft ;-)

Overall it gives to Linux what it deserves right now: a very good alternative

--
A duck's quack does not echo and no one knows why

Why such support for servers? by creature · 1999-05-12 05:40 · Score: 2

Why not push Linux where it can make a huge impact like Workstation class systems? I'm not talking about over glorified desktops, but real workstation systems. Intel finally has a workstation class processor in the Xeon series of processors, and I've been running Linux on one for a while now. It honestly is the closed thing to a SPARCworkstation Intel has come out with. Lot of Universities and companies could replace their older workstations with new Xeon based Linux systems easily.

Ohwell, Server market is good, but I think Linux could make better in roads into the Workstation market.

10 of 50 comments (clear)