Intel chipsets are the most stable?

[Michael Brown]

This kinda proves the point - you don't have to have
all-(INTC|AMD|NVDA|insert your favorite vendor here) system to have a
certain level of stability.
Yet I would not rely on the server you described for any
mission-critical (or simply important enough) application.

Heck, neither would I. If my job depended on the system being up I'd pretty
much start again from scratch As it is, if the server fell over, pretty
much noone would care about waiting a few hours till I got home and fixed
it.

And again,
the workload you describe is not exactly stressful. Ever tried that
system for something that is both CPU and IO - intence?

CPU intensive is easy: we're talking about a K6/2-400 here remember A VNC
session often pins the CPU to max with a bit of HDD activity, and burning a
CD at 52x gets it up to about 80% CPU usage on the outer edge of the CD (and
is obviously hitting the I/O subsustem pretty hard). And there's a virus
scan and tripwire scan that runs every week (both max out the CPU and hit
the disk pretty hard). But it's not under continual high load, no.

But then,
again, a compact car has no towing capability to speak of, but for
pizza delivery it's just the right thing, be it Honda Civic, or Toyota
Corolla, or even (gulp!) Chevy Metro.

Absolutely. I've probably rebuilt a dozen or so "ancient" computers ranging
from 486/100's up to the K6/2. Most have been given away to friends and
ended up in a cupboard somewhere running Linux with things like a DHCP
server, squid, etc. The only one that I know of that has had problems is a
486/100 that was powerd down to be moved and never powered back up again
(either motherboard or CPU died, didn't investigate too much due to the lack
of similar a motherboard or CPU). They handle the job nicely and don't cost
anything but time to make; even a low-end Duron/Celeron/C3 is complete
overkill for these sort of duties. But they're obviously not suited to
running anything too important or complex.

 

[Tony Hill]

Your pals over at anandtech actually did a comparison test between two
$200 CPU's The Athlon64 3200+ (on socket 939 therefor dual channel
capable) and the P4 530.
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2249&p=14

Oh especially interesting is the game tests. AMD won 10 out of 10.....

Considering every single store I looked at showed the Athlon64 3200+
(particularly the socket 939 version) being noticeably more expensive
than the 3.0GHz P4 530, I'm not sure I really agree with their choice
in chips. In fact, everywhere I looked had the Athlon64 3200+ and the
3.2GHz P4 540 being almost exactly the same price, only issue being
that the 3200+ in socket 939 is still VERY rare.

That being said, even if they had compared the P4 540 vs. the 3200+, I
don't think the gaming tests would have changed much.

 

[Ykalon]

Considering every single store I looked at showed the Athlon64 3200+
(particularly the socket 939 version) being noticeably more expensive
than the 3.0GHz P4 530, I'm not sure I really agree with their choice
in chips. In fact, everywhere I looked had the Athlon64 3200+ and the
3.2GHz P4 540 being almost exactly the same price, only issue being
that the 3200+ in socket 939 is still VERY rare.

That being said, even if they had compared the P4 540 vs. the 3200+, I
don't think the gaming tests would have changed much.

The A64 3200+ for 939 is still new, the price will drop soon enough.

 

[Frank]

The A64 3200+ for 939 is still new, the price will drop soon enough.

Most things relating to the AMD 64 will happen soon. I don't
give much thought to what may or may not happen soon.
I pay attention to what is happening now. (price/operating system/
whatever else.

 

[George Macdonald]

Considering every single store I looked at showed the Athlon64 3200+
(particularly the socket 939 version) being noticeably more expensive
than the 3.0GHz P4 530, I'm not sure I really agree with their choice
in chips. In fact, everywhere I looked had the Athlon64 3200+ and the
3.2GHz P4 540 being almost exactly the same price, only issue being
that the 3200+ in socket 939 is still VERY rare.

The socket 939s, in 3000+, 3200+ and 3400+, just became available in the
past few days at NewEgg, all "OEM" chips for the moment except for the
previously available 3500+ "retail" which took a dive in price to $296.
Whooopppeeee!

Only the 3000+ and 3200+ are the Winchester core and pricing is a little
odd today: the OEM 3400+ costs more ($5.) than the OEM 3500+.

Rgds, George Macdonald

"Just because they're paranoid doesn't mean you're not psychotic" - Who, me??

 

[Johannes H Andersen]

It's in the Marketing, once you start advertising on TV/radio as much as
Intel has done it's no longer just commercials, it's brainwashing, so
just feel sorry for those people where it fits! ;p

But why make such a meal out of which processor is in your box? I don't
think you need to feel sorry for me, in fact I rather resent it. I
think we have cleared up that differences in performance, if any, are
small and not one-sided. Secondly, there might be factors such as
availability of motherboards and how easy they are to build with...
I'm all for diversity in the market place, it benefits everybody.

 

[Johannes H Andersen]

Most things relating to the AMD 64 will happen soon. I don't
give much thought to what may or may not happen soon.
I pay attention to what is happening now. (price/operating system/
whatever else.

 

[Johannes H Andersen]

Most things relating to the AMD 64 will happen soon. I don't
give much thought to what may or may not happen soon.
I pay attention to what is happening now. (price/operating system/
whatever else.

Precisely. Intel will have this and that AMD beating chip out soon...
and besides, my dad will beat your dad...

 

[Frank]

Precisely. Intel will have this and that AMD beating chip out
soon...
and besides, my dad will beat your dad...

True, True, True. I am having a senior moment, but doesn't AMD
license the initial i386 technology from Intel anyway???????

 

[Ykalon]

True, True, True. I am having a senior moment, but doesn't AMD
license the initial i386 technology from Intel anyway???????

Lots of cross-licensing going on. Think Intel licensed 3DNow from AMD
and they definitely licensed EMT64 from AMD. They will likely license
Hypertransport too and AMD will probably license Hyperthreading.

 

[David Maynard]

Lots of cross-licensing going on. Think Intel licensed 3DNow from AMD
and they definitely licensed EMT64 from AMD. They will likely license
Hypertransport too and AMD will probably license Hyperthreading.

Off hand I don't know of any Intel processors with 3DNow in them so it
raises the question why would they have licensed it?

 

[keith]

True, True, True. I am having a senior moment, but doesn't AMD
license the initial i386 technology from Intel anyway???????

Initial '386?? AMD licensed the 8086/8 *design* from Intel. Up until the
'486 it was a symbiotic relationship. ...then many law suits, and AMD got
to keep the '486. After that the designs have pretty much gone down their
own paths. AMD has really had the upper hand for three or four years,
while Intel has been playing around with Itanic.

 

[keith]

Lots of cross-licensing going on. Think Intel licensed 3DNow from AMD
and they definitely licensed EMT64 from AMD.

Sure, and AMD licensed MMX, SSE, and SSE2 (if not all its offspring).
....though Intel has never implemented 3DNOW.

They will likely license Hypertransport

....not much different than PCI. Hypertransport licensing is pretty much a
matter of paying dues to the SIG.

too and AMD will probably license Hyperthreading.

What license? I don't believe Intel even owns the relevent patents. A
friend has some of the key ones.

 

[Ed]

Lots of cross-licensing going on. Think Intel licensed 3DNow from AMD
and they definitely licensed EMT64 from AMD. They will likely license
Hypertransport too and AMD will probably license Hyperthreading.

OVER AT THE US trademark and patent office, there's a slightly elderly
patent successfully filed by AMD back in 1999 which indicates that the
firm could provide hyperthreading for its processors, if it should so
wish.
http://www.theinquirer.net/Default.aspx?article=6773

Ed

 

[Tony Hill]

OVER AT THE US trademark and patent office, there's a slightly elderly
patent successfully filed by AMD back in 1999 which indicates that the
firm could provide hyperthreading for its processors, if it should so
wish.
http://www.theinquirer.net/Default.aspx?article=6773

"Hyperthreading" is just Intel's name for simultaneous
multi-threading. While Intel may have some patents related to SMT,
on-chip multithreading is definitely not a uniquely Intel thing. If
nothing else, IBM's Power5 already implements multithreading (can't
remember for sure if it's SMT, CMT or some other form) and Sun is
implementing CMT in their upcoming Niagara processor.

All that being said, I'm not sure that AMD has much use for SMT at
this stage of the game. If you compare the design of the Athlon64 to
the P4, you'll notice that AMD's chip has a much shorter pipeline and
significantly less memory latency. These two factors really take away
from the potential performance benefits of SMT, and even on the P4
those benefits aren't exactly earth-shattering. In fact, as often as
not hyperthreading ends up making things a little slower.

AMD's plan (and Intel's plan now as well) for dual-core chips is MUCH
more interesting in my books.

 

[keith]

"Hyperthreading" is just Intel's name for simultaneous
multi-threading. While Intel may have some patents related to SMT,
on-chip multithreading is definitely not a uniquely Intel thing. If
nothing else, IBM's Power5 already implements multithreading (can't
remember for sure if it's SMT, CMT or some other form) and Sun is
implementing CMT in their upcoming Niagara processor.

The IBM *star PowerPCs out of ROvhester (AS/400 lab) had *MT something
like eight years ago. Whether it be SMT or CMT or ?MT, who knows.
....even the folks over on c.a. can't decide on the definitions, so I'm not
going there. ;-)

All that being said, I'm not sure that AMD has much use for SMT at this
stage of the game. If you compare the design of the Athlon64 to the P4,
you'll notice that AMD's chip has a much shorter pipeline and
significantly less memory latency. These two factors really take away
from the potential performance benefits of SMT, and even on the P4 those
benefits aren't exactly earth-shattering. In fact, as often as not
hyperthreading ends up making things a little slower.

....and I'm not sure anyone's explained why that is. Perhaps Intel's
implementation sucks? ...overhead doesn't make it worthwhile? THis is an
interesting question that no one has addressed very well. It's great
marketeering though! ;-)

AMD's plan (and Intel's plan now as well) for dual-core chips is MUCH
more interesting in my books.

Ok. I don't see it as all that "interesting". Obvious, yes.
OTOH, AMD's integrated memory controller is *both*. ;-)