2 socket/4 core AMD FX system coming

[Yousuf Khan]

It is disappointing as the Conroe is good for the home market, so how does
releasing a server chip help?

It's not a server chip, it's a home chip with previously unknown
multiprocessor capabilities.

Yousuf Khan

 

[Mark]

It's not a server chip, it's a home chip with previously unknown
multiprocessor capabilities.

Yousuf Khan

Firstly, it is not a home chip if it is the extreme edition. Second, you
stated that "This looks like it's a slightly modified 2-way Opteron system
(modified because it won't require server-grade registered RAM), that's
gotten so popular in the server world." Again, what use is a chip that would
be popular/useful in the server world? Home computing has different
requirements, and the AMD response isn't satisfying those home requirements.

 

[YKhan]

Firstly, it is not a home chip if it is the extreme edition.

Yeah, you're splitting hairs here. It's a consumer chip if you prefer.

Second, you
stated that "This looks like it's a slightly modified 2-way Opteron system
(modified because it won't require server-grade registered RAM), that's
gotten so popular in the server world." Again, what use is a chip that would
be popular/useful in the server world? Home computing has different
requirements, and the AMD response isn't satisfying those home requirements.

Well, "slightly modified" in the sense that it's not a 2-way Opteron,
it's a 2-way Athlon. And it's using Hypertransport, which is popular in
the server world, because of its low-latency. And home computing
doesn't have that much different requirements than server computing --
multiprocessing is one of those areas which can be used in homes quite
obviously. Does everything have to be spelled out in detail? It sounds
like you're looking for a legal briefing.

Yousuf Khan

 

[Mark]

Yeah, you're splitting hairs here. It's a consumer chip if you prefer.


Well, "slightly modified" in the sense that it's not a 2-way Opteron,
it's a 2-way Athlon. And it's using Hypertransport, which is popular in
the server world, because of its low-latency. And home computing
doesn't have that much different requirements than server computing --
multiprocessing is one of those areas which can be used in homes quite
obviously. Does everything have to be spelled out in detail? It sounds
like you're looking for a legal briefing.

I would completely disagree with the view that home computing and server
computing have the same requirements. While multiprocessing can be used in
homes, how often is it used - particularly in the sense that multiprocessing
performance becomes the bottleneck. The reality is that home computers tend
to be used by one person at a time, and we are fundamentally serial
processors. Even if you have a few other things running in the background,
these are usually not strenous and the most important requirement is what is
in the foreground - often a game - so single threaded performance is
crucial. On the other hand, a server is naturally amenable to multithreading
as many people may be accessing the same maching at the same time. I can't
believe that you think that server needs are the same as home computing
needs.

 

[Tony Hill]

I wonder if power and ground are now the new "reserved" pins? You don't
want it used, and while you're not using it, you want it to be kept in
logical ON state, so you call it power. Similarly, other pins which you
want to remain in the logical OFF state, you call those ground. Makes
you wonder.

I suppose stranger things have happened, but I kind of doubt it. It
just doesn't make sense to waste pins with hidden, secret features
that you have no plans on using. Each and every pin costs money, so
generally you want to keep the excess to a minimum.

AMD is introducing Socket S2 or something for Turion laptops. It's 640
pins, so they've blown off hundreds of pins. And apparently it's still
capable of dual-channel DDR2 operation, despite the fact that it's got
less pins than the older AMD Socket 754 DDR single-channel-only
connector. Hell it's got 300 pins less than Socket AM2, which is also
dual-channel DDR2. Again makes you wonder.

Laptop processors are going to come in with a much lower power
requirement, and therefore they'll be able to get away with fewer
power and ground pins. Still seems like a rather small number of pins
to me though, so I'm going to hold out on too much judgment until more
concrete details are revealed.

 

[nobody]

I would completely disagree with the view that home computing and server
computing have the same requirements. While multiprocessing can be used in
homes, how often is it used - particularly in the sense that multiprocessing
performance becomes the bottleneck. The reality is that home computers tend
to be used by one person at a time, and we are fundamentally serial
processors. Even if you have a few other things running in the background,
these are usually not strenous and the most important requirement is what is
in the foreground - often a game - so single threaded performance is
crucial. On the other hand, a server is naturally amenable to multithreading
as many people may be accessing the same maching at the same time. I can't
believe that you think that server needs are the same as home computing
needs.

I look at my desktop at home and see 3 instances of DevStudio.NET,
Query Analyser, Enterprise Manager, Word, Excel, 3 instances of IE, 2
notepads, Firefox with 5 tabs, and CISCO VPN, IIS, and SQL Server in
the background - just the right load for my dual Opteron rig. It's 2
years old, and when I will have to install Vista, I'll probably need a
new rig. If the dual Athlon setup will be noticeably cheaper than
Opty, I'll definitely go for it - I couldn't care less about
registered RAM. Oh, yeah, I also encode DVDs quite often, and it
seems NeroVision can use both CPUs. For what I'm doing, this setup
would beat Conroe hands down, and Woodcrest, while competitive, will
be priced multiples of that. Not every home user is limited to a
single instance of a game ;-)
NNN

 

[YKhan]

I suppose stranger things have happened, but I kind of doubt it. It
just doesn't make sense to waste pins with hidden, secret features
that you have no plans on using. Each and every pin costs money, so
generally you want to keep the excess to a minimum.

Yeah, but there are other economic factors here too. AMD already had
the Socket 940 for Opterons. The Socket 939 in all likelihood could use
the same socket manufacturing process with just minor modifications, so
they could very likely come off of the same assembly line.

Laptop processors are going to come in with a much lower power
requirement, and therefore they'll be able to get away with fewer
power and ground pins. Still seems like a rather small number of pins
to me though, so I'm going to hold out on too much judgment until more
concrete details are revealed.

Sorry, it's Socket S1 not S2, and it's got 638 pins not 640. It is
dual-channel though.

www.hardwarezone.com® :: Articles - Coolest Hardware, Hottest Reviews
http://www.hardwarezone.com/articles/view.php?cid=1&id=1904&pg=3

 

[Nate Edel]

stated that "This looks like it's a slightly modified 2-way Opteron system
(modified because it won't require server-grade registered RAM), that's
gotten so popular in the server world." Again, what use is a chip that would
be popular/useful in the server world? Home computing has different
requirements, and the AMD response isn't satisfying those home requirements.

Uh, barely. The one major change in requirements is multi-socket systems
(and on the AMD side, registered/ecc ram). In the case of the Xeon on
Intel's side, that's all the difference is EITHER (at least for the
base-level Xeons) and except for the multi-socket support most of that is on
the northbridge on the Intel side.

 

[Nate Edel]

these are usually not strenous and the most important requirement is what is
in the foreground - often a game - so single threaded performance is
crucial. On the other hand, a server is naturally amenable to multithreading
as many people may be accessing the same maching at the same time. I can't

Dude, most Windows programs are multithreaded these days. Office, Explorer,
many others. Not to mention such things as virus checkers and firewalls, all
of which can take advantage of multiprocessors.

Granted, historically the price premium (ignoring the brief generation of
multiable Celerons that led to the Abit BP6 and a few similar boards) for
duals has been too high to appeal to most home users (or serious hobbyists,
more importantly) but look at how fast the adoption of dual cores has been:
if it's available afforably, they will come, and fast.

This will clearly not be for general home users, cost wise (heck, they're
buying Semprons and Celerons to begin with!), but my guess is if the price
premium is reasonable, serious hobbyists/gamers will be all over it.

Oh, and games are getting multithreaded very quickly as well.

 

[The little lost angel]

computing have the same requirements. While multiprocessing can be used in
homes, how often is it used - particularly in the sense that multiprocessing
performance becomes the bottleneck. The reality is that home computers tend
to be used by one person at a time, and we are fundamentally serial
processors. Even if you have a few other things running in the background,
these are usually not strenous and the most important requirement is what is
in the foreground - often a game - so single threaded performance is
crucial.

Until you realize that you're getting 50% (or higher if your base
latency is good) higher latencies on a single core when playing online
(which most games do have nowadays) because you don't have a second
core (real or virtual) to offload your firewall, virus scanner, P2P
stuff and background downloads to.

 

[max]

Dude, most Windows programs are multithreaded these days. Office, Explorer,
many others. Not to mention such things as virus checkers and firewalls, all
of which can take advantage of multiprocessors.

So is there a big performance increase going to a multi-core processor
for standard home Windows use? I thought there wasn't, but I'll admit
I haven't followed those benchmarks much.

max

 

[GSV Three Minds in a Can]

So is there a big performance increase going to a multi-core processor
for standard home Windows use? I thought there wasn't, but I'll admit
I haven't followed those benchmarks much.

Depends entirely on what this 'standard home user' does. I just made the
transition from XP3000+ (2.2Ghz) to XP4400x2 (2.2Ghz) and I notice a
detectable improvement in overall system =responsiveness= (i.e. when
encoding audio, or running big Excel VBA program or whatever, the system
still responds PDQ to requests like ctrl-alt-del, or windows keyboard
button). Actual 'flat out speed' on any one task is not a lot faster (a
bit, but then the new CPU has more cache, the disk is faster, etc.), but
I can now run two at once. 'Oblivion' (the game) plays really well, but
a lot of that is also down to the more memory and faster graphics card.

So back to your Q .. 'big performance increase' - no. 'Detectably more
responsive system' - yep.

 

[Mark]

Depends entirely on what this 'standard home user' does. I just made the
transition from XP3000+ (2.2Ghz) to XP4400x2 (2.2Ghz) and I notice a
detectable improvement in overall system =responsiveness= (i.e. when
encoding audio, or running big Excel VBA program or whatever, the system
still responds PDQ to requests like ctrl-alt-del, or windows keyboard
button). Actual 'flat out speed' on any one task is not a lot faster (a
bit, but then the new CPU has more cache, the disk is faster, etc.), but I
can now run two at once. 'Oblivion' (the game) plays really well, but a
lot of that is also down to the more memory and faster graphics card.

So back to your Q .. 'big performance increase' - no. 'Detectably more
responsive system' - yep.

Unfortunately, this doesn't answer the question we are discussing. The issue
is whether a 4-core AMD FX system would be preferable to a dual-core Conroe
processor for home users.

 

[GSV Three Minds in a Can]

Bitstring <[email protected]>, from the wonderful person
Mark <[email protected]> said

Unfortunately, this doesn't answer the question we are discussing. The issue
is whether a 4-core AMD FX system would be preferable to a dual-core Conroe
processor for home users.

We won't know that until we see them in action, will we. (and define
'preferable', and 'home users' a bit closer). Once you get past 2 cores
it is less clear cut IMO, and I might be more interested in other
criteria (like 'quiet' and 'energy efficient').

 

[Yousuf Khan]

I would completely disagree with the view that home computing and server
computing have the same requirements. While multiprocessing can be used in
homes, how often is it used - particularly in the sense that multiprocessing
performance becomes the bottleneck. The reality is that home computers tend
to be used by one person at a time, and we are fundamentally serial
processors. Even if you have a few other things running in the background,
these are usually not strenous and the most important requirement is what is
in the foreground - often a game - so single threaded performance is
crucial. On the other hand, a server is naturally amenable to multithreading
as many people may be accessing the same maching at the same time. I can't
believe that you think that server needs are the same as home computing
needs.

The people who don't strain their home computers hard enough to require
multiprocessing, won't need multiprocessing. Those who do strain them
hard enough to require multiprocessing, need multiprocessing. Simple as
that! There's not one single requirement that describes all home users.

Those that do require multiprocessing, will initially go for single
multi-cores, then if they find even that's not enough, they will go one
step beyond and go for multi-processing & multi-cores. There's also the
added benefit that if this technology works for FX's, they'll work for
other lower-ranked AM2 processors too, such as X2's and even Semprons.
Who knows maybe a dual-Sempron might have some cost advantages over a
single X2 in some cases? Gives the consumer some level of flexibility.
Of course, the first crop of 4X4 motherboards will be aimed at the
high-end, but perhaps in a few months, they'll have economy 4X4
motherboards too?

Yousuf Khan

 

[Keith]

The people who don't strain their home computers hard enough to require
multiprocessing, won't need multiprocessing. Those who do strain them
hard enough to require multiprocessing, need multiprocessing. Simple as
that! There's not one single requirement that describes all home users.

Wow, Yousuf, that's deep! ;-)

Those who do stress their macines may benefit from or *want*
multi-processing, but that is different from "needing" it. It's a good
thing for anyone who can chew gum and walk at the same time, though not
required.

Those that do require multiprocessing, will initially go for single
multi-cores, then if they find even that's not enough, they will go one
step beyond and go for multi-processing & multi-cores.

Huh? You lost me with that one.

There's also the
added benefit that if this technology works for FX's, they'll work for
other lower-ranked AM2 processors too, such as X2's and even Semprons.
Who knows maybe a dual-Sempron might have some cost advantages over a
single X2 in some cases?

I doubt it. That's not the way these things work.

 

[Yousuf Khan]

Huh? You lost me with that one.

You start with single-core. You then graduate to dual-core. If dual-core
is not enough, then you go with dual-core + multiprocessing.

Yousuf Khan

 

[The little lost angel]

You start with single-core. You then graduate to dual-core. If dual-core
is not enough, then you go with dual-core + multiprocessing.

I think the confusion's with the using of multi-processing to denote
the multiple package CPU as opposed to multi-processing with single
package multiple core CPU.

So the progression is Single Core -> Dual Core -> Quad Core ->
Multiple Dual/Quad Cores

 

[Yousuf Khan]

I think the confusion's with the using of multi-processing to denote
the multiple package CPU as opposed to multi-processing with single
package multiple core CPU.

So the progression is Single Core -> Dual Core -> Quad Core ->
Multiple Dual/Quad Cores

How about "multiple socket" to eliminate the confusion?

 

[Nate Edel]

So is there a big performance increase going to a multi-core processor
for standard home Windows use? I thought there wasn't, but I'll admit
I haven't followed those benchmarks much.

Either multicore or hyperthreading will make a noticeable difference in
responsiveness on a home desktop, but in general standard home Windows use
will not stress out a modern machine. Although standard advice on memory I
think can be a bit low, except for Via C3s or similar, I don't think CPU is
really an issue.

For serious home power users, though - either work from home or hobbyist
multimedia, etc - a multicore can make a big difference. DVD encoding
while doing serious work can slow down the machine a lot, for one example,
and letting one core crank on that can really help with both
time-to-complete and performance of your other applications.

Gaming is another matter, and I'm not a heavy-duty enough gamer (only just
getting to play Temple of Elemental Evil, for example) to bother to keep up
on it.