No point in investing anymore $$$ in socketA?

I think it was Wes who stated that AMD doesn't have any plans to come out
w/faster Athlons (say = 3.0Ghz). So, is it pretty much pointless to put any
money into socketA motherboards at this point?
Would it be better to wait and see what happens w/the 64-bit platforms than say
buy an Nforce2 motherboard now?
-Bill (remove "botizer" to reply via email)

Wblane wrote:
I think it was Wes who stated that AMD doesn't have any plans to come
out w/faster Athlons (say = 3.0Ghz).

Steady on... I'd suspect that the Barton/Tbred cores are pretty much
exhausted at ~2.25GHz (there's been no announcement or even talk of a 3400+
so far, and although people can generally run at 2.3GHz, heat production is
probably a concern)

However, AMD have said that the XP will probably live another year or so...
I'm not sure on their plans, but it's quite possible that they can do a
simple die-shrink - move the production to the new fabs (Fab 30, Dresdon is
planning to move to 90nm next year), which would probably allow greater
speeds.

So, is it pretty much pointless
to put any money into socketA motherboards at this point?
Would it be better to wait and see what happens w/the 64-bit
platforms than say buy an Nforce2 motherboard now?
-Bill (remove "botizer" to reply via email)


Depends... if you buy a nForce2 now, and stick a processor in there clocked
at ~2.2-2.3GHz, you're talking top of the range. 64bit is all well and
good, but you're unlikely to reap the benefits in most applications.

Price/performance for the home market will lie with Socket A for a while.

If you want more performance now, buy an nForce2... I wouldn't bother with
64bit right now... prices are too high for the sort of gains you'll see in
your average application.

The major benefits of 64bit right now are a memory space 4GB, which you are
unlikely to need.

Ben
--
I'm not just a number. To many, I'm known as a String...

On Fri, 31 Oct 2003 19:31:14 +0000, Wblane wrote:

I think it was Wes who stated that AMD doesn't have any plans to come
out w/faster Athlons (say = 3.0Ghz). So, is it pretty much pointless to
put any money into socketA motherboards at this point?

If you're buying a new system I'd say it was well worth it. If you've
already got a socket A system that works, that's a hard choice. I replaced
one of my old KT7-Raid boards with a new board that will do 200MHz+ FSB,
DDR400 ram, etc. (sis746fx). Honestly, it wasn't worth the trouble. And I
wouldn't have done it if I'd known how much work it was going to be. It's
just not that much of a difference to me. But I was building a system to
give away and needed an extra board.

Would it be better to wait and see what happens w/the 64-bit platforms
than say buy an Nforce2 motherboard now?

I'm going to let the people that really need the speed buy into this
first. I could get by with a 600MHz Duron. I'll wait until I just can't
take it from that little voice inside me that keeps saying, get it, you
know you want too.:-)

--
Abit KT7-Raid (KT133) Tbred B core CPU @2400MHz (24x100FSB)
http://mysite.verizon.net/res0exft/cpu.html

I have to agree with Wes, I bought an NF7-S motherboard after hearing all
the wonderful stuff about the nf2 chipset.

I lost about 100mhz overclock going from my stable and fast KT333 Epox
8k5a2+ motherboard to the Abit NF7-S. I also have pc2700 which doesn't
clock much higher than 190mhz with decent memory timings.

3dmark benchmarks were slightly higher but not enough to justify a mobo swap
at $180can.

I returned the Abit mobo and kept my Epox... I'm still hitting 17,500/ 6600
in 3dmark 2001/2003 so I ain't worried too much about a few points.

"Wes Newell" wrote in message
news
On Fri, 31 Oct 2003 19:31:14 +0000, Wblane wrote:

I think it was Wes who stated that AMD doesn't have any plans to come
out w/faster Athlons (say = 3.0Ghz). So, is it pretty much pointless to
put any money into socketA motherboards at this point?

If you're buying a new system I'd say it was well worth it. If you've
already got a socket A system that works, that's a hard choice. I replaced
one of my old KT7-Raid boards with a new board that will do 200MHz+ FSB,
DDR400 ram, etc. (sis746fx). Honestly, it wasn't worth the trouble. And I
wouldn't have done it if I'd known how much work it was going to be. It's
just not that much of a difference to me. But I was building a system to
give away and needed an extra board.

Would it be better to wait and see what happens w/the 64-bit platforms
than say buy an Nforce2 motherboard now?

I'm going to let the people that really need the speed buy into this
first. I could get by with a 600MHz Duron. I'll wait until I just can't
take it from that little voice inside me that keeps saying, get it, you
know you want too.:-)

--
Abit KT7-Raid (KT133) Tbred B core CPU @2400MHz (24x100FSB)
http://mysite.verizon.net/res0exft/cpu.html


---
Outgoing mail is certified Virus Free.
Checked by AVG anti-virus system (http://www.grisoft.com).
Version: 6.0.532 / Virus Database: 326 - Release Date: 10/27/2003

LBJGH wrote:
I have to agree with Wes, I bought an NF7-S motherboard after hearing
all the wonderful stuff about the nf2 chipset.

I lost about 100mhz overclock going from my stable and fast KT333 Epox
8k5a2+ motherboard to the Abit NF7-S. I also have pc2700 which
doesn't clock much higher than 190mhz with decent memory timings.

3dmark benchmarks were slightly higher but not enough to justify a
mobo swap at $180can.

I returned the Abit mobo and kept my Epox... I'm still hitting
17,500/ 6600 in 3dmark 2001/2003 so I ain't worried too much about a
few points.

It begs the question "What were you expecting from a new motherboard"

If you wanted the faster FSB, then you need the faster RAM (to be able to
take full advantage of it)... you did pretty well with 190MHz, but an
increase of 14% in memory bandwidth is hardly groundbreaking. 20% if you'd
have hit 200MHz.

Ben
--
I'm not just a number. To many, I'm known as a String...

The below is a major fallacy that people associate w/the x86 architecture. A
lie that Apple has publicly promulgated.

The maximum size of a SEGMENT is 4GB, but "the virtual address space can
consist of 2^14 unique segments in memory, each of which has a maximum size of
4GB. " "By combining the 14-bit segment selector w/the 32-bit offset, we get a
46-bit virtual address. Therefore the 80386DX's virtual address space can
contain 2^46 equals 64 TB." This quote is from Treibel and Singh's excellent
The 8088 and 8086 Microprocessors textbook. Treibel and Singh both having PHd's
in EE. Minimally, since EACH of the 6 segment selector registers can address
4GB we can have 24GB of immediately accessible, physical memory.

The major benefits of 64bit right now are a memory space 4GB, which you are
unlikely to need.

Ben
--
I'm not just a number. To many, I'm known as a String...



-Bill (remove "botizer" to reply via email)

This is all I needed to hear. The system I was thinking of upgrading is running
a Duron-Morgan at 1456 on a KT133 based board. It's using a Geforce 2/GTS video
card so I think I'll look into spending my $$$ there rather than a new
motherboard+CPU.

If you're buying a new system I'd say it was well worth it. If you've
already got a socket A system that works, that's a hard choice. I replaced
one of my old KT7-Raid boards with a new board that will do 200MHz+ FSB,
DDR400 ram, etc. (sis746fx). Honestly, it wasn't worth the trouble. And I
wouldn't have done it if I'd known how much work it was going to be. It's
just not that much of a difference to me. But I was building a system to
give away and needed an extra board.


-Bill (remove "botizer" to reply via email)

Wblane wrote:
The major benefits of 64bit right now are a memory space 4GB, which
you are unlikely to need.

The below is a major fallacy that people associate w/the x86
architecture. A lie that Apple has publicly promulgated.

The maximum size of a SEGMENT is 4GB, but "the virtual address space
can consist of 2^14 unique segments in memory, each of which has a
maximum size of 4GB. " "By combining the 14-bit segment selector
w/the 32-bit offset, we get a 46-bit virtual address. Therefore the
80386DX's virtual address space can contain 2^46 equals 64 TB." This
quote is from Treibel and Singh's excellent The 8088 and 8086
Microprocessors textbook. Treibel and Singh both having PHd's in EE.

You might want to get your quoting right ... the 8086 had 16-bit segments
and 16-bit offsets, which were combined in the form segment*16+offset, which
gave a 20-bit (physical, linear, and virtual) address space. The 64kb
accessable above the 1 meg mark wrapped around and was in fact just the
lowest 64kb of memory. So I'm guessing the above did not come out of a
8086/8088 book

Minimally, since EACH of the 6 segment selector registers can address
4GB we can have 24GB of immediately accessible, physical memory.

Sorta. But no Read the Intel docs, and you'll see that these virtual
segments wrap at the 4 gig physical mark during the translation from virtual
to linear addresses. So you're always limited to 4 gig of linear address
space (so only 4 gig of unique memory in your virtual address space).
However, you can get up to 64 GB (GB, not TB) using paging extensions (P6
and higher IIRC) which enable you to map from a larger 36-bit physical
address space into the 4gig linear address space of your process. But you
can never have more than 4 gig mapped in at any one point in time. For a far
better explanation, read chapter 3 of volume 3 of the Intel IA32 Software
Developer's Manual.

--
Michael Brown
www.emboss.co.nz : OOS/RSI software and more
Add michael@ to emboss.co.nz - My inbox is always open

The book covers the 8088/8086 through the Pentiums. The 80386DX protected
memory model is still used in modern x86 CPU's. I wasn't talking about real
mode. They weren't combined either the segment register was shifted left by
4-bits then added to the 16-bit offset. Real mode doesn't have a virtual
address space.

You might want to get your quoting right ... the 8086 had 16-bit segments
and 16-bit offsets, which were combined in the form segment*16+offset, which
gave a 20-bit (physical, linear, and virtual) address space. The 64kb
accessable above the 1 meg mark wrapped around and was in fact just the
lowest 64kb of memory. So I'm guessing the above did not come out of a
8086/8088 book

So all six segments which have segment selectors loaded in segment registers
have to have their selector base address within the same 4GB physical address
space?

Why does the Introduction to the 386 DX Microprocessor from the 386 DX
Microprocessor Programmer's Reference Manual clearly state: "the processor can
address up to 4 gigabytes of physical memory and 64 Terabytes of virtual
memory." Misprint or lie?

Sorta. But no Read the Intel docs, and you'll see that these virtual
segments wrap at the 4 gig physical mark during the translation from virtual
to linear addresses. So you're always limited to 4 gig of linear address
space (so only 4 gig of unique memory in your virtual address space).
However, you can get up to 64 GB (GB, not TB) using paging extensions (P6
and higher IIRC) which enable you to map from a larger 36-bit physical
address space into the 4gig linear address space of your process. But you
can never have more than 4 gig mapped in at any one point in time. For a far
better explanation, read chapter 3 of volume 3 of the Intel IA32 Software
Developer's Manual.


-Bill (remove "botizer" to reply via email)

Wblane wrote:
[...]
Sorta. But no Read the Intel docs, and you'll see that these
virtual segments wrap at the 4 gig physical mark during the
translation from virtual to linear addresses. So you're always
limited to 4 gig of linear address space (so only 4 gig of unique
memory in your virtual address space). However, you can get up to 64
GB (GB, not TB) using paging extensions (P6 and higher IIRC) which
enable you to map from a larger 36-bit physical address space into
the 4gig linear address space of your process. But you can never
have more than 4 gig mapped in at any one point in time. For a far
better explanation, read chapter 3 of volume 3 of the Intel IA32
Software Developer's Manual.

So all six segments which have segment selectors loaded in segment
registers have to have their selector base address within the same
4GB physical address space?

Yes. The descriptor for each selector has only a 32-bit value for the base
address (in the linear memory space) of that selector, so can only start
within the first 4 gig. Because of the wrapping at the 4-gig mark for the
linear address space, only 4 gig is accessable at any one time.

Why does the Introduction to the 386 DX Microprocessor from the 386 DX
Microprocessor Programmer's Reference Manual clearly state: "the
processor can address up to 4 gigabytes of physical memory and 64
Terabytes of virtual memory." Misprint or lie?

Neither, it marketing There are 64TB of virtual memory, but only 4 GB of
"unique" memory.

--
Michael Brown
www.emboss.co.nz : OOS/RSI software and more
Add michael@ to emboss.co.nz - My inbox is always open