PDA

View Full Version : Opteron series 100 vs athlon 64 fx


Martin
June 29th 04, 08:03 AM
Anyone got a nice article that compares the two?

I know opteron is for servers, and fx for multimedia/gaming, but I'd like to
see a comparison of the two, including value for money.

Thanks
Martin

Wes Newell
June 29th 04, 08:10 AM
On Tue, 29 Jun 2004 08:03:39 +0100, Martin wrote:

> Anyone got a nice article that compares the two?
>
The Opteron and 940 pin FX's are basically the same, except I believe the
multiplier is not locked on the FX's. Both require reg. ram.

The 939 FX's don't need registered ram.

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

Martin
June 29th 04, 08:45 AM
"Wes Newell" > wrote in message
news:[email protected] .net...
> On Tue, 29 Jun 2004 08:03:39 +0100, Martin wrote:
>
> > Anyone got a nice article that compares the two?
> >
> The Opteron and 940 pin FX's are basically the same, except I believe the
> multiplier is not locked on the FX's. Both require reg. ram.
>
> The 939 FX's don't need registered ram.
>
> --
> Abit KT7-Raid (KT133) Tbred B core CPU @2400MHz (24x100FSB)
> http://mysite.verizon.net/res0exft/cpu.htm

Martin
June 29th 04, 08:48 AM
So the price difference between
Opteron146 2GHz 1MB S940 Box (191exVAT)
and
Athlon 64 FX-53 1MB S940 (589 ex VAT)
is just down to clock speed 2GHz vs 2.4 GHz??

Martin

"Wes Newell" > wrote in message
news:[email protected] .net...
> On Tue, 29 Jun 2004 08:03:39 +0100, Martin wrote:
>
> > Anyone got a nice article that compares the two?
> >
> The Opteron and 940 pin FX's are basically the same, except I believe the
> multiplier is not locked on the FX's. Both require reg. ram.
>
> The 939 FX's don't need registered ram.
>
> --
> Abit KT7-Raid (KT133) Tbred B core CPU @2400MHz (24x100FSB)
> http://mysite.verizon.net/res0exft/cpu.htm

General Schvantzkoph
June 29th 04, 04:01 PM
On Tue, 29 Jun 2004 08:48:10 +0100, Martin wrote:

> So the price difference between
> Opteron146 2GHz 1MB S940 Box (191exVAT)
> and
> Athlon 64 FX-53 1MB S940 (589 ex VAT)
> is just down to clock speed 2GHz vs 2.4 GHz??

Yes the price difference is mostly due to clock speed although there
are also marketing issues at work. The Opteron 1xx is aimed at the bottom
of the server market whereas the FX is aimed at the very top of the
desktop market. The 64 bitness of the Opteron is a significant
differentiator in the server and workstation markets because Linux is
available in a 64 bit flavor. This allows AMD to compete for all segments
of the server market, something they could never do before. In the desktop
area the only OS that really counts is Windows which is still in beta for
the 64 bit version. Until MS releases 64bit Windows the only reason to buy
an AMD64 for ordinary desktop use is for top performance so AMD is only
offering the fastest speed grades for the FX. When MS finnaly ships the 64
bit version of XP I expect you'll see reasonably priced FX chips.

Martin
June 29th 04, 06:27 PM
Interesting insight.

I would have expected some aspects like 3DNow and SSE2 instructions to be
missing from the Opteron, but I can't clearly see if it's in or out.



"General Schvantzkoph" > wrote in message
...
> On Tue, 29 Jun 2004 08:48:10 +0100, Martin wrote:
>
> > So the price difference between
> > Opteron146 2GHz 1MB S940 Box (191exVAT)
> > and
> > Athlon 64 FX-53 1MB S940 (589 ex VAT)
> > is just down to clock speed 2GHz vs 2.4 GHz??
>
> Yes the price difference is mostly due to clock speed although there
> are also marketing issues at work. The Opteron 1xx is aimed at the bottom
> of the server market whereas the FX is aimed at the very top of the
> desktop market. The 64 bitness of the Opteron is a significant
> differentiator in the server and workstation markets because Linux is
> available in a 64 bit flavor. This allows AMD to compete for all segments
> of the server market, something they could never do before. In the desktop
> area the only OS that really counts is Windows which is still in beta for
> the 64 bit version. Until MS releases 64bit Windows the only reason to buy
> an AMD64 for ordinary desktop use is for top performance so AMD is only
> offering the fastest speed grades for the FX. When MS finnaly ships the 64
> bit version of XP I expect you'll see reasonably priced FX chips.
>

Ben Pope
June 29th 04, 06:36 PM
Martin wrote:
> Interesting insight.
>
> I would have expected some aspects like 3DNow and SSE2 instructions to be
> missing from the Opteron, but I can't clearly see if it's in or out.

They'll be there. The usefulness of the instructions/architecture extends
way beyond desktop multimedia apps.

Ben
--
A7N8X FAQ: www.ben.pope.name/a7n8x_faq.html
Questions by email will likely be ignored, please use the newsgroups.
I'm not just a number. To many, I'm known as a String...

General Schvantzkoph
June 29th 04, 07:13 PM
On Tue, 29 Jun 2004 18:36:46 +0100, Ben Pope wrote:

> Martin wrote:
>> Interesting insight.
>>
>> I would have expected some aspects like 3DNow and SSE2 instructions to be
>> missing from the Opteron, but I can't clearly see if it's in or out.
>
> They'll be there. The usefulness of the instructions/architecture extends
> way beyond desktop multimedia apps.
>
> Ben

Most importantly they don't want to have to maintain two different
processor cores. The differences between the different flavors mostly come
down to the cache size and packaging. The three Opterons are nearly
identical except the 1xx doesn't do any coherency checking on the
hypertransports, the 2xx does it on 1 bus and the 8xx does it on all
three. I'd guess that the difference comes down to a couple of enable bits
which are configured either on one of the metal layers or possible even
inside of the package. The 754 only has a single DDR interface but it's
possible that it's the same die as the Opterons and that it's strictly a
packaging difference, or if it is a die difference it's the same basic
core with one interface ripped off. The 939 is a slightly tweeked version
but the changes are likely very small and will eventually appear in the
Opteron when they do the next spin.

Ben Pope
June 29th 04, 07:24 PM
General Schvantzkoph wrote:
> On Tue, 29 Jun 2004 18:36:46 +0100, Ben Pope wrote:
>
>> Martin wrote:
>>> Interesting insight.
>>>
>>> I would have expected some aspects like 3DNow and SSE2 instructions to
>>> be missing from the Opteron, but I can't clearly see if it's in or out.
>>
>> They'll be there. The usefulness of the instructions/architecture
>> extends way beyond desktop multimedia apps.
>>
>> Ben
>
> Most importantly they don't want to have to maintain two different
> processor cores. The differences between the different flavors mostly come
> down to the cache size and packaging. The three Opterons are nearly
> identical except the 1xx doesn't do any coherency checking on the
> hypertransports, the 2xx does it on 1 bus and the 8xx does it on all
> three. I'd guess that the difference comes down to a couple of enable bits
> which are configured either on one of the metal layers or possible even
> inside of the package. The 754 only has a single DDR interface but it's
> possible that it's the same die as the Opterons and that it's strictly a
> packaging difference, or if it is a die difference it's the same basic
> core with one interface ripped off. The 939 is a slightly tweeked version
> but the changes are likely very small and will eventually appear in the
> Opteron when they do the next spin.

Agreed. I was gonna mention that, but was not sure of the differences in
the core between the XP and the Athlon64s. I assumed they were different
enough to warrant such modifications (but that they wouldn't bother removing
the features - why, anyway?). Of course, the FX and Opteron, are, as we
know, practically identical.

Ben
--
A7N8X FAQ: www.ben.pope.name/a7n8x_faq.html
Questions by email will likely be ignored, please use the newsgroups.
I'm not just a number. To many, I'm known as a String...

Martin
June 29th 04, 08:18 PM
In fact
http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/23932.pdf
confirms the instructions are there.

Thanks all
Martin

"Ben Pope" > wrote in message
...
> General Schvantzkoph wrote:
> > On Tue, 29 Jun 2004 18:36:46 +0100, Ben Pope wrote:
> >
> >> Martin wrote:
> >>> Interesting insight.
> >>>
> >>> I would have expected some aspects like 3DNow and SSE2 instructions to
> >>> be missing from the Opteron, but I can't clearly see if it's in or
out.
> >>
> >> They'll be there. The usefulness of the instructions/architecture
> >> extends way beyond desktop multimedia apps.
> >>
> >> Ben
> >
> > Most importantly they don't want to have to maintain two different
> > processor cores. The differences between the different flavors mostly
come
> > down to the cache size and packaging. The three Opterons are nearly
> > identical except the 1xx doesn't do any coherency checking on the
> > hypertransports, the 2xx does it on 1 bus and the 8xx does it on all
> > three. I'd guess that the difference comes down to a couple of enable
bits
> > which are configured either on one of the metal layers or possible even
> > inside of the package. The 754 only has a single DDR interface but it's
> > possible that it's the same die as the Opterons and that it's strictly a
> > packaging difference, or if it is a die difference it's the same basic
> > core with one interface ripped off. The 939 is a slightly tweeked
version
> > but the changes are likely very small and will eventually appear in the
> > Opteron when they do the next spin.
>
> Agreed. I was gonna mention that, but was not sure of the differences in
> the core between the XP and the Athlon64s. I assumed they were different
> enough to warrant such modifications (but that they wouldn't bother
removing
> the features - why, anyway?). Of course, the FX and Opteron, are, as we
> know, practically identical.
>
> Ben
> --
> A7N8X FAQ: www.ben.pope.name/a7n8x_faq.html
> Questions by email will likely be ignored, please use the newsgroups.
> I'm not just a number. To many, I'm known as a String...
>
>

Thomas A. Horsley
June 29th 04, 11:00 PM
>I would have expected some aspects like 3DNow and SSE2 instructions to be
>missing from the Opteron, but I can't clearly see if it's in or out.

Nope, everything is on opteron (in fact the amd64 architecture manuals
don't make any distinction between them). The only thing "missing" is
Intel's "prescott new instructions".
--
>>==>> The *Best* political site <URL:http://www.vote-smart.org/> >>==+
email: icbm: Delray Beach, FL |
<URL:http://home.att.net/~Tom.Horsley> Free Software and Politics <<==+

Martin
June 30th 04, 07:34 AM
"Thomas A. Horsley" > wrote in message
...
> >I would have expected some aspects like 3DNow and SSE2 instructions to be
> >missing from the Opteron, but I can't clearly see if it's in or out.
>
> Nope, everything is on opteron (in fact the amd64 architecture manuals
> don't make any distinction between them). The only thing "missing" is
> Intel's "prescott new instructions".

Prescott new instructions as outlined in
http://www.thejemreport.com/modules.php?op=modload&name=News&file=article&sid=109
??

Does the FX also omit these?

Although apparently not a functional enhancement, it will still lead to CPU
specific software in Linux/Windows. Would I lose out on software
availability if I bought AMD instead of Intel?

Thanks
Martn

Wes Newell
June 30th 04, 08:24 AM
On Wed, 30 Jun 2004 07:34:33 +0100, Martin wrote:

>
> "Thomas A. Horsley" > wrote in message
> ...
>> >I would have expected some aspects like 3DNow and SSE2 instructions to be
>> >missing from the Opteron, but I can't clearly see if it's in or out.
>>
>> Nope, everything is on opteron (in fact the amd64 architecture manuals
>> don't make any distinction between them). The only thing "missing" is
>> Intel's "prescott new instructions".
>
> Prescott new instructions as outlined in
> http://www.thejemreport.com/modules.php?op=modload&name=News&file=article&sid=109
> ??
>
> Does the FX also omit these?
>
> Although apparently not a functional enhancement, it will still lead to CPU
> specific software in Linux/Windows. Would I lose out on software
> availability if I bought AMD instead of Intel?
>
In short, no. Now think about it. Do you really think anyone would write
software that would require the customer to have a Prescott or newer CPU?
Answer, no.

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

Martin
June 30th 04, 08:36 AM
"Wes Newell" > wrote in message
news:[email protected] .net...
> On Wed, 30 Jun 2004 07:34:33 +0100, Martin wrote:
>
> >
> > "Thomas A. Horsley" > wrote in message
> > ...
> >> >I would have expected some aspects like 3DNow and SSE2 instructions to
be
> >> >missing from the Opteron, but I can't clearly see if it's in or out.
> >>
> >> Nope, everything is on opteron (in fact the amd64 architecture manuals
> >> don't make any distinction between them). The only thing "missing" is
> >> Intel's "prescott new instructions".
> >
> > Prescott new instructions as outlined in
> >
http://www.thejemreport.com/modules.php?op=modload&name=News&file=article&sid=109
> > ??
> >
> > Does the FX also omit these?
> >
> > Although apparently not a functional enhancement, it will still lead to
CPU
> > specific software in Linux/Windows. Would I lose out on software
> > availability if I bought AMD instead of Intel?
> >
> In short, no. Now think about it. Do you really think anyone would write
> software that would require the customer to have a Prescott or newer CPU?
> Answer, no.
>
In 9 months time I think it will matter less who was first, and more who has
which instructions.

Are you saying no one will write code to take advantage of SSE3 in
Prescott??

rstlne
June 30th 04, 01:03 PM
> > In short, no. Now think about it. Do you really think anyone would write
> > software that would require the customer to have a Prescott or newer
CPU?
> > Answer, no.
> >
> In 9 months time I think it will matter less who was first, and more who
has
> which instructions.
>
> Are you saying no one will write code to take advantage of SSE3 in
> Prescott??
>
>

What he is saying is that x86-64 (iAMD64) programs will run on either chip
even if they cant use every feature of said chip.

Tom Horsley
June 30th 04, 05:59 PM
> Does the FX also omit these?
>
> Although apparently not a functional enhancement, it will still lead to CPU
> specific software in Linux/Windows. Would I lose out on software
> availability if I bought AMD instead of Intel?

All the amd64 family chips have identical instruction sets as near as I
can tell, but you can "lose out" in the other direction as well,
the Intel amd64 clone ("nocona" or ia32-e or whatever it is called these days)
doesn't have the AMD 3dNow instructions, so they are both "missing"
pieces :-).

Martin
June 30th 04, 08:03 PM
"Tom Horsley" > wrote in message
...
> > Does the FX also omit these?
> >
> > Although apparently not a functional enhancement, it will still lead to
CPU
> > specific software in Linux/Windows. Would I lose out on software
> > availability if I bought AMD instead of Intel?
>
> All the amd64 family chips have identical instruction sets as near as I
> can tell, but you can "lose out" in the other direction as well,
> the Intel amd64 clone ("nocona" or ia32-e or whatever it is called these
days)
> doesn't have the AMD 3dNow instructions, so they are both "missing"
> pieces :-).
>
>
Fair point, though interestingly, Intel have said it's 64-bit x86 processors
will run 64 bit *OSes* developed for AMD64.
(http://www.linuxdevices.com/news/NS3581750843.html)

Wes Newell
June 30th 04, 08:12 PM
On Wed, 30 Jun 2004 08:36:08 +0100, Martin wrote:

> "Wes Newell" > wrote in message
> news:[email protected] .net...
>> In short, no. Now think about it. Do you really think anyone would write
>> software that would require the customer to have a Prescott or newer CPU?
>> Answer, no.
>>
> In 9 months time I think it will matter less who was first, and more who has
> which instructions.
>
They all have the same basic instruction set.

> Are you saying no one will write code to take advantage of SSE3 in
> Prescott??

No. I'm saying the software will still run on older cpu's. No software
company will write software that requires SSE3 just to run. That's what I
said, and that's what I meant. I didn't say that no one would take
advantage of the SSE3 (or any other extra instructions) in CPU's. And what
does it matter, AMD will have SSE3 instructions too in newer model cpu's
for what it's worth. And it isn't worth much from what I've read about it.

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

Martin
June 30th 04, 09:16 PM
"Wes Newell" > wrote in message
news:pan.2004.06.30[email protected] .net...
> On Wed, 30 Jun 2004 08:36:08 +0100, Martin wrote:
>
> > "Wes Newell" > wrote in message
> > news:[email protected] .net...
> >> In short, no. Now think about it. Do you really think anyone would
write
> >> software that would require the customer to have a Prescott or newer
CPU?
> >> Answer, no.
> >>
> > In 9 months time I think it will matter less who was first, and more who
has
> > which instructions.
> >
> They all have the same basic instruction set.
>
> > Are you saying no one will write code to take advantage of SSE3 in
> > Prescott??
>
> No. I'm saying the software will still run on older cpu's. No software
> company will write software that requires SSE3 just to run. That's what I
> said, and that's what I meant. I didn't say that no one would take
> advantage of the SSE3 (or any other extra instructions) in CPU's. And what
> does it matter, AMD will have SSE3 instructions too in newer model cpu's
> for what it's worth. And it isn't worth much from what I've read about it.
>
> --
Okay

Thomas A. Horsley
June 30th 04, 11:26 PM
>Fair point, though interestingly, Intel have said it's 64-bit x86 processors
>will run 64 bit *OSes* developed for AMD64.

I would imagine so. I have a hard time imagining any operating system
having a use for any of the 3DNow or SSE instructions,
--
>>==>> The *Best* political site <URL:http://www.vote-smart.org/> >>==+
email: icbm: Delray Beach, FL |
<URL:http://home.att.net/~Tom.Horsley> Free Software and Politics <<==+

rstlne
July 1st 04, 10:17 AM
"Thomas A. Horsley" > wrote in message
...
> >Fair point, though interestingly, Intel have said it's 64-bit x86
processors
> >will run 64 bit *OSes* developed for AMD64.
>
> I would imagine so. I have a hard time imagining any operating system
> having a use for any of the 3DNow or SSE instructions,
> --

Their Current chip doesnt run on Linux or Windows xp 64 ..
Redhat explains the linux problem
Microsoft nearly states that they will fix the problem in software for it to
run.

Peter Dickerson
July 2nd 04, 06:49 PM
"rstlne" > wrote in message
news:HEQEc.911$K%[email protected]
>
> "Thomas A. Horsley" > wrote in message
> ...
> > >Fair point, though interestingly, Intel have said it's 64-bit x86
> processors
> > >will run 64 bit *OSes* developed for AMD64.
> >
> > I would imagine so. I have a hard time imagining any operating system
> > having a use for any of the 3DNow or SSE instructions,
> > --
>
> Their Current chip doesnt run on Linux or Windows xp 64 ..
> Redhat explains the linux problem
> Microsoft nearly states that they will fix the problem in software for it
to
> run.

Apparently Nocona doesn't implement a 64-bit IOMMU, which the OSes assume
since AMD 64-bit CPUs have it. The sw fix is probably straightforward but
means that AMD will have a significant performance advantage for systems
with >4G of physical memory because it can avoid copying data.

Peter

Gregory I. Hayes
July 19th 04, 12:17 AM
If the the locked multiplier would be something to look at also. I bought a
FX-51 which ran at 2 GHZ. I have overclocked it to 2.4 GHZ ( without any
special coolers or anything, just changed a couple of bios settings) so
effectively I now have a FX-53. I wouldn't have been able to do this on an
Opteron if the multiplier was locked.

Gregory I. Hayes

"Martin" > wrote in message
...
> So the price difference between
> Opteron146 2GHz 1MB S940 Box (191exVAT)
> and
> Athlon 64 FX-53 1MB S940 (589 ex VAT)
> is just down to clock speed 2GHz vs 2.4 GHz??
>
> Martin
>
> "Wes Newell" > wrote in message
> news:[email protected] .net...
> > On Tue, 29 Jun 2004 08:03:39 +0100, Martin wrote:
> >
> > > Anyone got a nice article that compares the two?
> > >
> > The Opteron and 940 pin FX's are basically the same, except I believe
the
> > multiplier is not locked on the FX's. Both require reg. ram.
> >
> > The 939 FX's don't need registered ram.
> >
> > --
> > Abit KT7-Raid (KT133) Tbred B core CPU @2400MHz (24x100FSB)
> > http://mysite.verizon.net/res0exft/cpu.htm
>
>

JK
August 8th 04, 01:12 AM
Opteron motherboards are expensive, and use 333 registered ECC ddr ram.
Athlon 64(939) motherboards use regular ddr 400 ram.

Martin wrote:

> Anyone got a nice article that compares the two?
>
> I know opteron is for servers, and fx for multimedia/gaming, but I'd like to
> see a comparison of the two, including value for money.
>
> Thanks
> Martin

John R Weiss
August 8th 04, 05:12 AM
Some use ECC, some not... But yes, it's still more expensive than
non-registered RAM.

If you're not going to use more than 1 CPU, I'd go with the cheapest system
overall (MB, CPU, RAM) for the desired clock speed. The cores are
essentially the same.

"JK" > wrote...
> Opteron motherboards are expensive, and use 333 registered ECC ddr ram.
> Athlon 64(939) motherboards use regular ddr 400 ram.
>
>> I know opteron is for servers, and fx for multimedia/gaming, but I'd like
to
>> see a comparison of the two, including value for money.