PDA

View Full Version : Athlon XP 3200+ Question


Travis King
December 6th 04, 02:35 AM
Is there a 333MHz version of the 3200+? I thought the 3200+ was just a
400MHz FSB processor. I am always seeing adds in the paper saying that it's
a computer that comes with an Athlon XP 3200+ with PC-2700 DDR memory...
Now if it only comes in a 400MHz FSB version, isn't that ripping off the
user because you'd almost have to drop the FSB to 333MHz unless the RAM was
really, really good, right?

Cuzman
December 6th 04, 04:21 AM
"Travis King" > wrote in message
news:[email protected]_s53...

" Is there a 333MHz version of the 3200+? "


Apparently, it was only shipped for HP's OEM use.
http://www.fact-index.com/l/li/list_of_amd_athlon_xp_microprocessors.html

Travis King
December 6th 04, 04:32 AM
OK, and that's the brand that I was seeing that in...
"Cuzman" > wrote in message
...
> "Travis King" > wrote in message
> news:[email protected]_s53...
>
> " Is there a 333MHz version of the 3200+? "
>
>
> Apparently, it was only shipped for HP's OEM use.
> http://www.fact-index.com/l/li/list_of_amd_athlon_xp_microprocessors.html
>
>
>

Travis King
December 6th 04, 05:22 AM
OK... I just went on eMachines's website... They have one with a 3200+.
They say that this computer has a 400MHz FSB @ 2.2GHz , but then they turn
right around and say that the computer has PC-2700 memory... Now that one
they either messed up their typing or they're misleading the person.
"Travis King" > wrote in message
news:[email protected]_s52...
> OK, and that's the brand that I was seeing that in...
> "Cuzman" > wrote in message
> ...
>> "Travis King" > wrote in message
>> news:[email protected]_s53...
>>
>> " Is there a 333MHz version of the 3200+? "
>>
>>
>> Apparently, it was only shipped for HP's OEM use.
>> http://www.fact-index.com/l/li/list_of_amd_athlon_xp_microprocessors.html
>>
>>
>>
>
>

Wes Newell
December 6th 04, 07:01 AM
On Mon, 06 Dec 2004 04:22:36 +0000, Travis King wrote:

> OK... I just went on eMachines's website... They have one with a 3200+.
> They say that this computer has a 400MHz FSB @ 2.2GHz , but then they turn
> right around and say that the computer has PC-2700 memory... Now that one
> they either messed up their typing or they're misleading the person.

And why do you say that? the FSB clock can be at 200MHz (not 400) and the
memory bus can be 166MHz without a problem. It could even be at 133MHz.
with a 200MHz FSB. Not ideal, but still well within specs.

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

Robert Klute
December 6th 04, 07:45 AM
On Mon, 06 Dec 2004 04:22:36 GMT, "Travis King" >
wrote:

>OK... I just went on eMachines's website... They have one with a 3200+.
>They say that this computer has a 400MHz FSB @ 2.2GHz , but then they turn
>right around and say that the computer has PC-2700 memory... Now that one
>they either messed up their typing or they're misleading the person.

Two separate buses. The memory sits off the integrated DDR memory
controller, not the 'front side bus'. On the 'front side' AMD uses the
HyperTransport interconnect.

John Smithe
December 6th 04, 10:30 AM
"Travis King" > wrote in
news:[email protected]_s53:

> Is there a 333MHz version of the 3200+? I thought the 3200+ was just
> a 400MHz FSB processor. I am always seeing adds in the paper saying
> that it's a computer that comes with an Athlon XP 3200+ with PC-2700
> DDR memory... Now if it only comes in a 400MHz FSB version, isn't that
> ripping off the user because you'd almost have to drop the FSB to
> 333MHz unless the RAM was really, really good, right?
>
>

The 3200+ can probably go slower than its maximum spec provides for. So an
400MFz FSB XP3200+ can run its FSB bus at 333MHz, but it will be running
slower than it is capable of.

Travis King
December 6th 04, 09:58 PM
John, that's exactly what I wanted to tell everybody else. Thanks for
understanding it the way I do! I've had to run a processor a time or two
below the FSB because although the memory was advertised to run PC-3000 (and
yes PC-3000 exists) it wouldn't even run the 2800+ @ 166 (333MHz FSB), I had
to drop it down to at least 150 (300MHz FSB) to get the RAM to run stable,
so you also have to check the reliability before you buy your RAM. If you
ask me, I think the manufacturers are trying to find any way possible to
save a little bit of money even if it cuts into performance...
"John Smithe" > wrote in message
...
> "Travis King" > wrote in
> news:[email protected]_s53:
>
>> Is there a 333MHz version of the 3200+? I thought the 3200+ was just
>> a 400MHz FSB processor. I am always seeing adds in the paper saying
>> that it's a computer that comes with an Athlon XP 3200+ with PC-2700
>> DDR memory... Now if it only comes in a 400MHz FSB version, isn't that
>> ripping off the user because you'd almost have to drop the FSB to
>> 333MHz unless the RAM was really, really good, right?
>>
>>
>
> The 3200+ can probably go slower than its maximum spec provides for. So an
> 400MFz FSB XP3200+ can run its FSB bus at 333MHz, but it will be running
> slower than it is capable of.

Travis King
December 6th 04, 10:07 PM
Regardless who's right, they'd have to be cutting into performance at least
a little... (right?) I always thought that the FSB you set the computer to
was half of what was advertised. For example, I thought that when you set
up a 333MHz FSB processor, you always cut it into half which would be 166.
I've always wondered where the other 166 (or 167) goes... Now explain why
with a P4 800MHz FSB is supposed to be set to 200... Is Intel just trying
to make it look like a giant FSB over AMD when it is technically the same,
or does it have to do something with their so-called "hyper-threading
technology"? That's probably a big reason why Intel processors don't
perform much faster than AMD - because the actual FSB's the same. (If I
recall right, the FSB is much more important to performance than the
multiplier.) I have always bought AMD because Intel's prices can sometimes
be double that of AMD's and perform nearly the same.
"John Smithe" > wrote in message
...
> "Travis King" > wrote in
> news:[email protected]_s53:
>
>> Is there a 333MHz version of the 3200+? I thought the 3200+ was just
>> a 400MHz FSB processor. I am always seeing adds in the paper saying
>> that it's a computer that comes with an Athlon XP 3200+ with PC-2700
>> DDR memory... Now if it only comes in a 400MHz FSB version, isn't that
>> ripping off the user because you'd almost have to drop the FSB to
>> 333MHz unless the RAM was really, really good, right?
>>
>>
>
> The 3200+ can probably go slower than its maximum spec provides for. So an
> 400MFz FSB XP3200+ can run its FSB bus at 333MHz, but it will be running
> slower than it is capable of.

John Smithe
December 7th 04, 10:24 AM
"Travis King" > wrote in
news:[email protected]_s52:

> Regardless who's right, they'd have to be cutting into performance at
> least a little... (right?)

Yes.

> I always thought that the FSB you set the
> computer to was half of what was advertised. For example, I thought
> that when you set up a 333MHz FSB processor, you always cut it into
> half which would be 166.

Correct, you set your 'physical' bus speed to half the rated, or effective,
FSB speed. I've tried to explain this more below. Note that for AMD
Athlon64 and FX processors there is no FSB but there is a memory bus which
will be rated the same way that the FSB is rated. This is because for these
processors AMD has integrated the memory controller into the CPU.

> I've always wondered where the other 166 (or
> 167) goes...

It doesn't go anywhere. See below.

> Now explain why with a P4 800MHz FSB is supposed to be
> set to 200... Is Intel just trying to make it look like a giant FSB
> over AMD when it is technically the same, or does it have to do
> something with their so-called "hyper-threading technology"?

AMD does the same thing. It has nothing to do with hyperthreading. See
below.

> That's
> probably a big reason why Intel processors don't perform much faster
> than AMD - because the actual FSB's the same. (If I recall right, the
> FSB is much more important to performance than the multiplier.)

Yea, they're effectively the same (FSB) ratings wise. But, by integrating
the memory controller into the CPU, AMD has significantly reduced memory
latency. Latency, for these purposes, is the amount of time it takes for
memory to return data to the CPU, once the CPU has begun asking for data.
There is a delay, so to speak, and AMD has reduced the delay in their CPU's
dramatically by integrating the memocry controller into the CPU. Intel
CPU's still have the memory controller integrated into the Northbridge
chip.

> I
> have always bought AMD because Intel's prices can sometimes be double
> that of AMD's and perform nearly the same.

Some additional background information: Todays memory is typically 'DDR',
or 'Double Data Rate'. What that means is that the memory performs two data
transfers per memory access cycle, or bus cycle. Additionally, many (but
not all) modern memory-busses/motherboards/CPUs are 'Dual Channel' so that
there are two memory busses running in parallel. Therefore, since memory is
DDR, a 2X multiplication, and 'Dual Channel', another 2X mutiplication,
you effectively get 4X data transfers per bus cycle. So, whatever your
physical bus speed is, is multiplied by 4 to get an effective bus speed.
For example, a FSB speed (in the Intel world) or a memory bus speed (in the
AMD world) of 800MHz is derived from a physical bus speed of 200MHz
multiplied by four to account for DDR and Dual channel. In other words, the
bus is running 200MHz but your getting an 'Effective' bus speed of 800MHz
and so you see a FSB rating of 800MHz. This is not a gimmick. 'Dual
Channel' and 'Double Data Rate' produce real and significant memory
performance improvements. Both AMD and Intel take advantage of 'DDR' and
'Dual Channel' memory technologies. I have concentrated my
investigation of AMD mostly on Athlon64 and derivatives so you must see
what part of this applies to an XP3200+ yourself. Specifically, is it DDR,
and is it Dual Channel? Also check the motherboard spec's. Motherboard and
CPU spec's must be compatible. I guess there is a complication for single
channel CPU's. If your running a 200MHz bus you'll still want PC3200 even
if you don't have a Dual Channel system. If you just have DDR, then
multiply the physical bus frequemcy by 16 to get the PCXXXX rating you will
need to run full speed.

You must ensure that memory stick spec's, motherboard spec's, and cpu
spec's are all compatible. If you want to run PC3200 the CPU and/or memory
controller/northbridge, the memory, and the motherboard must all support
PC3200. If one of these is not, then your system will not run at PC3200
speeds. If they all support PC3200, then something is wrong if you cannot
run at PC3200 speeds.


>Is there a 333MHz version of the 3200+? I thought the 3200+ was just a
>400MHz FSB processor. I am always seeing adds in the paper saying that
>it's
>a computer that comes with an Athlon XP 3200+ with PC-2700 DDR memory...
>Now if it only comes in a 400MHz FSB version, isn't that ripping off the
>user because you'd almost have to drop the FSB to 333MHz unless the RAM
>was really, really good, right?

I suspect what's going on is that the system builder is saving some money
buy using the slower ram. But, at least they are telling you that they are
using the slower memory. I don't know what the XP3200+ FSB is rated at, so
I don't know if this is slower than the XP3200+ is capable of running.
(3200+ does not mean PC3200 memory speed.) Yes, you will have to run the
FSB at 166MHz for this memory to work "unless the RAM was really, really
good". You'd essentailly be overclocking the ram if you got it to work at
PC3200 speeds. Perhaps you could replace the ram with PC3200. Be careful if
you buy one of these systems. You may not be able to adjust the FSB speed
to even try to go PC3200. Many system builders limit what can be adjusted
in the BIOS and do not support overclocking by any means so you'd be stuck
with the speed they set. For example, I ahve an HP system and it cannot be
overclocked because the BIOS and the motherboard cannot be adjusted in any
way that would change to clocks seen by the memory or the cpu.

Wes Newell
December 7th 04, 11:23 AM
On Mon, 06 Dec 2004 21:07:36 +0000, Travis King wrote:

> Regardless who's right, they'd have to be cutting into performance at least
> a little... (right?) I always thought that the FSB you set the computer to
> was half of what was advertised. For example, I thought that when you set
> up a 333MHz FSB processor, you always cut it into half which would be 166.
> I've always wondered where the other 166 (or 167) goes... Now explain why
> with a P4 800MHz FSB is supposed to be set to 200... Is Intel just trying
> to make it look like a giant FSB over AMD when it is technically the same,
> or does it have to do something with their so-called "hyper-threading
> technology"?

Why would you cut the FSB speed in half? Simple answer, you don't. What
you are calling a 333MHz FSB is really only 166Mhz to start with. The
double value (quad in Intels case) is just marketing BS. See The Real
Front Side Bus in sig link.

> That's probably a big reason why Intel processors don't
> perform much faster than AMD - because the actual FSB's the same. (If I
> recall right, the FSB is much more important to performance than the
> multiplier.)

FSB speed doesn't have much impact on performance at all. It's the
bandwidth that may have a performance affect. The reason I say may is
because if your app never needs more than what your system can give,
nothings lost in terms of performance. There is a performance diference,
but since most ram accesses are handled by cache anyway, it's not a big
deal. Even if you double FSB clock speed, actual performance will only
increase abot 10% (assuming you run the cpu's at the same speed. CPU clock
speed is what you need to concern yourself with for the majority of
performance increase. It's used 100% of the time, not just 10% as the FSB
and the higher the core speed, the faster the cache ram speed too.
Oh, btw, the so called intel 800MHZ FSB is twice as fast (data wise) as
the AMD. Think about that when people tell you the FSB speed makes a big
difference.:-)

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

Wes Newell
December 7th 04, 11:59 AM
On Tue, 07 Dec 2004 09:24:06 +0000, Somebody wrote:

> 'Dual Channel' and 'Double Data Rate' produce real and significant
> memory performance improvements.

Real, yes. Significant, no. Unless you consider 1-3% significant. And in
some case single channel actually is faster due to dual channel overhead.
In short, at the bus, the increase is large, in the system at doesn't
amount to much overall performance at all.

> Both AMD and Intel take advantage of
> 'DDR' and 'Dual Channel' memory technologies.

He's got an XP. and the cpu doesn't know dual channel from the hooker in
Vegas. The chipset must rpovide dual channel, and the cpu still gets the
data over the same FSB.

> I have concentrated my investigation of AMD mostly on Athlon64 and
> derivatives so you must see what part of this applies to an XP3200+
> yourself. Specifically, is it DDR, and is it Dual Channel? Also check
> the motherboard spec's. Motherboard and CPU spec's must be compatible. I
> guess there is a complication for single channel CPU's.

Since there is no cpu channel to the ram, how does it get complicated?

> If your running a 200MHz bus you'll still want PC3200 even if you don't
> have a Dual Channel system. If you just have DDR, then multiply the
> physical bus frequemcy by 16 to get the PCXXXX rating you will need to
> run full speed.
>
Hmmm, so I need PC2128 for a 133Mhz FSB. Just being picky here. but as you
probably already think I'm an asshole, why not:-)

> You must ensure that memory stick spec's, motherboard spec's, and cpu
> spec's are all compatible. If you want to run PC3200 the CPU and/or
> memory controller/northbridge, the memory, and the motherboard must all
> support PC3200. If one of these is not, then your system will not run at
> PC3200 speeds. If they all support PC3200, then something is wrong if
> you cannot run at PC3200 speeds.
>
Just going to clarify that you can still use the PC3200 ram (or faster)
even if the board doesn't support it. It will just run at lower speeds.

>>Is there a 333MHz version of the 3200+? I thought the 3200+ was just a
>>400MHz FSB processor. I am always seeing adds in the paper saying that
>>it's
>>a computer that comes with an Athlon XP 3200+ with PC-2700 DDR memory...
>>Now if it only comes in a 400MHz FSB version, isn't that ripping off the
>>user because you'd almost have to drop the FSB to 333MHz unless the RAM
>>was really, really good, right?
>
> I suspect what's going on is that the system builder is saving some
> money buy using the slower ram. But, at least they are telling you that
> they are using the slower memory. I don't know what the XP3200+ FSB is
> rated at, so I don't know if this is slower than the XP3200+ is capable
> of running. (3200+ does not mean PC3200 memory speed.) Yes, you will
> have to run the FSB at 166MHz for this memory to work

No you won't. You can run the FSB at 200MHz and the memory bus at 133MHz
(PC2100) if you want to. They are 2 different buses.

The intention of this rep[ly was not to offend anyone, so don't take
offense. Name changed to hopefully avoid a flame war. And if you just have
to comment about how you think I'm an ass, do it in email and don't waste
the bandwidth here. If I made a technical mistake some wher in this reply,
please feel free to correct it. But I may hunt you down. (That's a joke)

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

John Smithe
December 8th 04, 10:54 AM
Wes Newell > wrote in
news:[email protected] net:

> Real, yes. Significant, no. Unless you consider 1-3% significant. And
> in some case single channel actually is faster due to dual channel
> overhead. In short, at the bus, the increase is large, in the system
> at doesn't amount to much overall performance at all.

I think you are correct for CPU's with integrated memory controllers but
not for CPU's with external memory controllers. I wonder if anyone has
benchmarks confirming your assertion. If you are correct then a dual
channel system gets good numbers on memory bandwidth benchmarks but those
benchmarks don't translate into system performance improvements. Could a
test be run where a dual channel system is tested with both channels active
and tested again with only one channel active? Have you seen a test or
enchmarks that confirms yor assertion? Could you please post it here?

>> If your running a 200MHz bus you'll still want PC3200 even if you
>> don't have a Dual Channel system. If you just have DDR, then multiply
>> the physical bus frequemcy by 16 to get the PCXXXX rating you will
>> need to run full speed.
>>
> Hmmm, so I need PC2128 for a 133Mhz FSB. Just being picky here. but as
> you probably already think I'm an asshole, why not:-)

So help me out here. What PCxxx would you need? Assume DDR SDRAM.
and 133MHz FSB. Wouldn't this be PC2100?


>> You must ensure that memory stick spec's, motherboard spec's, and cpu
>> spec's are all compatible. If you want to run PC3200 the CPU and/or
>> memory controller/northbridge, the memory, and the motherboard must
>> all support PC3200. If one of these is not, then your system will not
>> run at PC3200 speeds. If they all support PC3200, then something is
>> wrong if you cannot run at PC3200 speeds.
>>
> Just going to clarify that you can still use the PC3200 ram (or
> faster) even if the board doesn't support it. It will just run at
> lower speeds.

Absolutely.

>>>Is there a 333MHz version of the 3200+? I thought the 3200+ was just
>>>a 400MHz FSB processor. I am always seeing adds in the paper saying
>>>that it's
>>>a computer that comes with an Athlon XP 3200+ with PC-2700 DDR
>>>memory... Now if it only comes in a 400MHz FSB version, isn't that
>>>ripping off the user because you'd almost have to drop the FSB to
>>>333MHz unless the RAM was really, really good, right?
>>
>> I suspect what's going on is that the system builder is saving some
>> money buy using the slower ram. But, at least they are telling you
>> that they are using the slower memory. I don't know what the XP3200+
>> FSB is rated at, so I don't know if this is slower than the XP3200+
>> is capable of running. (3200+ does not mean PC3200 memory speed.)
>> Yes, you will have to run the FSB at 166MHz for this memory to work
>
> No you won't. You can run the FSB at 200MHz and the memory bus at
> 133MHz (PC2100) if you want to. They are 2 different buses.

Is this true for both Intel and AMD? Is this true for CPU's with an
integrated memory controller? So a BIOS (or other appropriate software)
would have options for settings the FSB speed and additional separate
options for settings memory bus speed. Are the two ever tied together?

John Smithe
December 8th 04, 11:18 AM
John Smithe > wrote in
. 1:

>> Real, yes. Significant, no. Unless you consider 1-3% significant. And
>> in some case single channel actually is faster due to dual channel
>> overhead. In short, at the bus, the increase is large, in the system
>> at doesn't amount to much overall performance at all.
>
> I think you are correct for CPU's with integrated memory controllers
> but not for CPU's with external memory controllers. I wonder if anyone
> has benchmarks confirming your assertion. If you are correct then a
> dual channel system gets good numbers on memory bandwidth benchmarks
> but those benchmarks don't translate into system performance
> improvements. Could a test be run where a dual channel system is
> tested with both channels active and tested again with only one
> channel active? Have you seen a test or enchmarks that confirms yor
> assertion? Could you please post it here?
>

Let me re-iterate that I'm wondering about CPU's with external memory
controllers. Again, I think you are correct for CPU's with integrated
memory controllers.

Travis King
December 8th 04, 10:01 PM
John Smithe wrote:
> John Smithe > wrote in
> . 1:
>
>>> Real, yes. Significant, no. Unless you consider 1-3% significant.
>>> And in some case single channel actually is faster due to dual
>>> channel overhead. In short, at the bus, the increase is large, in
>>> the system at doesn't amount to much overall performance at all.
>>
>> I think you are correct for CPU's with integrated memory controllers
>> but not for CPU's with external memory controllers. I wonder if
>> anyone has benchmarks confirming your assertion. If you are correct
>> then a dual channel system gets good numbers on memory bandwidth
>> benchmarks but those benchmarks don't translate into system
>> performance improvements. Could a test be run where a dual channel
>> system is tested with both channels active and tested again with
>> only one channel active? Have you seen a test or enchmarks that
>> confirms yor assertion? Could you please post it here?
>>
>
> Let me re-iterate that I'm wondering about CPU's with external memory
> controllers. Again, I think you are correct for CPU's with integrated
> memory controllers.
My BIOS only has one adjustment for my Athlon XP 2400+ - the FSB of the
processor with the multiplier for the total clock speed.

Wes Newell
December 8th 04, 10:16 PM
On Wed, 08 Dec 2004 09:54:31 +0000, John Smithe wrote:

> Wes Newell > wrote in
> news:[email protected] net:
>
>> Real, yes. Significant, no. Unless you consider 1-3% significant. And
>> in some case single channel actually is faster due to dual channel
>> overhead. In short, at the bus, the increase is large, in the system
>> at doesn't amount to much overall performance at all.
>
> I think you are correct for CPU's with integrated memory controllers but
> not for CPU's with external memory controllers. I wonder if anyone has
> benchmarks confirming your assertion. If you are correct then a dual
> channel system gets good numbers on memory bandwidth benchmarks but those
> benchmarks don't translate into system performance improvements. Could a
> test be run where a dual channel system is tested with both channels active
> and tested again with only one channel active? Have you seen a test or
> enchmarks that confirms yor assertion? Could you please post it here?
>
Yes there have been many test. Overall system performance varies from 0-3%
increase with dual over single. Here's one if I can find it.

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2149&p=5

> So help me out here. What PCxxx would you need? Assume DDR SDRAM.
> and 133MHz FSB. Wouldn't this be PC2100?
>
That would be the minimum requirement, but if you're buying new, get
something much faster for about the same price PC4400 ram will run at
PC2100 speeds but PC2100 ram won't even come close to PC3200 speeds and
most won't run PC2700 speeds. Juat for peace of mind I always buy ram at
least 1 speed faster than I need. And I've never had a ram problem at
the lessor speed.

>> No you won't. You can run the FSB at 200MHz and the memory bus at
>> 133MHz (PC2100) if you want to. They are 2 different buses.
>
> Is this true for both Intel and AMD? Is this true for CPU's with an
> integrated memory controller? So a BIOS (or other appropriate software)
> would have options for settings the FSB speed and additional separate
> options for settings memory bus speed. Are the two ever tied together?

Hmm... Yes, Yes, Yes. And the last depends on what yuo mean by tied
together. Some older boards won't let you set a memory bus speed lower
than the FSB speed, but may let you set it faster. Others may not let you
set it at all and it will run at the same speed of the FSB. Most newer
boards allow you to set it at least slower and some faster. Overclocking
the FSB outside the 33MHz boudaries almosy always affects the memory bus
too. IOW's if I have (and I have) a FSB of 210MHz and the memory bus set
at 166MHz, the actual memory bus speed 176MHz. And I'm pretty sure this
increase in memory bus speed occurs even if you use the BySpd setting. A
setting that I mostly never use because it can be interpreted wrong by the
bios writer.

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

Wes Newell
December 8th 04, 10:24 PM
On Wed, 08 Dec 2004 10:18:10 +0000, John Smithe wrote:

> Let me re-iterate that I'm wondering about CPU's with external memory
> controllers. Again, I think you are correct for CPU's with integrated
> memory controllers.

Dual channel controlers in the chipsets provide almost no increase in
performance. One person said he got about 1% over single channel when he
tested his NF2 chipset board. Dual channel controllers built in to the cpu
offer greater perfomnce increase, but not by a lot, system wise. Consider
that about 90% of all ram accesses byt the cpu is handled in the cpu cache
and this nay become clearer. Giing dual over single a 5% performance
increase on these would be very generous imo. It's probably closer to 3%.

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

Travis King
December 9th 04, 12:50 AM
"Wes Newell" > wrote in message
news:[email protected] .net...
> On Wed, 08 Dec 2004 09:54:31 +0000, John Smithe wrote:
>
>> Wes Newell > wrote in
>> news:[email protected] net:
>>
>>> Real, yes. Significant, no. Unless you consider 1-3% significant. And
>>> in some case single channel actually is faster due to dual channel
>>> overhead. In short, at the bus, the increase is large, in the system
>>> at doesn't amount to much overall performance at all.
>>
>> I think you are correct for CPU's with integrated memory controllers but
>> not for CPU's with external memory controllers. I wonder if anyone has
>> benchmarks confirming your assertion. If you are correct then a dual
>> channel system gets good numbers on memory bandwidth benchmarks but those
>> benchmarks don't translate into system performance improvements. Could a
>> test be run where a dual channel system is tested with both channels
>> active
>> and tested again with only one channel active? Have you seen a test or
>> enchmarks that confirms yor assertion? Could you please post it here?
>>
> Yes there have been many test. Overall system performance varies from 0-3%
> increase with dual over single. Here's one if I can find it.
>
> http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2149&p=5
>
>> So help me out here. What PCxxx would you need? Assume DDR SDRAM.
>> and 133MHz FSB. Wouldn't this be PC2100?
>>
> That would be the minimum requirement, but if you're buying new, get
> something much faster for about the same price PC4400 ram will run at
> PC2100 speeds but PC2100 ram won't even come close to PC3200 speeds and
> most won't run PC2700 speeds. Juat for peace of mind I always buy ram at
> least 1 speed faster than I need. And I've never had a ram problem at
> the lessor speed.
>
>>> No you won't. You can run the FSB at 200MHz and the memory bus at
>>> 133MHz (PC2100) if you want to. They are 2 different buses.
>>
>> Is this true for both Intel and AMD? Is this true for CPU's with an
>> integrated memory controller? So a BIOS (or other appropriate software)
>> would have options for settings the FSB speed and additional separate
>> options for settings memory bus speed. Are the two ever tied together?
>
> Hmm... Yes, Yes, Yes. And the last depends on what yuo mean by tied
> together. Some older boards won't let you set a memory bus speed lower
> than the FSB speed, but may let you set it faster. Others may not let you
> set it at all and it will run at the same speed of the FSB. Most newer
> boards allow you to set it at least slower and some faster. Overclocking
> the FSB outside the 33MHz boudaries almosy always affects the memory bus
> too. IOW's if I have (and I have) a FSB of 210MHz and the memory bus set
> at 166MHz, the actual memory bus speed 176MHz. And I'm pretty sure this
> increase in memory bus speed occurs even if you use the BySpd setting. A
> setting that I mostly never use because it can be interpreted wrong by the
> bios writer.
>
> --
> Abit KT7-Raid (KT133) Tbred B core CPU @2400MHz (24x100FSB)
> http://mysite.verizon.net/res0exft/cpu.htm
Does it help if I tell you that I have an Asus A7V333 motherboard? (It's
the older revision with 1.xx) The BIOS is updated to 1017.

Wes Newell
December 9th 04, 07:45 AM
On Wed, 08 Dec 2004 23:50:49 +0000, Travis King wrote:

> Does it help if I tell you that I have an Asus A7V333 motherboard? (It's
> the older revision with 1.xx) The BIOS is updated to 1017.

Only that it only has official upport for a 133MHz FSB, although it will
do considerably more than that unless they botched the bios. I doubt that
it's possible to get the FSB to the 200Mhz required by 3200+, but you can
try.. Most boards with this chipset crap out beyond 185MHz FSB.

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

Travis King
December 10th 04, 12:55 AM
I have tried to overclock before. I think when I set it to either 140 or
150 (can't remember for sure) the screen would go black and I'd have to
change it back down to a lower number. I also get restarts anything above
133 even with the VCORE up high. Actually, even if I had the multiplier up
a little, it would get unstable. Maybe limitations of the board. My board
with revision 1.xx doesn't run 333MHz FSB processors - 266MHz is the maximum
unless you have the revision 2, which 333 MHz is the maximum then.
Wes Newell wrote:
> On Wed, 08 Dec 2004 23:50:49 +0000, Travis King wrote:
>
>> Does it help if I tell you that I have an Asus A7V333 motherboard?
>> (It's the older revision with 1.xx) The BIOS is updated to 1017.
>
> Only that it only has official upport for a 133MHz FSB, although it
> will do considerably more than that unless they botched the bios. I
> doubt that it's possible to get the FSB to the 200Mhz required by
> 3200+, but you can try.. Most boards with this chipset crap out
> beyond 185MHz FSB.

John Smithe
December 10th 04, 12:46 PM
Just wanted to say thanks to Wes.

Brian2090
January 11th 05, 03:56 AM
Travis King wrote:
> Regardless who's right, they'd have to be cutting into performance at least
> a little... (right?) I always thought that the FSB you set the computer to
> was half of what was advertised. For example, I thought that when you set
> up a 333MHz FSB processor, you always cut it into half which would be 166.
> I've always wondered where the other 166 (or 167) goes... Now explain why
> with a P4 800MHz FSB is supposed to be set to 200... Is Intel just trying
> to make it look like a giant FSB over AMD when it is technically the same,
> or does it have to do something with their so-called "hyper-threading
> technology"?

Intel's scheme involves RDRAM which is quadrupled from the memory clock
speed to the CPU FrontSideBus speed. i.e., 133MHz to 533MHz FSB or for
the faster Pentium 4's 200MHz to 800MHz FSB. Since RDRAM has a narrower
bus width, the data flow is ultimately similar to DDR SDRAM into an
ATHLON with 400MHz FSB.


That's probably a big reason why Intel processors don't
> perform much faster than AMD - because the actual FSB's the same. (If I
> recall right, the FSB is much more important to performance than the
> multiplier.) I have always bought AMD because Intel's prices can sometimes
> be double that of AMD's and perform nearly the same.

I agree... in the Athlon line up, at equivalent clock speeds the AMD
chips blow away Intel Pentiums. This is also true at the equivalent
price levels for both cpu's. It should be no surprise when a more
expensive chip design running at a higher clock rate can outperform a
bargain option at a lower clock rate. The fact that this is not always
true, leaves me wondering how Intel convinces its customers to stay loyal.

> "John Smithe" > wrote in message
> ...
>

Brian M. (Seattle, WA)