What Core 2 To Buy Now?

Thanks for the good info.

BTW, this is what prompted my post:

http://www.tomshardware.com/2007/09/...ore/page6.html


thx

bob




"Phil Weldon" wrote in message
...
'Bob Johnson' wrote:
| I guess for me, it will get down to faster CPU vs cache vs $$$. The 2160
| should do 3.0 ghz but I'm not sure if the 6320 will. From what I read,
CPU
| speed will make up for Lack of L2 cache in most cases. Is that true?
_____
No.

CPU clock speed and L2 cache are not comparable. In fact, the faster the
CPU clock speed, the more important a large L2 cache is. The more cores a
CPU has, the more important a large L2 cache is.

If you plan to overclock to ~ 3.0 GHz, then your CPU clock speed will be
higher than any Core 2 Duo stock speed. Intel spends a lot of effort to
achieve a design that balances CPU clock speed and L2 cache size.

I would certainly avoid any Core 2 Duo with a 1 MByte L2 cache.

An E6320 will certainly overclock by 50% or more, but the motherboard will
have to handle a higher FrontSide Bus speed than would be necessary if you
were instead overclocking an E4300 or E4400 to the same CPU clock speed.

It is your choice - you aren't interested in games that require a lot of
processing power. If your audio applications don't have to do processing
in
real time, then you likely don't even need a Core 2 Duo; the only penalty
a
smaller L2 cache will extract in an increased total time required to
process
a file. If your audio applications must do processing in real time, then
a
smaller L2 cache might require you choose lower quality audio. Why don't
you look up the recommended system requirements for your audio processing
programs?

Phil Weldon





"Bob Johnson" wrote in message
...
| Thanks, Phil
|
| I guess for me, it will get down to faster CPU vs cache vs $$$. The 2160
| should do 3.0 ghz but I'm not sure if the 6320 will. From what I read,
CPU
| speed will make up for Lack of L2 cache in most cases. Is that true?
|
|
| thx
|
| bob
|
|
|
| "Phil Weldon" wrote in message
| ...
| 'Bob Johnson' wrote:
| | I'll be using a Gigabyte P35 DS-3? MB, Gskill (the red stuff) ram
and
a
| | 7900GT vid card.
| |
| | I still would like to know if the L2 cache makes *that* much
| difference??
| _____
|
| The short, simple answer to your question "I still would like to know
if
| the
| L2 cache makes *that* much difference??" is that the size of the L2
cache
| can make a HUGE difference. But that difference the depends on the
| applications and the mix of applications. Image processing, audio
| processing, and similar transforms use large chunks of repetitive code
and
| thus benefit greatly from a large L2 cache. Standard office
applications
| benefit less. The smaller the L2 cache, the greater the benefit of
| doubling
| its size. The higher the clock speed of the CPU, the more difference
the
| size of the L2 cache makes (a good reason to get a larger L2 cache if
you
| plan to overclock.)
|
| Sometimes the difference in processor cost is trivial (example - a
Core
2
| Duo E6320 1.86 GHz 4 MB L2 cache costs $5 US more than a Core 2 Duo
E6300
| 1.86 2 MB L2 cache - $176.90 vs. $171.80 US.) I recently bought a
Dell
| Inspiron 1520 and spent an extra $75 US to get a Core 2 Duo 2.2 GHz
with
a
| 4
| MByte L2 cache rather than a Core 2 Duo 2.0 GHz with a 2 MByte L2
cache.
|
| Why have an L2 cache? Well, a CPU needs much more data (including
| instructions) than the memory bus can supply. A Core 2 Duo CPU is
'super
| scalar'; each core can execute more than one instruction per clock
cycle.
| Two cores together can execute as many as six instructions per clock
| cycle.
| Even discounting the effect of memory latencies a 800 MHz FSB can
deliver,
| at most, about 3 bytes per CPU clock cycle, less than one tenth the
amount
| a
| 2.0 GHz Core 2 Duo might require. Without the L1 and L2 caches a CPU
| would
| spend most of its time waiting for data.
|
| The on die L1 cache can supply data as fast as the CPU can use it.
But
| the
| L1 cache is very small - a larger percentage of the data the CPU needs
| will
| not already be loaded into the L1 cache. When the data required is
not
in
| the L2 cache, then the CPU will obtain from the L2 cache which is much
| larger. The L2 cache operates at the CPU clock speed, but there is a
| latency that makes it slower to supply data than the L2 cache. But
this
| delay is MUCH smaller than retrieving data from main memory. Caches
work
| because there is a larger probability that the next instruction and/or
| other
| data needed is in a memory location near the location of the current
| instructions and other data. The L2 caches store data in chunks that
are
| large enough to include long loops of repetitive instructions as well
as
| arrays of data. The larger the L2 cache, the higher the probability
that
| the next data needed will already be in the L2 cache. The probability
| depends on the type of application and how it is programmed. You can
use
| the web to find the cache 'hit' probability (the probability that the
data
| requested is already in the cache) dependency on the cache size. This
| will
| be different for different applications. There are standard chunks of
| code
| that have different mixes of instructions. These standard chunks are
used
| to show the performance of caches. This information is available on
the
| web
| for various processors.
|
| Phil Weldon
|
| "Bob Johnson" wrote in message
| . ..
| | Thanks, Phil
| |
| | I'll be using a Gigabyte P35 DS-3? MB, Gskill (the red stuff) ram
and
a
| | 7900GT vid card.
| |
| | I still would like to know if the L2 cache makes *that* much
| difference??
| |
| | thx
| |
| | bob
| |
| |
| |
| | "Phil" wrote in message
| | ...
| | Bob Johnson wrote:
| | The C2 2160 looks like a dream according to Toms' and other
reviews.
| Are
| | all samples going to OC to 3.0 w/ a good MB and stock cooling? Or
is
| it
| | luck of the draw and getting a good "stepping"? My guru
| | says the lack of L2 cache is a real big deal and that a slower
speed
| | but higher cache would be better in real world day to day use?
| |
| |
| | Is a 4300 or similiar going to be as good for a few bucks more?
| |
| |
| | I'm not into much game playing but will be doing Digital audio
| | recording (Pro Tools, Sonar, etc) Ripping CD's, Burning DVD's if
that
| | matters.
| |
| | I've gotten 2.7G's with a 2160 on a crappy mobo and 2.9G with an
4300
| on
| a
| | decent mobo. There's no tangible difference except that the 2160
| setup
| | was about $250 less than the 4300 setup, separated by 3 months.
| |
| | I would go for the 2160, or 2180 for simplicity and avoiding FSB
| walls.
| |
| |
| | --
| | Phil
| |
| |
| |
|
|
|
|

Bob Johnson wrote:
Thanks for the good info.

BTW, this is what prompted my post:

http://www.tomshardware.com/2007/09/...ore/page6.html

That's good and all that but nowhere that I saw in that article did they
mention the E4xxx series of CPUs. They benched against E6x50 CPUs with their
4MB L2 and higher price but not against the E4xxx that fall in between the
two and have 2MB L2. For instance, my E4500, on a P35 board, runs at 3.2GHz
effortlessly at _below_ stock voltage. I wonder where *it* would cone in
their benchmarks? It's nowhere near the price of the E6x50's. Seems strange
that it's not included in the tests, they mention Pentium 4s and Ds, no
"Allendales" though.

Yeah, just double-checked, couldn't see mention of an Allendale anywhere.
Odd.
--
TTFN,

Shaun.

"Phil Weldon" wrote in message
...
'Bob Johnson' wrote:
I guess for me, it will get down to faster CPU vs cache vs $$$. The
2160 should do 3.0 ghz but I'm not sure if the 6320 will. From what
I read, CPU speed will make up for Lack of L2 cache in most cases. Is
that true?
_____
No.

CPU clock speed and L2 cache are not comparable. In fact, the
faster the CPU clock speed, the more important a large L2 cache is. The
more cores a CPU has, the more important a large L2 cache is.

If you plan to overclock to ~ 3.0 GHz, then your CPU clock speed
will be higher than any Core 2 Duo stock speed. Intel spends a lot
of effort to achieve a design that balances CPU clock speed and L2
cache size. I would certainly avoid any Core 2 Duo with a 1 MByte L2
cache.

An E6320 will certainly overclock by 50% or more, but the
motherboard will have to handle a higher FrontSide Bus speed than
would be necessary if you were instead overclocking an E4300 or
E4400 to the same CPU clock speed. It is your choice - you aren't
interested in games that require a
lot of processing power. If your audio applications don't have to
do processing in
real time, then you likely don't even need a Core 2 Duo; the only
penalty a
smaller L2 cache will extract in an increased total time required to
process
a file. If your audio applications must do processing in real time,
then a
smaller L2 cache might require you choose lower quality audio. Why
don't you look up the recommended system requirements for your audio
processing programs?

Phil Weldon





"Bob Johnson" wrote in message
...
Thanks, Phil

I guess for me, it will get down to faster CPU vs cache vs $$$. The
2160 should do 3.0 ghz but I'm not sure if the 6320 will. From what
I read, CPU speed will make up for Lack of L2 cache in most cases. Is
that true? thx

bob



"Phil Weldon" wrote in message
...
'Bob Johnson' wrote:
I'll be using a Gigabyte P35 DS-3? MB, Gskill (the red stuff) ram
and
a
7900GT vid card.

I still would like to know if the L2 cache makes *that* much
difference??
_____

The short, simple answer to your question "I still would like to
know if the
L2 cache makes *that* much difference??" is that the size of the
L2 cache can make a HUGE difference. But that difference the
depends on the applications and the mix of applications. Image
processing, audio processing, and similar transforms use large
chunks of repetitive code and thus benefit greatly from a large L2
cache. Standard office applications benefit less. The smaller
the L2 cache, the greater the benefit of doubling
its size. The higher the clock speed of the CPU, the more
difference the size of the L2 cache makes (a good reason to get a
larger L2 cache if you plan to overclock.)

Sometimes the difference in processor cost is trivial (example - a
Core
2
Duo E6320 1.86 GHz 4 MB L2 cache costs $5 US more than a Core 2
Duo E6300 1.86 2 MB L2 cache - $176.90 vs. $171.80 US.) I recently
bought a
Dell Inspiron 1520 and spent an extra $75 US to get a Core 2 Duo
2.2 GHz
with
a
4
MByte L2 cache rather than a Core 2 Duo 2.0 GHz with a 2 MByte L2
cache. Why have an L2 cache? Well, a CPU needs much more data
(including
instructions) than the memory bus can supply. A Core 2 Duo CPU is
'super scalar'; each core can execute more than one instruction
per clock cycle. Two cores together can execute as many as six
instructions per clock cycle.
Even discounting the effect of memory latencies a 800 MHz FSB can
deliver, at most, about 3 bytes per CPU clock cycle, less than one
tenth the amount a
2.0 GHz Core 2 Duo might require. Without the L1 and L2 caches a
CPU would
spend most of its time waiting for data.

The on die L1 cache can supply data as fast as the CPU can use it.
But the
L1 cache is very small - a larger percentage of the data the CPU
needs will
not already be loaded into the L1 cache. When the data required is
not
in
the L2 cache, then the CPU will obtain from the L2 cache which is
much larger. The L2 cache operates at the CPU clock speed, but
there is a latency that makes it slower to supply data than the L2
cache. But this delay is MUCH smaller than retrieving data from
main memory. Caches work because there is a larger probability
that the next instruction and/or other
data needed is in a memory location near the location of the
current instructions and other data. The L2 caches store data in
chunks that are large enough to include long loops of repetitive
instructions as well as arrays of data. The larger the L2 cache,
the higher the probability that the next data needed will already
be in the L2 cache. The probability depends on the type of
application and how it is programmed. You can use the web to find
the cache 'hit' probability (the probability that the data
requested is already in the cache) dependency on the cache size. This
will be different for different applications. There are standard
chunks of code
that have different mixes of instructions. These standard chunks
are used to show the performance of caches. This information is
available on the web
for various processors.

Phil Weldon

"Bob Johnson" wrote in message
. ..
Thanks, Phil

I'll be using a Gigabyte P35 DS-3? MB, Gskill (the red stuff) ram
and
a
7900GT vid card.

I still would like to know if the L2 cache makes *that* much
difference?? thx

bob



"Phil" wrote in message
...
Bob Johnson wrote:
The C2 2160 looks like a dream according to Toms' and other
reviews. Are all samples going to OC to 3.0 w/ a good MB and
stock cooling? Or is it luck of the draw and getting a good
"stepping"? My guru says the lack of L2 cache is a real big deal
and that a slower
speed but higher cache would be better in real world day to day
use? Is a 4300 or similiar going to be as good for a few bucks more?


I'm not into much game playing but will be doing Digital audio
recording (Pro Tools, Sonar, etc) Ripping CD's, Burning DVD's
if that matters.

I've gotten 2.7G's with a 2160 on a crappy mobo and 2.9G with an
4300
on
a
decent mobo. There's no tangible difference except that the
2160 setup was about $250 less than the 4300 setup, separated by
3 months. I would go for the 2160, or 2180 for simplicity and
avoiding FSB
walls. --
Phil

Thanks. Shaun

What is the best 4xxx chip to get now? Price/performance/ocing?

Is your 3.2 @ stock cooling? What board?

thx

bob



"~misfit~" wrote in message
...
Bob Johnson wrote:
Thanks for the good info.

BTW, this is what prompted my post:

http://www.tomshardware.com/2007/09/...ore/page6.html

That's good and all that but nowhere that I saw in that article did they
mention the E4xxx series of CPUs. They benched against E6x50 CPUs with
their 4MB L2 and higher price but not against the E4xxx that fall in
between the two and have 2MB L2. For instance, my E4500, on a P35 board,
runs at 3.2GHz effortlessly at _below_ stock voltage. I wonder where *it*
would cone in their benchmarks? It's nowhere near the price of the
E6x50's. Seems strange that it's not included in the tests, they mention
Pentium 4s and Ds, no "Allendales" though.

Yeah, just double-checked, couldn't see mention of an Allendale anywhere.
Odd.
--
TTFN,

Shaun.

"Phil Weldon" wrote in message
...
'Bob Johnson' wrote:
I guess for me, it will get down to faster CPU vs cache vs $$$. The
2160 should do 3.0 ghz but I'm not sure if the 6320 will. From what
I read, CPU speed will make up for Lack of L2 cache in most cases. Is
that true?
_____
No.

CPU clock speed and L2 cache are not comparable. In fact, the
faster the CPU clock speed, the more important a large L2 cache is. The
more cores a CPU has, the more important a large L2 cache is.

If you plan to overclock to ~ 3.0 GHz, then your CPU clock speed
will be higher than any Core 2 Duo stock speed. Intel spends a lot
of effort to achieve a design that balances CPU clock speed and L2
cache size. I would certainly avoid any Core 2 Duo with a 1 MByte L2
cache.

An E6320 will certainly overclock by 50% or more, but the
motherboard will have to handle a higher FrontSide Bus speed than
would be necessary if you were instead overclocking an E4300 or
E4400 to the same CPU clock speed. It is your choice - you aren't
interested in games that require a
lot of processing power. If your audio applications don't have to
do processing in
real time, then you likely don't even need a Core 2 Duo; the only
penalty a
smaller L2 cache will extract in an increased total time required to
process
a file. If your audio applications must do processing in real time,
then a
smaller L2 cache might require you choose lower quality audio. Why
don't you look up the recommended system requirements for your audio
processing programs?

Phil Weldon





"Bob Johnson" wrote in message
...
Thanks, Phil

I guess for me, it will get down to faster CPU vs cache vs $$$. The
2160 should do 3.0 ghz but I'm not sure if the 6320 will. From what
I read, CPU speed will make up for Lack of L2 cache in most cases. Is
that true? thx

bob



"Phil Weldon" wrote in message
...
'Bob Johnson' wrote:
I'll be using a Gigabyte P35 DS-3? MB, Gskill (the red stuff) ram
and
a
7900GT vid card.

I still would like to know if the L2 cache makes *that* much
difference??
_____

The short, simple answer to your question "I still would like to
know if the
L2 cache makes *that* much difference??" is that the size of the
L2 cache can make a HUGE difference. But that difference the
depends on the applications and the mix of applications. Image
processing, audio processing, and similar transforms use large
chunks of repetitive code and thus benefit greatly from a large L2
cache. Standard office applications benefit less. The smaller
the L2 cache, the greater the benefit of doubling
its size. The higher the clock speed of the CPU, the more
difference the size of the L2 cache makes (a good reason to get a
larger L2 cache if you plan to overclock.)

Sometimes the difference in processor cost is trivial (example - a
Core
2
Duo E6320 1.86 GHz 4 MB L2 cache costs $5 US more than a Core 2
Duo E6300 1.86 2 MB L2 cache - $176.90 vs. $171.80 US.) I recently
bought a
Dell Inspiron 1520 and spent an extra $75 US to get a Core 2 Duo
2.2 GHz
with
a
4
MByte L2 cache rather than a Core 2 Duo 2.0 GHz with a 2 MByte L2
cache. Why have an L2 cache? Well, a CPU needs much more data
(including
instructions) than the memory bus can supply. A Core 2 Duo CPU is
'super scalar'; each core can execute more than one instruction
per clock cycle. Two cores together can execute as many as six
instructions per clock cycle.
Even discounting the effect of memory latencies a 800 MHz FSB can
deliver, at most, about 3 bytes per CPU clock cycle, less than one
tenth the amount a
2.0 GHz Core 2 Duo might require. Without the L1 and L2 caches a
CPU would
spend most of its time waiting for data.

The on die L1 cache can supply data as fast as the CPU can use it.
But the
L1 cache is very small - a larger percentage of the data the CPU
needs will
not already be loaded into the L1 cache. When the data required is
not
in
the L2 cache, then the CPU will obtain from the L2 cache which is
much larger. The L2 cache operates at the CPU clock speed, but
there is a latency that makes it slower to supply data than the L2
cache. But this delay is MUCH smaller than retrieving data from
main memory. Caches work because there is a larger probability
that the next instruction and/or other
data needed is in a memory location near the location of the
current instructions and other data. The L2 caches store data in
chunks that are large enough to include long loops of repetitive
instructions as well as arrays of data. The larger the L2 cache,
the higher the probability that the next data needed will already
be in the L2 cache. The probability depends on the type of
application and how it is programmed. You can use the web to find
the cache 'hit' probability (the probability that the data
requested is already in the cache) dependency on the cache size. This
will be different for different applications. There are standard
chunks of code
that have different mixes of instructions. These standard chunks
are used to show the performance of caches. This information is
available on the web
for various processors.

Phil Weldon

"Bob Johnson" wrote in message
. ..
Thanks, Phil

I'll be using a Gigabyte P35 DS-3? MB, Gskill (the red stuff) ram
and
a
7900GT vid card.

I still would like to know if the L2 cache makes *that* much
difference?? thx

bob



"Phil" wrote in message
...
Bob Johnson wrote:
The C2 2160 looks like a dream according to Toms' and other
reviews. Are all samples going to OC to 3.0 w/ a good MB and
stock cooling? Or is it luck of the draw and getting a good
"stepping"? My guru says the lack of L2 cache is a real big deal
and that a slower
speed but higher cache would be better in real world day to day
use? Is a 4300 or similiar going to be as good for a few bucks
more?


I'm not into much game playing but will be doing Digital audio
recording (Pro Tools, Sonar, etc) Ripping CD's, Burning DVD's
if that matters.

I've gotten 2.7G's with a 2160 on a crappy mobo and 2.9G with an
4300
on
a
decent mobo. There's no tangible difference except that the
2160 setup was about $250 less than the 4300 setup, separated by
3 months. I would go for the 2160, or 2180 for simplicity and
avoiding FSB
walls. --
Phil

|
| What is the best 4xxx chip to get now? Price/performance/ocing?
|
| Is your 3.2 @ stock cooling? What board?
_____

There is no ONE answer to "What is the best 4xxx chip to get now?
Price/performance/ocing?" question.

The overclockability varies from one CPU to another EVEN if the model number
and stepping ARE THE SAME. Even if the each CPU is tested in the exact same
system. There is no guarantee that a E4500 you purchase will have the same
overclocking headroom as the E4500 ~misfit~ has.

Ambient room temperature can make or break an overclock. If the ambient
room temperature is 20 C, then you will very likely be able to get a
significantly higher overclock than if the room ambient temperature is 30 C.

I have an E4300 installed in an EVGA 680i chipset motherboard, 2 GBytes
DDR2-1066 SLI memory, and an Antec 550 Neo modular power supply. I get an
easy overclock to 2.7 GHz (50%) with the CPU core voltage set below
specifications. 3.0 GHz is possible with a core voltage higher than
specifications. I have a Thermaltake i7 air cooler (with heatpipes) and a
room ambient temperature of 22 C. I can not overclock my E4300 any higher
(with stable operation) on air cooling because the core temperatures reach
above 70C running Orthos small FFTs to stress the CPU. Further voltage
increases would be of no help because that would increase the temperature
even more.

Again, price/performance ratio is not much help when you are purchasing only
one CPU, and are building a system for specific tasks.

If you stretch out your decision making process long enough, the consider
buying a 45 nm 'Penryn' Core 2 Duo early next year. It should run with a
lower voltage, at a lower temperature, and the 'Penryn' CPU series will have
a higher specified speed.

If you wish to go ahead and build the system this month, then here is what I
would consider:

* You can't make a mistake choosing a CPU more powerful than you need.
Choosing a CPU that is LESS powerful than you need IS a mistake.

* Don't even consider a CPU with a 1 MB L2 cache.

* Intel CPUs with a higher CPU clock multiplier are easier to overclock (it
takes a smaller FSB speed increase to get a CPU clock speed increase.

* Consider how long you will be using this system build; the more capable
CPU you install, the longer the useful life of the system.

* Consider the total cost of your new system build. An extra $40 US or
$100 US may not seem as significant when compared against the total system
cost rather than compared to just the CPU cost.

I would suggest you choose among
1. E4500 2.2 GHz 800 MHz FSB 2 MB L2 cache $132.50 US
2. E6550 2.33 GHz 1333 MHz FSB 4 MB L2 cache $169.00 US
3. E6320 1.86 GHz 1066 MHz FSB 4 MB L2 cache $176.90 US
3. Q6600 2.4 GHz 1066 MHz FSB 4 MB L2 cache $275.80 US

** You with overclocking you should be able to get about the same top speed
for all three of these CPUs.

** The E6320 will perform better because it has twice the L2 cache and
because it will have faster transfers between memory and the CPU (faster
FSB).

** The Q6600 has the same advantages at the E6320 PLUS twice as many
processors. For some applications that means a better than a 50% processing
performance increase.

** The E6550 has a stock speed higher than the E4500 or E6320 AND has a
higher speed FSB so it starts off at stock with higher performance than the
E4500 or E6320. On the other hand, since the FSB starts at 1333 MHz, the
CPU clock multiplier is lower (X 7) and you will need a pretty high FSB
speed to overclock (1830 MHz) and that will require a pretty good
motherboard.

Why don't you post the purchase list for your new system build for comments.
All these components interact, and it may be that the newsgroup can suggest
a sweet spot for your total system.

Phil Weldon


"Bob Johnson" wrote in message
. ..
| Thanks. Shaun
|
| What is the best 4xxx chip to get now? Price/performance/ocing?
|
| Is your 3.2 @ stock cooling? What board?
|
| thx
|
| bob
|
|
|
| "~misfit~" wrote in message
| ...
| Bob Johnson wrote:
| Thanks for the good info.
|
| BTW, this is what prompted my post:
|
| http://www.tomshardware.com/2007/09/...ore/page6.html
|
| That's good and all that but nowhere that I saw in that article did they
| mention the E4xxx series of CPUs. They benched against E6x50 CPUs with
| their 4MB L2 and higher price but not against the E4xxx that fall in
| between the two and have 2MB L2. For instance, my E4500, on a P35 board,
| runs at 3.2GHz effortlessly at _below_ stock voltage. I wonder where
*it*
| would cone in their benchmarks? It's nowhere near the price of the
| E6x50's. Seems strange that it's not included in the tests, they mention
| Pentium 4s and Ds, no "Allendales" though.
|
| Yeah, just double-checked, couldn't see mention of an Allendale
anywhere.
| Odd.
| --
| TTFN,
|
| Shaun.
|
| "Phil Weldon" wrote in message
| ...
| 'Bob Johnson' wrote:
| I guess for me, it will get down to faster CPU vs cache vs $$$. The
| 2160 should do 3.0 ghz but I'm not sure if the 6320 will. From what
| I read, CPU speed will make up for Lack of L2 cache in most cases. Is
| that true?
| _____
| No.
|
| CPU clock speed and L2 cache are not comparable. In fact, the
| faster the CPU clock speed, the more important a large L2 cache is.
The
| more cores a CPU has, the more important a large L2 cache is.
|
| If you plan to overclock to ~ 3.0 GHz, then your CPU clock speed
| will be higher than any Core 2 Duo stock speed. Intel spends a lot
| of effort to achieve a design that balances CPU clock speed and L2
| cache size. I would certainly avoid any Core 2 Duo with a 1 MByte L2
| cache.
|
| An E6320 will certainly overclock by 50% or more, but the
| motherboard will have to handle a higher FrontSide Bus speed than
| would be necessary if you were instead overclocking an E4300 or
| E4400 to the same CPU clock speed. It is your choice - you aren't
| interested in games that require a
| lot of processing power. If your audio applications don't have to
| do processing in
| real time, then you likely don't even need a Core 2 Duo; the only
| penalty a
| smaller L2 cache will extract in an increased total time required to
| process
| a file. If your audio applications must do processing in real time,
| then a
| smaller L2 cache might require you choose lower quality audio. Why
| don't you look up the recommended system requirements for your audio
| processing programs?
|
| Phil Weldon
|
|
|
|
|
| "Bob Johnson" wrote in message
| ...
| Thanks, Phil
|
| I guess for me, it will get down to faster CPU vs cache vs $$$. The
| 2160 should do 3.0 ghz but I'm not sure if the 6320 will. From what
| I read, CPU speed will make up for Lack of L2 cache in most cases. Is
| that true? thx
|
| bob
|
|
|
| "Phil Weldon" wrote in message
| ...
| 'Bob Johnson' wrote:
| I'll be using a Gigabyte P35 DS-3? MB, Gskill (the red stuff) ram
| and
| a
| 7900GT vid card.
|
| I still would like to know if the L2 cache makes *that* much
| difference??
| _____
|
| The short, simple answer to your question "I still would like to
| know if the
| L2 cache makes *that* much difference??" is that the size of the
| L2 cache can make a HUGE difference. But that difference the
| depends on the applications and the mix of applications. Image
| processing, audio processing, and similar transforms use large
| chunks of repetitive code and thus benefit greatly from a large L2
| cache. Standard office applications benefit less. The smaller
| the L2 cache, the greater the benefit of doubling
| its size. The higher the clock speed of the CPU, the more
| difference the size of the L2 cache makes (a good reason to get a
| larger L2 cache if you plan to overclock.)
|
| Sometimes the difference in processor cost is trivial (example - a
| Core
| 2
| Duo E6320 1.86 GHz 4 MB L2 cache costs $5 US more than a Core 2
| Duo E6300 1.86 2 MB L2 cache - $176.90 vs. $171.80 US.) I recently
| bought a
| Dell Inspiron 1520 and spent an extra $75 US to get a Core 2 Duo
| 2.2 GHz
| with
| a
| 4
| MByte L2 cache rather than a Core 2 Duo 2.0 GHz with a 2 MByte L2
| cache. Why have an L2 cache? Well, a CPU needs much more data
| (including
| instructions) than the memory bus can supply. A Core 2 Duo CPU is
| 'super scalar'; each core can execute more than one instruction
| per clock cycle. Two cores together can execute as many as six
| instructions per clock cycle.
| Even discounting the effect of memory latencies a 800 MHz FSB can
| deliver, at most, about 3 bytes per CPU clock cycle, less than one
| tenth the amount a
| 2.0 GHz Core 2 Duo might require. Without the L1 and L2 caches a
| CPU would
| spend most of its time waiting for data.
|
| The on die L1 cache can supply data as fast as the CPU can use it.
| But the
| L1 cache is very small - a larger percentage of the data the CPU
| needs will
| not already be loaded into the L1 cache. When the data required is
| not
| in
| the L2 cache, then the CPU will obtain from the L2 cache which is
| much larger. The L2 cache operates at the CPU clock speed, but
| there is a latency that makes it slower to supply data than the L2
| cache. But this delay is MUCH smaller than retrieving data from
| main memory. Caches work because there is a larger probability
| that the next instruction and/or other
| data needed is in a memory location near the location of the
| current instructions and other data. The L2 caches store data in
| chunks that are large enough to include long loops of repetitive
| instructions as well as arrays of data. The larger the L2 cache,
| the higher the probability that the next data needed will already
| be in the L2 cache. The probability depends on the type of
| application and how it is programmed. You can use the web to find
| the cache 'hit' probability (the probability that the data
| requested is already in the cache) dependency on the cache size.
This
| will be different for different applications. There are standard
| chunks of code
| that have different mixes of instructions. These standard chunks
| are used to show the performance of caches. This information is
| available on the web
| for various processors.
|
| Phil Weldon
|
| "Bob Johnson" wrote in message
| . ..
| Thanks, Phil
|
| I'll be using a Gigabyte P35 DS-3? MB, Gskill (the red stuff) ram
| and
| a
| 7900GT vid card.
|
| I still would like to know if the L2 cache makes *that* much
| difference?? thx
|
| bob
|
|
|
| "Phil" wrote in message
| ...
| Bob Johnson wrote:
| The C2 2160 looks like a dream according to Toms' and other
| reviews. Are all samples going to OC to 3.0 w/ a good MB and
| stock cooling? Or is it luck of the draw and getting a good
| "stepping"? My guru says the lack of L2 cache is a real big
deal
| and that a slower
| speed but higher cache would be better in real world day to day
| use? Is a 4300 or similiar going to be as good for a few bucks
| more?
|
|
| I'm not into much game playing but will be doing Digital audio
| recording (Pro Tools, Sonar, etc) Ripping CD's, Burning DVD's
| if that matters.
|
| I've gotten 2.7G's with a 2160 on a crappy mobo and 2.9G with an
| 4300
| on
| a
| decent mobo. There's no tangible difference except that the
| 2160 setup was about $250 less than the 4300 setup, separated by
| 3 months. I would go for the 2160, or 2180 for simplicity and
| avoiding FSB
| walls. --
| Phil
|
|
|
|
|

Bob Johnson wrote:
Thanks. Shaun

What is the best 4xxx chip to get now? Price/performance/ocing?

Is your 3.2 @ stock cooling? What board?

Hi Bob,

My CPU is an E4500, (2.2GHz, 11 x 400) although there is rumoured to be an
E4600 released soon (some on-line stores even list it). My 3.2 isn't at
stock cooling, although it will run it at stock, full load temps (Orthos and
TAT together) were a little high for my liking, reaching 72°C. I bought and
fitted a Thermaltake Mini Typhoon which dropped my temps around 15°C.
Further fine-tuning, dropping vcore and testing stability (a time-consuming
process) shaved off another 5°C. (So, extrapolating, that would be around
68°C under full load with stock cooler).

My mobo is rather nice, despite a few teething problems. It's an Asus P5K-E
WiFi-AP. P35 northbridge, ICH9R southbridge. Firewire and two eSATA
connectors. It seems OC-friendly wth the FSB range in BIOS going up to 800
(!) I like that you can manually change the bootstrap setting which not only
changes the latency of the Northbridge, allowing you to reach target OCs by
removing FSB "walls" but also allows you a wider range of RAM speed options
as RAM speed is a function of a divider applied to the FSB.

It's a middle-priced "Bearlake" P35 board from Asus and the cheapest one
that uses the same PCB as their more expensive models in the range.
Everything that I've read and experienced (other than one glitch*) tell me
it's a very good board. That said, there are probably similar offerings from
Gigabyte and MSI etc. in the same price range. (It would be pointless
talking price as I'm in New Zealand). This board will support 45nm CPUs and
4 and maybe more cores leaving me a considerable upgrade option when prices
on these CPUs drops in a couple years or less.

*The glitch being that the vcore in BIOS isn't correct. It took me a while
to discover this but CPU-Z and Asus' own Windows hardware monitor finally
convinced me. This could however be unique to my particular board.

Last night I played a little more with OCing this setup and had Orthos
running just fine with the CPU at 3.4GHz. The full-load temps with my Mini
Typhoon were high 60's (compared with 52 @ 3.2). It took 1.40v (CPU-Z
readout, version 1.41 as 1.40 doesn't give vcore correctly) for stability at
3.4GHz. I also posted/booted into windows at 3.5GHz with Orthos failing
after 7 minutes, reporting an error, at the same vcore and temps. Perhaps
more vcore would help....

However, as we're yet to get into high summer here and I can expect ambient
temps to rise, up to 10°C or more yet (no AC and I really have no use for
more computing power), I've decided to continue to run at 3.2 for now, at a
relatively balmy 52°C at maximum load and 1.33v vcore (*Not* 'default' as,
with my board, it was too low even for stability at default speed. My
glitch). Bear in mind that the maximum load temps I'm getting are with
Orthos and TAT running and are highly unlikely to be ever seen under any
other circumstances.

I can't praise the Thermaltake Mini Typhoon's cooling performance enough
although, that said, I did have to modify it (grind off corners, and lap) to
get it to fit my board. However, the figures seem to show that this CPU/mobo
combo will do 3.2GHz with the stock HSF at not unreasonable core temps
long-term unless you're in a very hot environment. (Tom's overclocking
reviews don't mention ambient temps that I've noticed and they're all
conducted on an open testbench. No case to impede cooling and what's the bet
they're in a nice, cool, air conditioned room? Hardly a real-world situation
and should be taken with a pinch of salt.)

I'm on a tight budget, the reason I went with the E4500 over a 2160 or
similar was the 2MB L2 cache which both cores have access to. From what I
could see, the performance increase of the larger, more accessable L2 was
well worth the extra few bucks.

Luck,
--
Shaun.


"~misfit~" wrote in message
...
Bob Johnson wrote:
Thanks for the good info.

BTW, this is what prompted my post:

http://www.tomshardware.com/2007/09/...ore/page6.html

That's good and all that but nowhere that I saw in that article did
they mention the E4xxx series of CPUs. They benched against E6x50
CPUs with their 4MB L2 and higher price but not against the E4xxx
that fall in between the two and have 2MB L2. For instance, my
E4500, on a P35 board, runs at 3.2GHz effortlessly at _below_ stock
voltage. I wonder where *it* would cone in their benchmarks? It's
nowhere near the price of the E6x50's. Seems strange that it's not
included in the tests, they mention Pentium 4s and Ds, no
"Allendales" though. Yeah, just double-checked, couldn't see mention of
an Allendale
anywhere. Odd.
--
TTFN,

Shaun.

"Phil Weldon" wrote in message
...
'Bob Johnson' wrote:
I guess for me, it will get down to faster CPU vs cache vs $$$.
The 2160 should do 3.0 ghz but I'm not sure if the 6320 will.
From what I read, CPU speed will make up for Lack of L2 cache in
most cases. Is that true?
_____
No.

CPU clock speed and L2 cache are not comparable. In fact, the
faster the CPU clock speed, the more important a large L2 cache
is. The more cores a CPU has, the more important a large L2 cache
is. If you plan to overclock to ~ 3.0 GHz, then your CPU clock speed
will be higher than any Core 2 Duo stock speed. Intel spends a lot
of effort to achieve a design that balances CPU clock speed and L2
cache size. I would certainly avoid any Core 2 Duo with a 1 MByte
L2 cache.

An E6320 will certainly overclock by 50% or more, but the
motherboard will have to handle a higher FrontSide Bus speed than
would be necessary if you were instead overclocking an E4300 or
E4400 to the same CPU clock speed. It is your choice - you aren't
interested in games that require a
lot of processing power. If your audio applications don't have to
do processing in
real time, then you likely don't even need a Core 2 Duo; the only
penalty a
smaller L2 cache will extract in an increased total time required
to process
a file. If your audio applications must do processing in real
time, then a
smaller L2 cache might require you choose lower quality audio. Why
don't you look up the recommended system requirements for your
audio processing programs?

Phil Weldon





"Bob Johnson" wrote in message
...
Thanks, Phil

I guess for me, it will get down to faster CPU vs cache vs $$$.
The 2160 should do 3.0 ghz but I'm not sure if the 6320 will.
From what I read, CPU speed will make up for Lack of L2 cache in
most cases. Is that true? thx

bob



"Phil Weldon" wrote in message
...
'Bob Johnson' wrote:
I'll be using a Gigabyte P35 DS-3? MB, Gskill (the red stuff)
ram
and
a
7900GT vid card.

I still would like to know if the L2 cache makes *that* much
difference??
_____

The short, simple answer to your question "I still would like to
know if the
L2 cache makes *that* much difference??" is that the size of the
L2 cache can make a HUGE difference. But that difference the
depends on the applications and the mix of applications. Image
processing, audio processing, and similar transforms use large
chunks of repetitive code and thus benefit greatly from a large
L2 cache. Standard office applications benefit less. The
smaller the L2 cache, the greater the benefit of doubling
its size. The higher the clock speed of the CPU, the more
difference the size of the L2 cache makes (a good reason to get a
larger L2 cache if you plan to overclock.)

Sometimes the difference in processor cost is trivial (example -
a
Core
2
Duo E6320 1.86 GHz 4 MB L2 cache costs $5 US more than a Core 2
Duo E6300 1.86 2 MB L2 cache - $176.90 vs. $171.80 US.) I
recently bought a
Dell Inspiron 1520 and spent an extra $75 US to get a Core 2 Duo
2.2 GHz
with
a
4
MByte L2 cache rather than a Core 2 Duo 2.0 GHz with a 2 MByte L2
cache. Why have an L2 cache? Well, a CPU needs much more data
(including
instructions) than the memory bus can supply. A Core 2 Duo CPU
is 'super scalar'; each core can execute more than one
instruction per clock cycle. Two cores together can execute as many
as six
instructions per clock cycle.
Even discounting the effect of memory latencies a 800 MHz FSB can
deliver, at most, about 3 bytes per CPU clock cycle, less than
one tenth the amount a
2.0 GHz Core 2 Duo might require. Without the L1 and L2 caches a
CPU would
spend most of its time waiting for data.

The on die L1 cache can supply data as fast as the CPU can use
it. But the
L1 cache is very small - a larger percentage of the data the CPU
needs will
not already be loaded into the L1 cache. When the data required
is
not
in
the L2 cache, then the CPU will obtain from the L2 cache which is
much larger. The L2 cache operates at the CPU clock speed, but
there is a latency that makes it slower to supply data than the
L2 cache. But this delay is MUCH smaller than retrieving data
from main memory. Caches work because there is a larger
probability that the next instruction and/or other
data needed is in a memory location near the location of the
current instructions and other data. The L2 caches store data in
chunks that are large enough to include long loops of repetitive
instructions as well as arrays of data. The larger the L2 cache,
the higher the probability that the next data needed will already
be in the L2 cache. The probability depends on the type of
application and how it is programmed. You can use the web to
find the cache 'hit' probability (the probability that the data
requested is already in the cache) dependency on the cache size.
This will be different for different applications. There are
standard chunks of code
that have different mixes of instructions. These standard chunks
are used to show the performance of caches. This information is
available on the web
for various processors.

Phil Weldon

"Bob Johnson" wrote in message
. ..
Thanks, Phil

I'll be using a Gigabyte P35 DS-3? MB, Gskill (the red stuff)
ram
and
a
7900GT vid card.

I still would like to know if the L2 cache makes *that* much
difference?? thx

bob



"Phil" wrote in message
...
Bob Johnson wrote:
The C2 2160 looks like a dream according to Toms' and other
reviews. Are all samples going to OC to 3.0 w/ a good MB and
stock cooling? Or is it luck of the draw and getting a good
"stepping"? My guru says the lack of L2 cache is a real
big deal and that a slower
speed but higher cache would be better in real world day to
day use? Is a 4300 or similiar going to be as good for a few
bucks more?


I'm not into much game playing but will be doing Digital audio
recording (Pro Tools, Sonar, etc) Ripping CD's, Burning DVD's
if that matters.

I've gotten 2.7G's with a 2160 on a crappy mobo and 2.9G with
an
4300
on
a
decent mobo. There's no tangible difference except that the
2160 setup was about $250 less than the 4300 setup, separated
by 3 months. I would go for the 2160, or 2180 for simplicity
and avoiding FSB
walls. --
Phil

Phil Weldon wrote:
What is the best 4xxx chip to get now? Price/performance/ocing?

Is your 3.2 @ stock cooling? What board?
_____

There is no ONE answer to "What is the best 4xxx chip to get now?
Price/performance/ocing?" question.

The overclockability varies from one CPU to another EVEN if the model
number and stepping ARE THE SAME. Even if the each CPU is tested in
the exact same system. There is no guarantee that a E4500 you
purchase will have the same overclocking headroom as the E4500
~misfit~ has.

Good point. I seem to be lucky with silicon. Both my Barton and T'bred-B
were great OCers. And my Celeron Mendicinos, Coppermines and Tualatins.
Actually, it can't all be luck can it? g Actually, I know it can, I've
seen others not so lucky. I've been asked once or twice to reproduce my OCs
for others, on their machines and, despite spending lots of time on some,
haven't often been able to.

Ambient room temperature can make or break an overclock. If the
ambient room temperature is 20 C, then you will very likely be able
to get a significantly higher overclock than if the room ambient
temperature is 30 C.

I'm glad someone else considers this factor. I pretty much never hear
reference to it.

Cheers,
--
Shaun.

Ambient room temperature can make or break an overclock. If the
ambient room temperature is 20 C, then you will very likely be able
to get a significantly higher overclock than if the room ambient
temperature is 30 C.

I'm glad someone else considers this factor. I pretty much never hear
reference to it.

Cheers,
--
Shaun.

It is something that those of us who have been doing this a long time
probably take as a given, but we should mention more often to the newer
folks starting into OCing. It is a very important point to mention.


ed

Ed Medlin wrote:
Ambient room temperature can make or break an overclock. If the
ambient room temperature is 20 C, then you will very likely be able
to get a significantly higher overclock than if the room ambient
temperature is 30 C.

I'm glad someone else considers this factor. I pretty much never hear
reference to it.

It is something that those of us who have been doing this a long time
probably take as a given, but we should mention more often to the
newer folks starting into OCing. It is a very important point to
mention.

Very true. For myself, I have a thermometer on my desk and all temps are
considered as ambient + xx° as well as an empirical figure. That way, I know
what the likely hottest temp I'm going to get in my office is and can
predict core temps in the worst-case scenarios. (No AC in my home, if we get
a heatwave I've been known to down-clock for the duration) Also I factor a
couple extra degrees leeway in for when the dust needs cleaning out of the
HS fins. g

Cheers,
--
Shaun.

Bob Johnson wrote:

BTW, this is what prompted my post:

http://www.tomshardware.com/2007/09/...ore/page6.html

The benchmarks look pretty good to me, despite what people are saying about
the lacking L2. The degree of overclock is definitely not a given. My particular
E4500 will not overclock as Shaun's.

Fishface wrote:
Bob Johnson wrote:

BTW, this is what prompted my post:

http://www.tomshardware.com/2007/09/...ore/page6.html

The benchmarks look pretty good to me, despite what people are saying
about the lacking L2. The degree of overclock is definitely not a given.
My particular E4500 will not overclock as Shaun's.

Condolences. How high will it go? What mobo and cooler? (Also, I 'had' to
buy a more powerful PSU with a higher amperage CPU 12V rail).
--
TTFN,

Shaun.