Comparison of AMD XP 2500+ vs. 754-pin 3500+ vs. 939-pin 3000+?

[Random Person]

Hello all. I was thinking about buying a new AMD 64-bit 939-pin 3000+
"Venice" processor for my computational college work (involves Finite
Element Analysis).

I currently have 2 systems I use for the FE calculations. All the
processors are at stock speeds.

AMD XP "Barton" 2500+ 333MHz FSB
Abit NF7 motherboard
2x512MB Crucial DDR400 (underclocked to 333MHz - a waste!), 2.5-3-3-7

AMD64 754-pin 3500+ (Winchester, I think)
1GB RAM
(sorry about the sparse info here, it is a Linux box and I don't know
how to extract hardware info like via SiSoft Sandra in Windows)


The system I'm considering is a AMD64 939-pin 3000+ "Venice" one, with
1GB RAM. Being a skint student, I've calculated that the 3000+ one
offers the best FLOP per penny.

How would you rank the speeds of the three systems? I know speeds are
application-dependent, but I'm looking for a general idea (or if you
have specific info w.r.t. CPUs for FE analysis...). Would the ranking
be:
1) AMD64 939-pin 3000+ (say ~10% faster than #2?)
2) AMD64 754-pin 3500+ (say ~50% faster than #3?)
3) AMD32 Socket A 2500+

Any chance you can assign percentages to those ranks? (e.g. 754-pin
processor is ~50% faster than Socket A one)

The problem with me getting a 939-pin CPU is that it seems the
motherboards come with a PCI-E graphics card slot, which means I will
have to discard my half-decent Sapphire Radeon 9800 Pro AGP8X card
which I saved a LOT of money for.

Do you think the most cost effective solution would be for me to try to
overclock my Barton to DDR400 FSB? It would then match my RAM's maximum
FSB. How would you rank the processors assuming the overclock is
successful?

Is it possible for me to at least test if a DDR400 CPU FSB overclock
would work before spending money on the better heatsinks? For example,
my Radeon 9800 Pro artifacts insanely the moment I overclock it a
*bit*.

Thanks.

 

[General Schvantzkoph]

Hello all. I was thinking about buying a new AMD 64-bit 939-pin 3000+
"Venice" processor for my computational college work (involves Finite
Element Analysis).

I currently have 2 systems I use for the FE calculations. All the
processors are at stock speeds.

AMD XP "Barton" 2500+ 333MHz FSB
Abit NF7 motherboard
2x512MB Crucial DDR400 (underclocked to 333MHz - a waste!), 2.5-3-3-7

AMD64 754-pin 3500+ (Winchester, I think)
1GB RAM
(sorry about the sparse info here, it is a Linux box and I don't know
how to extract hardware info like via SiSoft Sandra in Windows)

To get the CPU info for a Linux box do

cat /proc/cpuinfo

For everything else do

lspci

--------------------------------------------------------------
Are you sure that your current A64 is a 3500+ and not a 3400+. The 3500+
is a 939 pin part with a 1/2M cache, the 3400+ is a 754 pin part with a 1M
cache. If you already have a 3400+ you already own the best
price/performing member of the A64 family. I have a 3400+, a 3800+ and a
4400+ system. The 3400+ is nearly twice as fast as my 3800+ (939 with 1/2M
cache) on some applications and nearly as fast as my 4400+ (939, dual
core, 1M cache on each processor) on single threaded jobs. On single
threaded jobs the 4400+ is about 10% faster then the 3400+ which gives you
an idea of what the difference is between single and dual memory buses.
The thing that makes the biggest difference is cache size, you want to get
a processor with 1M of cache. If your current system is a 3400+ then
that's what you have, do a cat /proc/cpuinfo to find out.

In any event a 3000+ will be slower then either a 3400+ or a 3500+ system.
If your current system is a 939 pin 3500+ the best upgrade would be to
replace the 3500+ with a 4400+. If your current system is a 3400+ then
don't do anything except add memory.

 

[Random Person]

Hi General. You're right, my processor is a 3400+ one. I'm pretty sure
it is a 754-pin one, I remember being unhappy about buying it (I had
little control as the college does the buying for you).

The processor has a 512kB cache (see below)...

Output of "cat /proc/cpuinfo":
processor : 0
vendor_id : AuthenticAMD
cpu family : 15
model : 12
model name : AMD Athlon(tm) 64 Processor 3400+
stepping : 0
cpu MHz : 2403.093
cache size : 512 KB
fpu : yes
fpu_exception : yes
cpuid level : 1
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge
mca cmov pat pse36 clflush mmx fxsr sse sse2 pni syscall nx mmxext lm
3dnowext 3dnow
bogomips : 4718.59
TLB size : 1088 4K pages
clflush size : 64
cache_alignment : 64
address sizes : 40 bits physical, 48 bits virtual
power management: ts fid vid ttp

---------------------------------------------------

 

[General Schvantzkoph]

Hi General. You're right, my processor is a 3400+ one. I'm pretty sure
it is a 754-pin one, I remember being unhappy about buying it (I had
little control as the college does the buying for you).

The processor has a 512kB cache (see below)...

Output of "cat /proc/cpuinfo":
processor : 0
vendor_id : AuthenticAMD
cpu family : 15
model : 12
model name : AMD Athlon(tm) 64 Processor 3400+
stepping : 0
cpu MHz : 2403.093
cache size : 512 KB
fpu : yes
fpu_exception : yes
cpuid level : 1
wp : yes
flags : fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge
mca cmov pat pse36 clflush mmx fxsr sse sse2 pni syscall nx mmxext lm
3dnowext 3dnow
bogomips : 4718.59
TLB size : 1088 4K pages
clflush size : 64
cache_alignment : 64
address sizes : 40 bits physical, 48 bits virtual
power management: ts fid vid ttp

---------------------------------------------------

You could upgrade it to a 3700+ with a 1M cache (they come in versions
with 1/2M or 1M). For your application I don't know how important cache
size is. My compute intensive application is NCverilog, when it has
recordvars off (not dumping intermediate results) the cache size has a
huge impact, 2 to 1, when I have recordvars on, which increases the size
of the data structures immensely, the cache size doesn't matter nearly as
much because the working set exceeds the 1M cache. You are doing finite
element analysis which is very different. Depending on how they did the
data blocking it's possible that 1/2m cache is enough in which case you
won't get much of a bang by going with a bigger cache. On the other hand
if they blocked for a larger cache then you would see a huge impact. The
only way to find out is to try it for yourself. The other thing is that
finite element processing should lend itself to parallel processing. If
this is code that you wrote, or is opensource that you can modify, then
you should upgrade it to run multithreaded. If it's commercial code it
might already support parallel processing, you'll have to find out. If you
have a way to get to a parallel version of your key application then your
best choice would be a dual core 4400+ system (dual core 1M cache on each
core). I got mine from MonarchComputer for $1800, 4400+, 4G of RAM, 250G
SATA drive, GeForce 6200 graphics, DVD RW.

 

[Captin]

Hello all. I was thinking about buying a new AMD 64-bit
939-pin 3000+
"Venice" processor for my computational college work (involves
Finite
Element Analysis).

I currently have 2 systems I use for the FE calculations. All
the
processors are at stock speeds.

AMD XP "Barton" 2500+ 333MHz FSB
Abit NF7 motherboard
2x512MB Crucial DDR400 (underclocked to 333MHz - a waste!),
2.5-3-3-7

AMD64 754-pin 3500+ (Winchester, I think)
1GB RAM
(sorry about the sparse info here, it is a Linux box and I
don't know
how to extract hardware info like via SiSoft Sandra in
Windows)


The system I'm considering is a AMD64 939-pin 3000+ "Venice"
one, with
1GB RAM. Being a skint student, I've calculated that the 3000+
one
offers the best FLOP per penny.

How would you rank the speeds of the three systems? I know
speeds are
application-dependent, but I'm looking for a general idea (or
if you
have specific info w.r.t. CPUs for FE analysis...). Would the
ranking
be:
1) AMD64 939-pin 3000+ (say ~10% faster than #2?)
2) AMD64 754-pin 3500+ (say ~50% faster than #3?)
3) AMD32 Socket A 2500+

Any chance you can assign percentages to those ranks? (e.g.
754-pin
processor is ~50% faster than Socket A one)

The problem with me getting a 939-pin CPU is that it seems the
motherboards come with a PCI-E graphics card slot, which means
I will
have to discard my half-decent Sapphire Radeon 9800 Pro AGP8X
card
which I saved a LOT of money for.

Do you think the most cost effective solution would be for me
to try to
overclock my Barton to DDR400 FSB? It would then match my
RAM's maximum
FSB. How would you rank the processors assuming the overclock
is
successful?

Is it possible for me to at least test if a DDR400 CPU FSB
overclock
would work before spending money on the better heatsinks? For
example,
my Radeon 9800 Pro artifacts insanely the moment I overclock
it a
*bit*.

Thanks.

For a few different reasons I would go with the 939 socket and the
Venice core 3000 chip.
1: There is hardly any difference in price between building a 754
socket and 939 socket system anyway.
2: The door is wide open to upgrade the 939 socket processor down the
track if you wanted to.
3: The build cost can be kept down because there are numerous boards
available that support AGP 8X and DDR ram for the 939 pin.

I have not mentioned performance of the Venice core because I can
honestly say I don’t think the Venice core shows any gains over the
earlier cores unless you overclock them.It’s mainly compatibility and

when you want to sell it it may be worth a few bucks more than a 754
socket system?

I have 2 AMD 64 systems and the pick for grunt is the 754 socket ,
3400 with a $80 motherboard.
The newer one is a Venice core 3200 chip , Asus A8V Deluxe board and
both have the identical 256 Mb Radeon 9700 Pro video cards and 1G of
corsair value ram.
The 754 socket system still has two WD IDE drives, the Venice core has
two WD SATA drives

 

[Captin]

Hello all. I was thinking about buying a new AMD 64-bit
939-pin 3000+
"Venice" processor for my computational college work (involves
Finite
Element Analysis).

I currently have 2 systems I use for the FE calculations. All
the
processors are at stock speeds.

AMD XP "Barton" 2500+ 333MHz FSB
Abit NF7 motherboard
2x512MB Crucial DDR400 (underclocked to 333MHz - a waste!),
2.5-3-3-7

AMD64 754-pin 3500+ (Winchester, I think)
1GB RAM
(sorry about the sparse info here, it is a Linux box and I
don't know
how to extract hardware info like via SiSoft Sandra in
Windows)


The system I'm considering is a AMD64 939-pin 3000+ "Venice"
one, with
1GB RAM. Being a skint student, I've calculated that the 3000+
one
offers the best FLOP per penny.

How would you rank the speeds of the three systems? I know
speeds are
application-dependent, but I'm looking for a general idea (or
if you
have specific info w.r.t. CPUs for FE analysis...). Would the
ranking
be:
1) AMD64 939-pin 3000+ (say ~10% faster than #2?)
2) AMD64 754-pin 3500+ (say ~50% faster than #3?)
3) AMD32 Socket A 2500+

Any chance you can assign percentages to those ranks? (e.g.
754-pin
processor is ~50% faster than Socket A one)

The problem with me getting a 939-pin CPU is that it seems the
motherboards come with a PCI-E graphics card slot, which means
I will
have to discard my half-decent Sapphire Radeon 9800 Pro AGP8X
card
which I saved a LOT of money for.

Do you think the most cost effective solution would be for me
to try to
overclock my Barton to DDR400 FSB? It would then match my
RAM's maximum
FSB. How would you rank the processors assuming the overclock
is
successful?

Is it possible for me to at least test if a DDR400 CPU FSB
overclock
would work before spending money on the better heatsinks? For
example,
my Radeon 9800 Pro artifacts insanely the moment I overclock
it a
*bit*.

Thanks.

If an AMD 64 has a clock speed of 2.4G and 512Kb of L2 cache it sounds

like a 939 pin 3500?
The 3400 I own has a clock speed of 2.2G and 1Mb of L2 cache.
It is ironic for me that the AMD 64 system that I first built on the
cheap ,
( I’m really an Intel man), with a cheap 754 socket board has ended
up in the office at work working 24/7 with out a hitch , yet I’ve been
around the merry go round with the 939 socket systems.
I would not use an Asus motherboard with the next one......
Comes back to there is nothing wrong so to speak with the 754 socket
other than the 939 is the future and trendy with the extras. The 754
socket is yesterdays hero
Budget build for me would be the Venice core 3000 with $100 gigabyte
board....The option is there to both overclock the chip a little even
with cheap ram and or upgrade the chip later.