Fooled by the Crafty 64-3000+ hype...

  • Thread starter Thread starter b-d0nk!
  • Start date Start date
B

b-d0nk!

Hi all, I recently put together an AMD64 nForce3 rig, very pimpy with a
6800GT and meaty components, as it sits, it's gaming nirvana, 1600x1200 fps
goodness with any eye candy desired in any game played... I spin in my chair
maniacly laughing at the poor people and their less-worthy riggery! However,
my mocking won't last long as I tell you the story of how I was blindly
duped into purchasing that underpowered silicon turd, the AMD64 3000+!

200Mhz, 210Mhz, 240Mhz, 250Mhz, 275Mhz! The seemingly openended bus rates
this chip could do...it was as if you just typed the number you wanted, lo
it shall be! I couldn't spend the money fast enough...those slag off
bastards NewEgg, turned me inside out by getting me my junk toot sweet. And
when it finally came all together....meh..9x. No one told me you couldn't
bump up the bloody multiplier, or that at 9x, the 3000x was crippled. After
throwing that p.o.s. in the bitbucket, I held a FedEx driver hostage until
the 3200+ with it's 10x multiplier sat in it's place and finally began to
demonstrate it's abilities. Like many, I was coming from the ancient epoch
know as "Barton", where these fast 32-bit dinosaurs ruled the landscape. The
3000+ was no more impresive than that chip in anuy circumstances save two:
Installing WindowsXP was very fast and WindowsXP booted much faster than
Win2K, and also the chip was a minimum of 10c cooler in the same case with 2
less (80mm) fans than before! Amazing.

Here's the point: Unless you have Godly Ram of Speediness, any attempt to
violate your ram by bus are going to result in absurdly loose timings and
potentially lessening the hyper transport multiplier, or even the dastardly
CPU multiplier, making the 3000+ perform unacceptably slow. Is true! Though
I do not have actual Game benchmarks, since I said "too damn slow, instead
of, just how damn slow is this?" ... I can report what differences were
noticed by dropping the 3200+ in place of the 3000+, both with the same
overclocked timings, given below (reported by nTune)

CPU = 3200+/3000+ (2.15Ghz 1.935Ghz)
RAM/HT = 215Mhz (x2 439.6Mhz /x5 1.074Ghz)
AGP = 81Mhz
PCI = 33Mhz
RAM = CAS 2.5 tRAS 5 tRCD 3 tRP 2 (2-3-5-2 @ 400Mhz) doesn't go 410+ unless
CAS 2.5
Auto Voltage for All (!)

As reported by nTune, the scores of it's benchmarks are as follows:

System Performance (3200/3000) = 240 / 216
Memory Performance (3200/3000) = Read 320/311 Write 268/261 Latency 152/152
Graphics Performance 400Mhz/1.1Ghz (3200/3000) = Bandwith 355/322

It would be nice if nMonitor supported the nforce3 Ultra I'm using, but I am
lucky nTune lite works. It has one button testing and results with on the
fly changes that bypass BIOS protections, very handy.

I played Far Cry, Rome:Total War and Call of Duty.

Far Cry is played at 1920x1440 with no FSAA and bilinear filtering and x1
Ansiotropic filtering. With the 3000+ it was...meh. The MAX FPS never seemed
to rise above unacceptable, some retardation of the mouse sesitivoty helped,
so there would be less catch up, but it was a disappointment. Granted, the
Barton 3200/400 wasn't playable at that resolution either, but with a 200Mhz
boost and a 1x multiplyer, it's a whole different experience, I enable
trlinear filtering now, and get adequate FPS, but under 60, one area that
expereinced a slowdown was exterior to "Carrier", granted the scenic vista
was extraordinary, but the 3200 made it playable, I suspect the 3000+ would
not have.

Rome: Total War showed no difference in the campaign map (also played at
1920x1440), but the battle map was a seemingly different experience. It is,
currently, as if each additional unit to the map has zero imapact on
performance of the other units, also, when you "hotclick" on a card and you
magically whisk to that unit, the FPS while flying over the battle field is
a sustained 40+. Also, when a volley of archers, say 500+, has fired it's
flaming missiles, you can pause the game and count each one in it's sooty
arc to target...all the while maneuvering the camera, 40+ fps, phenomenal.
The 3000+ did not reach 30+ in these conditions, but it played the battles
as well.

Call of Duty didn't demostrate dramatic differences like the other two
games, but I am able to run 1600x1200 with any level of FSAA and ansio I
choose with 60+ FPS with the 3200+, I had limits with the 3000+, but they
weren't dealbreaking...I think 4xFSAA and 8xAniso had some slowdowns so I
had to drop to 4xAnsio in those levels. Cry me a river...!

Ok, I have a point, the point is, you would think that the 3200+ has just a
200Mhz boost over the 3000+, that the 3000+, at 1.8Ghz is a better chip cuz
it can overclock just well, if not more, and is cheaper....right? right?
Wrong, the 9x multiplier, in my opinion based on the quasi-rigid
measurements by building the box myself, seems to cripple the 3000+ more
than it's product riating would have you think. By rasing the multiplier to
10x, you get a better results from an overlclocked system, more than just
what you would expect. If you are coming from a fast Barton, a stock AMD64
3000+ will not satisfy you, the multiplier is too low...just keep that in
mind when you are shopping. Peace!
 
It might have worked like this:

clock 265
ht 3x = 795 Slow HT won't affect it.
cpu default 9x = 2385 = 3800+ almost
memory 166 = 217 most ddr400 could do that
volts = 1.5

Then too, it might not have. Just one possibility.

My 3200+, nforce 3 ultra:

clock 240
ht 4x = 1000
cpu default 10x = 2400 = 3800+
memory 166 = 200
volts = 1.5

Like you say, a bit easier and cleaner.

It can go faster, if I want to endure a bunch of testing and some cmos
resets and re-setups.

Smiley :-/
MS Smiley :-\

Send spam to the FTC at
(e-mail address removed)
Thanks, robots.
 
PS I didn't totally read your post but I remember the agp was overclocked.
Best to lock the agp and pci busses. If your motherboard can't do that, best
to get another.

--
Ed Light

Smiley :-/
MS Smiley :-\

Send spam to the FTC at
(e-mail address removed)
Thanks, robots.
 
Is true, also, the price difference the 3200 and 3000....checking....about
$36. After spending $600...
 
Hi all, I recently put together an AMD64 nForce3 rig, very pimpy with a
6800GT and meaty components, as it sits, it's gaming nirvana, 1600x1200 fps
goodness with any eye candy desired in any game played... I spin in my chair
maniacly laughing at the poor people and their less-worthy riggery! However,
my mocking won't last long as I tell you the story of how I was blindly
duped into purchasing that underpowered silicon turd, the AMD64 3000+!

200Mhz, 210Mhz, 240Mhz, 250Mhz, 275Mhz! The seemingly openended bus rates
this chip could do...it was as if you just typed the number you wanted, lo
it shall be! I couldn't spend the money fast enough...those slag off
bastards NewEgg, turned me inside out by getting me my junk toot sweet. And
when it finally came all together....meh..9x. No one told me you couldn't
bump up the bloody multiplier, or that at 9x, the 3000x was crippled. After
throwing that p.o.s. in the bitbucket, I held a FedEx driver hostage until
the 3200+ with it's 10x multiplier sat in it's place and finally began to
demonstrate it's abilities. Like many, I was coming from the ancient epoch
know as "Barton", where these fast 32-bit dinosaurs ruled the landscape. The
3000+ was no more impresive than that chip in anuy circumstances save two:
Installing WindowsXP was very fast and WindowsXP booted much faster than
Win2K, and also the chip was a minimum of 10c cooler in the same case with 2
less (80mm) fans than before! Amazing.

Here's the point: Unless you have Godly Ram of Speediness, any attempt to
violate your ram by bus are going to result in absurdly loose timings and
potentially lessening the hyper transport multiplier, or even the dastardly
CPU multiplier, making the 3000+ perform unacceptably slow. Is true! Though
I do not have actual Game benchmarks, since I said "too damn slow, instead
of, just how damn slow is this?" ... I can report what differences were
noticed by dropping the 3200+ in place of the 3000+, both with the same
overclocked timings, given below (reported by nTune)

CPU = 3200+/3000+ (2.15Ghz 1.935Ghz)
RAM/HT = 215Mhz (x2 439.6Mhz /x5 1.074Ghz)
AGP = 81Mhz
PCI = 33Mhz
RAM = CAS 2.5 tRAS 5 tRCD 3 tRP 2 (2-3-5-2 @ 400Mhz) doesn't go 410+ unless
CAS 2.5
Auto Voltage for All (!)

As reported by nTune, the scores of it's benchmarks are as follows:

System Performance (3200/3000) = 240 / 216
Memory Performance (3200/3000) = Read 320/311 Write 268/261 Latency 152/152
Graphics Performance 400Mhz/1.1Ghz (3200/3000) = Bandwith 355/322

It would be nice if nMonitor supported the nforce3 Ultra I'm using, but I am
lucky nTune lite works. It has one button testing and results with on the
fly changes that bypass BIOS protections, very handy.

I played Far Cry, Rome:Total War and Call of Duty.

Far Cry is played at 1920x1440 with no FSAA and bilinear filtering and x1
Ansiotropic filtering. With the 3000+ it was...meh. The MAX FPS never seemed
to rise above unacceptable, some retardation of the mouse sesitivoty helped,
so there would be less catch up, but it was a disappointment. Granted, the
Barton 3200/400 wasn't playable at that resolution either, but with a 200Mhz
boost and a 1x multiplyer, it's a whole different experience, I enable
trlinear filtering now, and get adequate FPS, but under 60, one area that
expereinced a slowdown was exterior to "Carrier", granted the scenic vista
was extraordinary, but the 3200 made it playable, I suspect the 3000+ would
not have.

Rome: Total War showed no difference in the campaign map (also played at
1920x1440), but the battle map was a seemingly different experience. It is,
currently, as if each additional unit to the map has zero imapact on
performance of the other units, also, when you "hotclick" on a card and you
magically whisk to that unit, the FPS while flying over the battle field is
a sustained 40+. Also, when a volley of archers, say 500+, has fired it's
flaming missiles, you can pause the game and count each one in it's sooty
arc to target...all the while maneuvering the camera, 40+ fps, phenomenal.
The 3000+ did not reach 30+ in these conditions, but it played the battles
as well.

Call of Duty didn't demostrate dramatic differences like the other two
games, but I am able to run 1600x1200 with any level of FSAA and ansio I
choose with 60+ FPS with the 3200+, I had limits with the 3000+, but they
weren't dealbreaking...I think 4xFSAA and 8xAniso had some slowdowns so I
had to drop to 4xAnsio in those levels. Cry me a river...!

Ok, I have a point, the point is, you would think that the 3200+ has just a
200Mhz boost over the 3000+, that the 3000+, at 1.8Ghz is a better chip cuz
it can overclock just well, if not more, and is cheaper....right? right?
Wrong, the 9x multiplier, in my opinion based on the quasi-rigid
measurements by building the box myself, seems to cripple the 3000+ more
than it's product riating would have you think. By rasing the multiplier to
10x, you get a better results from an overlclocked system, more than just
what you would expect. If you are coming from a fast Barton, a stock AMD64
3000+ will not satisfy you, the multiplier is too low...just keep that in
mind when you are shopping. Peace!

Nice story, but everyone knows only the FX allows multipliers UP and
DOWN. Athlon 64's only do DOWN, it's partly how Cool n Quite works.
Ed
 
Where are the FSAA settings on COD you speak of?

b-d0nk! said:
Hi all, I recently put together an AMD64 nForce3 rig, very pimpy with a
6800GT and meaty components, as it sits, it's gaming nirvana, 1600x1200
fps goodness with any eye candy desired in any game played... I spin in my
chair maniacly laughing at the poor people and their less-worthy riggery!
However, my mocking won't last long as I tell you the story of how I was
blindly duped into purchasing that underpowered silicon turd, the AMD64
3000+!

200Mhz, 210Mhz, 240Mhz, 250Mhz, 275Mhz! The seemingly openended bus rates
this chip could do...it was as if you just typed the number you wanted, lo
it shall be! I couldn't spend the money fast enough...those slag off
bastards NewEgg, turned me inside out by getting me my junk toot sweet.
And when it finally came all together....meh..9x. No one told me you
couldn't bump up the bloody multiplier, or that at 9x, the 3000x was
crippled. After throwing that p.o.s. in the bitbucket, I held a FedEx
driver hostage until the 3200+ with it's 10x multiplier sat in it's place
and finally began to demonstrate it's abilities. Like many, I was coming
from the ancient epoch know as "Barton", where these fast 32-bit dinosaurs
ruled the landscape. The 3000+ was no more impresive than that chip in
anuy circumstances save two: Installing WindowsXP was very fast and
WindowsXP booted much faster than Win2K, and also the chip was a minimum
of 10c cooler in the same case with 2 less (80mm) fans than before!
Amazing.

Here's the point: Unless you have Godly Ram of Speediness, any attempt to
violate your ram by bus are going to result in absurdly loose timings and
potentially lessening the hyper transport multiplier, or even the
dastardly CPU multiplier, making the 3000+ perform unacceptably slow. Is
true! Though I do not have actual Game benchmarks, since I said "too damn
slow, instead of, just how damn slow is this?" ... I can report what
differences were noticed by dropping the 3200+ in place of the 3000+, both
with the same overclocked timings, given below (reported by nTune)

CPU = 3200+/3000+ (2.15Ghz 1.935Ghz)
RAM/HT = 215Mhz (x2 439.6Mhz /x5 1.074Ghz)
AGP = 81Mhz
PCI = 33Mhz
RAM = CAS 2.5 tRAS 5 tRCD 3 tRP 2 (2-3-5-2 @ 400Mhz) doesn't go 410+
unless CAS 2.5
Auto Voltage for All (!)

As reported by nTune, the scores of it's benchmarks are as follows:

System Performance (3200/3000) = 240 / 216
Memory Performance (3200/3000) = Read 320/311 Write 268/261 Latency
152/152
Graphics Performance 400Mhz/1.1Ghz (3200/3000) = Bandwith 355/322

It would be nice if nMonitor supported the nforce3 Ultra I'm using, but I
am lucky nTune lite works. It has one button testing and results with on
the fly changes that bypass BIOS protections, very handy.

I played Far Cry, Rome:Total War and Call of Duty.

Far Cry is played at 1920x1440 with no FSAA and bilinear filtering and x1
Ansiotropic filtering. With the 3000+ it was...meh. The MAX FPS never
seemed to rise above unacceptable, some retardation of the mouse
sesitivoty helped, so there would be less catch up, but it was a
disappointment. Granted, the Barton 3200/400 wasn't playable at that
resolution either, but with a 200Mhz boost and a 1x multiplyer, it's a
whole different experience, I enable trlinear filtering now, and get
adequate FPS, but under 60, one area that expereinced a slowdown was
exterior to "Carrier", granted the scenic vista was extraordinary, but the
3200 made it playable, I suspect the 3000+ would not have.

Rome: Total War showed no difference in the campaign map (also played at
1920x1440), but the battle map was a seemingly different experience. It
is, currently, as if each additional unit to the map has zero imapact on
performance of the other units, also, when you "hotclick" on a card and
you magically whisk to that unit, the FPS while flying over the battle
field is a sustained 40+. Also, when a volley of archers, say 500+, has
fired it's flaming missiles, you can pause the game and count each one in
it's sooty arc to target...all the while maneuvering the camera, 40+ fps,
phenomenal. The 3000+ did not reach 30+ in these conditions, but it played
the battles as well.

Call of Duty didn't demostrate dramatic differences like the other two
games, but I am able to run 1600x1200 with any level of FSAA and ansio I
choose with 60+ FPS with the 3200+, I had limits with the 3000+, but they
weren't dealbreaking...I think 4xFSAA and 8xAniso had some slowdowns so I
had to drop to 4xAnsio in those levels. Cry me a river...!

Ok, I have a point, the point is, you would think that the 3200+ has just
a 200Mhz boost over the 3000+, that the 3000+, at 1.8Ghz is a better chip
cuz it can overclock just well, if not more, and is cheaper....right?
right? Wrong, the 9x multiplier, in my opinion based on the quasi-rigid
measurements by building the box myself, seems to cripple the 3000+ more
than it's product riating would have you think. By rasing the multiplier
to 10x, you get a better results from an overlclocked system, more than
just what you would expect. If you are coming from a fast Barton, a stock
AMD64 3000+ will not satisfy you, the multiplier is too low...just keep
that in mind when you are shopping. Peace!
 
b-d0nk! said:
Hi all, I recently put together an AMD64 nForce3 rig, very pimpy with a
6800GT and meaty components, as it sits, it's gaming nirvana, 1600x1200 fps
goodness with any eye candy desired in any game played... I spin in my chair
maniacly laughing at the poor people and their less-worthy riggery! However,
my mocking won't last long as I tell you the story of how I was blindly
duped into purchasing that underpowered silicon turd, the AMD64
3000+!

Realistically there is barely any difference between the 3000+ and
3200+ at stock speeds. Even with the overclock numbers you gave there
is not too much of a difference. Enough to make benchmarks look better
but it would not be noticeable in any game. Knowing how to overclock
correctly might have helped ;) But it might not have helped too much.
Some of the last Winchestors were a gamble. Some barely overclocked
while others made it over 2.6Ghz. The new Venice and San Diego so far
are doing fairly well. Quite a few go over 2.8Ghz easily. Yes I've
seen over 2.7Ghz with a 9x multiplier. You just need decent ram and
know how to tweak the BIOS settings.

Eric
 
I have just realized with my 3200+ Winchester that I can put the ht at 3 and
the memory at 133 and even with the memory at 160 pcmark 04 only scores 2%
slower than with it at 200.

So I can inch my way up with the clock to find what the processor can do.

Could have got a 3000+. The memory reaches 200 again when the processor
would be at 2790.


--
Ed Light

Smiley :-/
MS Smiley :-\

Send spam to the FTC at
(e-mail address removed)
Thanks, robots.


So that leaves me room
 
I have just realized with my 3200+ Winchester that I can put the ht at 3 and
the memory at 133 and even with the memory at 160 pcmark 04 only scores 2%
slower than with it at 200.

So I can inch my way up with the clock to find what the processor can do.

Could have got a 3000+. The memory reaches 200 again when the processor
would be at 2790.


I'm glad you found something to play with.

Let us know when you figure it all out.

Dying to know how many more frames per second all the time and effort
brings you.

Ya'd think the engineers at amd and nvidia would have used any of this
if it was worthwhile ;)
 
I got to 255 at 1.55v and ran 15 minutes of prime 95 ok.
I went to 260 at 1.6v and it failed.

Not going further for awhile.

Dropped back down to 1.5v clock=240 ht=4 mem=166, which gives a memory speed
of 200.


--
Ed Light

Smiley :-/
MS Smiley :-\

Send spam to the FTC at
(e-mail address removed)
Thanks, robots.
 
Without looking back through this thread, I could be repeating something,
but ...

After some time with my 3200+, I actually wish I'd bought a 3000+.
Especially with plain pc3200 memory.

For higher end Winchester overclocking, you could set your memory at 133 and
just wind 'er up.

Such as

Clock 300
cpu 9x 2700
memory 193
ht 3x 900

For not so high-end overclocking, you could get stuck with your memory at
175 the step above where it wouldn't exceed 217 (for example) and you had to
turn it down a notch, but I found that that's only minus 2% in pcmark 2004,
with the memory running dual channel.

That's assuming things are happy at high clocks. Some combinations of
settings may not boot.
 
Ed Light said:
Without looking back through this thread, I could be repeating something,
but ...

After some time with my 3200+, I actually wish I'd bought a 3000+.
Especially with plain pc3200 memory.

For higher end Winchester overclocking, you could set your memory at 133
and
just wind 'er up.

Such as

Clock 300
cpu 9x 2700
memory 193
ht 3x 900

For not so high-end overclocking, you could get stuck with your memory at
175 the step above where it wouldn't exceed 217 (for example) and you had
to
turn it down a notch, but I found that that's only minus 2% in pcmark
2004,
with the memory running dual channel.

That's assuming things are happy at high clocks. Some combinations of
settings may not boot.

The bright side - you don't have to do any of that...

My chum just bought a 3500+ system that runs at 2Gig out of the box. Should
be 2.2!
 
Back
Top