Aquamark 03

[methylenedioxy]

It's "cache" not cash, as in the paper stuff you an exhange for more useful
items.

And Windows does NOT need to know about it to make use of it.
etc,

Incorrect, it checks for the amount of cache.


And before bothering, read this to find out that it's nonsense:
http://support.microsoft.com/support/kb/articles/Q183/0/63.ASP

Ben

Actually, although his spelling is appalling, XP DOESN'T check for the extra
cache, well known fact you need to modify the registry for the Bartons extra
to be acknowledged, I don't care if you throw a knowledgebase article at me,
the IT pros all around the world advise to modify the registry. After all,
MS didn't believe that there was any threat to the Remote Procedure call,
they were warned for many many months and did nothing, so suddenly we all
trust them?, don't think so.....
And your link provides nothing there at all, just talks about 256k cache,
nothing more.
Like the previous poster tried to describe, the XP operating system doesn't
see anything more than 256, just like the link you provided says the same
thing. Its a reg hack to get it going....

 

[Ben Pope]

Actually, although his spelling is appalling, XP DOESN'T check for the
extra cache, well known fact you need to modify the registry for the
Bartons extra to be acknowledged, I don't care if you throw a
knowledgebase article at me, the IT pros all around the world advise to
modify the registry. After all, MS didn't believe that there was any
threat to the Remote Procedure call, they were warned for many many
months and did nothing, so suddenly we all trust them?, don't think
so.....

How is that any kind of comparison? Can we stick to the topic in hand.

And your link provides nothing there at all, just talks about 256k cache,
nothing more.
Like the previous poster tried to describe, the XP operating system
doesn't see anything more than 256, just like the link you provided says
the same thing. Its a reg hack to get it going....

Would you like to explain to me exactly how significant performance gains
can be had if Windows knows how big the cache is? You didn't reply last
time I brought this up - you seem to know all about it, but don't provide an
explanation.

The CPU populates the cache. The CPU has it's won algorithm for working out
which bit of data to replace with the new bit.

The operating system cannot populate the cache. The operating system is
unlikely to know the contents of the cache.

If you care to refute any of those points, please do so, with an
explanation. And also explain HOW an operating system can increase
performance from knowing the size of the cache. I'm not being
argumentative - I am genuinely intrerested and you seem to know all about
it.

Ben

 

[Mangyrat]

so what if my spelling sucks.
atleast i try to help.

 

[GTD]

This is not related to the hardware; it is only useful for computers
with direct-mapped L2 caches. Pentium II and later processors do not
have direct- mapped L2 caches. SecondLevelDataCache can increase
performance by approximately 2 percent in certain cases for older
computers with ample memory (more than 64 MB) by scattering physical
pages better in the address space so there are not so many L2 cache
collisions. Setting SecondLevelDataCache to 256 KB rather than 2 MB
(when the computer has a 2 MB L2 cache) would probably have about a
0.4 percent performance penalty.


This is straight from the MSKB Article.

http://www.sandpile.org/impl/k7.htm
This explains that Bartons do NOT have direct-mapped cache.

 

[methylenedioxy]

I'm only going by a video editing point of view, when enabling the extra
cache there is an extra increase in speed/processing for video apps. As for
games etc, the extra cache makes no difference whatsoever. It's only vid
editing and the only way I know of utilising the extra cache and having
windows enable it is by the reg hack....

 

[patrickp]

so what if my spelling sucks.
atleast i try to help.

I would generally agree with that, Mangyrat; there is far too much useful
contribution being made online by people whose spelling is terrible (even
after allowing for regional variations and typos!) to worry about it. All
being uptight about it does is make you miss out on a lot of useful help and
information.

However, in this case, I think Ben might have been justified - your spelling
caused you to effectively substitute one word for another, which could have
been a bit confusing. You need to watch out for that! ;-)

patrickp

 

[OverKlocker]

Okay at an oc of 200fsb and 200 memory and running at 2.2ghz with 3.7 cats
and running gfx card at 385mhz core and 338 mem I still only get 41212
So I think somethign isn't quite right, this cannot possibly be....Anyone
else noticed such discrepancies? Is this yet another benchmark we can't
trust?
http://arc.aquamark3.com/arc/arc_view.php?runID=481520328

maybe he's msconfiging his system. i do for benchmarks. turning off
anti-virus, temp moniters, etc. can help kick up a score a little

 

[Ben Pope]

so what if my spelling sucks.
atleast i try to help.

Your help is appreciated, but in this case it seems you are peddling
misinformation too.

Ben

 

[Ben Pope]

I'm only going by a video editing point of view, when enabling the extra
cache there is an extra increase in speed/processing for video apps. As
for games etc, the extra cache makes no difference whatsoever. It's only
vid editing and the only way I know of utilising the extra cache and
having windows enable it is by the reg hack....

So you've tested this?

I assume you encoded a particular file 3+ times with and without the setting
and took the average and found that with the option it took less time?
Significantly? Or could it be put down to experimantal error? If we're
talking about less than 1% I suggest that you'd need to be _very_ careful
about what else your machine is doing and do the experiment 8-10 times to
even begin to call the results conclusive.

Ben

 

[fish]

Doesn't CMedia make the SoundStorm for nvidia?

A little tweaking ( raising fsb to 210x11.5 ):

http://arc.aquamark3.com/arc/arc_view.php?runID=514796860

got me 45857...

The P4c really kills with the faster fsb, but I really can't stomach most P4
mobos. It would be okay, if only they had a SoundStorm equivalent. You can
say what you want about SBLive/Audigy(2) - I don't like them, and the built
in sound isn't good enough...

I use my SoundStorm with a Dolby Digital decoder, the positional audio is
great - as is all audio. The Audigy2, and I have one, has greater fidelity,
but I really don't like it...

Did you know that most sound cards that send out Dolby surround over digital
connectors actually reverse the left and right channels? Most people don't
because it doesn't happen with the analog sound outputs, usually, but the
SoundStorm is the only solution that was actually designed to work properly
with DD.

 

[Ben Pope]

A7N8X Deluxe
Barton 2500+ @ 220MHz*10.5
Corsair XMS3200LL @ 220MHz, 5-2-2-2
Crucial 9800 Pro @450/360MHz.

Default driver settings:

CPU: 8635
GPU: 6480
Total: 47035




Ben

 

[John David Carter]

Doesn't CMedia make the SoundStorm for nvidia?

CMedia makes the AC'97 audio codec that is used with the nVidia
APU/SoundStorm ( v 2.1 I think ). In fact, CMedia is trying to get people
to support AC'97 v 2.3 - The codec supporting real-time *software* Dolby
Digital encoding...

The nVidia APU/SoundStorm enables real-time *hardware* Dolby Digital
encoding, so virtually no CPU overhead...

It is an integral part of the nVidia MCP-T, in the nForce2 chipset...

 

[Skid]

So you've tested this?

I assume you encoded a particular file 3+ times with and without the setting
and took the average and found that with the option it took less time?
Significantly? Or could it be put down to experimantal error? If we're
talking about less than 1% I suggest that you'd need to be _very_ careful
about what else your machine is doing and do the experiment 8-10 times to
even begin to call the results conclusive.

Ben

At the risk of spoiling a good argument, I tried it the first time I heard
about it. At least on a P4 with 512k L2, it made absolutely zero difference
in anything -- including ripping and burning DVDs. I put it back to zero.
The way I read Microsoft's article -- they did design the operating system
after all -- the OS gets the cache size from HAL and this registry kludge is
a backup that doesn't do squat for any cpu newer than the original Pentium.

 

[Thomas]

Seems as though FSB speed counts for alot...

The vidcard is very important too With a CPU scoring over 10k i get only
42xxx points This is a Radeon 9700 at 340/310, scoring 5300 points...

Thomas

 

[Lester Piglet]

I've just tried the hack on my pc with 2500 barton, it gave me an extra 100
or so CPU points in Aqua03. Further testing required.

 

[methylenedioxy]

I've just tried the hack on my pc with 2500 barton, it gave me an extra 100
or so CPU points in Aqua03. Further testing required.

It works, why else put extra cache on a cpu if it isn't utilised? Of course
it works, it is designed for video editing though, not burning vids/dvd's.
I'm talking about coding and there is a huge boost in performance.

 

[John David Carter]

The vidcard is very important too With a CPU scoring over 10k i get only
42xxx points This is a Radeon 9700 at 340/310, scoring 5300 points...

Of course, if the GPU is the limiting factor...

I mean, there *is* some difference between a 9700 ( pro? ) at 340/310 and a
9800pro at 425/410...

My non-DDRII memory 9800pro only does 410/360...

Sigh...

Why are we so stuck on numbers? Isn't fast enough fast enough? I just
really wanted to see how fast over fast enough I was, but I do enjoy a good
tweak...

I *know* that even if I get a 9800pro DDRII 256mb card, I can't match the
people with oc'ed P4cs running at 4Ghz...

Still, speed *is* addictive...

And we are *more* than way fast enough!

 

[fish]

I agree. Although after playing games at 1600 X 1200 for so long, I find the
jagged edges at 1024 X 768 sort of unacceptable.

I WANT to run FSAA and/or AF and at that my average frame rate is only about
35 with a low of 10 and a high of 125.

That's hardly fast enough - its actually NOT fast enough but it will do and
I'm running an overclocked 9700 Pro and an Overclocked Pentium 4 at almost
3.3GHz (1Ghz FSB)!

Anything much less is hardly playable without FSAA and AF no less with.

 

[fish]

Hey John,

My office PC has a GF4 Ti4600 (45.23 driver(Win-XP))on an NF2 board,
XP2600+, 2 sticks of Corsair XMS2700. Nothing overclocked but its tweaked
and runs very fast. I'm using a Promise FastTrack TX2000 with mirrored
ATA100 IBM drives.

Aquamark3 scores very poorly!!
Score: 18329
CPU: 5807
GFX: 2175
Average FPS: 18.33
TPS: 6108 K

Games based on DX9 will be totally unplayable at 1024 X 768, medium quality
and with NO FSAA/AF. Maybe at 800 X 600 and still with no FSAA/AF.

Crazy?!

 

[Thomas]

Hey John,

My office PC has a GF4 Ti4600 (45.23 driver(Win-XP))on an NF2 board,
XP2600+, 2 sticks of Corsair XMS2700. Nothing overclocked but its
tweaked and runs very fast. I'm using a Promise FastTrack TX2000 with
mirrored ATA100 IBM drives.

Aquamark3 scores very poorly!!
Score: 18329
CPU: 5807
GFX: 2175
Average FPS: 18.33
TPS: 6108 K

Games based on DX9 will be totally unplayable at 1024 X 768, medium
quality and with NO FSAA/AF. Maybe at 800 X 600 and still with no
FSAA/AF.

Crazy?!

The CPU score is a bit poor for a 2600+, i guess, but the GF4Ti series is
simply not DX90 compliant, and will score less good with a DX90 game...

And what are you doing, running games on office PC's? :-D

Thomas.