Printable View
just to be accurate, the celery uses the same deep pipelined architechture as the P4 so when prefetching fails flushing the whole pipeline and restarting takes a shedload longer on a P4 than a ath 64 or ath xp. That meaning cache is FAR more important to a P4 type cpu than a athlon 64 or xp. THis is why the diff between 256/512 is a lot less than on p4, and with the other gaps.
Oh, you've been so inaccurate to this point. Why would you want to change?
While we're being accurate, the Newcastle is slower and the Clawhammer is faster. How's that for accuracy?:ROTF:
I'll tell you what, it's time for you Newcastle lovers to either admit defeat or show me a 3D benchmark that even comes CLOSE to my benches. Put up or shut up!
I'll issue this challenge to everyone who wants to take it:
I'll beat ANY (even a 3800+) Newcastle in 3DMark2001 with a S754 2800+ ClawHammer at equal or even 100MHz lesser CPU clocks! Aircooled cards only please, as I'm using air on my videocard.
I am 100% sure I can do this. The cache difference IS that great!:toast:
Any takers?:wiggle: Macci? Opp? Pkrew...:D
hmm...
if someone will buy me a 3800+ NC ill race ya :D
yes , i'm ok , XT stock 560/520 and 2400mhz.
XP SP1 DX9 , 5/2/2/2.
i'm waiting for your score ;)
It's 520/560, not 560/520 misteroadster.
What core/mem clocks did you run?
I'm too busy tonight. I won't be benching until tomorrow. We need to use the exact same driver, DX ver and OS. I have to configure a HDD just for the bench.
yes , XT stock ;) cat 4.7 ?
i can't do it today , i'll do it tomorrow.
of course we have to do 3DM2K1 and 2K3 :toast:
You should run the cards stock to take away that variable.
Exactly.
He runs on better than air cooling but if his card clocks are stock it shouldn't be much of an advantage, if any.
misteroadster:
You need to run your HTT at 300 and your multi at 8X.
If I am to run the 2800+ at 2400MHz I have to do this. I only have 8X max! With a 3200 desktop I could run a 10 multi and lower my HTT and memory.
Or how about this, I'll run at 2400MHz and you can run at whatever you want.
Gator::Quote:
Originally posted by LilGator
And when did I say 'huge' ? :confused:
As for 3DM01, S754 1MB is much faster
Give me that 9700Pro and i'll rip your score to shreds , with my 'inferior' S754 setup...
Storm::
The 3700+ tears the 3800+ a new a$$hole when it comes to games and gaming benchmarks
------------------------------------------------------------------
much faster
rip to shreds
tears a new a$$hole
Yep HUGE doesn't quite measure up ;)
time for dinner with the family.
I suggest getting FSB as close as possible and set AGP to 66mhz also.Quote:
Originally posted by misteroadster
yes , i'm ok , XT stock 560/520 and 2400mhz.
XP SP1 DX9 , 5/2/2/2.
i'm waiting for your score ;)
240X10 is a good value :)
Ok, 3200+ vs 3800+ it is!
Oh, ok. I thought you had the 3800+. My mistake!
3200+ vs 3500+ but the next 3300+ will do the same ;)
in fact , i think we have to measure the bandwith impact , and i propose 2 benchs of each 3DM
- 10X200
- 10X240
are you ok ?
I believe the 3200+ stock is close to the 3500+ in 3D benches. We could do benches at equal clocks (240x10) and then wind them up to the max (keeping stock VC clocks of course) just to see which one overclocks better.
What are the idle temps on your cooler? I think your CPU cooler is comparable to mine.
10x200 is the next specification of 3300+
it's a good test.
we don't have the same cooling , it ll be not comparable with overclocking.
My CH5 don' t allow me to desynch , i can' t hit more than 250 2225 or 260 2,5 225
My 3500+ is an early sample 0401 , it don't oc good :/
Yeah we can do 10x200 if you want. Gotta run.
I'll be on later!;)
it will be interresting. :toast:
You guys should both run the same OS untweaked. Run each test individually and in the same order. There's less variablitly that way. When the first compare is done you should set your cpu clocks to try and match score so we have some idea how much difference there is. That way people can take the chip overclocking potential into consideration when buying a cpu
Yeah that could help me! I don't know if I should upgrade to 3700+ or somth or go 939 NC.
if the 3200+ scales you don't need to spend the extra money on the 3700+ . Can't go wrong either way
We really need a dedicated official S754 Clawhammer vs. S939 NewCastle thread :)
We kinda hijacked this one :D
i always knew than S754 1Mo with big FSB was better than FX940 , because ECC is a bottleneck for the FX.
but 754 didn't scales as much than FX , that's the reason why opp didn't bench with them i think.
Now the 939 FX is the absolute king.
Tomorrow we'll know what 939 NC can do , i think at 10X200 i'll win, but with 240X10 i don't give my hand to cut.
Of course 754 1Mo is the best overclock king because at stock frequency (200 fsb) there is not enough bandwith to stress the cpu , and when you give it 260 2225 there is a BIG gain.
With 939 or 940 there 's not a big increase of perf when you increase FSB.
here's 3200+ and FX53 940 OC compares :
Of course when i said FX55 it was FX53 at 13X200 FX55 is 14X200 and FX 57 is 14X200.
you can compare FX55 and 3200@2600 , as you can see the 3200+@260X10 :banana: :banana: :banana: :banana: the FX 13X200 at 2K1
If you compare FX57 and FX57 single (200FSB), there is a gain about 400 pts with DC , it's not very much.
http://pageperso.aol.fr/misteroadste...FX53VS3200.JPG
For those who speak french , the entire test wrote by me :
http://www.think-hardware.com/articl...=43&page=page1
Storm , Do you have doom 3 ? or far cry ? they are interesting cpu benchs too at 1024X768 no AA no ANISO.
i can't do 8X300 because i can't desynch my CH5.Quote:
Originally posted by StormPC
Exactly.
He runs on better than air cooling but if his card clocks are stock it shouldn't be much of an advantage, if any.
misteroadster:
You need to run your HTT at 300 and your multi at 8X.
If I am to run the 2800+ at 2400MHz I have to do this. I only have 8X max! With a 3200 desktop I could run a 10 multi and lower my HTT and memory.
Or how about this, I'll run at 2400MHz and you can run at whatever you want.
if you can't have 10X , i'll bench at 8X200 and 8X240.
after that we ll try to do our best at 2400mhz.
I will use DX9c , Nforce 4.27.
I will not use A64 tweaker.
You can't desynch your CH-5 ??
There is no synch on A64, its based on a divider from the CPU...
?