PDA

View Full Version : P4 Celeron versus "real" P4 in XFX6800 based system


wizlor
11-04-04, 06:22 AM
Hi

New to the forum but clearly lots of great information here. Hoping you may help with the following.

I have a Biostar SFF system in which I am running

- P4 Celeron 2.8
- 1 Gig Ram 3200 on 1 stick
- 40Gig HDD
- CDRW

I purchased an XFX 6800 128Meg Dual DVI card, installed it and applied Forceware 66.81 drivers. 3DMark05 returns a score of 1048 which is pretty lousy. I was hoping to get closer to 3000+. I have all the AA and AF etc all turned off, AGP apreature size at 128

It has been suggested to me that replacing the Celeron P4 by a real P4 (not a Prescott, simple Northwood) might yield a 200% increase in performance. Is this realistic or is my poor score due to something else I have not configured correctly??

Thanks

MUYA
11-04-04, 06:26 AM
No it is very real I am afriad, heck even the top of the line P4 or k8s sometimes hold back the modern 4 quad (16 pixel pipeline) GPU powered cards

john19055
11-04-04, 07:51 PM
Replaceing the Celeron with a P4 will get you better benchmarks and higher game play because of the extra cache on the P4,I think the L2 cache on the celeron is only 128k.

MustangSVT
11-04-04, 07:55 PM
If your Celeron is one of the older ones (i.e. 400MHz FSB, 128KB L2 cache), then yes it is very realistic if you upgrade to something like a P4 Northwood 3.0C (800MHz FSB) that you'll probably have double overall performance (including gaming).

Cali3350
11-04-04, 08:03 PM
Try out your games first. If games play ok, dont do anything. However, theyll play much, much better with a P4.

-=DVS=-
11-04-04, 10:27 PM
All celerons are a joke , they intended for office PCs not games, Celeron vs real Pentium is like night and day !
If ya do some searching for reviews you will see how slow it is.

This is a bit dated (no AMD 64CPUs) fastes they have is 2Ghz P4 Celeron , but it gives you idea , it much slower then 1.5Ghz real Pentium 4, so 2.8Ghz Celeron is arround that level.
http://www.tomshardware.com/cpu/20030217/images/image022.png

55illinois
11-04-04, 10:35 PM
I run a 1.5GHz P3 based Celeron (yes, that says P3 - I know) with 512 MB RAM and a 6800GT. I know the vidcard is overkill but if you got it for the price I did (think $0) then you would be using it too. And I get about 3400 in 3Dmark05, so I'm thinking your score is a bit low. Something else is going on there.

Here's a link (http://service.futuremark.com/compare?3dm05=64715) for proof. Shockingly, mine seems to be the highest 3dmark05 score for a Celeron based system of any speed in FM's database (including P4 based Celerons) :D :D

einstein_314
11-04-04, 10:47 PM
I run a 1.5GHz P3 based Celeron (yes, that says P3 - I know) with 512 MB RAM and a 6800GT. I know the vidcard is overkill but if you got it for the price I did (think $0) then you would be using it too. And I get about 3400 in 3Dmark05, so I'm thinking your score is a bit low. Something else is going on there.

Here's a link (http://service.futuremark.com/compare?3dm05=64715) for proof. Shockingly, mine seems to be the highest 3dmark05 score for a Celeron based system of any speed in FM's database (including P4 based Celerons) :D :D

You have a 6800GT. Wizlor has a 6800NU. That will make somewhat of a difference. Although, he should be getting higher than that.

tristancarton
11-04-04, 11:24 PM
wizlor you should def get more than that. probably close to or over 3k. heck my p3 tualatin 1133 system with a 9800 pro gets 22xx.

reinstall your chipset drivers then install your forceware drivers again.

hey 55illinois nice celeron chip.(tualatin based p3/celeron chips are great chips) How does that 6800gt run for you?(games etc..) I am running a 9800 pro in my dual tualatin p3 1133 system. Never thought I would find a friend with a more outrageous combination. (cpu limited)

hyakki
11-05-04, 12:33 AM
is your system running 4x agp (if its not running 8x agp more it can effect performace)? also i must agree Pentium 4 celerons are very bad
i only had 1 celeron in my life and would never use another one throw it in the trash heap and get a real p4

as for your score it is kinda on the low end ..but if your system is running in 2x or 4x agp mode it would be a impact ...check to make sure your latest chipset drivers are installed
via , sis , Intel ...etc

msxyz
11-05-04, 01:10 AM
Don't confuse P4 Celeron with P3 Celerons... The P4 architecture, due its deep pipelines, need more cache and lower latencies to be kept competitive. Unfortunately, the P4 L1/L2 cache latencies are almost twice those of a P3.

The P2/P3 Celerons, despite the fact that Intel tried to push back their performance with a slow bus, were quite efficient. "Tualatin" Celerons were "real" Pentium 3 with a tweaked bus for increased performance and they had 256KB of L2 cache (I think Intel upped also the associativity of L2 cache). I still have one running in my system at 1.33GHz (133*10). It beats a 1.7GHz P4 Willamette in PIFast and in most games, due to the still superior FPU of the P6 architecture and its much shorter pipeline which is also able to perform more intructions per clock. It also consumes half the power of a P4 (29W at 1.33 GHz)

wizlor
11-05-04, 06:42 AM
Thanks for all the replies, the general consensus on this thread is clearly that the Celeron P4 sucks the big one for gaming. I have seen this view on other boards where I have posted the same enquiry.

Nevertheless, some of you feel that a 1k score on 3DMark05 for my setup is indicative of another problem. I would tend to agree with you. I should be getting somewhere in the 2-2.5k range at least.

The exact CPU I have is Celeron 2.8GHz 256Kb S478 533 FSB. Purchased a few weeks ago.

I will check everything and re-install all the drivers I can and report back.

Keep the advice coming :-)

Wiz

msxyz
11-05-04, 09:31 AM
So wou have a Celeron model "D". It's based on the Prescott core, with 256KB of L2 cache and a bus of 133MHzx4. It's better than the old P4 Celeron, yet slower than a Pentium 4.

With the Prescott core, Intel made the pipeline even deeper and the latency higher. In other word it repeated the same mistakes over again. :P

wizlor
11-05-04, 10:41 AM
Thanks,

I realised while checking my order that I incorrectly listed my RAM here as 3200. It is in fact 2100 with 2.5CL (not sure what that means, I think it relates to the timing numbers...big mystery to me).

Would that be a significant difference with respect to my 3DMark05 score?

Thanks

Wiz

jAkUp
11-05-04, 12:02 PM
It will make a difference, but keep in mind 3DMark05 is very video card intensive. You will see a huge difference in actual games, probably at least double the performance in most situations. Your 3Dmark score should rise about a thousand points or so.

tristancarton
11-05-04, 12:30 PM
Thanks,

I realised while checking my order that I incorrectly listed my RAM here as 3200. It is in fact 2100 with 2.5CL (not sure what that means, I think it relates to the timing numbers...big mystery to me).

Would that be a significant difference with respect to my 3DMark05 score?

Thanks

Wiz
no. it matters a little but but def not 2k (in 3dmark 2005). there is no way your system is running correctly if it reports a 1000 point score for 3dmark 2005 with a 2.8 ghz celeron and a 6800 card. yes the celerons are slow. yes they are crappy for gaming. however you should def. be getting much higher than that. again i state reinstall your chipset drivers than nvidia forceware. i guarantee you that is the problem.

update after you do this. if you don't understand what chipset drivers are post your motherboard and I will find the drivers you need to install.

55illinois
11-05-04, 03:13 PM
Yeah - this thing is a Tualatin Celeron (256KB L2, same as older P3s) rated at 1.3GHz that I'm running at 1.5. It kicks ass, especially considering how old it is. It's nearing retirement age however, Doom3 really choked it pretty badly. But I still get some nice 3dMark scores!

I've got one of those 1.0A Tualatin celerons running at 1.33 GHz in another machine. I think I paid a total of like $85 for the two processors combined. I think I got my money's worth out of them :)

Don't confuse P4 Celeron with P3 Celerons... The P4 architecture, due its deep pipelines, need more cache and lower latencies to be kept competitive. Unfortunately, the P4 L1/L2 cache latencies are almost twice those of a P3.

The P2/P3 Celerons, despite the fact that Intel tried to push back their performance with a slow bus, were quite efficient. "Tualatin" Celerons were "real" Pentium 3 with a tweaked bus for increased performance and they had 256KB of L2 cache (I think Intel upped also the associativity of L2 cache). I still have one running in my system at 1.33GHz (133*10). It beats a 1.7GHz P4 Willamette in PIFast and in most games, due to the still superior FPU of the P6 architecture and its much shorter pipeline which is also able to perform more intructions per clock. It also consumes half the power of a P4 (29W at 1.33 GHz)

-=DVS=-
11-05-04, 07:33 PM
3Dmark 03 and 05 , get minor influence from CPU , moustly Video benchmark thats why scores are decent , id recommend 3Dmark 2001Se , shows overall system speed much better , Offcourse nothing beats game testing ;)

MustangSVT
11-05-04, 07:50 PM
Considering he has a Celeron 2.8 D, that's a completely new story. He should get about 3000 to 3500 at stock in 3DMark2003. The celeron D he has is a lot faster than older celeron's, it's about as good as Barton 2600+ or a Sempron 2800+ as far as I know. Yes, a P4 Northwood 2.8C or 3.0C would be faster, but not 50-100% faster, more like 10 to 15% faster in the best case. Anyway, install motherboard chipset drivers from intel 865pe/875p (whichever you have).

msxyz
11-06-04, 03:31 AM
Considering he has a Celeron 2.8 D, that's a completely new story. He should get about 3000 to 3500 at stock in 3DMark2003. The celeron D he has is a lot faster than older celeron's, it's about as good as Barton 2600+ or a Sempron 2800+ as far as I know. Yes, a P4 Northwood 2.8C or 3.0C would be faster, but not 50-100% faster, more like 10 to 15% faster in the best case. Anyway, install motherboard chipset drivers from intel 865pe/875p (whichever you have).Back IT, 3DMark05 should not be THAT dependant from the CPU so I agree there's something else wrong.

Oh, BTW a difference between a Celeron D and an equally clocked Northwood P4 with 533 MHz bus can be as high as 10% in 3DMark01

wizlor
11-06-04, 01:25 PM
Chaps, what can I say except thanks for all the advice!

I re-installed the intel chipset driver and hey presto << 2850>> scored on 3DMark05!

Now I guess its basically working and I can work on fine tuning it

Cheers for all the help!

Wiz

coldpower27
11-06-04, 03:10 PM
Well the Celeron D are "ok" with clock rates of 2.4GHZ - 2.93GHZ and based on the Prescott core, the performance ranges from around Athlon XP 2000+ to Athlon XP 2600+ levels.

wizlor
11-06-04, 04:43 PM
2992 with performance settings instead of quality :-) Need to break the magic 3000 now!! Tips Tips Tips!! Turned FastWrite off with RivaTuner but no impact from that.

Wiz