Celeron vs. Pentium

[SodaCose]

I know that the basic difference between the two is the cache size, but how much exactly does this affect performance? Clearly at the same clock speeds a Pentium is going to kick the crap out a Celeron, but is there an approximate ghz conversion formula? Like speed times cache size divided by lunar cycle or something? Say between a p4 2.0 and a Celeron D 3.0? I suppose it may depend on what kinda tasks you're trying to accomplish though.

Just wondering.

Oh yeah, there's FSB's to worry about too. Just assume they are the same.

 

[lithium726]

the celeron D would rape the p4 2.0

what are you looking to do with the machine? that will help us decide what is best.

 

[SodaCose]

Actually I was reading around after I wrote this and found someone had said a Celeron 3.0 is closer to a 2.6-2.8 p4. Is that true? Also Celerons run at lower vCore (usually) right? Does that mean there's more potential for OC'ing?

I'm really just thinking hypothetically. I'm looking into upgrading (from my p4 2.0) and I'm trying to decided whether to go for a Celeron D now (skt 478) or save for awhile longer and upgrade my mobo and cpu. I do some causual gaming, and music recording along with all my school related activities.

Thanks for the reply.

 

[lithium726]

that depends on teh p4 - if its a p4-B, id believe it. a C, i wouldnt. the 800mhz FSB really helps the p4.

the old celerons, which had 128kb and the 400 mhz FSB, just sucked major ass - it is not true of the Celeron D's. with the 533 bus and the 256k cache, with the addition of SSE3, the are decent performers, basically a willamette on steriods from the tweaks on the core, new instruction set, and higher FSB. of course, a pentium4 (and sempron, and athlon 64) will still kick the crap out of it, but the celeron is once again a viable contender, like it was back in teh p3 days.

due the archtechture of the p4, the 128k really killed the northwood celerons. the additional cache helps the chip tremendously. if you have the money to go with a p4, i most definatly would - more cache, higher bus speed, and hyperthreading would justify the price to me, not to mention a northwood pentium4 would run cooler than the celeron D, as it is based on the prescott core, and intel encountered a lot of power leakage with teh die shrink to 90, cuasing temps to rise, especially on s478 - on 775 its less of a problem, but still there.

for your uses, i would look for a used p4 2.6/2.8C, and OC it. should hit around 3.2-3.4ghz, and youll be golden for a good while. you will have the benfit of hyperthreading, and the system will be relitivly cool (compared to teh celly D)

 

[chrisf6969]

Here's some benchmarks of a Celeron Northwood with only 128Kb of L2 cache.

http://www6.tomshardware.com/cpu/20021016/celeron-07.html

The new Celeron D has 256Kb of L2 cache, which improves performance A LOT, b/c so many programs were designed with that size cache in mind. (even P3's had 256Kb of cache). Look here:
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2093&p=5
shows a huge performance.

You've got tons of headroom to OC since the bus is only 533 and celerons are low speed grades. So if you OC, you're on a faster bus, so even more of an improvement.

Like the computer I just build is at 160FSB (640 quad pump) and ram is at DDR400, 2.66GHz Celeron is at 3.2Ghz easily on default voltage. (only tested to 3.4 @ default voltage and it ran a little warm with stock HSF)

 

[USMC2Hard4U]

Just for the Record I would Never use a Celeron.

I dont care, I would rather have a lower clocked P4 than a Celeron.

Cache is important. It improves not only the performance of programs, but overall system responsiveness and smoothness. Also the Lack of hyperthreading. HT may not be a big deal to alot of people, but on the Netburst architecture, it dosent hurt to have it.

Celeron is garbage. All they are is Pentium cores that didnt make the grade to become pentiums...so they disabled HT and Lasered the cache off. To hell with that.

 

[Scali]

not to mention a northwood pentium4 would run cooler than the celeron D, as it is based on the prescott core

I'm not sure about that.
Sure, a Northwood P4 may run slightly cooler than a Prescott P4 (most of the extra heat is under control by now... wouldn't be surprised if Prescotts actually ran cooler than the old Northwoods).
But a Celeron is obviously very different from a Pentium. It has far less cache, which means far less transistors, which means far less heat.
Celerons have traditionally run extremely cool, and overclocked like crazy.

 

[delusion_2005]

Just for the Record I would Never use a Celeron.

I dont care, I would rather have a lower clocked P4 than a Celeron.

Cache is important. It improves not only the performance of programs, but overall system responsiveness and smoothness. Also the Lack of hyperthreading. HT may not be a big deal to alot of people, but on the Netburst architecture, it dosent hurt to have it.

Celeron is garbage. All they are is Pentium cores that didnt make the grade to become pentiums...so they disabled HT and Lasered the cache off. To hell with that.

Agreed man, i would use a lower clock p4, amd xp or amd semprom then a celeron, celeron to me are used for basic thing like WP, net surfing and stuff.

 

[SodaCose]

Here's some benchmarks of a Celeron Northwood with only 128Kb of L2 cache.

http://www6.tomshardware.com/cpu/20021016/celeron-07.html

The new Celeron D has 256Kb of L2 cache, which improves performance A LOT, b/c so many programs were designed with that size cache in mind. (even P3's had 256Kb of cache). Look here:
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2093&p=5
shows a huge performance.

those links are exactly what I was looking for, thanks.

 

[chrisf6969]

If you're an avid gamer P4's are the way to go. The extra cache really helps.

Though for a large % of the market a Celeron D is fine. Most people don't game, most people surf, email and do simple office apps (word, excel, etc).

And even for a gamer a cheap celeron OC'd can cut it. Obviously a P4 OC'd would be even better. But for some people paying twice as much isn't an option.
$70-90 for Celeron vs. $140-200+ for P4's
That $80-100 is important and could upgrade their videocard which is usually the biggest bottleneck.

The 515 is an interesting processor. I wish it had HT.
I could be a cool processor to play with. (high mulitplier, 775pin, etc)

 

[pxc]

I upgraded my nephew's Duron 1.5GHz system to a Celeron D 320 @ 3.3GHz 2 months ago. It's only as fast as a P4 2.6-2.7GHz in some things (and as fast as a P4 3.3GHz in others), but for a total of $82 with a new motherboard and new retail cpu + heatsink, no P4 touches it for the price.

 

[splinter3k]

I tried to get off cheap with the Celeron D and be satisfied with the game performance, below is a snip from another post I wrote regarding this....

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

Here's the evolution of a system built on the cheap that now performs properly, save your self time and money and just skip to the more expensive parts...

Original Build Celeron D 2.66/ 2x 512MB DDR 3200/Intel GMA900 w/DVI card (I was low on funds at the time)

-Quake III 1024x768, details medim to high, 71 fps (with sound enabled)

-NOLF 2 didn't run fraps but overall meh at 1024x768 medium detail, some hitching pretty much every few steps. No T&L on graphics so Celeron D is doing everything, yuckage.

I was too depressed to test anything else with this config.

First bump upgrade Celeron D 2.66/ 2x512MB DDR 3200/ Leadtek GeForce 6600

-Quake III 1024x768, details medium to high, 217fps
-Quake III 1280x1024, details max, 199fps

-3D Mark 2001SE default settings 10,197

-WoW, 1024x768, details medium to high, 28-50fps, very playable but not as pretty as it could be

-Max Payne 2, didn't run fraps but overall very good at 1280x1024 high detail, some hitching in busy areas

-NOLF2, didn't run fraps but overall very good at 1280x1024 high detail, some hitching in busy areas

Bump to better graphics Celeron D 2.66/ 2x512MB DDR 3200/ LeadTek 6600GT

-Quake III 1280x1024, details max, 217
-3D Mark 2001SE default settings 12,653

-WoW, 1024x768, details medium to high, 28-50fps, very playable but outside of moving a couple sliders to slightly higher detail still underperforming

-Max Payne 2, didn't run fraps but overall very good at 1280x1024 high detail, less hitching in busy areas

-NOLF2, didn't run fraps but overall very good at 1280x1024 high detail, very smooth, obviously enough power to play this title without AA/AF, occaisional choppy sound

Final gamesome build PIV 530j/ 2x 512MB DDR 3200 /Leadtek GeForce 6600GT

-All games tested at 1280x1024, max details, sound enabled, 4x AA and 8x AF via in game control panel unless not available, then done via ForceWare

-Quake III 274fps

-3D Mark 2001SE 16,908 (no AA or AF enabled)

-Max Payne 2 buttery smooth and gorgeous

-NOLF 2 drool worhty and choppy sound now gone

-WoW, super smooth at 1280x1024, AA only at 2x, AF at 50% via in game settings, looks far better and hangs rock steady in the mid-40s to 60

EDIT add 3D Mark 2003/2005 scores, definitely worth the price of the upgrades

-3D Mark 03 8396

-3D Mark 05 3560

 

[chrisf6969]

Thats why I said, the extra money is best spent on a videocard. (the biggest bottleneck)

#1 - don't even bother benchmarking with integrated graphics.

#2,3- good upgrade, but if you knew you were an avid gamer shoulda saved up in the 1st place for a 6600GT or 6800nu, or 6800GT

#4- another nice bump, but you're comparing a 3Ghz / 800Mhz FSB processor to a 2.66Ghz/ 533FSB processor. Like I said if you're a gamer you can get by with an OC'd Celeron. You could have easily OC'd your celeron to probably 3.2Ghz+ and gotten similar results.

P.S. AA/AF is more of a function of the videocard power than processor power. Granted you need a decent # of fps before you turn it on, but for most games I play, I can enable 2x or 4x AA almost "for free" very little hit to performance.

 

[splinter3k]

#1, 2,3

Umm yea, I know that, of course when I built the box I was tight on funds.

#4- another nice bump, but you're comparing a 3Ghz / 800Mhz FSB processor to a 2.66Ghz/ 533FSB processor. Like I said if you're a gamer you can get by with an OC'd Celeron. You could have easily OC'd your celeron to probably 3.2Ghz+ and gotten similar results.

I think you are overstating the OCs effect, look at the Celeron D benchmarks out there, the stuff at Anand and Xbit shows a incremental performance bump for each clock speed bump. Some games show going from a Celeron D 2.66 > 2.88 > 3.06 only yeilding 1-2 fps per each bump, versus taking the same bumps with a PIV and getting at least 7-10fps for each clock speed bump. Going from a Celeron D 2.66 > 3.2 will likely get you 3-8fps, going from a PIV 2.66 > 3.2 will likely get you in excess of 30fps. A Celeron D at 3.2 is still going to be a bottleneck for the 6600GT or better.

P.S. AA/AF is more of a function of the videocard power than processor power. Granted you need a decent # of fps before you turn it on, but for most games I play, I can enable 2x or 4x AA almost "for free" very little hit to performance.

True enough, but I am not willing to sacrifice detail and resolution to get AA, I want it all.

 

[Pongi]

What kind of motherboard do you have SodaCose? Do you plan on upgrading just the CPU or the motherboard/CPU/ Possibly RAM?

 

[chrisf6969]

I think you are overstating the OCs effect, look at the Celeron D benchmarks out there, the stuff at Anand and Xbit shows a incremental performance bump for each clock speed bump. Some games show going from a Celeron D 2.66 > 2.88 > 3.06 only yeilding 1-2 fps per each bump, versus taking the same bumps with a PIV and getting at least 7-10fps for each clock speed bump. Going from a Celeron D 2.66 > 3.2 will likely get you 3-8fps, going from a PIV 2.66 > 3.2 will likely get you in excess of 30fps. A Celeron D at 3.2 is still going to be a bottleneck for the 6600GT or better.

OC'ing increases the FSB, which helps bandwidth. Which is what having a smaller L2 hurts. It can't completely make up for it, but it helps.

And obviously an OC'd P4 > OC'd Celeron

My point is OC'ing a Celeron D helps it to be competitive with a stock P4. (and many of them are hitting in the mid 3Ghz range... 3.2-3.5+ Ghz fairly easily) So bang for buck its very good.

They will always be slower than a P4 clock for clock. But if you can OC a Celeron D to say 3.4Ghz for $70-80 and its about equivalent to maybe a 3.0Ghz P4 thats a pretty good deal in my book.
(depending on benchmark, etc.)

 

[SodaCose]

What kind of motherboard do you have SodaCose? Do you plan on upgrading just the CPU or the motherboard/CPU/ Possibly RAM?

I have a Biostar P4VTG; it's a piece o' crap: no vCore manip, can't lock agp/pci, skt 478, probably due for an upgrade. So, probably the mobo and cpu.

but is there a way to flash my Biostar to enable vCore/agp-lock?

 

[Met-AL]

All they are is Pentium cores that didnt make the grade to become pentiums...so they disabled HT and Lasered the cache off. To hell with that.

They use a laser? Hmmm......

 

[pxc]

They use a laser? Hmmm......

Not necessarily. Intel is able to control a lot about the processor (multiplier(s), voltage, bus speed, EDB, EM64T, HT, etc) by blowing "fuses" on the core (older) and using the still uncracked software method (newer).

In the past they had disabled the x87 unit in the 486DX to make the 486SX by destroying it with a laser.

 

[Scali]

They use a laser? Hmmm......

Not necessarily... It's cheaper to adapt the design and remove the cache and HT altogether. That way you reduce the die size, so you can get more dies out of the same wafer, and also get higher yields.
So some Celerons are faulty Pentiums, while others are entirely new cores. Same goes for AMDs Durons and Semprons.
I'm not sure in which category the current Celeron D's fit.

 

[chrisf6969]

Not necessarily... It's cheaper to adapt the design and remove the cache and HT altogether. That way you reduce the die size, so you can get more dies out of the same wafer, and also get higher yields.
So some Celerons are faulty Pentiums, while others are entirely new cores. Same goes for AMDs Durons and Semprons.
I'm not sure in which category the current Celeron D's fit.

Source for this information?

B/c to my knowledge all RECENT Celerons... P3(coppermine) on were all just the same die with the extra cache disabled by whatever method.

Back in the day the early celeron with integrated cache was a different die, than the P2's with off-die L2 cache. But thats ages ago.

 

[Scali]

Source for this information?

B/c to my knowledge all RECENT Celerons... P3(coppermine) on were all just the same die with the extra cache disabled by whatever method.

Back in the day the early celeron with integrated cache was a different die, than the P2's with off-die L2 cache. But thats ages ago.

I have as much source as you do
As I say, I don't know what the current Celerons are, but I know that both Intel and AMD have used separate dies in the past... Which you apparently agree on, so I don't have to bother looking for the ancient article where I read it

I did bother to look up the Celeron D at least, and that one (and apparently its predecessor) are indeed disabled P4s... according to this article: http://www.xbitlabs.com/articles/cpu/display/celeron-d_2.html

 

[chrisf6969]

Not necessarily... It's cheaper to adapt the design and remove the cache and HT altogether. That way you reduce the die size, so you can get more dies out of the same wafer, and also get higher yields.
So some Celerons are faulty Pentiums, while others are entirely new cores. Same goes for AMDs Durons and Semprons.
I'm not sure in which category the current Celeron D's fit.

Well, your original post about different dies... is misleading and not true.
Almost all cheaper processors have been cut down versions of the more expensive mainstream. There has never been a cheaper version with the HT and cache just removed.

In fact EVERY P4 has had HT, just they never enabled it until the 3.06.

While it would save die space, they would need to create another mask which is expensive to have 300mm die mask made AND redesign the chip, bug test it, etc...
Instead of one complete design, that plans ahead to have components that can be disabled.

Also there is always a percentage of chips which will have faults in their cache. So if they didn't used them for cut down processors like the Celeron or Sempron they would have to throw them away. Then if they had a different die for the cheaper processors and they had faults.... then those chips would be useless.

Things are moving towards a dynamic software based cache enabling/disabling. Look at the new Itaniums with 24Mb of cache. Their cache and cache errors will be monitered and controlled (disable broken parts) with software. Down the line when we might have Pentium 6's with 8mb of cache, etc... their cache might be controlled the same way.

 

[Scali]

Well, your original post about different dies... is misleading and not true.
Almost all cheaper processors have been cut down versions of the more expensive mainstream. There has never been a cheaper version with the HT and cache just removed.

Yes, we know that now, I said I wasn't sure... I corrected myself.

In fact EVERY P4 has had HT, just they never enabled it until the 3.06.

Every Northwood yes... Not sure about Willamette.

While it would save die space, they would need to create another mask which is expensive to have 300mm die mask made AND redesign the chip, bug test it, etc...
Instead of one complete design, that plans ahead to have components that can be disabled.

That depends on the volume too. AMD chose to use a separate die for its Durons, probably because they sold so much more of them than of the Athlon.

Also there is always a percentage of chips which will have faults in their cache. So if they didn't used them for cut down processors like the Celeron or Sempron they would have to throw them away. Then if they had a different die for the cheaper processors and they had faults.... then those chips would be useless.

Yes, but if demand is far higher than the amount of faulty chips, it'd be a shame to waste all that wafer-space.

 

[dnas]

Just for the Record I would Never use a Celeron.

I dont care, I would rather have a lower clocked P4 than a Celeron.

Cache is important. It improves not only the performance of programs, but overall system responsiveness and smoothness. Also the Lack of hyperthreading. HT may not be a big deal to alot of people, but on the Netburst architecture, it dosent hurt to have it.

Celeron is garbage. All they are is Pentium cores that didnt make the grade to become pentiums...so they disabled HT and Lasered the cache off. To hell with that.

Ok, answer this: which processor would perform better, and 2.4G Celeron D or a 2.0G 0.18 micron Pentium 4??

 

[chrisf6969]

Ok, answer this: which processor would perform better, and 2.4G Celeron D or a 2.0G 0.18 micron Pentium 4??

Looks like possibly you're stock on a 400FSB and tight budget by your choices??

Why not 2.8Ghz Celeron 400FSB $108 vs. 2.0 P4 $119? (that would be about a draw - depending on application)

If that was a Celeron D (instead of crappy NW Celeron 128Kb) it would beat the P4 in pretty much every application. So if you're not stuck on a 400FSB, get a Celeron D their cheap and pretty fast, and overclock like crazy for the price.

If you're not on a budget, and just want an easy plug & play upgrade, there's a "secret" P4 2.8Ghz with 400FSB available here - $220. Most people thought the highest P4 made was the rare 2.5 and 2.6Ghz.

 

[lithium726]

i think he was referring to the fact taht USMC said that he would take a pentium over a celeron in any case.

 

[chrisf6969]

i think he was referring to the fact taht USMC said that he would take a pentium over a celeron in any case.

Well in that case how about a 478 pin P4 Williamette 1.6Ghz

or even worse....

a 423 pin P4 1.3 Williamette on SDram!

vs. 3.06Ghz Celeron D

Blanket statements like "P4's rule, I'd never buy a Celeron" fuel the ignorance of the average joe. (non-tech person) I had a stupid friend who wanted me to build him a PC, and he would have rathered a Pentium 3, than a 775pin Celeron D (OC'd at that!). Ignorance made him think Pentium > Celeron all the time.


Williamette's had a 256 cache, were on generally crappy mobos with Sdram, etc... but ignorant people would get all excited b/c its a "P4"!!
I'd rather a 1.2Ghz Pentium 3 than any 423pin socket P4 system with sdram. B/c in that case a P3 > P4, etc.

 

[dnas]

i think he was referring to the fact taht USMC said that he would take a pentium over a celeron in any case.

Yes, that's what I meant. The 2.0G 0.18 micron Pentium 4, it has a 400MHz fsb, 256k of L2.

The 2.4G Celeron D has 533MHz fsb, 256k of L2, plus the multimedia extensions.

The Celeron D would blow that P4 out of the water, so saying "Just for the Record I would Never use a Celeron. I dont care, I would rather have a lower clocked P4 than a Celeron."

In this case, it's a lower clocked P4 vs a celeron D, and clearly the Celeron D is a much better choice.