celeron-celeron D

[demons9872]

ok guys i have a small pc for just playing with it has a 1.7ghz p4 celeron in it i wanna get a new celeron d the oine im looking @ is 2.93ghz thinks its worth the 90 bucks to upgrade to that and o/c a lil thanxs

 

[xonik]

The Celeron D has signficantly higher performance per clock, and as long as your applications or personal interest justifies it, go for it. At least check out some benchmarks before you jump on it, though.

 

[pxc]

Unless your motherboard supports 533MHz/800MHz Prescott CPUs, a Celeron D isn't going to work in it.

But yeah, the Celeron D CPUs are great.

Original Celeron (Willamette core 0.18u): 400MHz FSB, 128KB L2, overclocked very poorly
Celeron (Northwood core 0.13u): 400MHz FSB, 128KB L2 cache
Celeron D (Prescott Core 0.09u): 533MHz FSB, 256KB L2 cache

 

[Dan_D]

Unless your motherboard supports 533MHz/800MHz Prescott CPUs, a Celeron D isn't going to work in it.

But yeah, the Celeron D CPUs are great.

Original Celeron (Willamette core 0.18u): 400MHz FSB, 128KB L2, overclocked very poorly
Celeron (Northwood core 0.13u): 400MHz FSB, 128KB L2 cache
Celeron D (Prescott Core 0.09u): 533MHz FSB, 256KB L2 cache

You summed it up. A older mainboard that came out around the time of the Pentium 4 1.7 is not going to be able to do it.

 

[BigBadBiologist]

yeah, you might want to figure out what mobo you have before proceeding any further.

 

[demons9872]

my mobo does support it i have have a 2.6 in it now @ 3.3ghz its pretty fast but thought the 533 fsba nd cache would make a difference

 

[Talonz]

ok guys i have a small pc for just playing with it has a 1.7ghz p4 celeron in it

my mobo does support it i have have a 2.6 in it now @ 3.3ghz

Care to explain?

 

[pxc]

Care to explain?

I guessed that he bought a NW Celeron 2.6GHz, but who knows.

 

[kadaj]

the extra cache has minnimal effect on performance, its the extra FSB step(which translates to higher memory bandwidth) that is the real enhancer.

fast but thought the 533 fsba nd cache would make a difference

Prescot Celeron's outpeform North Wood Celerons usualy by 30-60%

 

[xonik]

the extra cache has minnimal effect on performance, its the extra FSB step(which translates to higher memory bandwidth) that is the real enhancer.


Prescot Celeron's outpeform North Wood Celerons usualy by 30-60%

That's not true at all. Look at the difference between the Pentium 4 'A' family and its Celeron equivalent,

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=1927&p=5

Note that while the Celeron has 128 kB of L2, they are otherwise identical, including FSB, and look at that staggering difference.

 

[kadaj]

That's not true at all. Look at the difference between the Pentium 4 'A' family and its Celeron equivalent,

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=1927&p=5

Note that while the Celeron has 128 kB of L2, they are otherwise identical, including FSB, and look at that staggering difference.

prescots incure a 4x latency hit on their cache compared to Northwoods, and I dont have to mention the vastly extended pipeline.
Thus what I orginaly said, the extra cache on the Prescot Celerons has minnimal effect.

 

[xonik]

On the contary, the increased number of cache misses means that more cache is needed to cover these issues.

Consider this: the Celeron D has a 33% faster front side bus frequency, yet it consistently performs 20-25% better per clock. With a 4X cache latency hit, as you call it, the Celeron D's performance is still only peforming 5-7% slower than its theoretical performance increase, yet even the Northwood B didn't perform 33% better than the Northwood A of identical architecture but different FSB--it was more like 20%, if I recall correctly.

AnandTech has this to say about the Celeron D vs. Northwood Celeron:

The improvements in Prescott's core weren't enough to help it keep up with Northwood as a Pentium 4, even with a double-sized L2 cache. With a 128kb cache, much of Northwood's strong points were ripped away, causing plenty of costly pipeline stalls as we mentioned in last year's budget CPU roundup. With the Prescott based Celeron, Intel's architectural enhancements aimed at avoiding pipeline stalls really had a chance to shine. This is due in no small part to the increased returns on performance when smaller caches are doubled in size.

It's clear that at this L2 cache size, Prescott is able to avoid enough potential pipeline stalls that the increased impact of refilling a 31-stage pipeline is much less significant than Northwood's constant struggle to keep itself busy. It's hard to tell how much of this is due to the improved branch prediction and scheduling as opposed to the fact that the extra 8kb of L1 cache that the Prescott has is a more significant percentage of the overall cache size on Celerons than Pentiums, but either way, the performance advantage over Northwood is there.

Source:

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2093&p=14

 

[kadaj]

exactly, but in theory 256kb on a prescot is equivilent to 128kb on a northwood. Its still 1/4 of the size of the Pentium 4(latency on NW processors is arround 1T and on prescots its about 4-5T), and with a 50% longer pipeline, and slower cache, its not going to have a vastly noticalbe effect(even though its doubled). Espicaly when you consider Prescot core's dont leave Northwood cores in the Dust until about 3.5-3.6ghz. And the Celeron D is still sub 3ghz.
Memory bandwidth is just as important as cache to netburst processors, and a 33% increase from a horribly slow 100mhz goes a long way, but when you add a mear 128kb of slower cache to a longer pipelined arch, its not going to have as noticable effect as the FSB would.
Celeron D has the same problems as the NW celeron had.. they're just hidden behind a faster FSB.

 

[Nitro16]

What you need to remember is that with cache size (and indeed FSB Speed) you eventually hit a point of diminishing returns. Although the 256Kb L2 Cache on the Celeron-D is a quarter of the amount on the Prescott (as with the Northwood P4 / Northwood Celeron), the jump from 256KB to 1MB is always going to be much less than 128Kb to 512KB.

Of course, this is entirely dependent on the design of the processor, look at AMDs Sempron/Athlon 64 Clawhammer, the difference between 256KB L2 and 1MB L2 there is minor. Its easy to say theoretically you will get a certain speed increase, but that is never the case. With Netburst, the core is so badly crippled at 128KB L2 that the increase to 256KB provides huge increases.

As for FSB speed, going from 400MHz to 533MHz does provide an increase, but its more in the range of 7-12%.

 

[markku]

also, what crippled the original 128k celeron, was its L2 associativity being 2-way, while it's 8-way with normal p4's.
mobile celeron has 256k l2 cache, 2-way associativity, but it's actually not much better than its 128k desktop counterpart. it's better, but not much.
but celeronD must have at least 4- or 8-way..
also doubled L1 cache on D helps /me thinks

 

[pxc]

From a different page of the same AnandTech article listed above:

They tested a NW Celeron and Celeron D at the same clock speed and FSB speed (Celeron D lowered to 400MHz FSB), leaving the only difference the cache sizes (L1 and L2 doubled in size) and changes in the prescott core (improved branch predictor, improved scheduler, improved execution core, SSE3): http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2093&p=5

"As the following graph shows, over all our benchmarks, the new Celeron D outperforms the Northwood based Celeron even when clock speeds and FSB speeds are equal. This leaves only the core improvements and extra L1/L2 cache as variables in performance difference.
...
For these benchmarks, both the Celeron D and the Celeron were run at a 100 MHz FSB with a 20x multiplier."

The cache and other improvements (despite the 31 stage pipeline and slower caches) make a huge difference in the Celeron D.