3GHz P4-Celeron unfairly getting bad rap?

[MadRat]

Review here

I noticed the 3GHz Celeron has no L2 cache in all of the pics. WTF, they disabled the L2 cache to reach 3GHz then bitch about the results?? The tests were also conducted with underclocked DDR333 running at 150MHz/DDR300. The 3GHz machine is not hardly the king of performance here when its doubly crippled. A thorough review would have used a functional L2 cache and also shown the difference between SDR, DDR, and RDRAM settings at a variety of ratioes in order to give an accurate picture of the Celerons capabilities. As it is these reports are pretty lame.

 

[Pink0]

No the rap is well deserved. They didn't disable any cache in tomshardware's reviews and it still performed like a POS being beaten by the 1.6a. Tomshardware isn't exactly pro-amd so I trust them when They say it's a POS.

 

[mechBgon]

It occurred to me that with the 2.0 Cel4 being the first 0.13-micron Celeron of its kind, it's possible the ID software doesn't know what it is. I can't read Russian, but I tried converting the site to English and it doesn't look like they purposely disabled the L2... or am I incorrect? Do they come out and say they disabled the L2 themselves?

 

[MadRat]

The NW-Celeron is being made on a process that is prior to the latest NW-P4 core. Its unthinkable it could hit 3GHz being that the L2 cache wouldn't of made it without exotic cooling. Even kryogenically-cooled P4 parts wouldn't regularly hit an L2-enabled 3GHz without the new stepping. Most likely the latest stepping has an improved L2 cache design allowing for a stable cache at high speeds.

Their own screenshots are pretty conclusive to being a disabled L2 cache being that the Celeron microcodes for BIOS tags have been out for quite some time before the NW-Celeron hit the scene.

 

[Elminster]

the 2.0 celeron is on the new stepping.

 

[mechBgon]

MadRat, it sounds like you expected the Celeron to be closer in performance to the P4. Ok, here's a question, then: what is the difference between a 2.0A Northwood P4 and a 2.0 Celeron? It really is simply the amount of cache, right? So does anyone here have a 2.0A Northwood that they would do some tests on, to see where it stands compared to the numbers the reviewers got for the Celeron 2.0@2.0? Just pop into BIOS, switch off the L2, and run 3DMark2001SE, and we'll have something to go by.

I was looking at the results that Tech-report.com got with the Cel 1.7, and it was giving them about 50% of the performance of a 1.7A P4 in some benchmarks such as Comanche 4, while doing well in other benchmarks, so that lends some plausibility to the results in the review that you're critiquing. Check out the graphs on the gaming benchies here.

 

[MadTom]

However, I hope that Intel will give the Celeron 256KB of L2. Especially when all P4 NW will be rated for 533MHz bus. Then it could be "the" Celeron. Still, for ripping DVD's it's great (at 2.66 of course).

 

[adrian12]

Originally posted by: MadTom
However, I hope that Intel will give the Celeron 256KB of L2. Especially when all P4 NW will be rated for 533MHz bus. Then it could be "the" Celeron. Still, for ripping DVD's it's great (at 2.66 of course).

at that price?

 

[MadRat]

<<At that price?>>

Thats what I was thinking, too.

Durons are going for under $40. XP1600+ is delivered to your door for $55.

 

[AkumaX]

they're still 128kb

 

[MadRat]

The 3GHz Celeron looks good at THG. It looked sad in the review with the L2 cache disabled.

 

[First]

Originally posted by: MadRat
The 3GHz Celeron looks good at THG. It looked sad in the review with the L2 cache disabled.

MadRat, no where in the THG review does it mention that the Celeron's cache(s) were disabled. Where are you getting these ideas?

 

[mechBgon]

MR is referring to the Russian review, where a CPU ID program shows a blank box for the L2 cache.

 

[First]

Originally posted by: mechBgon
MR is referring to the Russian review, where a CPU ID program shows a blank box for the L2 cache.

Ah, I see. But still, it doesn't change the fact that a 1600+ or a 1.6A would significantly outperform a 2GHz Celeron for less than or equal to $.

 

[MadRat]

But if you need a cheap high performance multimedia chip for encoding the 2GHz-to-3GHz overclock of the Celeron puts the cost benefit into an unmatched category.

One should consider that some people actually do use one processor exclusively for one task.

On the same token I wouldn't dare touch a P4 in alot of the tasks where the Celeron performed poorly simply because the P4 just isn't strong in that area as it stands.

 

[DragonMasterAlex]

well I just bought the Soyo Dragon LITE (RAID, 533Mhz FSB, lotsa goodies!) and Celeron 2.0 for $159 from Fry's tonight, with a $10 rebate on the board. It works great with my PC2100, I'm using it to reply to you now, and coupled with my Geforce4 and a pair of 10 gig 7200 RPM drives in RAID, it flat out rocks, IMHO. For a budget price it kicks ass, and now when the P4's get to a reasonable price level, I have a quick and dirty upgrade path without all the work

DMA

 

[ScrewFace]

Intel should have continued with the Tualatin Celerons instead of Pentium4 Celeron's as the Tualatins beat the snot outta Pentium4 Celerons and overclock well.

 

[Jhhnn]

P4's are heavily dependent on cache and memory bandwidth. That's pretty much obvious when you consider that a p4 needs twice the cache and 25% higher clock speeds to compete with the K7's across the board. With very long pipelines, the penalty for a cache miss is large, and there's a lot of misses with only 128K cache.

The earlier p3 design didn't have those very strong dependencies, so the s370 128K cache celerons didn't suffer as badly. Other than for special purpose machines, the new celerons are pretty much woofers- At the same money, a K7 system will slap one into the weeds.

Intel is obviously dumping the s370 format RSN, and depending on brand name recognition marketing and magic numbers to carry sales. I probably would have done the same myself, there's really no reason to think it won't work out favorably.

 

[mrman3k]

Come on, pay $20 more and get a 2.4B P4!

 

[Jhhnn]

I'm not quite sure where that price comparison came from- The 2gig celeron is $109, the 2.4gig P4 is $186, and an xp2000 is $105 - all retail packaged, three year warranty, straight from pricewatch....

 

[DragonMasterAlex]

well I got the 2.0 celly AND the Dragon Lite board from Soyo for $159 with a $10 MIR from Fry's, and I am so far quite impressed; it whips the piss out of my Athlon 750 or Duron 1200 by a longshot, and hey, when the P4's get reasonable, all I have to do is drop in the CPU So it gives me a nice upgrade path, which in itself is worth a bit!

DMA

 

[MadRat]

Is there an echo in here?

 

[dannybin1742]

the review at THG showed it was actually pretty good, but you had to run it at 3ghz or higher, i wouldn't mind building a box with it to primarily do divx ripping

 

[Zap]

Intel should have continued with the Tualatin Celerons instead of Pentium4 Celeron's as the Tualatins beat the snot outta Pentium4 Celerons and overclock well.

I don't think they could have, and they certainly did not want to. Anyone remember how much problems Intel had with going over 1GHz with the Coppermine P3? They shrunk it and added prefetch to make the Tualatin, but it seems to max out at 1.5GHz or so. Check the various overclocking databases. Not counting the ones using exotic cooling, with air cooling (required for the CPUs to sell to the general public) hitting 1.6GHz is difficult with a Tualatin. 1.5 seems okay. Pretty telling that Intel hit 1.4 and no higher, eh? I don't think the process can handle any higher _comfortably_.

As for Intel not wanting to... it would be to their benefit to have everything on one platform. Share manufacturing and share a common socket. Also, they know that still, to this day, MHz sells. No matter how powerful a CPU is... if tomorrow AMD could come out with a 500MHz CPU that can spank a P4 3.06GHz at a lower price, would they? Nope, because it wouldn't sell (unless they gave it a PR rating of 3000+). I'm not talking about all us enthusiasts here. I'm talking about the other 99.9% of the buying consumer, which a CPU manufacturer NEEDS the support of in order to stay in business.

I am personally interested in the Celeron 2.0 for a reason besides performance... let's see, run a large Zalman heatsink, drop the voltage... quiet computer? It should still be more powerful than the main "cool computing" competitor, the VIA C3.

Speaking of the Tualatin... I like them. If anyone's read some of my other posts mentioning my various systems, I have a BUNCH of socket 370 setups. A few Tualatin Celerons, P3 Coppermine, Celermines, VIA C3 in systems, even have a bunch of older Celerons (circa 300A-500MHz) sitting around collecting dust as bare CPUs. VIVA SOCKET 370!

 

[Y23KC]

I can tell you right now this processor is not that bad. I use my pretty much exclusively for mpeg2 encoding and it whips the socks off my old 1600+ dually setup. Of course, this is when it's overclocked to 3.1ghz . My 3dmark 2001 score is a little over 8600 w/ gf3 ti200 overclocked to stock gf3 speeds. This score is faster than when I had a 1900+ overclocked to 1860mhz with a 186mhz fsb. I think if you overclock this proc and use some good ram to get some more bandwidth, you should have a DECENT processor for the money. It's definately not as fast as my p4 @ 3.0ghz for gaming and such but would pretty much do what I needed it to do with ease.