Original Link: http://www.anandtech.com/show/913
Intel Celeron 1.7GHz - Pentium 4 Poweredby Anand Lal Shimpi on May 16, 2002 4:04 AM EST
- Posted in
It was the first Intel Celeron processors clocked at 266MHz that took the overclocking world to new levels. In the past, being able to attain a 50% increase in clock speed through overclocking was unheard of; however the maturing of Intel's manufacturing processes as well as the deliberately low clocking of the Celeron parts resulted in a performance gold mine for enthusiasts.
Unfortunately it is only the first iteration of a new Celeron core that actually promises this sort of overclocking potential. In order to understand why, you have to understand the Intel methodology for applying a single microprocessor architecture to multiple market segments. Let's take the Celeron in particular; once a higher end desktop core has been in production for a significant amount of time and yields have reached a certain level, Intel readies the core for use in the Celeron line. The core isn't simply branded as a Celeron without any modifications, generally speaking Intel performs one or more of the following alterations to the core:
- Lower L2 cache size/speed
- Lower core clock speed
- Lower FSB frequency
The beauty of this method for overclockers however is that in the event that Intel limits the core by reducing its clock and/or FSB frequency then the chips are almost guaranteed overclockers to higher levels. For example, when the first Celeron 266MHz processors hit the streets they could easily overclock to 400MHz just by increasing the FSB frequency from 66MHz to the 100MHz frequency of their Pentium II counterparts. Since the cores were relatively similar, it wasn't asking too much of the Celerons to run at the higher clock speed.
While we don't normally cover Intel's Celeron line, whenever there's an introduction of a new Celeron core we're here to exploit it's overclockability. Today Intel is introducing the first Celeron processors based on a Pentium 4-derived core on a 0.18-micron manufacturing process. The use of the 0.18-micron manufacturing process is a bit puzzling (you'll understand why later) but the end result is that an overclocked Celeron could easily become a poor man's Pentium 4.
Let's find out a bit more about the architecture of the chip before diving into a performance investigation, shall we?
Half a Willamette
The new Celeron core is based on a 128KB L2 version of the original Willamette core that the Pentium 4 debuted with in November of 2000. Unlike previous-generation Celerons, the Willamette-128 core is no different architecturally than the Pentium 4's old Willamette core. The cache organization and mapping algorithms are still the same, the only difference is that the Celeron core is only outfitted with a 128KB L2 cache instead of the 256KB cache present on the original Pentium 4.
Only having a 128KB L2 cache increases the Celeron's dependency on a high-speed memory bus. Luckily, the processor will work just fine on an 845 or 850 platform both of which offer a significantly larger amount of memory bandwidth than the i815 that the older Celerons were often paired with. With only a 128KB L2 cache, the new Celeron would appreciate the higher bandwidth i850 even more than the Pentium 4. Keep in mind that the short-lived Tualatin based Celeron processors had a 256KB L2 cache; there are situations where the new core can be outperformed by its predecessor.
Keep in mind that the first generation Pentium 4s didn't always fare so well against the Pentium III and especially the Tualatin based processors. Now with a smaller L2 cache, it will take even more for the Celeron to do well. For information on the NetBurst architecture behind the Celeron take a look at our one page explanation of its strengths and weaknesses.
The new Celeron also uses the now "old" 100MHz quad-pumped FSB, delivering a total of 3.2GB/s of FSB bandwidth. This won't be a limitation for quite a while as the Celeron won't be ramping up to clock speeds nearly as high or nearly as quickly as with the Pentium 4. The quad-pumped FSB fixes an age-old problem with the Celeron - a lack of FSB bandwidth, and when paired with DDR memory on an i845 or even RDRAM on an i850 the issue of low main memory bandwidth is nonexistent as well.
The introductory speed of the new processor is 1.7GHz running at 1.75V. Remembering that this is a 0.18-micron core helps explain the reason behind the high core voltage. Speaking of which, the new Celeron is only being introduced on the 0.18-micron process but will undoubtedly migrate down to 0.13-micron as quickly as possible. Intel still has a good amount of 0.18-micron manufacturing capacity and until those fabs get converted over for 0.13-micron production we won't see a significant push for a smaller Celeron (the 128KB L2 cache already makes it very cheap to produce).
Although the use of Intel's 0.18-micron severely limits the overclocking potential of the chip, remember that Intel got the Pentium 4 up to 2GHz on this process and that was with twice as much cache and significantly more transistors.
It's safe to assume that after the Celeron hits 2GHz Intel will transition it to a derivative of the Northwood core, most likely with a 256KB L2 cache. But until that time we have to deal with the warm-running 0.18-micron Celeron at 1.7GHz.
Since the new Celeron isn't made on a 0.13-micron process we didn't have any delusions of being able to overclock it to 3GHz marks, but we were able to do quite well with it.
We performed the simplest overclock possible by increasing the FSB from 100MHz to 133MHz which increased the clock speed from 1.70GHz to 2.26GHz. In order to maintain stability at that speed we had to boost the core voltage from 1.750V to 1.850V, an increase of less than 6%.
The overclocked CPU did not require any additional cooling aside from Intel's retail heatsink/fan; the CPU was an OEM part bought off the web from Newegg.
Unlike our usual Pentium 4 test bed, we chose an 845 with DDR SDRAM for the Pentium 4 and new Celeron to properly represent the target market for the new CPU.
Windows XP Professional Test Bed
Intel Celeron 1.2GHz
Intel Pentium III 1.2GHz
Intel Pentium 4 1.7GHz
Intel Pentium 4 2.0GHz
AMD Duron 1.3GHz
AMD Athlon XP 1600+ (1.40GHz)
EPoX 8K3A+ - VIA KT333 Chipset
Gigabyte P4 Titan DDR - Intel 845 Chipset
1 x 256MB DDR333 CAS2.5 Kingston DIMM
80GB Maxtor D740X
|Video Cards (Drivers)||
NVIDIA GeForce4 Ti 4600 (28.32)
Before we dive into the benchmarks we took an informal survey of "street" prices on the new Celeron CPU from various online vendors:
You can use these prices as a gauge for what you're paying for in terms of performance with the processors compared here.
Internet Content Creation & General Usage Performance
With this review we continue to use SYSMark 2002; SYSMark 2002 can be considered to be a much more memory bandwidth intensive version of the Winstone tests. The benchmark is split into two parts, Internet Content Creation which deals with content creation applications (Photoshop, Dreamweaver, etc...) and Office Productivity which is more general usage oriented (Word, Excel, Netscape, Anti-Virus, etc...).
The 2002 update changes things around a bit; first of all the benchmark's total scores are arrived at differently than in the 2001 benchmark. Windows Media Encoder no longer accounts for close to half of the Internet Content Creation test, rather only about 10%. There is also no need for a special Athlon XP SSE patch as the 2002 suite uses a version of the encoding dll that properly detects SSE support on all Palomino cores as well as Pentium 4 cores.
The rest of the benchmark is much more evenly distributed and it is much more memory bandwidth intensive than the old benchmark. The Internet Content Creation tests on average use about 600MB/s of bandwidth vs 300MB in SYSMark 2001. The Office Productivity tests are still stuck at around 580MB/s of memory bandwidth.
For more information on the tests and the applications used consult this whitepaper provided by BAPCo.
Internet Content Creation is clearly not the target market for the Celeron but it does quite well, only losing out to the Pentium 4 1.7GHz by 7%. In any situation where you're constantly streaming data from main memory to the CPU without reuse, the benefits associated with a larger L2 cache are diminished. Remember that one of the major principles of caching is that data that is requested once will probably be requested again, in the case of most streaming applications this is not the case and thus we do not see such a large penalty associated with the Celeron's smaller L2 cache.
When overclocked to 2.26GHz the processor is almost as fast as the Pentium 4 2.0GHz.
The most important performance metric to look at here is the Office Productivity suite which closely mimics daily computer usage; this is where business and home users alike will notice the true performance differences between these processors. Here the Pentium 4 1.7GHz is 11% faster than the new Celeron which itself can barely outperform the 1.2GHz Pentium III. Although the Duron provides healthy competition for the new processor it is the similarly priced Athlon XP 1600+ that ends up offering the most bang for your buck here to those that are unwilling to overclock.
When pushed to 2.26GHz the new Celeron once again rises to the top, bested only by the 2GHz Pentium 4.
Media Encoding Performance
What once was a very CPU intensive task is now fairly trivial. Because of the streaming nature of MP3 encoding, having a larger cache doesn't necessarily result in a tangible increase in performance. The reason we continue to stress MP3 encoding as a CPU benchmark is mainly because of the fact that MP3 encoding usually does play a role in larger projects such as MPEG-4 video encoding where you're ripping audio as well as video.
Performance when encoding audio is pretty much a function of clock speed for the Celeron and Pentium 4 since the architectures are identical (minus the cache size differences). Because of this, the overclocked Celeron can take the lead while the 1.7GHz part is on the heels of the Pentium 4. It is worth noting that since MP3 encoding is generally a very floating point intensive process the older Pentium III and Celeron processors do just fine due to their relative strength when it comes to FP tasks.
3D Rendering Performance
Next we have our usual two 3D rendering tests. We'll start off with rendering the first frame of the Waterfall.max scene (provided on the 3DSMAX CD) at 1024x768:
The vast majority of rendering tasks under 3D Studio MAX are not blessed with SSE2 optimizations so you end up paying the ultimate price when you attempt to render scenes on lower clocked NetBurst based processors. Because of Intel's decision to sacrifice raw x87 floating point execution power in favor of beefing up SSE2 execution potential, applications such as 3D Studio MAX don't perform all too well on the lower clocked processors. As we've seen with the newer Pentium 4s however, as clock speeds increase the performance scales quite well and you end up having very competitive processors. Unfortunately for the Celeron, it will be a while before it enjoys the high clock speeds of the current Pentium 4.
When it comes to raw FP execution power the Athlon has always been king and at a meager 1.4GHz it's spreading its wings wide.
The new Celeron does better under Maya than under 3D Studio MAX but still tends to lag behind the Pentium 4 by around 10%; overclocking helps level the playing field a bit.
3D Rendering Performance using SSE2
While 3D Studio MAX is SSE2 optimized, the level of optimization is nowhere near what NewTek reported with Lightwave upon releasing version 7.0b. The performance improvements offered by the new SSE2 optimized version were all above 20% using NewTek's supplied benchmarking scenes.
Even the SSE2 support of the new Celeron doesn't help it out under Lightwave 7.5; a larger cache is clearly necessary to help it out here.
3D Gaming Performance
When it comes to most 3D games there's generally very little performance to be found by heavily optimizing for SSE2 or 3DNow! on either of these processors and thus the performance is mostly dependent on the overall platform (e.g. FPU capabilities, chipset, memory latency/bandwidth, cache latency/bandwidth, etc...).
We'll start off with our favorite 3D gaming benchmark - the Unreal Performance Test 2002. For an explanation of what this test is and why it is so significant, be sure to read our 15-way GPU Shootout that we used to introduce the test. In short, the benchmark uses the current build of the Unreal Engine (that will power games such as UnrealTournament 2003 and Unreal II) and serves as a great indication for future performance in games that use the engine.
Once again we see that although it's a member of AMD's current flagship line, the Athlon XP 1600+ does quite well as a value processor. The small L2 cache of the Celeron hurts its performance in next-generation games such as Unreal Tournament 2003; if you look at any of our high-end desktop CPU comparisons you'll realize that without the 512KB L2 cache of the Northwood core, the Pentium 4 does not do well in many of these gaming tests.
Even overclocking the 1.7GHz Celeron to 2.26GHz can't bring the Celeron up to speed with a 1.7GHz Pentium 4.
Even under today's latest games the Celeron is in dire need of more cache. It's clear that for the money, the Athlon XP 1600+ is once again the best offering here.
By far the poorest showing of its performance comes under Comanche 4 where the new Celeron is unable to outperform the 256KB L2 equipped Tualatin-Celeron running at 1.2GHz.
Now that you've seen the performance of the Celeron in the various applications you'll quickly understand a major reason why it has been launched on a Willamette core (0.18-micron) with a 128KB L2 cache instead of as a Northwood (0.13-micron) with 256KB of L2 cache; the 1.7GHz Pentium 4s are still considered to be relatively new processors and with a 256KB L2 cache the Celeron would easily perform just as well as the Pentium 4s (since they would be the same chip but on a smaller process). In fact, as soon as the Celeron moves to a 0.13-micron Northwood derived core it will be time for all of the older Willamette Pentium 4s to finally say good-bye as they will be undercut by cheaper Celerons.
There are a few situations where the Celeron would clearly benefit from a 256KB L2 cache; the two areas that immediately come to mind are general usage/office productivity and 3D games, both of which are in desperate need of more cache in order to gain better performance on the Celeron platform. With that said, the Celeron in its current shape does perform well as a general use desktop processor and brings an appropriate level of performance at a very cheap price.
For overclockers, the Celeron has a bit of untapped potential but that will truly be unleashed when the 0.13-micron Celerons finally ship. For now, you can expect to hit anywhere between 2 and 2.3GHz on these processors with a bit of voltage tweaking. Keep in mind that at 2.3GHz these Celerons will run extremely hot so you may need to revamp your cooling methods at such high speeds.
If it weren't for AMD then there would be no choices outside of the new Celeron but luckily we have the Athlon XP. AMD's Duron is pretty much useless right now, it has never been a significant revenue generator for AMD and its days are clearly numbered. Instead the Athlon XP 1600+ actually ends up filling in the gap quite nicely; where the Celeron does well, the Athlon XP 1600+ does equally as well, and where it does poorly, the XP 1600+ takes the gold.
The new Celeron ends up being a logical next-step for Intel and a good one at that. The processor needs a 256KB L2 cache and it will eventually get one but it won't be able to trample all over the Duron, instead it will be forced to compete with the lower end of AMD's Athlon XP line. A 256KB L2 Celeron up against older Athlon XPs while AMD's Hammer takes the high-end, this should sound a lot like the Pentium 4 vs. Athlon comparison that has been going on for quite a while. If you remember, before the Pentium 4 got a 512KB L2 cache it was losing out to the Athlon XP on a fairly consistent basis. A price and clock speed war will determine how the future of the Celeron and AMD's low-end Athlons turns out.